i

Abstract

This thesis presents observational data and analysis of the OH masers in the star-

forming region W3(OH). The observation was performed with the VLBA in dual

circular polarization at all four ground state frequencies, with a velocity resolution of

~90 m s–1, a resolution of ~7 mas, and an rms noise in the images of ~15 mJy/beam.

The maser emission is fit by 2-dimensional elliptical Gaussians, which allows

for high precision measurements of position, dimensions and flux of each maser. 276

masers have been detected, and from these 83 Zeeman pairs identified, allowing the

velocity profile and magnetic field structure of the gas to be determined. The bulk of

the masers lie in a north-south distribution that contains a clear velocity gradient and

indicates ordered Keplerian rotation, placing a minimum mass for the central star at

~8 M . Proper-motion calculations support previous work in the literature, and

combined with the discovery of a rotating disc suggest that a cluster of masers not in

the disc may be in an orthogonal outflow. The proper motions suggest there may be

an element of expansion in the rotation – possibly the destruction of a disc is being

observed.

Deviations from Gaussian lineshapes were found to be extremely rare: major

deviations exist for just two cases. Polarization studies indicate that the magnetic field

orientation varies systematically over the field. New correlations are demonstrated

between the degree of polarization and the FWHM of the maser; and the Landé

splitting factor of the transition and the likelihood of finding Zeeman pairs.

In several parts of the region, the masers lie in arcs, where the major axis of

the maser usually lies parallel to the arc; these may represent propagating shocks. A

spectacular arc of maser emission occurs in the far south of the region at both 1665

MHz and 1667 MHz. Such co-propagation of maser frequencies is very rare and

exists only for pairs of lines where one line is 1665 MHz. Analysis of the two-point

correlation function shows that the dimensions of a typical maser cluster are ~600 AU.

ii

Acknowledgements

A huge “thanks” must go to Dr Malcolm Gray – my main collaborator and mentor.

For handing me the kind of dataset that most postgrads could only dream of, and

dragging me kicking and screaming against my will for a spectacular month in the US

– cheers to that man! I only hope I’ve done the data justice. A massive “thanks” goes

to Dr Martin Hardcastle – AIPS God, Unix guru, radio-data wizard, and all round

helpful guy. It would have been that much harder to get finished if he hadn’t been

around to constantly bother.

Thanks as well to the bods down in ‘my first home’ of Chemistry for not

booting me out, even after all these years! I’m thinking of Professor Mike ‘The

Patient’ Ashfold and Dr Peter ‘Always There’ Cox. Special thanks to Peter for

helping me get this thing written by reading it a dozen times! I like to think of myself

as Chemistry’s prodigal son. Hope I did you proud.

Thanks as well to the Physics department, Professor Mark Birkinshaw

especially, for taking me on all these years. I know I’ve taken advantage – as a

chemist, the Laws of Physics never did really apply to me though... A big thanks to

Dr Jeremy Yates and Professor David Field – for setting me on the road to this PhD; a

shame perhaps that the astro-chemistry crew didn’t stay together, perhaps we would

have done great things.

I must also thank Phil Diamond and other long forgotten people at the NRAO

facility in Socorro, NM, for their time and guidance through the long data reduction

sequence. Who knows what would have happened had I been left to wrestle with

AIPS alone…

iii

Declaration

I declare that the work in this thesis was carried out in accordance with the

Regulations of the University of Bristol. The work is original except where indicated

by reference in the text, and no part of the thesis has been submitted for any other

degree. This thesis has not been presented to any other University for examination

either in the United Kingdom or overseas.

Any views expressed in the thesis are mine, and in no way represent those of

the University of Bristol.

SIGNED:

Mark Michael Wright

DATE:

iv

Contents

Abstract ......................................................................................................................i

Acknowledgements...................................................................................................ii

Declaration ...............................................................................................................iii

Contents....................................................................................................................iv

List of Figures ..........................................................................................................ix

List of Tables....................................................................................................... xiiiii

Astronomical Units and Abbreviations ................................................................xv

Dedication .............................................................................................................xvii

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

OH Spectral Line Emission and Masing . . . . . . . . . . . . . . . . . . . . . . 4

2.1 The OH Molecule and its Rotational Spectrum ..............................................4

2.1.1 Angular Momentum in the OH Molecule ......................................................4

2.1.2 Hund’s Cases for OH ....................................................................................6

2.1.3 The Hyperfine Interaction in OH ..................................................................8

2.2 Magnetic Fields and Polarization in OH........................................................10

2.2.1 The Zeeman Effect .......................................................................................10

2.2.2 Selection Rules ............................................................................................13

2.3 Fundamentals of Maser Emission...................................................................14

2.3.1 The Boltzmann Distribution ........................................................................14

2.3.2 Population Inversions .................................................................................15

2.3.3 Collisional Factors......................................................................................17

2.3.4 Radiative Transfer: Optical Depth..............................................................18

2.3.5 Radiative Transfer: Source Function..........................................................19

2.4 Characteristics of Maser Radiation................................................................20

v

2.4.1 The Unsaturated Regime.............................................................................20

2.4.2 The Saturated Regime .................................................................................22

2.4.3 Brightness Temperature ..............................................................................24

2.4.4 Pumping ......................................................................................................25

OH Maser Environments..........................................................27

3.1 Star-Forming Regions......................................................................................27

3.1.1 Star Formation ............................................................................................27

3.1.2 Physical Characteristics of GMCs..............................................................28

3.1.3 Stability of Molecular Clouds .....................................................................29

3.1.4 The Collapse of Molecular Clouds and Birth of Stars ................................31

3.2 Chemical Evolution in Molecular Clouds ......................................................33

3.2.1 Gas-Phase Chemistry..................................................................................35

3.2.2 Grain-Surface Chemistry ............................................................................35

3.2.3 Shock Chemistry..........................................................................................36

3.2.4 Photodissociation Regions ..........................................................................38

3.2.5 The Role of Dust..........................................................................................40

3.3 Interpretation of OH Masers in Star-Forming Regions ...............................41

3.3.1 Physical Location........................................................................................41

3.3.2 Occurrence ..................................................................................................42

3.3.3 Environment ................................................................................................44

Data Reduction ..........................................................................46

4.1 Interferometry ..................................................................................................46

4.1.1 Detection in Radio Telescopes ....................................................................47

4.1.2 Aperture Synthesis.......................................................................................49

4.1.3 Visibility Data .............................................................................................50

4.1.4 Imaging........................................................................................................52

4.1.5 The Stokes Parameters ................................................................................54

4.2 Data Acquisition ...............................................................................................56

4.2.1 AIPS.............................................................................................................56

4.2.2 The Observation ..........................................................................................56

4.3 The AIPS reduction sequence .........................................................................59

4.3.1 Preliminary Flagging..................................................................................59

vi

4.3.2 Amplitude Calibration.................................................................................60

4.3.3 Phase Calibration .......................................................................................61

4.3.4 Bandpass Calibration..................................................................................61

4.3.5 Polarization Calibration .............................................................................62

4.3.6 Image Calibration .......................................................................................63

4.4 Measuring the Maser Emission.......................................................................66

4.4.1 Gaussian Fitting in AIPS (SAD)..................................................................66

4.4.2 Identification of Masers (HAPPY) ..............................................................73

4.5 Data Analysis ....................................................................................................75

4.5.1 Lineshape Fitting.........................................................................................75

4.5.2 Filtering the Masers ....................................................................................76

4.5.3 Zeeman Pairs...............................................................................................77

4.5.4 Demagnetisation..........................................................................................78

4.5.5 Magnetic Field Strengths ............................................................................79

4.5.6 Proper Motion .............................................................................................79

4.5.7 Two-Point Correlation ................................................................................80

Analysis of W3(OH) – 1665 MHz.............................................86

5.1 The Observational Data...................................................................................87

5.2 Comparison with Previous Work at 1665 MHz ............................................98

5.3 Maser Morphology .........................................................................................102

5.4 Maser Lineshapes ...........................................................................................103

5.5 Polarization .....................................................................................................110

5.6 Magnetic field Structure ................................................................................116

5.7 Velocity Distribution ......................................................................................121

5.8 Proper Motion ................................................................................................128

5.9 Statistical Relationships.................................................................................134

5.10 Two-Point Correlation .................................................................................137

Analysis of W3(OH) – 1667, 1612 and 1720 MHz ................ 139

6.1 1667 MHz ........................................................................................................139

6.1.1 The Observational Data ............................................................................139

6.1.2 Comparison with Previous Observations at 1667 MHz............................145

6.1.3 Comparison with Norris et al. (1982) .......................................................146

vii

6.1.4 Maser Morphology....................................................................................149

6.1.5 Maser Lineshapes......................................................................................153

6.1.6 Polarization...............................................................................................154

6.1.7 Magnetic Field and Velocity Structure .....................................................155

6.1.8 Statistical Relationships ............................................................................157

6.1.9 Two-Point Correlation ..............................................................................158

6.2 1612 MHz ........................................................................................................160





6.2.5 Polarization...............................................................................................167




6.3 1720 MHz ........................................................................................................173





6.3.5 Polarization...............................................................................................182


6.3.7 Proper Motion ...........................................................................................184



6.4 Comparisons between the Ground State Lines ...........................................186

6.4.1 Co-Propagating Masers............................................................................186

6.4.2 Polarization...............................................................................................196

6.4.3 Velocity Structure and Magnetic Field .....................................................198

6.4.4 Morphology ...............................................................................................199

Concluding Remarks ............................................................... 201

7.1 Data Acquisition .............................................................................................201

7.1.1 The Observation ........................................................................................201

viii

7.1.2 Data Reduction..........................................................................................201

7.2 Analysis ...........................................................................................................202

7.2.1 Comparisons with Previous Work.............................................................202

7.2.2 Morphology ...............................................................................................202

7.2.3 Lineshapes.................................................................................................203

7.2.4 Polarization and Magnetic Field ..............................................................204

7.2.5 Velocity Structure......................................................................................205

7.2.6 Proper Motion ...........................................................................................205


7.2.8 Co-Propagation.........................................................................................206

7.3 Future Work ...................................................................................................207

7.3.1 Observations..............................................................................................207

7.3.2 Code Enhancements ..................................................................................207

7.3.3 Modelling ..................................................................................................208

Appendix A............................................................................... 209

References................................................................................. 231

ix

List of Figures

2.1 Hund’s cases (a) and (b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

2.2 Energy levels of the two ladders of rotational states of OH. . . . . . . . . . . . . . . 7

2.3 Lambda doubling and hyperfine splitting of OH rotational levels. . . . . . . . . . 8

2.4 Magnetic field splitting of the 1665 MHz transition in OH. . . . . . . . . . . . . . . 11

2.5 Zeeman components of the 2Π3/2, J=3/2 ground state of OH in a magnetic

field. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

2.6 The saturated tubular maser . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .24

3.1 The popular ‘Burning Cigar’ model of sequential star-formation . . . . . . . . . .33

3.2 The state of molecules in the environment around massive YSOs . . . . . . . . . 37

3.3 Diagram of the transition from UCH II region (left) to dense molecular core

(right) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

3.4 The most important reactions of oxygen bearing molecules in PDRs. . . . . . . 40

4.1 The NRAO’s VLBA sites. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

4.2 Diagram of a ‘tied array’ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

4.3 The u,v coverage of the VLBA during the observation . . . . . . . . . . . . . . . . . . 52

4.4 Diagram illustrating the relationship between the Stokes parameters and the

polarization properties of the electromagnetic radiation . . . . . . . . . . . . . . . . . 55

4.5 Variation of Tsys for 1665 MHz LCP throughout the observation . . . . . . . . . . 58

4.6 Plot of amplitude versus u,v distance for 1665 MHz Stokes I channel 64;

W3(OH) only. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60

4.7 Complex bandpass calibration calculated for the RCP feed of the Brewster

antenna at 1665 MHz. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

4.8 The observational beam for 1720 MHz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

x

4.9 The variation of (off source) actual channel noise for 1612 MHz and 1665

MHz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

4.10 The variation of (off source) actual channel noise for 1667 MHz and 1720

MHz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

4.11 Two hypothetical Gaussian sources at separations of 7 (top) and 15 (bottom)

units, showing the effects of overlap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72

4.12 The bin counts (i.e. the undivided correlation) for each box and the

correlation for each box . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83

4.13 The correlation to a random set of sources with different numbers of

iterative averaging of the random field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

5.1 1665 MHz spectra. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .88

5.2 The 1665 Stokes I integrated flux map. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .89

5.3 The 1665 MHz maser spot map. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

5.4 Position angles of 2-Dimensional Gaussian fits to the 1665 MHz masers. . . . 101

5.5 The three most intense masers, found in the northwest of W3(OH) . . . . . . . . 102

5.6 The most intense 1665 MHz maser lineshapes showing slight deviations

from a Gaussian form. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103

5.7 The most intense 1665 MHz maser lineshapes showing Gaussian form. . . . . 104

5.8 Distinctive non-Gaussian maser lineshapes. . . . . . . . . . . . . . . . . . . . . . . . . . . 107

5.9 Contour plots of masers with non-Gaussian lineshapes. . . . . . . . . . . . . . . . . . 108

5.10 W3(OH) maser groups. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109

5.11 Stokes lineshapes of simple 1665 MHz masers with high linear flux

densities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

5.12 The two candidate 1665 MHz maser Zeeman triplets. . . . . . . . . . . . . . . . . . . 112

5.13 A possible 1665 MHz maser π component . . . . . . . . . . . . . . . . . . . . . . . . . . . 112

5.14 The variation of linear polarization with RA and DEC; and the variation

of circular and total polarization with DEC. . . . . . . . . . . . . . . . . . . . . . . . . . . 114

5.15 The Projection of the angle of linear polarization for masers with over 30

mJy/beam of linear flux intensity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115

5.16 1665 MHz Zeeman pairs in W3(OH), with maser groups labelled. . . . . . . . . 116

5.17 Variation in magnetic field strength with RA and DEC in W3(OH). . . . . . . . 119

5.18 Magnetic field strength of Zeeman pairs versus their ‘Zeeman ratio’. . . . . . . 121

5.19 Variation in maser velocity with RA and DEC in W3(OH) . . . . . . . . . . . . . . 122

xi

5.20 Variation in demagnetised velocity with RA and DEC in W3(OH) . . . . . . . . 123

5.21 Top, the relative proper motions measured by Bloemhof et al. (1992).

Bottom, the relative proper motions measured in the current data. . . . . . . . . . 129

5.22 Histogram of the pairwise change in separation of the 1665 MHz masers

converted into a velocity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130

5.23 Apparent change in separation of two masers on the sky as they rotate first

towards then away from the observer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130

5.24 Proper motions of 1665 MHz masers in W3(OH) with varying reference

velocities added . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132

5.25 Left, showing the 1665 MHz proper motions with a reference motion of

(4,–1) km s–1 (RA, DEC) added. Right, model proper motions in a rotating

disc diameter 6000 AU, inclined to the line of sight by 10° . . . . . . . . . . . . . . 133

5.26 Charts showing no relationships between various maser properties . . . . . . . . 135

5.27 Charts showing likely relationships between various maser properties. . . . . . 136

5.28 Top, the 1665 MHz masers and a random selection of points in the same

space. Bottom, the two-point correlation of the 1665 MHz masers . . . . . . . . 137

5.29 The two-point correlation of the 1665 MHz masers . . . . . . . . . . . . . . . . . . . . 138

6.1-1 1667 MHz spectra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140

6.1-2 The 1667 Stokes I integrated flux map . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141

6.1-3 The 1667 MHz maser spot map. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142

6.1-4 1667 MHz masers from the current new data overlaid with data from

Norris et al. 1982. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147

6.1-5 The 1667 MHz arc feature. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150

6.1-6 Position angles of the masers in the 1667 MHz arc feature. . . . . . . . . . . . . . 151

6.1-7 The 1667 MHz arc feature with minimised circular fit. . . . . . . . . . . . . . . . . 152

6.1-8 Lineshapes of masers over 2 Jy/beam at 1667 MHz . . . . . . . . . . . . . . . . . . . 153

6.1-9 Stokes parameters for 1667 MHz feature 13. . . . . . . . . . . . . . . . . . . . . . . . . 154

6.1-10 1665 MHz Zeeman pairs in W3(OH); divided up into three groups for

analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155

6.1-11 Variation in magnetic field strength and demagnetised velocity with RA

and DEC at 1667 MHz. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157

6.1-12 1667 MHz maser area versus magnetic field strength. . . . . . . . . . . . . . . . . . 158

xii

6.1-13 Top, the 1667 MHz masers and a random selection of points in the same

space. Bottom, the two-point correlation of the 1667 MHz masers. . . . . . . 159

6.1-14 The two-point correlation of the 1667 MHz masers . . . . . . . . . . . . . . . . . . . 159

6.2-1 1612 MHz spectra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161



6.2-4 The 6 most intense maser lineshapes at 1612 MHz. . . . . . . . . . . . . . . . . . . . 168

6.2-5 Zeeman pair 3 (features 22 and 27). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169


and DEC at 1612 MHz. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171




6.3-1 1720 MHz spectra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174



6.3-4 Movement of the Gaussian fit positions of the components of feature 12 . . 179

6.3-5 Lineshapes of 1720 MHz masers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181

6.3-6 Zeeman pair 3 – the most perfect Zeeman pair in W3(OH) . . . . . . . . . . . . . 183


and DEC at 1720 MHz. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184

6.3-8 Relative proper motion of the southern maser to the northern maser at

1720 MHz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185




6.4-1 The ground state masers in W3(OH), showing regions of distinct velocity . 188

6.4-2 The southern arc feature in both 1665 MHz and 1667 MHz emission . . . . . 194

6.4-3 Variation of Zeeman probability with Zeeman component separation. . . . . 197

6.4-4 Demagnetised velocities and magnetic field strengths at all four ground

state OH frequencies in W3(OH). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199

6.4-5 Velocity of all the masers in the ground state . . . . . . . . . . . . . . . . . . . . . . . . 199

6.4-6 The elongated north-south nature of masers at all three frequencies present

in the east of W3(OH) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200

xiii

List of Tables

2.1 Details of the lower rotational transitions of OH . . . . . . . . . . . . . . . . . . . . . . . . 9

2.2 Top, Zeeman shifts in OH levels. Bottom, Zeeman shifts in OH transitions . . 13

4.1 Relationship of the Stokes parameters to the polarized correlations . . . . . . . . . 54

4.2 The observed sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

4.3 The VLBA Antenna codes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

4.4 The beam dimensions at the four ground frequencies . . . . . . . . . . . . . . . . . . . . 67

4.5 Typical errors as output by SAD for masers of an example range of peak

fluxes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

5.1 The 1665 MHz Masers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91

5.2 1665 MHz masers compared at epochs 1978, 1986, 1996. . . . . . . . . . . . . . . . . 99

5.3 Failure reasons for 1665 MHz masers that failed to be fit with Gaussian

lineshapes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105

5.4 1665 MHz masers with high linear polarization. . . . . . . . . . . . . . . . . . . . . . . . . 110

5.5 1665 MHz Zeeman pairs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117

5.6 Magnetic field properties of the different maser groups . . . . . . . . . . . . . . . . . . 118

6.1-1 The 1667 MHz Masers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .143

6.1-2 Masers that have been identified in both 1977 (Norris et al. 1982) and

1997 epochs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148

6.1-3 1667 MHz Zeeman pairs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156

6.2-1 The 1612 MHz Masers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .164

6.2-2 Variation in intensity of the northern masers over recent years . . . . . . . . . . 166

6.2-3 1612 MHz Zeeman pairs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170

xiv

6.3-1 The 1720 MHz Masers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .177

6.3-2 1720 MHz Zeeman pairs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181

6.4-1 Maser overlaps among the ground state lines . . . . . . . . . . . . . . . . . . . . . . . . 190

6.4-2 The fluxes of all class A and B overlaps . . . . . . . . . . . . . . . . . . . . . . . . . . . .195

6.4-3 Polarization and Zeeman pair properties of the ground state OH maser

lines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196

xv

Astronomical Units and Abbreviations

Units

AU Astronomical unit = 150×106 km

pc Parsec = 206,265 AU, or 3×1013 km, or 3.26 ly

ly Lightyear = 10×1012 km, or 0.3 pc

M Mass of the Sun = 2×1030 kg

R Radius of the Sun = 7×105 km

L Luminosity of the Sun = 3.8×1026 W

' Arcminute = 1/60 of a degree

'' Arcsecond or as = 1/3600 of a degree

mas Milliarcsecond = 1/3,600,000 of a degree

Jy Jansky = 10-26 W m-2 Hz-1

cm-1 Wavenumbers (per cm) → Reciprocal wavelength

Abbreviations

CO Carbon monoxide

CVR Complete velocity redistribution

DMC Dwarf molecular cloud

EVN European VLBI Network

FIR Far infrared

FUV Far ultraviolet

GMC Giant molecular cloud

H I Atomic hydrogen region, H (also for O, C etc); because every H

contributes one free particle to the environment.

H II Ionised hydrogen region, H+ (also for O, C etc); because every H

contributes two free particles to the environment.

xvi

IRAS Infrared Astronomical Satellite source

ISM Interstellar medium

ISO International Space Observatory

LCP Left hand circular polarization

LEP Left hand elliptical polarization

MERLIN Multi-Element Radio Linked Interferometer Network

MHD Magneto-hydrodynamic wave

NVR Negligible velocity redistribution

OB/O/B star Stars of spectral type OB/O/B

OH Hydroxyl

PAH Poly-aromatic hydrocarbon

RA Right Ascension (east-west on the sky)

RCP Right hand circular polarization

REP Right hand elliptical polarization

SFR Star-forming region

SIS Single isothermal sphere

UCH II Ultra-compact H II region

VLA Very Large Array

VLBA Very Long Baseline Array

VLBI Very Long Baseline Interferometry

YSOs Young stellar object

ZAMS Zero-age main-sequence (star)

DEC Declination (north-south on the sky)

xvii

Dedication

This thesis is dedicated to those who made is possible. First and foremost I dedicate it

to the love of my life: my fountain of happiness, my Golden Bunny. You were always

there to lift me up when it was all getting me down; took care of everything so I could

work unhindered. Your effort is hidden in every word, in every picture…

In addition, thanks to the boys who made life fun: Dave, Pierre, Tino and

Jordi… Fine men indeed. When it comes to jokes, political discussion, musical

analysis, double-entendre, laughter and all round entertainment these guys simply

can’t be beat. You name it – they got it! And I’m not talking about diseases either. I

shall in fact miss you all quite a bit I reckon… [Sniff] Hey – yes it is okay for a guy

to get upset at times like this! Honourable mentions go to lab-mates through the ages:

Liz, Alex, Julia, John, Sarah, Kurt, Nathan, Will. You were all fun to meet and work

around, keep at it. A word for the original Flat 61 crew: Mark, Mark, Brendan and

Dor. Man, those were the days!

Finally, a word of thanks for the Internet – what a wonder of the world it is. I

love you ADS! Who would of thought that there was so much por^H^H^H information

out there just a click away?

Dave, that was the mother of all football matches – I can still feel in my legs!

Just five of one score and six:

You opened up our eyes,

Showed us what we could be if we tried.

Our lives are richer now…

Thank-you, E. Merck

1

Chapter 1

Introduction

The Background of Astronomical Masers

In 1965, Weaver et al. made an unexpected observation – peculiar emission lines in

space of unknown origin, at a frequency of 1665 MHz. The emission was at first

famously put down to an unknown interstellar species named “mysterium”, but was

soon identified as line emission from hydroxyl (OH) molecules in compact sources

within star-forming regions (SFRs) inside molecular clouds (Davis et al. 1967). More

discoveries followed, with water (H2O) emission (Cheung et al. 1969), methanol

(CH3OH) emission (Ball et al. 1970) and silicon monoxide (SiO) emission (Snyder

and Buhl 1974), all coming from within molecular clouds. These were termed

“masers”, since from their narrow line-widths and high brightness it was suspected

that these sources were a phenomenon of microwave amplification by stimulated

emission of radiation, although proof of their masing origin actually rested in the

narrow band coherence function (Litvak et al. 1966). Masers were then discovered

around highly evolved red-giant stars; first was OH emission (Wilson and Derek

1968), then H2O emission (Knowles et al. 1969) and SiO emission (Buhl et al. 1974).

Masers were soon discovered in external galaxies (Whiteoak and Gardner

1973), followed in 1982 with the discovery of OH emission from the extra-galactic

source IC 4553 (Arp 220), with an unrivalled luminosity – 106 times greater than any

previous source (Baan et al. 1982). This was termed a “megamaser” because of its

great luminosity, and since then, more megamasers have been found in other galaxies

at cosmologically significant distances (e.g. z~0.1). Since those early days, many

more masing molecules have been found in space, including NH3, SiS, CH, H2CO and

HCN; although fewer molecular species have been found in megamasers than galactic

masers – only H2O and formaldehyde (H2CO) have been added so far. The masers

Chapter 1: Introduction

2

have been discovered in diverse places, including super-nova remnants (SNRs),

accretion discs, and even in our own solar system in comet haloes. In essence,

wherever warm molecular gas exists in space, masers are likely to be found.

Masers are a remarkable feature of conditions in space. However, more than

just causing us to wonder at the variety of nature, masers are extremely useful probes

for us in our exploration of the universe. By their very nature, masers tend to exist in

those areas of the universe that are hidden from us in the optical radiation range – the

very places into which we are most desperate to see! Due to its particle size,

interstellar dust is very efficient at absorbing radiation at wavelengths shorter than ~1

µm, but becomes virtually transparent in the microwave and radio regions, and so our

radio-telescopes can peer deep into these otherwise hidden places where some of the

most interesting events in the universe are occurring. Because of their high brightness

and small physical size, masers can probe the smallest scales at the greatest distances:

they can be used to measure both the size of and distance to the phenomena they lie in,

such as circumstellar shells, accretion discs, and bipolar outflows. Since their

radiation is often point-like and unresolved, it can also be used to examine effects of

the interstellar medium (ISM) between them and us such as scattering and reddening.

Correlations between distance and maser size have been demonstrated by Diamond et

al. (1988) that agree well with scattering theory. By their very existence, masers can

give us valuable information about the conditions in the space they inhabit, such as

temperature, number density, and magnetic field strength and direction. Where more

than one ‘flavour’ of maser (different molecular species, or different frequency from

the same molecular species) coexists in the same space this can tell us even more

about the conditions, since computational models are more constrained when this

additional data is available.

Masers trace the bulk movement of the gas that they are in, and so their

Doppler velocity and proper motion will reveal to us gas velocity fields. In this way,

masers have been used to confirm the galactic rotation profile in stars near the galactic

centre (Habing et al. 1983) and distance to the centre of the Galaxy (Reid et al. 1988).

Study of the Zeeman effect in masers has helped to map the Galactic magnetic field,

as in the work of Reid and Silverstein (1990). The development of Very Long

Baseline Interferometry (VLBI) techniques has enabled unprecedented resolution,

with imaging of masers with resolution limits at high frequencies breaking through the

Chapter 1: Introduction

3

milliarcsecond (mas) barrier. Using VLBI, the small size and high brightness of

masers coupled with physical persistence means that their very small proper motions

can be measured on the timescale of just a few years. In recent years this has allowed

motion mapping of expanding shells of gas around Mira-type variables (Humphreys

1999 and references therein) and, directly relevant to this work, mapping of the

expansion of star-forming regions such as W3(OH) (Bloemhof et al. 1992, Gwin et al.

1992). It is this characteristic of masers that much of this thesis will deal with: the

movement of gas, and an attempt to add the third dimension to our two-dimensional

(projected) view of space. Of all the events in the universe that are hidden from us at

most wavelengths, perhaps the most interesting are those of gas in violent turmoil.

Contrast the motion of gas during formation and in the immediate aftermath of the

birth of new stars, with the motion of gas in the grip of super-massive, voracious black

holes – gas experiencing environments of cosmic birth and death. This thesis focuses

on the environment of stellar birth, and will use information from the observation and

modelling of masers to construct the clearest picture of the case study that represents

this environment.

4

Chapter 2

OH Spectral Line Emission and Masing

Before the characteristics of OH maser emission can be discussed and understood it is

necessary to examine how OH molecules give rise to microwave emission, and

therefore the basic structure and dynamics of the OH molecule must be described.

After this, the processes and mechanisms by which maser action arises are outlined.

The rotational spectrum of OH in its ground vibronic state is of primary interest to us

in this work, since this is where masing action arises. At the temperature of OH

masers in the area of study (~100 K) the fraction of molecules in excited electronic or

vibrational states will be practically zero – for example: the fraction of molecules in

the first state of the OH stretching vibration would be ~1.6 × 10–19 at thermal

equilibrium.

2.1 The OH Molecule and its Rotational Spectrum

Since the electrons in a molecule move very much more quickly than the nuclei, a

good approximation is that the nuclei can be assumed stationary whilst the electronic

motions are considered. This is called the Born-Oppenheimer Approximation (Born &

Oppenheimer 1927).

2.1.1 Angular Momentum in the OH Molecule

The electrons in a molecule can have both orbital and spin angular momentum, but

when electrons are paired, both of these quantities cancel out. The OH molecule has

one unpaired electron, so the total orbital angular momentum, L (quantum number L),

and total spin angular momentum, S (quantum number S), of the electrons are both

non-zero. As the molecule is not spherically symmetric (whereas an atom is) the

electrons can possess angular momentum around the molecular axis, but precession of

Chapter 2: Spectral Line Emission and Masing

5

the momentum axes means that the average angular momentum about directions

perpendicular to the molecular axis is zero. Consequently, only the projections of the

angular momenta L and S onto the internuclear axis are constant and well defined (i.e.

quantized). The unpaired electron occupies a 2pπ orbital on the oxygen atom which

gives it L = 1, and its constant projection onto the internuclear axis, Λ = 1. Thus, the

ground state of OH is a Π state (sigma and pi are derived from the Greek for s and p

atomic orbitals). The magnetic moment of the orbital motion (which is aligned with

the nuclear axis) and of the spin (S = ½, with projection onto the internuclear axis Σ)

of the electron interact and cause the spin of the electron to be orientated so that the

magnetic moment of the spin is also aligned along the internuclear axis. This is called

Spin-Orbit Coupling. The total electronic angular momentum along that axis is

labelled Ω, where

Ω = Λ + Σ (2.1)

This means that the rotational spectrum arises from two ladders of levels (also called

the fine structure): one where the spin and orbital angular momentum are aligned

parallel (Ω = ±3/2), and the other where the alignment is anti-parallel (Ω = ±1/2). Spin-

orbit coupling does not remove the two-fold degeneracy of these levels, i.e. the states

Ω = +1/2 and –1/2 have the same energy at this point, as do the Ω = +3/2 and –3/2.

The end-over-end rotation of the molecular framework has an angular

momentum R that is perpendicular to the molecular axis. This angular momentum is

quantized, and R can be zero or any positive integer. The sum of the molecular

rotational angular momentum and the electronic angular momentum is labelled J,

where

J = R + Λ + Σ (2.2)

The sum of the angular momentum excluding electronic spin is labelled N, where

N = R + Λ (2.3)


6

2.1.2 Hund’s Cases for OH

The coupling of R with L and S leads to two of Hund’s Cases (see figure 2.1). When

the magnetic field generated by the molecular rotation is not large (generally when the

rotation is not fast) both S and L are well resolved along the molecular axis and Ω is a

‘good quantum number’ even when the molecule is rotating. The condition where L

and S are both coupled to the molecular axis in the presence of rotation is called

Hund’s case (a); where J is formed from R and Ω. For certain light molecules like

OH, the field generated by the rotation becomes so large for the higher rotational

states that S becomes coupled to N rather then the molecular axis, so that J is formed

from S and N; this is Hund’s case (b). OH in its low rotational states is intermediate

between Hund’s cases (a) and (b).

Deviations from Hund’s cases are caused by the effect of a third vector

uncoupling L and S from their usual axes; in the case of OH this is the interaction of

the molecular rotation with the electronic motions. Note that these conditions violate

the Born-Oppenheimer approximation.

Ω

ΛΣ

J

R

N

S

L

J

R

L S

Λ

Figure 2.1: Hund’s case (a) and (b) at left and right respectively. Circles are atoms; arrows are

angular momentum vectors; dotted ellipses are precessional motions of the vectors. In Case (a), L

and S precess around the molecular axis faster than the molecular axis precesses about the total

angular momentum J. In case (b) precession of the molecular axis is still slower then that of L but

is now faster than that of N and S about J.


7

The effect of uncoupling S from the molecular axis causes a shift in the

rotational levels that increases as the decoupling gets stronger – i.e. as the molecular

rotation gets faster. The two ladders of rotational levels (The 2Π3/2 and 2Π1/2 states –

sometimes confusingly labelled F1 and F2 respectively) are shifted in opposite

directions so that they become separated by an energy A that is called the spin-orbit

coupling constant (or fine structure constant). The 2Π3/2 states are shifted by +½A and

the 2Π1/2 by –½A; A is –139.7 cm–1 for OH (Carrington 1974). The negative sign of A

means that the spins states of OH are inverted so that the 2Π3/2 state is at a lower

energy than the 2Π1/2 state (See figure 2.2).

The effect of slightly uncoupling L from the molecular axis is to lift the two-

fold degeneracy of the Ω states. This is known as Lambda Doubling, and the effect is

larger for faster rotation. Each state in the ladder is now a doublet approximately

symmetrical about the unperturbed level. The two states are of opposite symmetry or

parity. This is essentially a result of violation of the Born-Oppenheimer

approximation: the wave function can no longer be represented solely by the ground

electronic form – a small admixation of the first excited state is required, and this

results in opposite parity. Therefore, the two levels of the lambda doublet are actually

13/2

11/2

9/2

7/2

5/2

J =3/2

11/2

9/2

7/2

5/2

3/2

J =1/2

0

200

400

600

800

Equivalent energy aswavenumber / cm–1

2Π3/2 2Π 1/2

80 µm

53 µm

35 µm

Figure 2.2: Energy levels of the two ladders of rotational states of OH


8

different electronic states. The difference in energy of the doublet states in the 2Π3/2,

J=3/2 state is equivalent to radiation of 1666 MHz. Lambda doubling has been

studied extensively by van Vleck (1929) and Herzberg (1950).

2.1.3 The Hyperfine Interaction in OH

The hydrogen nucleus (a proton) has non-zero spin angular momentum I, and to

obtain the total molecular angular momentum this needs to be taken into account. The

total molecular angular momentum is labelled by the quantum number F, which is the

vector sum of J and the spin quantum-number, I, of the proton – which has value ½.

So, finally,

F = J + I (2.4)

Therefore, each level of the lambda doublet is split into two levels, of values J+½ and

J−½. This is known as the hyperfine interaction. Because of this and lambda

Figure 2.3: Lambda doubling and hyperfine splitting of OH rotational states, with ground state

transitions indicated.

7/2

3/2

J =1/2

5/2

J =3/2

−

+

+

Parity −

−

Parity +

−

+

−

+

F=1

1

2

2

1

1

0

F=0

3

1

2

4

3

4

1

2

2

3

2

3

Lambda and hyperfine splitting not to scale

a

b d

c

a = 1667 MHz b = 1665 MHz c = 1612 MHz d = 1720 MHz

0

100

200

Equivalent energy / cm-1


9

doubling, each J state is split into four levels (See figure 2.3). The difference in

energy of the hyperfine states is equivalent to radiation of 53 MHz.

Of greatest importance in the present work on OH maser emission are

transitions between the four levels (at zero magnetic field) in the lowest rotational

state, 2Π3/2, J=3/2 (see table 2.1). These transitions are called the hyperfine transitions

and are divided into two groups: Those in which the F quantum number does not

change, which are called strong or mainline transitions – at 1665 MHz and 1667

MHz; and those in which the F quantum number does change (i.e. the nuclear spin

flips over), called weak or satellite transitions – at 1720 MHz and 1612 MHz. The ∆F

= ± 1 transitions have reduced dipole intensity (Dipole transitions are enhanced where

∆F = ∆J; and within the ground state ∆J is clearly 0), and in local thermodynamic

equilibrium (LTE) the 1667 MHz transition is 1.8 times as strong as the 1665 MHz

transition and nine times as strong as the satellite transitions. A comprehensive source

of frequencies and molecular constants for OH is available in Destombes et al. (1977).

Table 2.1: Details of the lower rotational transitions of OH (Burdyuzha and Varshalovich 1973,

also Destombes et al. 1977).

State Energy above ground

state / cm-1

Transition in quartet (Finitial →

Ffinal)

Frequency of

transition / MHz

Equilibrium relative

line strength

2Π3/2,J=3/2 0 2→1 1612 1 1→1 1665 5 2→2 1667 9 1→2 1720 1 2Π3/2,J=5/2 84 3→2 6017 1 2→2 6031 14 3→3 6035 20 2→3 6049 1 2Π1/2,J=1/2 126 1→0 4660 1 1→1 4751 2 0→0 Forbidden Forbidden 0→1 4766 1 2Π1/2,J=3/2 188 2→1 7749 1 1→1 7761 5 2→2 7820 9 1→2 7832 1 2Π3/2,J=7/2 202 4→3 13434 1 3→3 13435 27 4→4 13441 35 3→4 13442 1


10

(2.7)

2.2 Magnetic Fields and Polarization in OH

Molecules with unpaired electrons and hence non-zero electronic angular momentum

(such as OH) have paramagnetic properties: Each unit of angular momentum gives

rise to a magnetic moment called the Bohr magneton, µΒ, where

µΒ = eh / 2mec (2.5)

Each molecular level F contains (2F + 1) magnetic sub-states, labelled mF, which are

the projection of F onto the magnetic field axis.

2.2.1 The Zeeman Effect

In the absence of a magnetic field, all of these sub-levels are degenerate, but an

external magnetic field causes the magnetic moment to precess, which lifts the

degeneracy. In the presence of a weak magnetic field of strength B the energy, E, of

an mF state becomes

E = E0 + B µB gF mF (2.6)

where E0 is the zero field energy for the particular F state and gF is the Landé Factor –

a coefficient characteristic of the state, which can be calculated from the various

quantum numbers that make up the total molecular angular momentum:

+

+−+++=

)1(2

)1()1()1(

FF

IIJJFFgg JF

The factor gJ has been measured by Radford (1961) for the 2Π3/2 state of OH to be:

0.935 for the upper levels of the lambda doublet and 0.936 for the lower levels. In the

upper levels, the factor in brackets is equal to exactly 3/4; in the lower, it is 5/4.

Clearly, any set of four levels will now be split into possibly very many levels indeed

if F > 2 for any of the levels, but fortunately the ground state of OH which will be

investigated in this work has values of F of just 1 and 2 (see figure 2.4). Transitions

between the magnetic components in the presence of a magnetic field therefore

produce a number of lines at similar frequencies known as Zeeman components. For


11

the mainline transitions, where F is the same for both states, the bracketed factor in

equation 2.7 evaluates to ±3/4 for F=2 (the 1667 MHz transition) and ±5/4 for F=1

(the 1665 MHz transition). For the satellite lines where F is different for the two

states the situation is more complex: since the bracketed factor is different for both

states its contribution to the energies of the transitions can take the values ±1/4, ±1/2,

±3/4, ±5/4 (see figure 2.5). For the main lines the ∆mF = 0 transitions are un-shifted

by the magnetic field; for the satellite lines the mF = 0 ↔ 0 transition is un-shifted.

The relative intensities of each of the Zeeman components are also dependent on the

quantum numbers F and mF of the levels involved, which gives rise to the pattern seen

in figure 2.5.

The results of the Zeeman effect in masers depend on the size of the Zeeman

splitting in relation to the maser linewidth, the stimulated emission rate and ‘cross-

relaxation’ effects through infrared transitions (Goldreich et al. 1973a,b). Zeeman

lines arising from ∆mF = 0 transitions are called π components. They are linearly

polarised and their electric vector oscillates along the magnetic axis. They are

suppressed if the line of sight is parallel to the magnetic field – they are best observed

when the field is perpendicular to the line of sight. Lines arising from ∆mF =±1

transitions are called σ components. They are elliptically polarised (in a left-handed

and right-handed sense) in general, with the electric field vector confined to a plane

perpendicular to the magnetic field. Under IEEE convention σ + components give rise

to left-hand circularly polarised radiation (LCP) and σ – give rise to right-hand

J=3/2

1665 MHz

F=1

σ+

–1

0

F=1

–

1

σ −

+

–1

0

1

mF

π

Figure 2.4: Magnetic field Zeeman splitting of the 1665 MHz transition in OH. Relative splittings

are not to scale.


12

circularly polarised radiation (RCP) if the magnetic field points away from the

observer. If the line of sight is parallel to the magnetic field they will appear as pure

circularly polarised lines; if the line of sight is perpendicular to the magnetic field they

will appear, relatively suppressed, as linearly polarised with the vector perpendicular

again to the field (see figure 2.5). Between these two idealised cases, the

σ components will appear as elliptically polarised (a superposition of linear and

circular), denoted as left and right-hand elliptically polarised radiation (LEP/REP).

The suppression or preferential amplification of π and σ maser components –

depending on the orientation of the magnetic field relative to the observer’s line of

sight – is sometimes called magnetic beaming (see chapter 2.4.2). Magnetic beaming

is the tendency of maser radiation to amplify preferentially along or perpendicular to

the magnetic field lines. The theory behind this can be found in e.g. Nedoluha and

Watson (1990a), Gray and Field (1994, 1995), and Field and Gray (1994).

1665 MHz 1667 MHz 1612 & 1720 MHz

Polarised parallel to magnetic field

-5/4 0 0

0 0 0

0

Polarised perpendicular to magnetic field

Right circular

Left circular

-5/4

-3/4

3/4 -3/45/4

5/4

3/4

1/4 -1/4

-3/4

-5/4

3/4 -3/45/4

1/4 -1/4 -2/4

2/4

3/4

-5/45/4

a)

b)

σ+ σ -

σ+

π

σ -

Figure 2.5: Zeeman components of the 2Π3/2, J=3/2 ground state of OH; where observations are in

case a) Parallel to magnetic field – all peaks are σ components; b) Perpendicular to magnetic field –

solid lines are π components, the others are σ components. Fractions are relative splittings as

indicated in table 2.2b)


13

2.2.2 Selection Rules

Selection rules for electric dipole transitions are: ∆Ω=0, ±1; ∆J=0, ±1; ∆F=0, ±1

(except F=0 ↔ 0, which is forbidden); ∆mF =0, ±1 (except 0 ↔ 0 when ∆F=0); and

the parity must change. Transitions with ∆Ω=±1 lie in the infrared region and have

been shown to be important for pumping of masers (see chapter 2.4.4) by Skinner et

al. (1997). Collisional transitions are governed by “propensity” rather than strict

selection rules. Changes in J of 2 are possible via quadrupolar interactions, but

changes of more than 2 are very unlikely – so: ∆J=0, ±1, ±2 is a “weak selection rule”.

The same is true for the F quantum number: ∆F=0, ±1, ±2. The parity need not

change in collisional transitions. Another aspect of collisional transitions is that at the

Parity of Level

F of Level mF of Level

Energy above ground / (MHz)

Shift (µB gF mF)

/ (kHz mG–1)

+ 2 0 1720 0 ±1 ±0.9814 ±2 ±1.9628 + 1 0 1665 0 ±1 ±1.6357 – 2 0 53 0 ±1 ±0.9828 ±2 ±1.9656 – 1 0 0 0 ±1 ±1.6379

Transition Transition / MHz

Total transitions

Splitting of Zeeman components interpreted as a Doppler motion / km s–1 mG-1

d 1720 9 (0) 0.113 (0.229) 0.342 0.571

a 1667 12 (0) 0.354

b 1665 6 (0) 0.590

c 1612 9 (0) 0.123 (0.243) 0.366 0.608

Relative splitting in figure 2.5 0 1/4 2/4 3/4 5/4

Tables 2.2a) and b): a) Top, showing Zeeman shifts in the ground state rotational levels of OH. b)

Bottom, showing Zeeman splitting in the hyperfine transitions between those levels. Rounding of

the shifts to 3 d.p. simplifies the table greatly: Each 1667 MHz Zeeman line is actually a close

quadruplet, and each 1665 MHz Zeeman a close doublet. Bracketed components are π components

– only visible under certain conditions (see figure 2.5)


14

temperature of the masing cloud (~100 K) large energy transitions may be very

unlikely due to the low average kinetic energy of the collision partners.

A more in-depth description of the origins of molecular energy level structure

and selection rules can be found in Townes and Schawlow (1955) and Gordy and

Cook (1984).

2.3 Fundamentals of Maser Emission

2.3.1 The Boltzmann Distribution

In a collision dominated system, the population distribution amongst molecular energy

states (translational, rotational, vibrational) is determined by the Boltzmann

distribution. In this distribution the number of molecules in a state i is related to the

number of molecules in a higher energy state j by the equation

nj = ni exp[−hνij / kT ] (2.8)

where ni, nj are the populations in the state i and j, and hνij is the energy gap between

the state i and j. As we saw in the previous sections, energy states can be degenerate.

To take this into account the degeneracy (statistical weight, or multiplicity), g, is

introduced into the equation, and now

Nj / gj = (Ni / gi) exp[−hνij / kT] (2.9)

where gi is the degeneracy of the ith level, and Ni is population summed over all such

states of i. Therefore Ni/gi = ni.

Under the conditions described above (negligible radiative processes) the

Boltzmann distribution will eventually be reached from any starting distribution,

provided that the coefficient for any excitation is related to that for the de-excitation

by the principle of microscopic reversibility (also called ‘detailed balance’). A gas

under the Boltzmann distribution at kinetic temperature Tk that is bathed in a

blackbody radiation field at temperature Tbb, where Tk = Tbb, is in complete

thermodynamic equilibrium (CTE). A gas under a Boltzmann distribution (or very

close approximation to the Boltzmann distribution) at kinetic temperature Tk that is


15

bathed in a radiation field at a temperature Td, where Tk ≠ Td is in local

thermodynamic equilibrium (LTE). The radiation and kinetic temperatures are said to

be decoupled.

The conditions in space are in many cases far from any type of equilibrium

(although the Maxwellian velocity distribution is unusually well maintained in the

interstellar medium). Regions of gas are often in the process of being heated, cooled,

expanded, compressed and are subjected to chemical reactions and non-uniform

radiation fields. This results in the breakdown of the Boltzmann distribution; most

importantly in the case of masers, this results in population inversions.

2.3.2 Population Inversions

If there are more molecules in an upper substate than a lower substate – that is, if

Nj / gj > Ni / gi (2.10)

then the population is inverted. In space, for masers to function, population inversions

are required, and they can and do arise naturally as is detailed later. The process by

which population inversions are formed in masers is called pumping (see chapter

2.4.4).

If a photon of energy hνij is incident on a molecule in energy state i, which is

separated from a higher energy state j by a dipole allowed transition of energy hνij,

then that photon can be absorbed by stimulated absorption and the molecule is

promoted from state i to j. Likewise if the molecule is already in the higher state j

then a photon of energy hνij can stimulate the molecule to fall down in energy to state

i and emit another photon of energy hνij by stimulated emission. The emitted photon

is exactly in phase with the first photon and moving in exactly the same direction. A

molecule in the higher state j can also fall to state i without the presence of any

applied electromagnetic field, so called spontaneous emission. This is caused by zero

point oscillations of the electromagnetic field. The photon will be in a random

direction and phase.

Consider a two state system bathed in a uniform radiation field; with

population Nj in a higher energy state j and a population Ni in a lower energy state i.


16

(2.12)

(2.15)

The rate of change of population Nj (or –Ni) due to stimulated absorption, RSA,

transferring molecules from state i to j is

RSA = BijρNi (2.11)

where Bij is the Einstein coefficient of stimulated absorption from i to j; and ρ is the

radiation energy density, which from Planck’s radiation law (for black body radiation

only – the Planck function can be labelled ρ (ν, T ) since it is a function of temperature

and frequency) is

118

3

3

−⋅=

/kThec

hν

νπρ

The units of ρ here are J m–3 Hz –1; but sometimes the radiation intensity is used

instead in equations, which has units W m–2 Hz –1 Sr–1. This takes out a factor of 4π

/c, effectively representing integration over spherical space.

Similarly, the rate of change of population Nj (or –Ni) due to stimulated

emission, RSE, transferring molecules from state j to i is

RSE = −BjiρNj (2.13)

where Bji is the Einstein coefficient of stimulated emission from j to i. The rate of

change of population Nj (or Ni) due to spontaneous emission, RS, transferring

molecules from state j to i is

RS = − AjiNj (2.14)

where Aji is the Einstein coefficient for spontaneous emission from j to i. The Einstein

coefficients are related by equations (2.15)

3

3

8 ji

jiji

j

iji

πhνcA

Bg

gB ==

(∴ A ∝ ν 3)


17

The relationship of Bji and Bij implies that if the substate populations (for state i that is

equal to Ni / gi) of states i and j are numerically equal, then the rate of stimulated

emission is equal to the rate of stimulated absorption. If the populations are higher for

i than for j then there will be an excess of stimulated absorption – an absorption line

will form in the radiation field. If the populations in the substates of j are larger

however, exposure to radiation of energy hνij will cause an excess of stimulated

emission, and the incident radiation will be amplified – a maser emission line will

form. Radiative transfer through a medium is very sensitive to changes in level

populations from their LTE values (which can lead to rapid variability, see chapter

2.4.1). The fact that A∝ ν 3 is important because it means that at microwave

frequencies of OH masers, spontaneous emission has a relatively small effect (and

therefore population inversion may be easier to form) because its probability is very

low relative to other processes, e.g. FIR transitions between rotational states which are

104 times more likely. The ratio of stimulated emission to spontaneous emission is

~1018 times larger for radio waves than for visible light.

2.3.3 Collisional Factors

Collisions can also change the populations of states. Collisional transitions are

governed by ‘propensity’ (as mentioned in chapter 2.2.2), so typically more collisional

transitions are allowed than dipole transitions. Collisional interactions depend on a

second order rate constant, k2nd (typically of ~10–10 cm3 s–1; cf. FIR transitions which

have a rate of about 10–3 to 1 s–1 in OH), and the concentration of the collision partner

– which can be taken to be the concentration of molecular hydrogen, [H2], as this is

typically 5 to 10 times larger than the next most abundant species He. This creates a

pseudo first-order rate constant, k. So the rate of change of population Nj (or Ni) due

to collisional processes transferring molecules from state i to state j, RCOL, is

RCOL = kij[H2]Ni (2.16)

where kij is the rate coefficient for the excitation transition from i to j. The rate for the

reverse process, i.e. transfer of molecules from state j to state i, R’COL, is

R’COL = kji[H2]Nj (2.17)


18

where the rate constants kij and kji are related by the principle of microscopic

reversibility so that:

kij / kji = gj / gi exp[− hνij / kT] (2.18)

The above equations demonstrate the theory for a two-level scenario. In a

multi-state environment every state is connected to a large number of other states,

above and below in energy, by all the allowed collisional and radiative transitions

between those states.

2.3.4 Radiative Transfer: Optical Depth

The Beer-Lambert law describes the changing intensity of radiation of frequency ν

passing through a simple absorbing medium

Iν = I0ν exp[− κν r] (2.19)

where I0 is the initial intensity, r is the distance travelled through the medium and κ is

the absorption coefficient of the medium. The product of κ and r is called the optical

depth, τ. τ is usually defined as a differential so that

dτν = −κν dr (2.20)

In this scheme, the optical depth is zero at the observer and becomes increasingly

positive backward along the distance r towards the source. However, for a line in

population inversion, κ at that frequency will be positive – since more radiation is

added to the radiation field along the path of the rays. In this case, the optical depth

becomes increasingly negative with distance.

If the optical depth at a frequency ν is low, photons of frequency ν are rarely

absorbed and their escape probability is high. If the optical depth is high, photons of

frequency ν are nearly always absorbed and their escape probability is low (e.g. there

may be an allowed transition at the frequency ν that absorbs them). The escape

probability of a photon emitted from anywhere within a medium is between 0 and 1,


19

(2.21)

(2.22)

(2.23)

where 0 represents absorption of all photons and 1 represents absorption of no

photons.

2.3.5 Radiative Transfer: Source Function

In any gas not at equilibrium incident radiation plays a role in determining level

populations, so the Beer-Lambert law is of little use because the astronomical medium

is itself adding and subtracting line radiation from the passing beam. Instead, the

equation of radiative transfer must be used, which (ignoring scattering – for the sake

of simplicity here) is

νννν εκ +Ir

I −=dd

where ε is the emissivity (emission coefficient) of the medium. The emissivity is

defined as the energy emitted per unit frequency, per unit volume, per unit solid angle,

per unit time. Note the similarity between κ and ε – they are opposites with respect to

energy change. In the case of a two level environment the absorption coefficient may

be represented by

)()BB( ννκν fc

hnn

jijijjii −=

and the emissivity by

)(4

νπνεν f

hn

jijijΑ=

where f (ν ) is the normalised line profile in units of inverse frequency. We assume

here that the absorption and emission line profile are equal; when the distribution of

molecular velocities is Maxwellian, the line profile is a Gaussian. The quotient of ε

and κ is called the source function, S, so that

S = εν / κν (2.24)


20

(2.25)

(2.26)

The source function becomes independent of frequency when the assumption of

equivalence of the line profiles holds. This is the assumption of complete velocity

redistribution (see chapter 2.4.2). The equation of radiative transfer may also be

written with respect to optical depth so that

SId

dI −= νν

ν

τ

This equation may be solved by integrating with respect to optical depth between

values of (0) and (−τ'), and this gives the formal solution to the equation of radiative

transfer:

')]'([exp]exp[0

0νννννν ττττ

ν

d S + I=Iτ

−−− ∫

Note that this is just the Beer-Lambert decay plus an integral term: the integral

contains the emission (and absorption) from rays generated elsewhere. The integral

term is an integro-differential equation, since S depends on ni and nj, which if not in

LTE, in turn depend on Iv. In this way, the specific intensity at any point is

determined by the radiation field behind the cloud and a contribution from within the

cloud, which is dependent on the molecular populations. This is called Non-local

Radiation Transfer, because the radiation intensity at any point depends on the

radiation intensity at all other points, and clearly solving such problems numerically

for modelling is complex.

2.4 Characteristics of Maser Radiation

Masers have negative absorption coefficients and optical depth, and so from equation

(2.26) it can be seen that this leads to exponential growth along the length of the

maser cloud. At this point, the maser is said to be unsaturated.

2.4.1 The Unsaturated Regime

The exponential amplification in the unsaturated regime has three main consequences

of beaming, line narrowing and rapid variability:

Small path length differences in the maser cloud are greatly distorted by

exponential amplification. In this way, maser spots are smaller than the clouds that


21

hold them, since the cores of the clouds will give rise to much of the radiation.

Radiation passing through a spherical cloud is amplified much more through the

longer paths near the centre of the sphere, so that the emergent radiation is

geometrically beamed. Geometrical beaming may be even more pronounced where

the maser is made of a beam through several clouds of gas, as in the scenario invoked

by Masheder et al. (1994) to explain the exceptionally small angular size of 1720

MHz masers in W3(OH). The angular size of the spherical cloud is reduced by a

factor of (1–τν)1/2 (Goldreich and Keeley 1972, Watson 1993). Maser radiation is

highly beamed for any geometry of cloud, and beaming increases with increasing path

length (Elitzur 1990a).

If the normalised line profile f (ν ) is Gaussian in form and constant throughout

the maser then the optical depth away from the line centre at a frequency ν ' is

τν = τν' exp[–(ν –ν' )2 / 2σ 2] (2.27)

where σ is the dispersion. The result of this is that the centre of the line is amplified

much more than the wings of the line, and the normalised Gaussian is therefore

thinner – by a factor of (1–τν)1/2 (Goldreich and Kwan 1974). Since the amplification

of maser radiation depends exponentially on both the molecular population inversion

and the path length via equations (2.19) and (2.22), any variation in either parameter

will be reflected exponentially. This means that the maser property of variability can

be much more rapid than the physical process which is causing the variability.

Maser amplification in the unsaturated regime can be represented by

Ir,ν = I0,ν exp[ γν r] (2.28)

In this representation, γν is the gain coefficient at the line centre. The exponential term

is known as the integrated gain, which is a measure of the ‘power’ or ‘strength’ of the

maser, and typically has values of ~105 to 1010 for strong galactic masers, and a much

lower ~2 to 10 for megamasers. From equation (2.19) it can be seen that the gain

coefficient and optical depth are related by –τν = γν r. In this way the gain coefficient

is found from


22

(2.29))(8

)(2

2

νπν

γν fAc/ggN-N ijij=

This indicates that the gain is proportional to the Einstein A coefficient as well as the

population inversion. The effect of this is that masing in the main lines of OH will be

~7 times larger (see table 2.1) than the satellite lines for an equal population inversion.

Exponential growth of the passing radiation can only continue up to a point.

Since the masing action itself removes molecules from the upper level, there comes a

point where the intensity of radiation is so great that the masing action causes

significant shifts in the molecular populations. At this point, the maser is said to be

saturating.

2.4.2 The Saturated Regime

In the saturated regime, equation (2.28) no longer holds; intensity growth is linear

with distance and is heavily dependent on the pumping processes that maintain the

population inversion (see chapter 2.4.4). Unlike the unsaturated regime, the gain is no

longer dependent on the Einstein A coefficient. In an unsaturated maser, masing in

other transitions in the same molecular system are unaffected by each other – they are

decoupled. Saturation can couple different masing transitions in an effect known as

competitive gain, which was investigated by Field (1985). In this way, competition

between co-propagating maser transitions can lead to one line dominating at the

expense of inversion in the other, and the ‘winner’ is not necessarily the transition

with the largest A coefficient.

Saturation can cause lines to rebroaden back towards their full thermal width.

This happens because as the line centre saturates first and the gain coefficient drops in

value, the line wings remain unsaturated for longer and maintain their high gain

coefficient. Maser lines have an inherent width, arising from thermal and turbulent

motion of the masing molecules, which greatly exceeds the homogeneous transition

width. Because the seed radiation of the maser has a relatively broad range of

frequencies and is incoherent, the total output of the maser is incoherent. However, in

saturated regions of the maser, local coherence effects develop and the resulting

radiation is partially coherent (Field and Richardson 1984). Velocity redistribution

also becomes important in the saturated regime; if the processes which maintain a

Maxwellian velocity distribution in the molecules (the molecular velocity relaxation


23

rate, Γ) are faster than the stimulated emission rate, then the velocity distribution

remains Gaussian. This scenario is called the case of complete velocity redistribution

(CVR). In the case of CVR, the lineshape does not rebroaden in the saturated regime,

as the gain coefficient remains largest at the line centre. The effect of velocity

gradients on the lineshape has been investigated in e.g. Field et al. 1994. If Γ is

slower than the stimulated emission rate then the case is of negligible velocity

redistribution (NVR). In the case of NVR, lineshapes do rebroaden, and may become

cropped in the line centre or even ‘hole burnt’ – where radiation in the line centre may

be reabsorbed. Field et al. 1994 find that CVR conditions should lead to an increase

in maximum amplification factor of ~3.

The question of maser geometry becomes important in the saturated regime.

The simplest form of maser to imagine and model is the symmetrical tube maser

(Goldreich and Keeley 1972). The situation is complicated here by the fact that rays

will be amplified in both directions along the maser, and the saturation after a point in

one direction will have a hindering effect on the growth of the maser in the other

direction. This situation is illustrated diagrammatically in figure 2.6. The saturated

linear maser contains an unsaturated core where growth is exponential and saturated

ends where growth is linear. As the maser tube is lengthened, the unsaturated zone

shrinks to a minimum size (independent of length) and the output intensity of the

maser becomes proportional to the cube of the length. Goldreich and Keeley (1972)

also studied the spherical maser. They showed that the apparent size of a saturated

spherical maser is approximately the beamed size of the unsaturated core of the maser

as mentioned in chapter 2.4.1. As the radius of the sphere is increased, the core radius

decreases steadily and the output intensity of the maser becomes proportional to the

cube of the radius. The effects of velocity gradients in spherical masers have been

studied by Litvak (1973), and Bettwieser and Misselbeck (1977), who show that the

apparent size of a spherical shell maser is smaller than the size of the gas density

distribution. The problem of competition between beams in different directions in

saturated masers has also been investigated extensively by Alcock and Ross (1985a,b;

1986a,b).

Saturation is very effective at quenching weak beams, but does not add

significantly to the main beam intensity. For this reason background emission is

unlikely to be a significant factor in the brightness of saturated masers whose


24

brightness may be 105 times that of background emission (e.g. those in star-forming

regions), although it will still be important in low gain masers (e.g. megamasers)

whose brightness is only a few times the background source. The effect of

background radiation is studied in Elitzur (1991).

Maser cloud geometry and internal cloud velocity fields are clearly very

important in the formation of masers; and in some cases may be the dominant factor

which differentiates different types of masers, rather than different molecular number

densities or temperatures (Elitzur 1992).

2.4.3 Brightness Temperature

Since maser emission is obviously non-thermal, maser brightness is in no way

indicative of their physical temperature. However, a useful comparison between

masers is their brightness temperature. The brightness temperature, TB, of a maser is

the temperature at which a blackbody would have to be to produce the same intensity

at the wavelength of the maser. Using the Rayleigh-Jeans approximation (hν « kT) to

the black-body function this is

kTB = ½ (c/ν)2 Iν (2.30)

Masers can have huge effective temperatures, with OH masers in star-forming regions

around 1010K, but 22 GHz H2O masers up to 1015K.

Maser clouds forming tube

Idealised maser tube

Unsaturated core

Saturated zone

Saturated zone

Figure 2.6: The saturated tubular maser. In reality masers may be made up of two or more clouds,

constantly moving in turbulent eddies – breaking and reforming the masers constantly.


25

2.4.4 Pumping

Pumping is the process by which population inversions form and are maintained.

Only relatively small deviations from the Boltzmann distribution are needed for

masing to occur, and there are many possible ways this could occur. The three main

methods of creating population inversions are by radiative processes, collisional

processes, or chemical processes. There may of course be combinations of all three,

and the relevance of each may depend very much on the type of location; in particular

the density and ambient radiation.

Purely radiative processes can lead to population inversions if radiative rates

are high compared to collisional rates (e.g. molecular densities are low) and if

transitions have widely varying optical thickness. Pumping in megamasers is likely to

be solely radiative because of the low densities predicted for megamaser regions.

Dominantly collisional processes can lead to population inversions if different

transitions have unequal cross-sections, e.g. the upper and lower lambda doublet

states. In general, higher collisional rates (e.g. high densities) will tend to quench

population inversions by re-thermalising the populations. Chemical and

photochemical processes (e.g. UV photodissociation, Andresen et al. 1984) can cause

population inversions by creating molecules in an excited state from which radiative

decay is slow, and collisional de-excitation is unlikely because of low densities.

Radiative pumps may be thought of as ‘converters’, which absorb photons of

frequency νp and convert them to photons of frequency ν with efficiency η, which is

dependent on the pump. Several photons, perhaps of different frequencies, might be

required for each maser photon. This can effectively boost the brightness temperature

of the maser up from the thermal brightness by a factor of η( νp / ν)2. For OH ground-

state transitions pumped by infrared photons at 35 µm this factor is ~107, which would

boost TB from ~100 K to 109 K. Beaming can increase TB further by a factor of 4π/Ω,

where Ω is the beam angle. Typical beam angles in SFRs are ~10-4 sr, which would

mean a potential boost up to 1014 K, though Gray et al. 1992 have shown that Ω for a

50 Jy OH maser needs to be ~2.5 10–5.

An important process in radiative pumping schemes is Line Overlap (Litvak

1969, Lucas 1980). Line overlap is a non-local effect arising because of Doppler

effects and/or thermal broadening. These effects can shift lines and allow photons

born in one transition to be absorbed in another. Most important for OH masers are


26

the effects of thermal overlap (studied by e.g. Pelling 1977, Guilloteau et al. 1981,

Flower and Guilloteau 1982) and velocity overlap (Bujarrabal et al. 1980, Cimerman

and Scoville 1980). Line overlap is useful for supplying additional photons into

frequency-space which may have been depleted locally by absorption and so can have

an additional effect to boost brightness temperatures. This effect can be appreciable

for the infrared transitions between J states in OH, such that three or more transitions

may overlap with velocities of just a few km s–1. Radiation transport in masers has

been investigated by e.g. Field and Gray (1988), in which they used a semi-classical

method (radiation is treated classically, interactions with masing molecules quantum

mechanically).

Chemical pumping schemes for OH have been studied extensively by

Andresen. Andresen et al. 1984 showed experimentally that photodissociation of cold

H2O molecules can lead to selective population of the lambda doublet states. This is

because the dissociation goes through an intermediate state in which the OH molecule

has the unpaired electron π-lobe perpendicular to the rotation plane of the molecule;

radiative decay of the transition state then tends to invert the 2Π3/2 lambda doublet and

anti-invert the 2Π1/2 lambda doublet. Andresen (1985) investigated this effect further

in SFRs, and Thissen et al. (1999) expanded on the mechanism using a theory

modelling melting ice mantles on grains of varying sizes as the source of H2O, and

therefore OH via UV radiation, as the direct source of population inversions.

However, models of chemical schemes have always failed to find sufficient H2O to

explain steady-state masers.

Excellent reviews of maser radiation can be found in Cohen (1989) and Elitzur

(1992), and pumping schemes in Elitzur (1982). Also useful are Cook (1977), Doel

(1990) and extremely an thorough and complex analysis of the physics behind masers

can be found in Elitzur (1992).

27

Chapter 3

OH Maser Environments

As outlined in the introduction, OH masers have been found in diverse locations in

space. One particular environment is examined in this work: Star-forming regions.

This environment is discussed next. After this, the occurrence of the four ground-state

transitions is discussed.

3.1 Star-Forming Regions

The Milky Way contains several thousand giant molecular clouds (GMCs). These

clouds contain most of the molecular material in the Galaxy (Weliachew and Lucas

1988) and have a total mass of about 109 M (Strong et al. 1988, Blitz 1993) – 1% of

the Galaxy’s total mass. Molecular clouds are the main sites of star formation in the

Galaxy (Zuckerman and Palmer 1974, Burton 1976), and when they are the sites of

star formation, they are known as star-forming regions (SFRs).

3.1.1 Star Formation

It is thought that high mass and low mass stars are formed through separate

mechanisms, known as bimodal star formation (Herbig 1962, Mezger and Smith

1977, Elmegreen and Lada 1977). Herbig observed that there were regions where star

formation was producing no stars of mass greater than ~2 M (e.g. Taurus), whilst

there were other areas that contained ‘young stellar objects’ (YSOs) of both high and

low mass. He suggested that star birth occurs in a cloud for a period of a few tens of

millions of years, with low mass stars forming first. Mezger and Smith suggested that

low mass and high-mass stars are formed in distinct clouds; with low mass stars

(named after their prototype as T Tauri stars) forming in clouds of all sizes throughout

Chapter 3: OH Maser Environments

28

the galactic disc, while high mass stars (labelled OB stars after their spectral type1) are

formed in large clouds in the Galactic spiral arms. This view has been reinforced by

observations, e.g. Solomon et al. (1985) who divided molecular clouds into ‘cold’

(<10 K) and ‘hot’ (>20 K). The cold clouds do not contain stars of earlier spectral

type than late B, and are distributed about the Galactic disc. The hot clouds tend to be

much larger, inhabit only the spiral arms and are associated with H II regions (e.g.

Stark 1985, Waller et al. 1987, Lo et al. 1987). Low mass stars may form as just a

few lightly clustered stars (the isolated mode), or in tight clusters (the clustered

mode); while massive stars form almost exclusively in clusters (Lada 1992, Lada et al.

1993). Stars form only from the densest parts of molecular clouds and the properties

of cloud ‘cores’ differ for regions of high and low-mass star formation (Myers 1986).

The average number of stars in the mass range 1 ≤ M / M ≤ 120 that are formed per

massive SFR varies between ~225 within the solar circle (the circle describing the

Sun’s orbit around the Galactic centre) to ~120 outside it (Casassus et al. 2000).

3.1.2 Physical Characteristics of GMCs

The primary method of detecting sites of molecular material in space is via carbon

monoxide (CO) emission lines – CO is known as a ‘tracer’ – direct measurement of

H2 via its quadrupolar spectrum is difficult because the Einstein A-values (hence line-

strengths) are very weak. From this it is known that most of the molecular mass is

contained in giant clouds with masses between 105 and 3×106 M and scale sizes

between 20 and 40 pc across, with average hydrogen densities of order 100 cm–3

(Solomon et al. 1979). These values are in dispute (e.g. Blitz and Thaddeus 1980)

largely because of disagreement over the conversion factor from CO intensity to H2

column depth (van Dishoeck and Black 1987).

Millimetre-wave radio maps have shown that GMCs are actually cloud

complexes made of smaller clouds called ‘clumps’, that often have a filamentary

structure (e.g. Pudritz 1986, Bally et al. 1987, Mizuno et al. 1995, Falgarone et al.

1 Stars are classified by their spectral type: the standard types are O, B, A, F, G, K, M, in order of

decreasing surface temperature. Types are subdivided into 10, so that the sequence runs O0, O1, …

O9, B0, B1 etc. Throughout most of a star’s life, while it is stable on the ‘main sequence’, stars that are

more massive will have a higher surface temperature. O and B stars are ‘early’ types; K and M are

‘late’ types. O and B stars have surface temperatures above 10,000 K, and so are ionising stars because

their photon flux in the Lyman continuum (<912 Å) is significant.


29

1998, Nakahama et al. 1998, Wiseman and Ho 1998, Allen and Shu 2000). The

clumps have masses of about 103 to 104 M in regions about 2 to 5 pc across, with

hydrogen densities of about 102 to 103 cm–3 (Sargent 1977, Evans 1978, Blitz 1978,

Rowan-Robinson 1979). Within the clumps are ‘cloud cores’, which because of their

higher density can be mapped in less common molecules like formaldehyde (H2CO)

(e.g. Evans and Kutner 1976), carbon sulphide (CS) (e.g. Linke and Goldsmith 1980),

cyanoacetylene (HC3N) (e.g. Avery 1980), and ammonia (NH3) (Ho and Townes

1983). Mapping dense cores in CO becomes less useful because the gas becomes

optically thick above a certain column density. The cores have sizes of about 0.1 pc,

masses of about 10 to 100 M , hydrogen densities of about 104 cm–3, and

temperatures of about 10 K. The dense cores are likely surrounded by massive

envelopes of ~100 M (Falgarone 1987). These cores are often detected as IRAS

sources (infrared sources detected by the Infrared Astronomical Satellite), and often

associated with bipolar outflows (Fuller and Myers 1987); both of which indicate that

they are the sites of low-mass star formation (Myers and Benson 1983). The

protostars emit radiation that is absorbed by surrounding dust. The reemission of the

infrared from warm dust can be identified by its spectral profile (e.g. Adams and Shu

1986, Adams et al. 1987, Myers et al. 1987a). There are also a smaller number of

denser clumps that appear to be forming bound clusters of stars. They have sizes of

about 0.3 to 0.6 pc (Loren et al. 1983), and may consist of closely associated cores of

higher densities of over 106 cm–3 (Snell et al. 1984).

Recent observations (Ward-Thompson 1999) with the Infrared Space

Observatory (ISO) have identified ‘pre-stellar’ cores – a dense core that is

gravitationally bound, but contains no embedded luminosity source. As such, this

represents the initial conditions of star formation. These sources were too cold to be

detected by IRAS, and they do not contain a significant quantity of warm dust.

3.1.3 Stability of Molecular Clouds

The stability and lifespan of GMC is of interest in predicting star-formation rates and

the evolution of galaxies. It is now thought that a substantial fraction of the Galactic

molecular material is in smaller (or dwarf) molecular clouds (DMCs), which are found

throughout the Galactic disc. These clouds probably live longer than one complete

rotation of the Galaxy, which means that they live for more than 100 million years.

Larger, hotter molecular clouds are likely assembled and broken up dynamically as


30

they cross a spiral arm, and so live for not more than 10 million years (Bash and

Peters 1976). (It has been suggested that GMCs actually define the spiral arms of

galaxies – in this case they obviously only live within a spiral arm.) This helps

explain their heterogeneous internal structure if they are only transient phenomena

(Blitz and Shu 1980). High-resolution observations of M83 and M51 (Allen et al.

1986, Lo et al. 1987) have shown that H I emission comes from downstream of the

molecular clouds; indicating that it is the result of dissociation of H2 after (and

probably by) OB stars, and also that GMCs are assembled from molecular rather than

atomic material (e.g. Kwan and Valdez 1983, 1986, Roberts and Stewart 1987). It

was originally thought that molecules could only form in the cores of dense clouds of

atomic material and dust! This is now known to be far from the mark.

DMCs must be in near pressure equilibrium in order to last so long. The

number of clouds observed means that they cannot be collapsing in free-fall or the

star-formation rate of the Galaxy would be higher than observed (Zuckerman and

Palmer 1974). Several mechanisms are suggested that could act to support a cloud

against its self-gravity: magnetic fields coupled to ions (Chandrasekhar and Fermi

1953, Mestel 1965, Spitzer 1968, Mouschovias 1976), rotation (Field 1978), and

turbulence (Norman and Silk 1980, Larson 1981). Magnetic fields are believed to

play the biggest part here. Much of the mass of most molecular clouds in our Galaxy

is in regions of moderate density, allowing photons from the interstellar radiation field

to maintain sufficient ionisation for magnetic fields to resist collapse (McKee 1989);

this gas is effectively in a photon dominated region (see chapter 3.2.4). Polarization

observations (e.g. Vrba et al. 1976) show that embedded magnetic fields are fairly

ordered over the size of clouds, and hence turbulence cannot dominate over magnetic

fields. Magnetic fields, unlike turbulence, are difficult to dissipate and so are a likely

source of stability in clouds. Typical field strengths in clouds are in the range of 10

µG to 150 µG (Heiles 1987, Shu et al. 1987). However, magnetic fields cannot

support gas along the field lines, yet clouds are not always flattened in the plane

perpendicular to the field lines; although it is hard to determine the 3-D structure of

clouds from 2-D images. The support along field lines is likely to come from

magneto-hydrodynamic waves (MHD waves) and outward propagating Alfvén waves

within the cloud (Shu et al. 1987, Mckee et al. 1993). MHD waves are also likely to

be the source of the large linewidths observed in tracers in the clouds, since the widths


31

cannot be the result of collapsing gas without exceeding the known star-formation rate

of the Galaxy of 3-5 M yr–1 (Scalo 1986, Evans 1999).

3.1.4 The Collapse of Molecular Clouds and Birth of Stars

Given that magnetic fields play the main role in supporting molecular clouds, there are

two scenarios to consider (Mestel 1985). These two cases are likely the basis of

bimodal star formation (Shu et al. 1986).

1. The supercritical case. In this case, magnetic fields cannot support the

clump against its self-gravity, even if the field lines remain frozen in the

fluid. In the absence of any other support a supercritical cloud will

contract rapidly and lead to efficient star formation, possibly with high

mass stars.

2. The subcritical case. In this case, indefinite gravitational collapse cannot

be induced by any amount of external pressure if the field lines remain

frozen. Subcritical clouds will evolve however because of ambipolar

diffusion – the slip of neutrals relative to the ions to which the magnetic

field is coupled. This eventually leads to a dense pocket, but the long

timescale explains why star-formation this way is inefficient.

For many years a simple model of cloud collapse was accepted – known as the single

isothermal sphere (SIS). In this scenario, as the cloud core gets ever denser, evolution

will tend to produce a 1/r2 density distribution (Bodenheimer and Sweigart 1968).

Eventually the growing concentration causes instability and gravitational collapse

begins. The collapse proceeds inside out, such that a wave of infall propagates at the

speed of sound into the cloud. This means that the mass infall rate is constant (Shu et

al. 1987). However, more recently this model has been successfully challenged –

despite its elegant aesthetic appeal – by e.g. Whitworth et al. (1996) who have argued

that in the centre of star-forming cores the density must be below that predicted by the

1/r2 model. They point out two major shortcomings of the SIS model: 1. It is unlikely

anything like an SIS can be formed in a typical star-forming region, given the innate

instability of the SIS and all paths leading to it – especially the extreme deviations

from spherical symmetry produced by ambipolar diffusion. 2. The SIS appears to

make binary star formation nearly impossible, which given the number of stars in

binary systems is a big problem. It is likely therefore, that such a simple scenario is


32

not the complete picture of stellar collapse. For a current review of this field, see

Vázquez-Semadini et al. (2000).

The increase in density at the centre continues until the protostar becomes

optically thick. This happens at about 10–3 to 10–2 M , and from there after the star

grows hydrostatically. From this point, the evolution of massive stars and low-mass

stars is different, because massive stars reach the main sequence (i.e. ZAMS – zero-

age main sequence, or stellar birth) more quickly – whilst still accreting from the

envelope (Kahn 1974, Yorke and Krugel 1977). Once the star’s luminosity rises,

infall may be halted by a combination of stellar winds, shocks and radiation pressure

on dust grains.

In the case of massive stars, which have high UV fluxes, a region of ionised

gas grows around the star, forming what is called an ultra-compact H II region (UCH

II region) (Spitzer 1978). These regions are signposts of star formation in much the

same way as IRAS sources, because they are readily identified by radio continuum

spectra (Habing and Israel 1979, Crawford and Rowan-Robinson 1986). The mean

free path of an ionising photon is long in the fully ionised gas and very short in neutral

gas so the transition between them is sharp – a zone known as an ionisation front. The

ionised region is also known as a Stromgren sphere. The density inside the region and

outside are roughly the same, but the temperature inside is ~104 K, while that outside

is ~100 K. This means that there is a pressure difference of about two orders of

magnitude across the ionisation front, and so the ionised region expands at the speed

of sound in the ionised gas. At ~10 km s–1 this is supersonic in the cool gas, so the

region grows outwards proceeded by a shock – which will be important for the

chemistry of masers (see chapter 3.2.3). The UCH II regions are relatively short lived

(a few times 104 years) and they soon become giant H II regions, although their exact

lifespan is unclear since they appear to be confined by the surrounding cocoon gas –

probably by magnetic pressure. Therefore, O stars spend ≤ 2% of their lifetime in the

UCH II region (Casassus et al. 2000).

It is now believed that all YSOs pass through a stage of outflow at the

beginning of their life (e.g. Lada 1985). These warm molecular outflows are rich in

phenomena, e.g. water masers. Outflows may be important in supplying the

turbulence and/or instability in molecular clouds, possibly eventually destroying the

cloud (Franco et al. 1990) – such that each generation of star formation influences the


33

next. Some have gone further and suggest that OB star formation is triggered by such

events, along with stellar winds, supernova shocks and ionisation fronts, so that star

formation may spread like a ‘disease’ or ‘fire’ (see figure 3.1) through molecular

clouds (e.g. Herbst and Assousa 1978, Ho et al. 1986, Cameron 1985).

The processes of star formation, including specifically massive-star formation

and cluster formation are covered extensively in the books by Hartmann (1998), Lada

and Kylafis (1999), Mannings et al. (2000); and in the review by Evans (1999).

3.2 Chemical Evolution in Molecular Clouds

The interstellar medium (ISM) – like the rest of the universe – is dominated by

hydrogen. The abundance of other elements is measured relative to hydrogen: for the

four most common elements, helium (He), oxygen (O), carbon (C) and nitrogen (N)

these abundances are 0.1, 6.8×10–4, 3.8×10–4 and 1.0×10–4 respectively. Outside of

clouds, all material is atomic and most atoms are ionised due to the high intensity of

UV photons permeating space from hot stars. Since the mean cross-section for

hydrogen ionisation is ~6×10–18 cm2, local condensations of gas with a column density

above 1018 cm–2 are effectively shielded, and the hydrogen is neutral. Lower energy

New OB SubgroupMaser Sources,

Infrared Sources,Compact H II regions,

MolecularCloud

Structure~20pc

Older Subgroup

Expanding H II Region

Younger Subgroup

Expanding Shock

Ionisation Front

Figure 3.1: The popular ‘Burning Cigar’ model of sequential star-formation, with the direction of

‘burning’ left to right.


34

UV photons that are not absorbed by hydrogen still penetrate though, and so more

easily ionised atoms, such as carbon are still ionised. The shielding of cloud gas is

measured by the visual extinction Av, which is very roughly equal to τv: the optical

depth in the optical region. Clouds of the above description are called diffuse and

have Av ≤ 1. Ionising photons cannot penetrate deeper into a cloud than Av ~3 to 5,

and beyond this point hydrogen forms H2 which provides additional self-shielding. At

this stage, molecular chemistry can begin.

Interstellar molecular clouds are now known to be home to a rich chemistry,

with more molecules discovered on a monthly basis. In Ohishi (1997) there were just

under 120 molecular species detailed. In recent years, there have been major

advances in the modelling and observation of numerous molecular species in

molecular clouds. It is known that the chemistry is dependent on the conditions inside

the cloud, and therefore observation of the chemistry has become an invaluable

diagnostic of the conditions in clouds. Some molecules are particularly sensitive to

different formation paths and are useful in distinguishing a variety of environments.

The conditions under which OH is formed or destroyed are of great importance to the

present work, and will be outlined in the next section.

In cold molecular clouds before star formation has begun, the chemistry is

dominated by low-temperature gas-phase ion-molecule and neutral-neutral reactions.

This leads to the prevalence of unsaturated molecules and carbon chains, and of small

radicals such as OH. In the depths of such clouds one of the main driving forces in

the chemistry are cosmic rays – which have the ability to penetrate deep into the

clouds. During the initial ‘cold-core collapse’ phase, the density increases such that

most molecules (except H2) accrete onto refractory grains (typically silicate/graphite)

and form an icy mantle. This phase is dominated by surface reactions, driven by UV

photons and cosmic rays. Energy released from such reactions can send molecules

back into the gas phase. As the star forms the energy released in the gravitational

collapse heats up the gas until it becomes a ‘hot core’, and molecules begin to

evaporate back into the gas phase – the most volatile first. Outflows from the YSO

cause high-temperature shocks to propagate through the gas into lower temperature

regions, and this provides a mechanism for the rapid vaporisation of huge amounts of

ices in grain mantles, sometimes even liberating silicon-containing compounds. The

hot-core phase is the richest in chemistry, and may last for ~105 years (van Dishoeck


35

and Blake 1998). In the final phase, the envelope or cocoon is dispersed by the YSO

winds and, in the case of massive stars, UV photons. This final phase is photon

dominated, and leads to the formation of photodissociation regions. The following

sections focus on the processes of relevance to OH.

3.2.1 Gas-Phase Chemistry

Hydrogen is much more abundant than any other element; only helium comes to

within a factor of 10 in abundance, but naturally helium plays only a small part in

reaction chemistry (molecular ions may form, e.g. HeH+). Because of this, reactions

with H or H2 dominate the chemical networks if they are exothermic. This is only the

case in molecular clouds for small ions; most reactions of neutrals or large ions with H

or H2 have significant energy barriers, and so do not proceed in the cold-core phase of

collapse. These reactions become significant though in the hot-core phase and in

shocked regions, and above ~230 K most of the oxygen is driven into water by the

reactions: O + H2 → OH + H; OH + H2 → H2O + H (Charnley 1997). One possible

route for the destruction of OH is by atomic sulphur (S): S + OH → SO + H; SO +

OH → SO2 + H. However, the lack of abundant SO2 in dark clouds indicates an

incomplete understanding of the above reactions and sulphur chemistry (Palumbo et

al. 1997). The basic gas-phase processes are reviewed in van Dishoeck (1988), while

the chemical-reaction networks are described in van Dishoeck (1998a,b) and Bakes

(1997).

3.2.2 Grain-Surface Chemistry

The chemistry on the surfaces of interstellar grains is discussed in depth by e.g.

Schutte (1996), and Williams and Taylor (1996). The main characteristic of grain-

surface chemistry at lower temperatures and densities is the formation of fully

hydrogenated species such as H2O, NH3 and CH4 because of the mobility of atomic

hydrogen on the cold grain surfaces. As the density increases the amount of gas-phase

atomic H drops enormously in favour of H2, and reactions with atomic oxygen

become important. Tunnelling reactions with H2 can occur competitively once [H2] »

[H]. At higher temperatures, the movement of heavier species such as radicals over

the surface becomes significant, allowing a complex and poorly understood chemistry

to occur (e.g. Caselli et al. 1993).


36

Photochemical reactions within the ice mantles can be triggered by UV

photons, the photons themselves coming either directly from the young star, or

scattered by dust in the inner gaseous envelope into the outer envelope (Spaans et al.

1995), or by cosmic ray produced photons (e.g. Gredel et al. 1989). Cosmic ray

processes are always present and generate a UV field of ~3 × 103 photons s–1 cm–2

inside dense clouds for a typical cosmic-ray rate (Cecchi-Pestellini and Aiello 1992).

This rate is ~105 times less than that from the general radiation field at the edge of a

cloud, but it may still be relevant for chemistry on timescales of millions of years.

An essential factor in promoting gas-grain chemistry is a mechanism to return

molecules to the gas phase. Without desorption, molecules accrete onto the grain

surfaces on a timescale of ~2 × 109 yS / nH years, where yS is the sticking coefficient –

thought to lie between 0.1 and 1.0, and nH is the number density per cm3 (Williams

1993). This rate would be sufficient to absorb most molecules from the gas phase in

less than a million years for a typical dark cloud density of 104 cm–3. Since this is not

consistent with observation, desorption mechanisms clearly exist. Thermal

evaporation only becomes significant at temperatures Td > 20 K, i.e. after the star has

formed. The energy of formation of some molecules may heat grains locally and

release some species (Willacy et al. 1994b), and cosmic ray heating will always be

present (Léger at al. 1985). Polar molecules such as H2O, which contain strong

hydrogen bonds when in condensed form, are difficult to remove from the surface of

grains in cold clouds. H2O is expected to evaporate at about 90 K, along with other

polar ices (Sandford and Allamandola 1993). The variation of grain mantle

compositions with temperature is shown in figure 3.2.

3.2.3 Shock Chemistry

The outflow of material from condensing stars at supersonic speeds is now thought to

arise from the need of accreting material to shed large amounts of angular momentum.

The exact means by which the outflow is produced is still unclear; in particular,

whether collimated jets or star/disc winds are responsible (Pelletier and Pudritz 1992,

Shu et al. 1995). When high velocity (10 to 100 km s–1) material collides with the

envelope of ambient medium shocks form that propagate into the medium. Shocks are

divided into two types: C type (for continuous, because the shock transition occurs

gradually) and J type (for jump, because the shock transition is quick – a couple of

mean-free paths), which depends on the shock velocity and magnetic field strength


37

(3.1)

and fractional ionisation in the pre-shock material (Draine and McKee 1993,

Hollenbach 1997). For typical dense cloud conditions C shocks occur below ~40 to

50 km s–1, and J shocks at velocities well above that.

When a shock passes through gas the bulk flow is converted into random

thermal motions. The temperature Ts behind a shock front in a single-fluid region

obeys the relation

2s2s ((2

mvkT1)+1)−

=γγ

where γ is the adiabatic index of the gas (ratio of heat capacities at constant volume

and pressure), m is the mean molecular mass of the gas and vs is the shock velocity.

For vs of 10 km s–1 the post shock temperature is ~2275 K for H and 3400 K for H2.

At these temperatures reactions with energy barriers, such as O + H2 → OH → H2O

proceed rapidly, driving all O into H2O (e.g. Draine et al. 1983). Reactions with

atomic hydrogen can reverse these reactions, so the balance between O, OH, and H2O

depends on the H/H2 ratio in the gas in the shock. For much higher vs, i.e. in J shocks,

the temperature may reach 105 K. At this temperature, not only are all molecules

destroyed but also UV photons are emitted by the gas, which can dissociate molecules

both before and after the shock (Hollenbach and McKee 1989, Neufeld and Dalgarno

Figure 3.2: The state of molecules in the environment around massive YSOs. From van Dishoeck

and Blake (1998).

Hot coreOutflows

SiO

T(gas)=200-1000 KT(dust)~90 K ~60 K ~45 K ~20 K

50000 AU

UV

1000 AU

TrappedCO

COice

2

COCONOice

2

2

ice

H2

H2ONH3

HCH3

O ice

2

CNS

CH

CH OH

OH

2

3

3

2

ComplexOrganics


38

1989a,b). As the temperature falls and H2 reforms the chemistry resembles that of C

shocks.

As the gas temperature falls further, (typically < 500 K) the rates of many

reactions become very slow and the molecular mixture is effectively ‘frozen’. In this

way shocks can produce high H2O abundances with all of the O not locked up in CO

being driven into H2O (e.g. Goldreich and Scoville 1976, Elitzur 1979); and

photodissociation can form OH so that OH abundances are often raised in areas of

high H2O abundance. Observations (primarily of the Orion shock) do indicate high

H2O/H2 abundances of at least 10–5 (e.g. Cernicharo et al. 1994, Zmuidzinas et al.

1995, Timmermann et al. 1996). Shocks are likely to be the most important factor in

the required chemistry for OH masers.

3.2.4 Photodissociation Regions

The photodissociation region (PDR) is taken to extend into clouds from the ‘atomic

surface layer’ into where FUV flux still affects the chemistry of e.g. O and C not

locked up in CO (see figure 3.3). In terms of extinction this is up to Av ~10. At this

depth hydrogen is molecular; and therefore all of the atomic and most of the molecular

gas in the Galaxy is in PDRs, since only dense cloud-forming cores have Av > 10.

PDRs are the source of a large amount of the IR radiation from the ISM (i.e. non-

stellar IR radiation). Incident radiation is absorbed primarily by dust grains and

polycyclic aromatic hydrocarbons (PAHs), and is then reemitted as PAH IR features

and far-infrared (FIR) continuum radiation from the cooling dust grains. About 0.1 to

1% of the FUV energy is converted to energetic photoelectrons (~1 eV) that are

ejected into the medium and heat the gas. The dust absorbs 102 to 103 times more

energy per unit volume than the gas, but remains at a lower temperature than the gas

due to the much higher efficiency of dust ‘greybody’ cooling (not quite blackbody).

The primary gas cooling routes are through the C II (158 µm) and O I (63 µm) spectral

lines.

PDR chemistry contains several important differences from normal ion-

molecule chemistry. Firstly, the high flux of FUV photons makes photoreactions very

important, as are reactions with atomic hydrogen. Vibrationally excited H2 ( *2H ) is

also common and can drive reactions that would not be possible with H2. If the gas

gets very warm (>500 K), the energy barrier to reactions of atoms and radicals with H2

can be overcome, and reactions with H2 can dominate. Finally, an important factor for


39

OH chemistry is that atomic O is kept abundant by the FUV photons, and so ‘burning’

reactions are effective. FUV pumping of H2 can lead to a vibrational temperature that

is much higher than the gas temperature. Astrophysical reactions with *2H are

covered in Wagner and Graff (1987). These non-equilibrium excitation conditions

can give rise to reaction rates that are enhanced over thermal rates (Gardiner 1977,

Dalgarno 1985); for example, rates of reaction of *2H with O, OH and C+ all show

large enhancements with vibrational excitation of H2 (e.g. Light 1978, Jones et al.

1986). However, vibrational excitation of OH does not enhance the reaction rate of

OH + H2. Chemical reactions can also be promoted by translational energy; in

particular, turbulence can lead to non-Maxwellian velocity fields and hence non-

Maxwellian reaction rates. The reaction O + H2 is not affected by this to a great

extent because of its relatively high activation energy. The enhancement of OH, H2O,

CH+, and HCO+ in turbulent PDRs has been studied by Falgarone et al. (1995).

The PDR interface layer consists mostly of neutral atoms or cations formed by

photodissociation reactions and ionisation reactions. OH is produced through

reactions of O with *2H and H2, and most of it is photodissociated again. A small

fraction reacts with C+ to form CO+, which then strips an electron from H to form CO

(Hollenbach and Tielens 1997). With increasing depth into the PDR, chemistry

involving small radicals such as OH and +nCH becomes more important. In warm

VA VAVA < 0.1∆

C

O2

+ +C

HH+

H2

C

O

COC

H+

UV Flux

~ 1 ~ 10

O

T~ 1000 K T~ 100 Kto 10 K

T~ 10000 K

[e] ~ 1

to 100 K

[e] ~ 10 [e] ~ 10-6-4

Slow MHD Shock Fast MHD Shock

Max penetrationof UV < 91 nm

Max penetrationof UV > 91 nmto optical

Figure 3.3: Diagram of the transition from UCH II region (left) to dense molecular core (right). The

entire area left of O2 is a PDR. [e] values are relative abundances of electrons; Av values are visual

extinction in magnitudes. Adapted from Hollenbach and Tielens (1997).


40

high-density PDRs, such as those formed near massive YSOs, reactions involving H2

dominate; while in cool low-density PDRs, such as those before gravitational collapse

has begun, reactions with FUV pumped *2H are important. The main reactions in

oxygen bearing molecules are shown in figure 3.4. PDR chemistry is covered in depth

in Sternberg and Dalgarno (1995).

3.2.5 The Role of Dust

Two main mechanisms couple the gas to the FUV photons of stars: FUV pumping of

H2 molecules and energetic electrons released via the photoelectric effect on small

dust grains and PAHs. Other heating mechanisms – such as collisions with warm

dust, cosmic ray heating and pumping via the FIR radiation field of the warm dust

(FIR pumping is however very important for the functioning of OH masers) – play

only a small role, save for in the depths of clouds (Tielens and Hollenbach 1985a).

Theoretical calculations indicate that small dust grains (<50 Å) and PAHs are more

efficient at heating the gas than larger grains. The distribution of grain sizes

Figure 3.4: The most important reactions of oxygen bearing molecules in PDRs. From Hollenbach

and Tielens (1997).

H2

H2 H2

H2

CH2

CH2H3+

O O

O

OH OH

CO

H

H

H

H

H

H

HOH

CC

e

e

2

2

2

2 2

2

3

O

O

O+

+

++

,

,

,

,

+

+

*

*

ν

ν

H

e

e

e

νO+


41

approximately follows a power law: the number of grains per unit size interval is

proportional to r –3.5 (Mathis, Rumpl, Nordsieck 1977); this is the ‘MRN’ distribution.

The function is normalised so that the mass function of the dust is 1% of the mass of

hydrogen nuclei. The MRN distribution also assumes that the dust is a

silicate/graphite mixture. The original MRN distribution only covered the range 0.005

µm (50 Å) to 1 µm; the more recent ‘extended’ MRN distribution covers the range

0.001 µm (10 Å) to 10 µm. For an MRN distribution stretching down to the sizes of

molecular PAHs, approximately half of the gas heating is due to grains with sizes

below 15 Å (Bakes and Tielens 1994). Grains this small are capable of being

significantly heated by the absorption of a single photon. The remaining half comes

mostly from grains of sizes 15 to 100 Å, with grains larger than 100 Å contributing

negligibly to the heating of gas.

More extensive coverage of the chemistry of molecular clouds can be found in

the reviews by Tielens and Whittet (1997), Hollenbach and Tielens (1997), van

Dishoeck and Blake (1998); and also in books edited by van Dishoeck (1997) and

Hartquist and Williams (1998).

3.3 Interpretation of OH Masers in Star-Forming Regions

3.3.1 Physical Location

OH masers in star-forming regions were the first astronomical masers discovered, and

are generally the strongest masers. Early interpretations suggested that every maser

was the site of a proto-stellar object, but as soon as higher resolution observations

showed the masers to be still unresolved and ever more numerous then it was clear

that this could not be the case. Zuckerman et al. (1965) were the first to suggest that

the masers were associated with compact H II regions – which had also been shown to

be abundant in star-forming regions. Cook (1966) suggested a model in which the

masers were located at the interface between the H II region and the surrounding warm

gas. This received much support, and was finally confirmed by Reid et al. (1980)

with their maps of W3(OH) showing the masers did indeed cover the UCH II region.

Subsequent observations demonstrated that the masers of other star-forming regions

also lay over the top of UCH II regions, most notably in the work of Gaume and Mutel

(1987). The relationship between OH masers in star-forming regions and H II regions


42

is now so well established that these regions are sometimes called H II/OH regions;

the existence of one of these phenomena is often a signpost to the other.

The exact nature of the zone around the compact H II region has been the

subject of much debate. Much of this debate has followed closely from the study of

the W3(OH) itself, and as the proposed models for the structure of W3(OH) have

come and gone so the generalised model of OH masers in star-forming regions has

followed. In this respect, an important aspect of the results in this thesis may well be

in advancing the generalised model of masers in this environment. The first model

proposed was that of Elitzur and de Jong (1978) who suggested that the maser zone

was an expanding compressed shell between the shock and ionised fronts. As is

discussed more extensively in chapter 5.7, alternative models have since been

suggested by e.g. Reid et al. 1980 who proposed that the masers were within in falling

material still accreting onto the star; and Bloemhof et al. (1992) who showed that the

general maser proper motion was divergent and proposed that the masers were in a

‘champagne flow’ around the outside of the UCH II region. This thesis provides

strong evidence (in chapters 5.7 and 5.8) that in fact the masers are in a rotating disc

of material, which may itself be expanding or disintegrating.

3.3.2 Occurrence

OH maser emission in the ground state main-lines has long been used as a tracer of

UCH II regions and YSOs in general, but many previous surveys have had large

position errors (> 10'') and so been unable to prove conclusively that emission at either

line was actually coming from the same region. The survey by Caswell and Hayes

(1987) suggested that of 1665 MHz masing sites, 90% had 1667 MHz emission, and

about 7% had 1612 MHz or 1720 MHz emission. The satellite lines rarely occurred

together in the same site. The 1665 MHz emission was usually (90% of cases)

stronger than 1667 MHz by a (median) factor of three, but sometimes weaker by a

factor of four.

Recently, with the development of the Australia Telescope Compact Array,

surveys have been undertaken in the southern hemisphere Galactic plane with a

resolution of ~1'' that have shown conclusively star-forming regions where masers of

more than one of the ground state lines are present. These surveys, in particular those

of Caswell (1998, 1999), provide results in sufficient number to be statistically

meaningful. Caswell (1998) surveyed over 200 SFR maser sites in the main-lines and


43

found that the masers were typically confined to regions less than ~1'', confirming

results from an earlier survey of 70 maser sites which found that the median size of

OH groups was 0.6'' (Forster and Caswell 1989). Where distances are known, the

linear size is rarely greater than 6000 AU. Care always needs to be taken in

interpreting surveys because of selection effects; e.g. the occurrence of 1665 MHz

maser emission with 1667 MHz emission may be biased in a 1665 MHz survey.

Where 1667 MHz emission was present in Caswell’s survey it was generally weaker

than the 1665 MHz emission, but was sometimes stronger. The velocity range of the

100 strongest sources had a median value of 9 km s–1, although this is likely to be a

slight overestimate of the true value because the observations (with two orthogonal

polarised receivers) could not tell the difference between RCP and LCP emission, and

hence could not identify Zeeman splitting in the spectra. The Zeeman splitting for a

typical 7 mG field is ~4 km s–1. In a later addition to this survey of over 200 masing

SFRs Caswell (1999) reported 11 1720 MHz maser sources and 11 1612 MHz

sources. At only two of these sites (i.e. 1%) were both satellite lines present together

(this highlights the special nature of W3(OH), where all the ground-state lines and

many excited-state lines are found). Caswell (1999) also confirmed that 1720 MHz

masers in SFRs are quite variable, and found that there was no correlation between the

nature of the underlying UCH II region and the presence of either of the satellite lines,

confirming the finding of Gaume and Mutel (1987). However, Moore et al. (1988)

found a correlation between the peak flux densities of 1665 MHz emission and

infrared emission at 60 µm. The relationship suggested that a threshold of flux was

needed before masers would be present; then above that threshold, brighter masers

were associated with higher 60 µm fluxes. Caswell (1999) did note that there

appeared to be a correlation between the presence of 6668 MHz methanol masers and

the satellite lines. At every site of 1720 MHz emission there was methanol emission,

while at half of the 1612 MHz maser sites there was no methanol emission. This

suggested that 1612 MHz maser emission might indicate a slightly earlier stage of the

evolution of the UCH II region, since methanol is believed to generally favour

younger UCH II regions (H2O masers have been identified as the earliest maser sign of

YSO activity).

Usually, the radial velocity distribution of the masers appears chaotic, as does

the spatial distribution. However, work in this thesis shows that – at least in the case


44

of W3(OH) – there is in fact underlying order in the velocity distribution of the

masers; the main barrier to noticing this is the effect of Zeeman splitting in a region of

varying magnetic field. This may be why methanol and H2O (which show no Zeeman

splitting) have been more often identified with ordered motion (e.g. discs or outflows).

3.3.3 Environment

Early observations appeared to eliminate radiative pump models of SFR masers, on

the grounds that maser photons exceeded the number of UV photons. However,

models which involved far-infrared pumping survived, and were boosted after

Thronson and Harper (1979) observed that there was much more radiation from very

cool dust than had been predicted by extrapolating ground based measurements. The

dust emission peaks at ~100 µm, corresponding to a temperature of 30 to 50 K, and

the photon rate exceeds the maser photon rate. The inclusion of line overlap has made

models much more successful in producing maser emission at dust temperatures

which match observations.

Recent modelling work has been conducted by e.g. Cesaroni and Walmsley

(1991), Gray et al. (1991, 1992) and Pavlakis and Kylafis (1996a,b,c). These have

mostly been using the Sobolev (or ‘large velocity gradient’) approximation, and been

successful at reproducing maser emission in the ground-state lines at least (although

not always at excited states). These models produce valuable diagnostic information

about the conditions in the masing areas; in general predicting that the gas kinetic

temperature lies in the range TK = 30 to 150 K; [H2] = 106 to 108 cm–1; [OH] / [H2] ~

10–5; and velocity shifts of up to a few km s–1. Dust temperatures, TD, are generally

assumed to be above TK. In particular, the modelling work of Gray et al. (1992)

makes the following predictions about the ground lines:

1. Bright 1665 MHz masers require FIR line overlap and form in accelerating

flows of gas where the velocity field exists over long enough distances to

produce shifts of a few km s–1. Higher kinetic temperatures increase the

necessary velocity shift required for 1665 MHz masers to dominate, and

therefore suggest the 1665 MHz dominant emission is confined to lower

temperatures than ~100 K. Powerful emission arises at [OH] > 20 cm–3, but

this threshold rises for higher TK. In general, powerful 1665 MHz emission

should dominate in regions below 75 K and [OH]>15 to 20 cm–3, with velocity

shifts greater than 1.5 km s–1. Saturation begins at very short path lengths of


45

about 1 AU under these conditions. These are probably the easiest conditions

to satisfy in SFRs, which explains why 1665 MHz masers are the most

common.

2. 1667 MHz masers can form with or without the presence of FIR line overlap.

In the absence of line overlap 1667 MHz masers form in the temperature range

50-75 K with [OH] ~ 60 cm–3. In the presence of line overlap, 1667 MHz is

enhanced in an accelerating flow for small velocity shifts, but weakened at

larger velocity shifts. This is in contrast to 1665 MHz emission, which may

dominate when the shift becomes too large for 1667 MHz emission. The range

for which 1667 MHz dominates increases as [OH] drops, such that 1667 MHz

dominates up to velocity shifts of 3 km s–1 at [OH] ~ 10 cm–3. Thus, 1667

MHz dominant emission could indicate lower OH densities.

3. 1720 MHz masers may also form with or without FIR line overlap. The

modelling suggests that in the absence of line overlap 1720 MHz emission is

indicative of lower TK than 1667 MHz emission. Maser emission is reduced at

[OH] > 100 cm–3. In the presence of line overlap, 1720 MHz emission can

become the dominant line at relatively low velocity shifts of less than 1 km s–1.

Line overlap also allows maser emission to continue at [OH] > 100 cm–3.

4. Strong 1612 MHz emission generally requires the presence of line overlap and

strong FIR radiation fields. In such conditions, 1612 MHz masers dominate at

colder temperatures, TK = 30 to 40 K for velocity shifts > 2 km s–1. The

masers may form in lower density gas with [OH] ~10 cm–3, and hence may

indicate colder, less dense flowing zones.

46

Chapter 4

Data Reduction

The data analysed in this thesis was observed on the Very Long Baseline Array

(VLBA), which is run by the National Radio Astronomy Observatory (NRAO) in the

USA (http://www.aoc.nrao.edu/vlba/html/VLBA.html). The VLBA is a system of ten

identical radio antennas controlled remotely from the Array Operations Center in

Socorro (New Mexico) that work together as the world’s most extended, dedicated,

full-time astronomical instrument. Construction started in February 1986 and was

complete by May 1993. The first observation using all ten antennas was on May 29,

1993, after a total construction cost of $85 million (cf. Hubble Space Telescope ~$1

billion!). Each antenna has diameter 25 m and weighs 240 tons, and are a very similar

model to that used in the Very Large Array (VLA). The locations of the 10 antennas

are shown in figure 4.1. The 10 antennas give 45 baselines, the longest (which

determines the telescope’s maximum resolution) of which is 8611 km (~5000 miles).

The VLBA can observe in 9 bands from 330 MHz to 43 GHz; these frequencies

translate to wavelengths from 90 cm to 7 mm. More comprehensive technical

specifications for the VLBA can be found in the VLBA Observational Status

Summary (Wrobel 2000).

4.1 Interferometry

The theory of interferometry is a complex blend of electronics/receiver technology

and Fourier transform mathematics, which is beyond the scope of this thesis. The next

section will outline the basic principles that are relevant to the data reduction. For the

dedicated reader, the books by Thompson et al. (1986) and Rohlfs (1986) cover the

area of interferometry much more thoroughly.

Chapter 4: Data Reduction

47

4.1.1 Detection in Radio Telescopes

Radio telescopes measure the passing of an electromagnetic wave, as opposed to

optical telescopes, which measure the absorption of photons. If a radio telescope of

area Ae detects a total energy E, in an integration time τ, in a receiver of bandwidth

∆ν, then the mean power is

P = E / τ (4.1)

And the mean power per unit frequency, or mean spectral density, is

Pν = P / ∆ν (4.2)

Mauna Kea Hawaii

Owen’s Valley California

Brewster Washington

North Liberty Iowa

Hancock New Hampshire

Kitt Peak Arizona

Pie Town New Mexico

Fort Davis Texas

Los Alamos New Mexico

St. Croix Virgin Islands

Figure 4.1: The NRAO’s VLBA sites.


48

(4.3)

(4.7)

The spectral flux density (usually shortened to simply ‘flux density’) is

e

2A

PS

ν∆=

For an astronomical object, values of flux density would always be extremely small,

so the unit of flux density – the Jansky – is defined as 10–26 W m–2 Hz–1. Flux density

is not an intrinsic property of the source, since it depends both on the power of the

source and the distance to the source. The surface brightness, B, of the source is the

flux density per unit solid angle, Ω, of the source. Brightness is independent of

distance to the source (neglecting transit effects such as scattering).

From the Nyquist Theorem, the noise power in a receiver can be specified in

terms of the temperature, Ts, of a matched resistive load producing equal power to an

equivalent noise-free receiver:

PN = k Ts ∆ν (4.4)

Contributions to the system temperature come from resistive losses in devices and

circuits, absorption losses, shot noise on bias currents in devices like transistors,

pickup of radiation from sky and ground; and losses in feeds, polarisers and

transmission lines ahead of the receiver.

The radio source power (i.e. signal) in the receiver is

PR = k Ta ∆ν (4.6)

Where Ta is the antenna temperature, and so

e

ak2A

TS =

The antenna temperature is converted into a measured flux via the gain, which is a

measure of how well a flux falling on the antenna is converted into an output signal,

measured as a temperature by comparison with a calibrated noise source. It is

measured in Jy / K. This is effectively a measure of the sensitivity of the antenna.


49

4.1.2 Aperture Synthesis

In a single dish radio telescope, diffraction broadens the response to a point source.

The angular size, θ, of the half-power level of the main lobe of the pattern is known as

the half-power beamwidth, and is related to the diameter, d, of the measuring

instrument and the wavelength of radiation, λ,

θ ≈ λ / d (4.8)

The projection of θ onto the sky is known as the beam. The relation of θ to λ is a

major problem for radio astronomy because it greatly reduces the resolution of any

single dish radio telescope that it is possible to construct. Antennas above 100 m in

size are prohibitively expensive and difficult to make, limiting single dish resolution

to just 0.1° at 18 cm. Aperture synthesis is the way around this.

A parabolic reflector ensures that the path length (and therefore time of

journey) is equal from the source to the detector for every element of the detecting

surface. The same effect could be achieved if the parabola were flattened out into a

plane and the path lengths of the signals were made equal afterwards by adding the

correct required delays to individual signal paths. This is called an ‘adding

interferometer’ or ‘tied array’; shown in figure 4.2. Different parts of the sky can be

viewed by varying the delays to the individual elements. This relies on two basic

assumptions: 1. The source is in the ‘far-field’ (i.e. that oncoming waves are plane

parallel) 2. The source is spatially incoherent (i.e. radiated signals from any two points

on the source are uncorrelated (There is some speculation that masers may be more

coherent than ordinary chaotic light, but since masers are rarely resolved this is not a

problem). Since the antennas are widely separated, there will be a time delay in the

detection of the wave at one antenna to the other. Because the earth is rotating, there

will also be a time derivative of the delay (the delay rate, or fringe rate). Additional

rates will be present from the motion of the earth around the sun.

In modern interferometers, the signals from individual elements are multiplied

together in pairs (or baselines), in a ‘correlator interferometer’. This creates an

interference pattern where maxima are separated by λ / d (the fringe spacing). In this

case, the amplitude of each fringe pattern contains information about the strength of

the source, and the relative phase contains information about the source position. It is


50

(4.9)

important to remember that that in some ways the phase of the detected signal is more

important than the amplitude, since the source distribution is generally more important

than the relative intensities.

An extended source can be treated as a collection of point sources, each of

which contributes to a fringe pattern. What is measured in an interferometer is a

superposition of sinusoidal functions – a Fourier series – from an unknown source

structure. This is called the visibility function. In order to determine the brightness

distribution of the source the function needs to be broken down into its component

parts to find all the amplitudes and phases of the source components. This is the task

of data reduction, since it is the visibility function that the telescope operators at the

VLBA hand to the observer after correlation in the correlator at the Array Operations

Center.

4.1.3 Visibility Data

If a radio source in the direction R radiates and produces a time variable electric field

E(R, t), then the correlation of the field at points r1 and r2 (the antenna locations) at

frequency ν is

)(*)(),( 2121 rrrr EEV =ν

Figure 4.2: Diagram of a ‘tied array’. The use of appropriate delays in the signal feeds can point

the array in different directions. ‘d’ is the single aperture diameter; D is the maximum baseline

length.

d

D

Plane parallel waves from SourcePlane parallel waves from Source

Delay

Correlator


51

(4.10)

From this, using the two basic assumptions mentioned in the previous section – that

the source is far field and spatially incoherent, the spatial coherence function can be

derived:

ΩIV i de)(),( c/)(221

21 rrssrr −−∫= νπνν

where Iν(s) is the observed intensity of the radiation field, s is the unit vector in the

direction of R, and dΩ is the solid angle subtended by the source. Note that the

function depends only on the separation vector r1-r2, and not on the absolute

positions. Therefore, more can be learnt about the correlation properties by holding

one point fixed and moving the other point around.

The baseline may be represented as a vector, (r1–r2), which has magnitude the

separation of the antennas and an angle that is the orientation of the baseline. The

projection of this vector onto the celestial sphere can be though of as a point in a

plane, called the aperture plane or u,v plane (The w axis comes out of this plane to

form a right handed coordinate system). u and v are measured in wavelengths at the

centre frequency of the band. As the earth rotates, the baseline projection on the

celestial sphere rotates and foreshortens. The locus of the length and direction of the

baseline projection is an ellipse, the parameters of which depend on the declination of

the source, the latitude of the midpoint of the baseline and the length and orientation

of the baseline. The superposition of all the ellipses from all the baselines is called the

u,v coverage; the u,v coverage for the VLBA observation is shown in figure 4.3. The

u,v coverage determines the scales of structure (‘spatial frequencies’) that can be

measured by the interferometer. A single dish of the aperture of the maximum

baseline would entirely fill the space in the u,v plane – so, clearly, the fewer gaps in

the coverage the better.

A visibility is a measurement of the phase and amplitude at a point in the u,v

plane. Since every time and baseline (antenna pair) has a corresponding point in the

u,v plane, it is possible to rewrite the spatial coherence function (equation 4.10) in

terms of u and v:

∫∫ +−= yxyx,Ivu,V vyuxi dde)()( )(2πνν


52

(4.12)

(4.13)

where Iν (x,y) is the intensity distribution of the source and Vν(u,v) is the visibility

function.

4.1.4 Imaging

Since equation 4.11 is a Fourier transform, it can be formally inverted:

∫∫ += vuvu,Vyx,I vyuxi dde)()( )(2πνν

Of course, we can see from e.g. figure 4.3 that we never have full u,v coverage from

an interferometer, but only samples on the u,v plane. The sampling can be described

by a sampling function, S(u,v), which is 1 where data have been taken and 0 where it

has not. Equation 4.12 now becomes

∫∫ += vuvu,Svu,Vyx,I vyuxi dde)()()( )(2D πνν

Meg

a W

avel

engt

h

Mega Wavelength-40 -20 0 20 40

50

40

30

20

10

0

-10

-20

-30

-40

-50

Figure 4.3: The u,v coverage of the VLBA during the observation.


53

(4.15)

where νDI (x,y) is known as the dirty image or dirty map. The dirty image is related to

the true intensity distribution, Iν (x,y), through the Fourier convolution theorem:

νDI (x,y) = Iν (x,y) ∗ Β (x,y) (4.14)

where the * denotes the convolution, and

∫∫ += vuvu,Syx,B vyuxi dde)()( )(2π

where B(x,y) is called the synthesised beam, dirty beam or point spread function, and

has unit intensity. Note that it is the Fourier transform of the sampling function. In

order to determine the true brightness distribution, the dirty beam must be

deconvolved from the dirty map. The most common deconvolution algorithm in radio

astronomy is ‘Clean’ (Högbom 1974), which considers a source to be a collection of

point sources. This process works iteratively:

1. The dirty beam and dirty map are determined from the visibility data.

2. The location of the peak flux in the dirty map is found, and the dirty beam is

subtracted at this point (in practice, cleaning gives better results by subtracting

just a small fraction of the dirty beam each time). Remember that the peak

flux is treated as a point source, and the dirty beam is the point source

response; so this has the effect of correctly removing all intensity associated

with the peak, i.e. side lobes and artefacts. This leaves behind a residual map,

and the position and intensity of the peak flux are stored as a clean component.

3. Go to 1, unless there are no points of flux greater than a specified noise level.

4. Convolve the clean components with the clean beam. The clean beam is the

full width at half-maximum (FWHM) of the main lobe of the dirty beam. The

size of the clean beam is dependent on equation 4.8.

5. Add the last residual map to the clean map.

When cleaning maps it is important to consider weighting of the u,v points.

Weighting individual antennas is not an issue for VLBA data because all the antennas

are identical, and sensitivities vary only slightly due to local conditions. Each track

(arc) in figure 4.3 contains the same number of measurements, yet clearly the outer

tracks are longer, and therefore the u,v points are more spread out. This means that


54

the u,v density is higher in the centre of the plane, an effect which can be seen in

figure 4.3, where the points in the outer arcs are more spread out. Natural weighting

just weights the u,v points according to σ–2, where σ is the visibility error. The leads

to the lowest noise, but means that the inner u,v range (lower resolutions) will

dominate – resulting in a larger beam. Uniform weighting corrects for the local

density of u,v points, effectively weighting up the outer parts of the u,v coverage.

This increases the noise slightly, but results in a smaller beam, and hence higher

resolution.

4.1.5 The Stokes Parameters

Recall from chapter 4.1.2 that the signals from antennas are multiplied together in

pairs. Radio telescopes are equipped with two separate complementary polarised

feeds; in the VLBA, these are in the form of oppositely circularly polarised receivers

(other telescopes may use orthogonal linearly polarised feeds). This gives a total of

four independent signal measurements of the source; when multiplied together in the

correlator they give four correlations: the same-hand L1L2 and R1R2 correlations and

the cross-hand L1R2 and R1L2 correlations (where subscripts 1 and 2 indicate feeds

from the two antennas). These four correlations are sufficient to define four rather

more useful quantities called the Stokes parameters (Stokes 1852), as is shown in

table 4.1. The relationship of the Stokes parameters to the actual properties of the

radiation is shown in figure 4.4.

Hence, from combinations of the Stokes parameters it is possible to calculate the full

polarization properties of the measured source intensity. From figure 4.4 it can be

seen that for pure circular polarization β = 45°, and so Q = U = 0. For pure linear

polarization β = 0, so V = 0. If V is negative then this represents LCP emission; if

Circularly Polarised

Feeds

Parameters to which

Instrument is Sensitive

L1 L2 I–V

R1 R2 I+V

L1 R2 Q+iU

R1 L2 Q–iU

Table 4.1: Relationship of the Stokes parameters to the polarized correlations.


55

(4.17)

(4.18)

(4.19)

(4.20)

positive then RCP emission. In this way the fraction of circular polarization present in

a source is

V / I = sin(2β) (4.16)

The total linearly polarised intensity is

22 UQ +

and so the fraction of linear polarization present is

)=+ β2cos(

22

I

UQ

Likewise, the total polarised flux is

222 VUQ ++

and so the total fraction of polarization present is

I

VUQ 222 ++

Figure 4.4: Diagram illustrating the relationship between the Stokes parameters and the

polarization properties of the electromagnetic radiation. The ray is propagating into the page. Note

that the double angles enter into the equations – this is because of the squaring involved in going

from amplitude to intensity.

a sin

(2β)

χ)(2(2β)

(2β) (2χ)β

χ

cosa β

V =

U =

Q =

I =

Plane ofReference

β

cos

cos

a

a

a

a

2

2

2

2

cos

sin

sin


56

The angle χ can be found from U/Q = tan(2χ). Note however that because the

signs of both Q and U are important then an arctangent function must be used which

uses the signs of both arguments to determine the quadrant of the return value (i.e.

arctanU /Q ≠ arctan–U / – Q).

4.2 Data Acquisition

4.2.1 AIPS

For the purposes of the following description, ‘data reduction’ has been defined to be

the processes involved from the receipt of the correlator output on magnetic tape to

the production of the map images. Under this interpretation, all of the data reduction

was carried out within the NRAO’s Astronomical Image Processing System (AIPS)

software suite of programs (http://www.cv.nrao.edu/aips/).

The AIPS suite is a mature project, started in 1978 in Vax/VMS environments,

but has been under continuous development since. It now consists of nearly 1,000,000

lines of FORTRAN code in 300 distinct applications or ‘tasks’. The tasks allow for

the complex mathematics outlined in the previous two sections to be carried out by the

observer without the extremely high level of mathematical understanding that would

be otherwise needed – although it should be pointed out that ‘user friendliness’ is most

certainly not one of the attributes of AIPS (AIPS is a world away from ‘point and

click’!)

4.2.2 The Observation

The observation itself (observer label BD31-W3OH) was on 2nd August

1996, from ~7 am until 8 pm (UTC), using all 10 antennas of the VLBA. The data

was reduced during a month-long visit to the Array Operations Center in Socorro,

where expertise in many of the more difficult aspects of calibration and imaging of

such data is readily available. In addition, powerful computers that were needed for

such a large dataset were available at the site, and probably reduced the computational

time required by a factor of ~5 over the computers available to us in Bristol at the

time. Since the reduction took ~600 hours of CPU time in Socorro, in Bristol this

would have taken half a year of CPU time – if indeed it had been possible at all.


57

The data on W3(OH) were recorded in dual circular polarization in a

bandwidth of 62.5 kHz centred on VLSR for W3(OH), –45 km s–1, assuming rest

frequencies of 1612.231 MHz, 1665.402 MHz, 1667.359 MHz, 1720.530 MHz for the

four lines. In addition, shorter observations were made in a wider band of 1 MHz.

These observations provide better signal to noise ratios for calibration where spectral

resolution is unimportant. Observed along with W3(OH) were several calibrator

sources which are vitally important in the data reduction. Table 4.2 shows the sources

observed and the reasons for their observation. Within AIPS, each antenna is given a

code, which is shown in table 4.3.

Each scan of the target source was 15 minutes long, and each scan of calibrator

sources was 10 minutes long. There were three scans of the target source for every

calibrator scan. Calibrator scans were spread out among the target scans, to ensure

that any effects that change during the time of the observation can be accounted for,

e.g. effects caused by the differing elevation of the source, or weather changes. Figure

4.5 shows the variation of Tsys of each antenna with time during the observation.

Several aspects are noticeable in the plot:

1. The calibrator scans can be seen evenly spread through the observation, and

have a lower Tsys than W3(OH) – the latter is because of the much lower flux

being collected from calibrator sources

2. The Tsys at Mauna Kea starts high and gradually falls over a few hours – this is

because the observation began at sunset over Mauna Kea, which is the most

western of the antennas

3. Likewise, the easternmost antennas (e.g. St. Croix) experience rises in Tsys

towards the end of the observation, where the sun was rising.

The recorded signals were correlated in full cross-polarization mode at the

VLBA correlator in Socorro, which allows for 128 channels across the 62.5 kHz

bandwidth. The resulting u,v data were then read into AIPS for reduction.


58

Source Position (RA, DEC – B1950) Scans Time / min Purpose

W3(OH) 02:27:03.8251 61:52:24.653 44 660 Target

1611+343 16:13:41.0643 34:12:47.909 1 10

1611CAL 16:13:41.0643 34:12:47.909 2 20 Pol. angle calibrator

3C84 03:19:48.1601 41:30:42.106 6 60 Fringe finder;

Bandpass calibrator;

Amplitude calibrator

3C84CAL 03:19:48.1601 41:30:42.106 7 70 Polarization calibrator

Table 4.2: The observed sources. Sources ending in ‘CAL’ are wide-band continuum observations.

1: BR - Brewster 6: MK - Mauna Kea

2: FD - Fort Davis 7: NL - North Liberty

3: HN - Hancock 8: OV - Owen's Valley

4: KP - Kitt Peak 9: PT - Pie Town

5: LA - Los Alamos 10: SC - St. Croix

Table 4.3: The VLBA Antenna codes.

1L BR353025

2L FD40353025

3L HN15010050

4L KP5550454035

5L LA605040

6L MK

Kel

vin 80

6040

7L NL3025

8L OV605040

9L PT403530

10L SC

TIME (HOURS)06 08 10 12 14 16 18 20

806040

Figure 4.5: Variation of Tsys for 1665 MHz LCP throughout the observation.


59

4.3 The AIPS Reduction Sequence

In principle, single polarization spectral line calibration has changed little from the

method described by Reid et al. (1980). However, dual polarization spectral line

reduction – also called spectral-line polarimetry – raises new problems. These are

covered in detail in Kemball et al. (1995); and the books by Zensus et al. (1995),

Taylor et al. (1999), and will only be summarised in the following sections.

4.3.1 Preliminary Flagging

The data were loaded into AIPS using the task FITLD, applying all the

standard digital and scaling corrections relevant to VLBA data. Then, the task

INDXR was run to index the u,v data. This speeds access to the u,v dataset by giving

offsets into the file for time ranges and source numbers.

Next, the data were examined in their rawest form to see if any obviously bad

data needed flagging. There are many ways of looking at the data at this point; for

example Tsys versus time (as in figure 4.5), which can be accomplished with the task

SNPLT; or plots of amplitude versus baseline length, which can be accomplished with

the task UVPLT, as in figure 4.7.

Even at this early stage in reduction, plots such as figure 4.6 are informative

and give an indication of what the source looks like. Figure 4.6 shows that as u,v

distance (baseline length) increases the flux drops steadily. This indicates that areas

of emission are being increasingly ‘resolved out’ by the longer baselines (i.e. the scale

of structure of such areas is larger than the beam of those baselines). There is a

suggestion that the trend stops at ~35 ‘megawavelength’ (6300 km) and becomes level

at this point. This indicates that there is probably not any structure beyond this point,

and therefore that most of the masers are probably of size ~5 mas.

Next, the task VLOG was run to segment and re-format the TSM calibration

file produced for the VLBA observation. This allowed for automatic flagging of data

by some tasks and provides additional calibration and sensitivity information for the

VLBA as determined from single-dish pointing observations. Editing and flagging

was assisted by viewing data by antenna (using SNPLT) to identify bad data. The task

EDITA allows for the examining and editing of antenna Tsys, gain amplitudes, phases,

delays, and rates.


60

4.3.2 Amplitude Calibration

At this point, the first stage of calibration begins: amplitude calibration.

Amplitude calibration uses measured antenna gains and Tsys values, as well as finding

corrections for voltage offsets in the samplers. First, the task ACCOR was used to

correct amplitudes in the cross correlation spectrum. These are caused by errors in

sampler thresholds and are fixed by using measurements of the auto correlation

spectrum (the 2-bit data correction). ACCOR creates a solution (SN) table that is

applied to the calibration (CL) table with the task CLCAL. The task CLCOR was

then used make the parallactic angle data correction. The correction for parallactic

angle needs to be applied for antennas that have alt-azimuth mountings because the

orientation of the polarised feeds changes with respect to the sky as the antenna tracks

the object. This is not an issue with polar mounted antennas, but all the VLBA

antennas are alt-azimuth. Then the task APCAL was used to take the Tsys and gain

curve data from the file produced by VLOG to produce an SN table, which was

applied with CLCAL. The task ANCAL applies similar corrections directly to the CL

table. Los Alamos was chosen as the reference antenna since the weather and

atmospheric conditions in the southwest of the USA are usually the best.

Jans

kys

Mega Wavelength0 5 10 15 20 25 30 35 40 45

60

50

40

30

20

10

0

Figure 4.6: Plot of amplitude versus u,v distance for 1665 MHz Stokes I channel 64; W3(OH) only.


61

4.3.3 Phase Calibration

The next stage was phase calibration, which removes errors and delays

introduced by the instrumentation. The amplitude and phase of the data can be

examined with the task POSSM. The aim was to produce flat phases as a function of

frequency (channel) in the data. Phase calibration is achieved by finding interference

fringes between the phases of the signals from the two antennas on each baseline.

This is needed because we don’t know absolutely: the frequency of the local

oscillators; the exact time; the fixed delays at the antennas; the exact location of the

antennas; the exact location of the sources; and the exact direction in which the

antennas were pointing during the observation. All of these effects produce slight

shifts and uncertainties in the data, not all of which are removed by the correlator and

so need to be removed by the observer. The task FRING was used (with 3C84 as the

calibrator source because of its simple point like appearance) to find fringes in the

selection of u,v data in order to find the group delay and phase rate calibration to be

applied to the whole data. This removes the slope of the phases through the channels

in the band (variation of the phase over the time of the observation will be removed

just before imaging). At this point, it was discovered that the Pie Town antenna was

not giving acceptable fringes on its baselines. If fringes cannot be found on the

calibrator source then it usually means the data will be useless, so the antenna was

dropped from the data. FRING creates an SN table containing the fringe frequency to

a few mHz and the delay to less than 1 ns, so that the data can be integrated over

timescales of a few minutes and used to make an image. The task SNCOR was used

to set the residual rates to zero and put a constant phase shift into the SN table that

was then applied with CLCAL.

4.3.4 Bandpass Calibration

After phase calibration, the bandpass was calibrated to correct for imperfect

electrical filters. The task CPASS (the successor to BPASS) was used for this, again

using 3C84 as a calibrator source (this time because of its flat continuum spectrum).

Figure 4.7 shows a complex bandpass plot calculated for the RCP feed at the Brewster

antenna at 1665 MHz (this is called a ‘complex’ bandpass because it includes a phase

correction as well as amplitude correction). The task SETJY was used to set the

measured flux from 3C84 to the correct absolute flux – determined from numerous

observations in the recent past. Finally, the task CVEL was used to remove residual


62

Doppler motions from the data. The rotation of the earth has already been removed in

the VLBA correlator, but the orbit of the earth around the sun and the orbit of the sun

around the centre of the Galaxy are removed at this point. The task CVEL applied all

the calibrations so far determined to the data and produced a new data file of corrected

data. At this point, the data were reinspected in plots of amplitude versus baseline,

and some more flagging was carried out with the task IBLED.

4.3.5 Polarization Calibration

Because the observation was intended to detect linear polarization angles, the

extra stage of linear polarization calibration was needed. For this the wide-band data

was copied out of the main u,v data with the task UVCOP. The wide-band data has a

lower noise level because of its greater bandwidth, and the spectral resolution is

unimportant for the polarization calibrations. FRING was run again on the wide-band

data to find fringes and produce an SN table that was applied as usual with CLCAL.

There are four separate factors that must be corrected for linear polarization

observations:

1. The Parallactic angle. This has already been corrected.

2. The time delay between the left and right feeds. This is introduced by e.g.

different path lengths of the LCP and RCP feeds. Without correcting this the

polarization angle is undefined. To correct for this a short selection of u,v data

that had a reasonable signal to noise ratio (SNR) for the cross-hand

polarizations (LR and RL) was copied out. FRING was run on this segment of

data to obtain corrections for the delays of each antenna with respect to the

Figure 4.7: Complex bandpass calibration calculated for the RCP feed of the Brewster antenna at

1665 MHz.

Channels0 20 40 60 80 100 120

1.6

1.4

1.2

1.0

0.8

0.6

0.4

0.2

0.0

420

-2-4

BR


63

reference antenna. The SN table produced was applied with CLCAL so that

the next run of FRING would use these corrections. The task SWPOL was

used to swap the polarization data streams in the u,v data file. This puts the

RL and LR polarizations into the parallel-hand polarization (LL and RR) slots

so that FRING can use them (FRING only works with parallel hand

polarizations usually). FRING was then run again as before. The task POLSN

was run on the SN table produced by FRING to produce another SN table

suitable for polarization calibration. This table was copied back to the main

dataset and applied to the data with CLCAL.

3. The polarization angle determination. Typically, the polarization angle as

determined by the VLBA is in error by a significant number of degrees

because of the phase offset between the right and left feeds. Comparison of a

calibrator source of known position angle is required. For this a map of the

calibrator source 1611+343 was made in images of Stokes I, Q and U. The

polarization position angle was measured to be –9°, but was known to be 32°

from recent VLA observations. This 41° discrepancy was corrected with the

task CLCOR.

4. Correction of feed impurities (‘D-terms’). There is a low level of leakage of

signal from one feed to the other due to imperfect polarization splitters, such

that without calibration an unpolarised source will appear to have a few

percent of polarised emission. The D-terms were corrected using the task

CLCOR by referencing the calibrator source 3C84, which is known to have

~0% left circular polarization. Finally, the task PCAL was used to calibrate

the data and determine the effective feed parameters for each frequency.

4.3.6 Image Calibration

The first part of image creation is to perform some more phase self-calibration.

The aim was to remove phase shifts over the length of the observation, caused

generally by atmospheric effects. Masers are good for self-calibration here because

they have bright point-like appearance, and the fact that they only exist in a few

channels is not important here (unlike in the earlier segment of phase calibration),

since the delay rates and complex bandpass have already been determined. For this a

channel was found with a single unresolved bright feature. It was known from

previous observations that such a feature existed in 1665 MHz channel 93 (–47.46 km


64

(4.19)

s–1). Therefore, FRING was run on just this channel, for the whole time range of the

observation to solve for the phase variations. The SN table containing this correction

was then copied to the whole dataset with SNCOR and applied with CLCAL.

More self-calibration was achieved by running the AIPS calibration task

CALIB on this same channel, specifying a point source model. This is an iterative

process that proceeds as follows:

1. Calibrate the phases with CALIB in phase solution mode. Create an SN table.

2. Fourier inversion of the u,v data and deconvolution of the image with the task

IMAGR, using the SN table as calibration.

3. Calibrate the phases again, using the image produced in 2. as the source model.

4. Go to 2, repeat three times.

5. Calibrate with CALIB again, this time with phase and amplitude solution

mode; but then copy the SN table back to the main data set.

After this the task SPLIT was used to split out only the W3(OH) data from the

u,v data. SPLIT applies all the calibrations and flagging and permanently alters the

data (up until this point all processes can be reversed by deleting the CL tables). Split

was used again to split out each of the four individual frequencies into separate u,v

data files.

Finally, IMAGR was run for each Stokes parameter for each frequency to

produce the final cleaned and calibrated map cubes. The first 4 channels and the last 5

channels were left out because they typically contain high noise because of the

bandpass limitations.

The theoretical resolution θ, of the VLBA is given by

maxkm

cm

2063~B

λθ ×

where λ = 18 cm and xmakmB =8611 km; which evaluates to 4.3 mas. For good images,

the cell size (pixel size) should be 25% of the beam size. In order to cover the whole

of the W3(OH) region to find any outlying masers a 4'' by 4'' field of view was

needed. In 1997, the maximum image size that was computationally feasible was

2048×2048 pixels (the Fourier inversion and deconvolution of all the data took 500

continuous hours of top-end CPU-time of the day and created 50 GB of map data!),


65

which would imply a cell size of 2 mas. For good sampling this would itself imply a

beam size of ~8 mas. Hence, the field of view desired could not be achieved without

compromising image quality at the maximum resolution. To achieve the field of view

with good image quality required that the resolution be decreased by a factor of 2 by

using a u,v ‘taper’ in IMAGR to weight down the u,v, data outside a circle of radius

20,000 kilolambda (3600 km). A declination shift south, of 1'', was also added

because the phase-reference feature in channel 93 – which ordinarily is the centre of

the image – was in the north of the map. The beam for 1720 MHz is shown in figure

4.8 (all of the beams are very similar), and the beam sizes are listed in table 4.4.

For weighting the u,v data AIPS uses a parameter called ‘robustness’ devised

by Daniel Briggs. This allows for a scale between –5 (uniform weighting) and +5

(natural weighting). An intermediate robustness of 0 was chosen for the imaging.

The observation was not phase-referenced, and so the absolute position of the

phase centre cannot be determined to accuracy greater than 0.1''. However,

observations of W3(OH) with the MERLIN array at 1720 MHz in December 1993

that were phase-referenced were available (Gray et al. 2001), and were used to

determine the absolute position of the phase centre, giving a position of: RA 02h 27m

03.s825 ± 0.s001, DEC +61° 52' 25.''089 ± 0.''01 (J2000). The MERLIN absolute

position has an intrinsic error of 10 mas associated with it, but additional uncertainties

arise from the possible motion of the 1720 maser with respect to the 1665 MHz

Mill

iAR

C S

EC

MilliARC SEC300 200 100 0 -100 -200 -300

300

200

100

0

-100

-200

-300

-400

Figure 4.8: The observational beam for 1720 MHz.


66

reference maser in the three years between the observations. As detailed in chapter

6.3.2 there is little doubt that the MERLIN reference maser has been correctly

identified; but the absolute motion of the maser is unknown. From chapters 5.8 and

6.3.7 the maximum likely speed of any maser is ~20 km s–1, or 2 mas year–1. Over 3

years this is a maximum likely motion of 6 mas. Alignment between the different

frequencies is excellent, and the error is of the order of 1 mas.

4.4 Measuring the Maser Emission

At this point, the data consisted of image cubes of 119 channels at 2048×2048 pixels

per Stokes parameter per frequency. Since masers are usually at or around the

resolution limit, fitting 2-dimensional Gaussians to the patches of maser emission

typically collects the flux very well and gives very accurate values for peak flux, total

flux and position.

4.4.1 Gaussian Fitting in AIPS (SAD)

For fitting the Gaussians the AIPS task SAD (search and destroy) was used. SAD

attempts to find all sources in a sub image whose peak is brighter than a given level

(CUTOFF). It searches in the image for ‘islands’ of pixels above this level, for which

it generates initial estimates of intensity and size of the emission. The Gaussian fitting

routine in the older AIPS task JMFIT is then used to determine the least square

Gaussian fit. SAD was run with the option to allow multiple peaks to be fit to islands

since many of the masers were close enough to be contiguous with their neighbours.

If the peak residual of subtracting a single Gaussian from an island of emission is

above a specified level (ICUT) then a two-Gaussian fit is attempted; and so on, up to 4

Gaussians. SAD can be forced to fit the clean beam (i.e. a point source) to every peak

discovered, but this fails badly for even slightly resolved and extended masers and so

was not used. Having fit all of the islands in the image SAD then creates a residual

image that can be inspected to see where Gaussian fitting failed. If the fitting routine

failed (by not converging), or if the fitting still left residuals that were above ICUT

then all fitting for that island is abandoned. For some complex areas of maser

emission, (especially in 1665 MHz) this was the case and these areas required extra

attention. SAD also deconvolves the clean beam from the fitted Gaussian component

size, and gives the true size of the source. Occasionally emission can be found which


67

(4.20)

has FWHM smaller than the clean beam – this is usually a result of nearby negative

noise flux, and so faint masers may be affected by this in noisy areas.

It was not generally possible to fit both the very high peaks and the very low

peaks with the same input parameters to SAD. The reason for this is centred on the

value of ICUT. The value of ICUT is not used directly by SAD, but rather is adjusted

in quadrature in the form

22 )GAIN(ICUT flux×+

where GAIN is another specified input parameter (typically 0.1) and flux is the peak

of the component in question. If ICUT is at a few times the rms noise of the map (i.e.

quite low) then if flux is large then this expression tends to GAIN×flux; if flux is low

then this tends to ICUT. Although this accounts for the widely varying fluxes of

peaks, it does sometimes fail, and in particular, it leads to very strong peaks with

slight deviations from Gaussian form being broken up into multiple smaller peaks that

may not represent reality. This is of course a limitation of fitting Gaussians to

physical objects that arise from gas clouds (and so will stray from Gaussian form).

Previous published data has shown attempts to retrieve every last mJy of emission

from extended masers by breaking them into numerous phantom masers, but this is

unrealistic: where masers stray from Gaussian form then Gaussian fitting will

necessarily only be an approximation, and so exhaustive efforts were not made to

retrieve every last mJy of flux in this way. Rather, where Gaussian fitting was

consistently failing on intense and slightly extended masers, ICUT was raised in an

attempt to allow a more approximate fit to one Gaussian. The result of this is evident

in comparisons with previous work (e.g. Garcia-Barretto 1988) where previously an

Frequency Maj. Axis Min. Axis Pos. Angle

(Hz) (mas) (mas) (mas)

1612 7.78 7.43 66.4

1665 7.6 7.48 -81.55

1667 7.55 7.44 -52.17

1720 7.52 7.38 84.66

Table 4.4: The beam dimensions at the four ground frequencies.


68

(4.21)

intense maser might be accompanied by myriad overlapping much weaker masers

(which in reality are probably just outlying emission of the main maser), in the current

data there would be a single slightly more intense maser.

The base level chosen for ICUT is of course of prime importance. For solid

guarantee of a real detection, the detection threshold for masers is usually considered

to be 5σ (i.e. 5×rms noise in the map). The theoretical noise in a VLBA image map

∆Im, can be found from the equation

intsm

)1(

1

tNN

SEFDI

νη ∆−⋅=∆

where ηs accounts for VLBI system inefficiency (~0.5 for the VLBA), N is the

number of antennas used, ∆ν is the channel width (488 Hz), tint is the total source

integration time (37200 s) and SEFD is the system equivalent flux density, which has

a value of 303 Jy at 1.6 GHz. Where dual polarizations were available, the noise in

each Stokes image is reduced by a factor of 1/ 2. For 9 antennas (after dropping Pie

Town) the theoretical noise in the maps was therefore 11.9 mJy/beam. From this, the

minimum threshold for ICUT could be 60 mJy/beam. Given that the most intense

masers have fluxes of ~70 Jy/beam, this gives a dynamic range of just over 6000, and

consequently dynamic range induced amplitude errors become significant in channels

with intense masers.

Figures 4.9 and 4.10 show the variation of actual channel noise for each of the

frequencies. Note the extra noise introduced by the dynamic range limited amplitude

errors at two peaks in 1665 MHz and a smaller peak in 1667 MHz. Note also that the

increase in noise in the 1665 MHz Stokes Q and U is larger than that in Stokes I and

V. Since in general the masers have <10% linear polarization this will have a major

impact on the relative level of linear polarization noise. The huge peak in noise in

channels 6 to 12 of the 1720 MHz data was caused by interference during the last two

hours of the observation – primarily at Mauna Kea – that was missed at the flagging

stage. Fortunately, no masers were expected to be in this part of the spectrum (see

single dish spectra, chapter 6.3.1). It was clear that the typical real noise in the maps

was about 15 mJy, so a value of 75 mJy was more appropriate for the 5σ value of

ICUT.


69

1612 MHz Stokes I rms Noise

0.000

0.005

0.010

0.015

0.020

0.025

0.030

0 20 40 60 80 100

Channel Number

rms

Noi

se /

mJy

1612 MHz Stokes V rms Noise

0.000

0.005

0.010

0.015

0.020

0.025

0.030

0 20 40 60 80 100

Channel Number

rms

Noi

se /

mJy

1612 MHz Stokes Q rms Noise

0.000

0.005

0.010

0.015

0.020

0.025

0.030

0 20 40 60 80 100

Channel Number

rms

Noi

se /

mJy

1612 MHz Stokes U rms Noise

0.000

0.005

0.010

0.015

0.020

0.025

0.030

0 20 40 60 80 100

Channel Number

rms

Noi

se /

mJy


0.000

0.005

0.010

0.015

0.020

0.025

0.030

0 20 40 60 80 100

Channel Number

rms

Noi

se /

mJy


0.000

0.005

0.010

0.015

0.020

0.025

0.030

0 20 40 60 80 100

Channel Number

rms

Noi

se /

mJy


0.000

0.010

0.020

0.030

0.040

0.050

0.060

0 20 40 60 80 100

Channel Number

rms

Noi

se /

mJy


0.000

0.010

0.020

0.030

0.040

0.050

0.060

0 20 40 60 80 100

Channel Number

rms

Noi

se /

mJy

Figure 4.9: The variation of (off source) actual channel noise for 1612 MHz and 1665 MHz. Flux

densities are all mJy/beam.


70


0.000

0.005

0.010

0.015

0.020

0.025

0.030

0 20 40 60 80 100

Channel Number

rms

Noi

se /

mJy


0.000

0.005

0.010

0.015

0.020

0.025

0.030

0 20 40 60 80 100

Channel Number

rms

Noi

se /

mJy


0.000

0.005

0.010

0.015

0.020

0.025

0.030

0 20 40 60 80 100

Channel Number

rms

Noi

se /

mJy


0.000

0.005

0.010

0.015

0.020

0.025

0.030

0 20 40 60 80 100

Channel Number

rms

Noi

se /

mJy


0.000

0.005

0.010

0.015

0.020

0.025

0.030

0 20 40 60 80 100

Channel Number

rms

Noi

se /

mJy


0.000

0.005

0.010

0.015

0.020

0.025

0.030

0 20 40 60 80 100

Channel Number

rms

Noi

se /

mJy


0.000

0.005

0.010

0.015

0.020

0.025

0.030

0 20 40 60 80 100

Channel Number

rms

Noi

se /

mJy


0.000

0.005

0.010

0.015

0.020

0.025

0.030

0 20 40 60 80 100

Channel Number

rms

Noi

se /

mJy

Figure 4.10: The variation of (off source) actual channel noise for 1667 MHz and 1720 MHz. The

1720 MHz noise peaks at ~700 mJy/beam – this explained in the text.


71

The development AIPS task MFQUV was used after SAD to search through

the Stokes Q, U and V cubes using the Gaussian data already determined from the

Stokes I cube. MFQUV reads the position of every Gaussian fit in Stokes I and then

measures the flux at the corresponding position in the other three Stokes cubes. This

has advantages over fitting Gaussians to every Stokes cube in turn (as well as being

much faster), because it allows measurements of flux to be made right down to the

noise. Only a small number of masers would have linear polarization fluxes above the

75 mJy threshold, but having already fit the maser in Stokes I the linear polarization

flux in Q and U can still be measured even if it is below the threshold for Gaussian

fitting. These measurements will however be subject to larger errors at lower signal to

noise levels. MFQUV is also useful because it avoids having to fit both positive and

negative flux Gaussians in the same map (which is not currently possible with SAD).

It should be noted though that where masers are overlapping MFQUV might return

incorrect values for fluxes. This is because Gaussian fitting will take account of

increase in flux caused by the overlap of adjacent masers, whereas simple

measurement of pixel-value flux will not – as is shown for the case of two Gaussian

sources in figure 4.11. In this way, polarised flux (from Stokes Q, U or V) may be

effectively counted twice. This may lead to a slight overestimate of the polarised

contributions in such cases – possibly taking the total of polarised flux above the

Stokes I flux (i.e. > 100% polarization), which is of course impossible. Fortunately

masers are rarely this close to each other, but this was occasionally the case where

weaker masers were merged with a larger maser, giving a measurement of >100%

polarization. The same principle can also give overestimated measurements in the

Stokes Q and U fluxes.

Another source of polarization values above 100% is where noise has

contributed differently to the different Stokes measurements, e.g. if noise reduces the

true Stokes I measurement, but increases Stokes Q, U or V then for highly polarised

masers polarization can again be >100%. This problem is greatest where signal to

noise ratios are lowest.

Errors are output from SAD, but they should be regarded as estimates. This is

because formal standard error estimates are meaningless since the objects are

normally well fit using only 1 to 4 clean beam areas – which means that the number of

parameters fit exceeds the number of truly independent samples. The errors are

calculated using the (signal free) rms noise in the maps using the method of Condon


72

(1997), which gives expressions from theory for two limiting cases: point source (the

beam area > 0.9 × the fitted Gaussian area) and expanded source (the beam area < 0.1

× the fitted Gaussian area). Intermediate cases are handled by interpolation between

the two limits. In general, the masers will be at or around the point source limit, and

so are covered fairly well by the point source theory.

Clearly, since many errors are related to the signal to noise ratio, for intense

masers the estimated errors are much smaller than for weaker masers. Typical

uncertainties for masers of different intensities are shown in table 4.5.

However, by far the largest cause of uncertainties are circumstances where

masers are closely overlapping. The Gaussian fitting of the masers is then at least

partly dependent on the parameters chosen for the fitting criteria, and may produce

significantly different results with different parameters. In such cases the error on

fluxes may be as high as 10%, and the error on positions may be nearly the size of the

beam (for resolved emission), i.e. ~8 mas. In general, the accuracy of fit parameters

for masers in complex areas of emission and well resolved extended masers should be

Gaussian Plots for Proximate Sources

0

5

10

15

20

0 10 20 30 40

Position

Hei

ght

Gaussian Plots for Proximate Sources

0

5

10

15

0 10 20 30 40 50

Position

Hei

ght

Figure 4.11: Two hypothetical Gaussian sources at separations of 7 (top) and 15 (bottom) units; the

dotted line is the sum. Without effective Gaussian subtraction, the measured heights are

overestimated when the sources are close. In the top plot the overestimations are 30% (left) and

27% (right); in the bottom plot the overestimations are just ~0% (left) and 1% (right).


73

treated with some caution, while isolated and unresolved masers have errors that

approach the theoretical minima.

4.4.2 Identification of Masers (HAPPY)

The data as output by SAD consists of Gaussian fits with accompanying attributes in

columns of text. At this point there were a very large number of such fits, a large

proportion of which were concentrated into just a few channels with intense maser

emission, and were caused by amplitude errors from dynamic range limitations going

over the detection threshold. It was decided, adopting the typical standard in this area,

that a candidate maser (called a ‘feature’ hereafter) would require detected emission in

three consecutive channels, each within a certain spatial separation (‘tolerance’) of its

neighbours. This is physically sensible because masers typically have FWHM values

of ~300 m s–1 and so will be spread over several channels, and it helps to eliminate

noise detections.

In order to process the Gaussian fits output by SAD, a FORTRAN program

called HAPPY was developed. HAPPY takes the strongest emission Gaussian fit in

the first channel and looks in the channel above for the nearest Gaussian spatially

within the tolerance. It then marks those two Gaussians as linked and moves on to the

next Gaussian in the first channel. Starting with the strongest emission Gaussian fit

ensures that the strongest masers are fit first preferentially to weaker masers. This

continues until the first channel has been completed, when the second channel is

Maser Peak Flux Integrated Flux Position Size Position angle Polarisation

Peak flux error error error error error error

(Jy/beam) (mJy/beam) (mJy) (mas) (mas) (degrees) (percent)

20 15 30 <0.01 0.01 <1 <0.1

5 15 30 0.01 0.03 1 0.3

2 15 35 0.03 0.10 5 0.8

0.5 15 40 0.10 0.50 10 3.0

0.1 15 60 0.50 1.50 20 15.0

Table 4.5: Typical errors as output by SAD for masers of an example range of peak fluxes. These

numbers assume that the noise in the maps is ~15 mJy/beam, as it is in most channels. In a small

number of channels in the 1665 MHz maps the noise is higher (see figure 4.10) – in these channels

the errors will be larger.


74

processed, and so on. At the end, HAPPY contains a list of features consisting of

linked Gaussians; the final stage is to filter the features according to the minimum

desired number of channels per feature – all those with less than the desired minimum

(usually 3 channels), and those still currently unlinked, are dropped and sent to the

rejected fits file.

The core algorithm of HAPPY was written in 1997 by Dr Malcolm Gray (now

at the UMIST) and consisted of ~600 lines of code and comments. This version of

HAPPY however was very limited in its abilities, in particular:

1. It could only handle Stokes I data and therefore could not identify polarization

of any type

2. It did not handle errors

3. It contained several serious bugs.

I built upon the core algorithm and greatly expanded it for the present work, such that

HAPPY now consists of ~1700 lines of code and comments. In particular, I added the

following functionality:

1. Polarization capability was added. This read in the output data of MFQUV

and used it to calculate where emission was LCP or RCP; the percentages of

circular, linear and total polarization; and linear polarization position angles

2. Processing of errors and weighting was added. This read in additional data

produced by SAD that is not in the standard SAD output, and fed it along with

the main data.

3. All bugs removed

4. Added output options for up to 4 different file formats, including: a summary

file which contained all the most important values of each maser found –

condensed into a table with one line per maser; a file output for post-

processing in IDL for maser lineshape fitting (see next section); and a rejection

file listing all Gaussian fits which failed to be collected into candidate features.

5. Peak brightness temperature calculated for each maser

6. Extensive crash-protection and program-error handling added.

The value chosen for the tolerance is of course of prime importance. The

value was in the end calculated empirically as follows: initially the tolerance was set

at twice the beam size (15 mas) and HAPPY run to see which features were generated.

The tolerance was then reduced by a small amount and HAPPY re-run to check how

this has altered the number and nature of features generated. There was little change


75

in the features down to tolerances of 6 mas, but between 6 mas and 4 mas there was

some change, and below 4 mas the maser features rapidly broke up and were dropped

because they failed to achieve 3 consecutive channels. In the end, the tolerance was

set at 4 mas, but the output of features from a tolerance of 6 mas was examined

manually as well to identify any masers in areas of complex emission that were being

lost at 4 mas but that ought to be kept (as indicated in the previous section, position

error in these extreme cases may be ~8 mas).

4.5 Data Analysis

The following sections describe the procedures used to extract the useful information

from the maser data.

4.5.1 Lineshape Fitting

The data required to fit a Gaussian lineshape to each maser feature was written into a

convenient file format by HAPPY. The code to fit the lineshapes was written in the

Interactive Data Language (IDL, by Research Systems, Inc.), which is a high level

language particularly suited to fitting, managing and manipulating data. A Gaussian

can always be found to go through 3 points, so the minimum required channels in a

feature to attempt fitting was 4. The fitting routine calls the popular ‘curvefit’

minimising routine (e.g. Bevington 1992), which finds the least squares fit of a

Gaussian to the data, obtaining the height, mean and standard deviation of the

Gaussian fit. The height is the maser peak flux density, the mean is the fit velocity

and the standard deviation is easily converted into the FWHM of the line. The code

then outputs a file that can be easily merged with the HAPPY summary output file to

display all the calculated information for each maser feature.

There are two possible flux values for fitting lineshapes to: the total integrated

flux of the Gaussian fit (in Jy) and the peak flux density (in Jy/beam). The latter is

better for investigating the properties of individual maser lineshapes because it reflects

the intensity at the centre of the maser, while the total integrated flux reflects the

intensity over the whole of the maser – and so is easily influenced where masers are

overlapping. Indeed, the total integrated flux is more useful where it is suspected that

a maser feature may in fact be an unresolved double feature, since this often shows up


76

in the lineshape as a ‘shoulder’ on the main peak, whereas the peak flux density

lineshape is less influenced by this unless the masers are extremely close.

4.5.2 Filtering the Masers

After generating the features by the method in chapter 4.4.2, two stages of refining

remained. The first was to manually ‘stitch’ any maser features back together that had

been broken up by the tolerance value, but which on visual inspection of the maps

should be considered a single feature. In addition, at this stage it was possible to

check the fitting produced by SAD to see if SAD had been over-zealous in attempts to

break up complex emission. Where it had, some re-runs of SAD with higher ICUT

values (as explained in chapter 4.5.1) were necessary to produce fits that most closely

matched visual inspection of the maps. In this sense it is clear that SAD and HAPPY,

either on their own or together, do not form a ‘black box’, and so the output of both

was checked at every stage to make sure that the generated features had a strong

foundation in the maps.

The second stage involved the removal of the dynamic range induced

amplitude errors from the feature list generated by HAPPY. These amplitude errors

affected channels with strong maser peaks, in particular: 1612 MHz channels 31, 32,

34, 35; 1720 MHz channels 58, 59, 67, 68; 1665 MHz channels 16, 17, 58, 59*, 60*,

61*, 62*, 63*, 64, 75, 76, 77, 78, 87, 88, 89, 90 (starred channels indicate badly

contaminated channels). The 1667 MHz maps did not have contamination with

amplitude errors. At this point HAPPY had identified 80 features at 1612 MHz, 534

at 1665 MHz, 40 at 1667 MHz and 34 at 1720 MHz. It was clear however, that a

great many of these features were of low flux (<300 mJy) and peaked in the noisy

channels. After a preliminary inspection of the masers outside the noisy channels, the

characteristics of real maser emission became apparent, and masers in the noisy

channels with the following attributes were labelled ‘suspicious’:

1. Features with deconvolved sizes (major axis or minor axis) of 0, i.e. their

apparent size was smaller than the beam and so could not be deconvolved.

2. Features with FWHM over 400 m s–1.

3. Features with apparent total polarization above 125%

4. Features with apparent total polarization below 70%

5. Features with apparent circular polarization above 125%

6. Features with apparent circular polarization below 70%


77

7. Features with apparent linear polarization above 25%

A great many features fell into the suspicious category, and all were then

inspected individually both in the images and in the output from HAPPY to see if they

were passable as real masers. For many the choice was fairly obvious – the noise

spots were often irregularly shaped, and many had much higher flux in the Stokes Q

and U planes (see noise levels in figure 4.9) which gave them impossible polarizations

(e.g. 500%!). In the images, the amplitude noise often follows symmetrical patterns,

especially ripples that centre on the intense maser causing the dynamic range

problems, and such patterns were visible in the early plots of the feature positions.

The majority of noise spots were eliminated easily, especially where they were a

member of two or more of the above categories, in particular categories 1 and 7.

Where the masers were trickier, say because they had only failed on one of the more

conservative categories, e.g. circular polarization below 70%, then individual contour

maps of the channels that contained the features were inspected. A useful

characteristic of dynamic range amplitude errors is that they mirror closely in velocity

the rise and fall of the intense emission that caused them. With this in mind, the

surrounding area in the image was checked for intense masers rising and falling in the

same channels as the suspect feature. Caution was needed here of course, because

there is no reason why a real maser could not just be mirroring another in the same

channels. Another feature of noise is that its shape is also sometimes related to the

location of the intense maser and this helped identification also. After a great deal of

analysis, the following numbers of features were removed from the data: 66 features at

1612 MHz, 337 at 1665 MHz, 6 at 1667 MHz and 26 at 1720 MHz. The remaining

features were therefore convincingly all real maser emission, and the process of actual

analysis of the masers could begin.

4.5.3 Zeeman Pairs

It has long been suspected that the similarity of RCP and LCP emission from maser

areas as detected by single dish observations was caused by the Zeeman effect

splitting and shifting the two spectra apart. Early VLBI observations finally

confirmed this by mapping individual masers of opposite polarization at the same

spatial location, separated only by frequency. The identification of Zeeman pairs

provides much more useful information than individual Zeeman components because

without pairs no conclusive values can be placed upon the true velocity of the masers.


78

This is because the magnetic field strength is both unknown and variable across the

source.

The maser features were examined for other masers closely coincident

spatially. At this stage polarization was not important, because π−rays could be in the

data and the polarization characteristics of these was not yet known because none had

ever been observed. Having cleaned the data of noise, those features which had been

rejected earlier by HAPPY for having just 2 adjacent channels were reincluded into

the data for the purposes of seeing if any were the weaker component of a pair. Their

inclusion here was justified, since in order to be a Zeeman component the feature

would have to lie within a very small area, within a certain range of velocity

(determined by the knowledge that the magnetic field ranges between ~1 and ~20 mG)

and be of opposite polarization. These three criteria reduced enormously the chances

of any noise emission making it back into the data as a Zeeman partner. In the event,

the spatial restriction proved extremely significant, and very few 2-channel features

made it back into the data (1 at 1612 MHz, 14 at 1665 MHz, 7 at 1667 MHz, 2 at 1720

MHz). Upon visual inspection, all the new additions passed examination. Naturally,

a tolerance value had to be adopted, within which Zeeman components could be

considered associated. It was decided that there would be two classes of Zeeman

components: class a would be the highest confidence associations where all

components were within 6 mas of each other; class b would be lower confidence

associations where all components were between 6 and 12 mas of each other. In only

two instances were candidates for π- rays or Zeeman triplets found, all the rest were

σ± Zeeman pairs. All Zeeman pairs where one component had only 2 channels were

declared to be quality b, regardless of their separation.

4.5.4 Demagnetisation

Having identified Zeeman pairs, the process of extracting the extra information

contained in them could begin. The first of these processes was ‘demagnetisation’.

This allows the determination of the true velocity of the maser from the velocity of its

two components. In order to do this the velocity of each component needs to be

determined. Where a maser had been successfully fit with a lineshape, the value of

the Gaussian mean was used, and the standard error of the Gaussian was used as the

error in the velocity. Where a fit had not succeeded, because either there were less

than four points in the lineshape or because the lineshape was not a Gaussian, then the


79

velocity of the peak channel was used. The value of the velocity in a channel is the

mid-point of that channel, and therefore the error was estimated at 0.5 channel widths.

Demagnetisation is then simply the averaging of the two component velocities. It

should be noted that the estimated errors on the peak of spots without a fit are not

standard errors in the statistical sense – the error represents simply ‘the range the peak

is most likely to fall in’. As such, it is not strictly valid to simply add them in

quadrature (multiplied by a factor of 0.5 because of the averaging) in order to obtain

the standard error in the demagnetised velocity. The errors on Gaussian fitted spectra

were typically about 0.5 channels in size also. Where these errors went over 0.5

channels in size they were capped at 0.5 channels – since it would have also been

possible to just use the peak channel and obtain a lower error.

4.5.5 Magnetic Field Strengths

The magnetic field strength in the maser is found by a similar method to the

demagnetised velocity. When the Zeeman components are more widely split than the

FWHM of each component, the magnetic field strength is related to the separation of

the peaks (see chapter 2.2.1). Using the component peaks and errors as in the

demagnetised velocity calculation, the magnetic field strength along with an error was

calculated for every Zeeman pair, using table 2.2.

4.5.6 Proper Motion

The absolute proper motion of the masers cannot be determined because of the

relatively large error in the absolute position determination of this and previous

observations compared to the presumed size of the proper motion. However, the

proper motion of the masers with respect to a reference maser can be calculated with

great precision if the reference maser can be identified in observations at two or more

epochs. Identification of such a maser in the data of Bloemhof et al. (1992) was easy

and that work had attempted a proper motion measurement that would also allow a

good comparison check. The reference feature of Bloemhof et al. was different to that

in the present work, but this was easily corrected by adding the position offset

between the two reference features to every maser in the current work. At this point,

the reference maser in maps of both epochs is the same and has position (0,0), but the

reference maser of the Bloemhof et al. data was in B1950 coordinates whilst the

present data was in J2000 coordinates. Precession, nutation and aberration cause first


80

and second order effects on maps made at different epochs. Since all offset

measurements are relative to the reference feature, the first order effects are

eliminated. However, the second order effects, called ‘differential precession’, distort

the coordinate axes because the corrections are only applied at the centre of the map.

In order to eliminate the differential precession the following steps were taken:

1. The Bloemhof et al. maser locations were made into full B1950 coordinates by

adding in the absolute position of their reference maser. The error in the

position of their reference maser is small compared to the size of the ~20' of

precession over the 19 years between the observations, and so will not affect

the results.

2. The complete set of B1950 coordinates were precessed to J2000 coordinates

using the Rutherford Appleton Laboratory’s COCO v2.2-3 Coordinate

Conversion software utility in high-resolution mode.

3. The new J2000 position of the Bloemhof et al. reference maser was subtracted

from all of the their masers to yield the J2000 offsets. This conversion was

found to have made a position correction, the magnitude of which was

dependant on the offset from the reference maser. Over the 2'' field of view

concerned the maximum positional correction was ~12 mas.

4. Masers from the data of Bloemhof et al. were identified in the current data, by

position, velocity and polarization attributes, and their proper motions

calculated. In general, there was no ambiguity in the identification of masers,

since the data of Bloemhof et al. had only listed masers that they had easily

identified in two previous epochs, and because the velocity of a maser is

accurately known and varies little with time – even when the maser is in a tight

cluster – and so is a good marker for each maser.

The effect of gravitational deflection by the Sun is order of 1 µas over a 2''

field of view and was ignored. The proper motions yielded were in the range ~3 to 13

mas over the 19 years, indicating the importance of removing the differential

precession.

4.5.7 Two-Point Correlation

The two-point correlation function is a useful analysis where there are a significant

number of sources distributed in an area of space. It gives an indication of the

distribution, and can indicate whether the distribution is random or shows structure


81

(4.21)

(e.g. clustering), and it can do this over the full range of scale lengths in the region

concerned. It does this by comparing the distribution of point-pair separations in the

real source map with point-pair separations in a random source map. The two-point

correlation ξ(r) is taken to be as follows

)()(

)(DR

DD

rn

rnr =ξ

where nDD(r) is the number of point pairs with separation r→r+dr in the real map

(called the ‘real-real’ data hereafter), and nDR(r) is the number of point pairs with

separation r→r+dr where one member of the pair is in the real map and one member

of the pair is in the random map (called the ‘real-random’ data hereafter). Clearly, if

the real data is random, then the correlation function will be a flat line at a correlation

of 1. Where the sources are more clustered than a random distribution, the correlation

will be greater than 1; where the sources avoid each other more than random data the

correlation will lie between 0 and 1.

The correlation was performed with only one maser from each Zeeman pair.

Since both masers in a Zeeman pair are effectively the same maser, only the position

of the stronger maser was entered into the correlation. If both masers had been

entered from every pair, then the correlation would have shown a huge spike below 12

mas scale lengths where the correlation between the components of each pair

dominated. The code to calculate the two-point function was written in IDL and

works as follows. Having read in the source positions, the program calculates the

separation of every single pairwise combination of sources. The program then creates

a random source list, and calculates the separation of every possible combination of

pairs containing one real source and one random source. The separations are then

‘binned’ into bins of a predetermined bin size, which determines the distribution of

separations. Finally, the correlation is the ratio of the real-real separation distribution

to the real-random separation distribution. In order to generate the correlation, some

limits had to be chosen. The first was the value of dr (bin size hereafter), and the

second was the geometry and size of the box within which the correlation would be

performed. Larger bin sizes lead to more samples per bin, and therefore a smoother,

better-averaged function. However, this smoothing also has the effect of eliminating


82

any fine structure that may be interesting. Small bin sizes keep the fine structure, but

create a more ‘noisy’ image. In the end, two bin sizes were chosen which gave the

best indication of both situations.

The chosen geometry of the box is a more complex issue, but is important

because it determines the random source distribution. To decide what would be

chosen, some tests on varying the size and shape of the box were performed. In order

to keep the distribution of random sources computationally simple, the geometry of

the box was chosen to be rectangular (as opposed to, say, circular). In order to see the

effect of choosing different size and orientations of rectangle, the two-point

correlation was performed for three test cases:

1. A ‘standard’ rectangular box just large enough to fit all the sources inside,

2400×2800 mas.

2. A ‘tall’ (vertically stretched) version of the standard box, 3000×4000

3. A ‘fat’ (horizontally stretched) version of the standard box, 4000×3000 mas.

Hence, the tall and fat boxes were the same but rotated 90°. Figure 4.12a shows the

bin counts (i.e. the undivided correlation) for each box with the 1665 MHz sources.

The random-random count (dotted line) is included, though it is not used in the

calculation, because it shows that the random-random data is identical for both the tall

and fat boxes – as would be expected. The real-real count is identical for all three

boxes. The real-random line can be thought of as related to the shape of both the real-

real and the random-random lines. Understanding why the real-random line for the

tall and fat boxes are different even though both the real-real and random-random

lines are identical is key to understanding how the box shape changes the correlation:

the change arises because the real source distribution is unsymmetrical and elongated

north-south – changing the box orientation allows different parts of the source to be

closer to the edge of the box, effectively changing the way the real-real and real-

random interact. Increasing the size of the box has a more obvious effect – it will

always increase the apparent correlation, since any source will appear more clustered

as the empty space around it grows ever larger. Figure 4.12b shows the correlations

from the three boxes. Clearly, they are very similar and qualitatively the same. Since

both the tall box and the fat box are larger than the standard box, they both show

correlations that suggest that the masers are more clustered. Apart from that, it is


83

Figure 4.12a) and b): a) Left, the bin counts (i.e. the undivided correlation) for each box with the

1665 MHz sources; solid line is real-real data, dashed line is real-random data and dotted line is

random-random data. b) Right, the correlation for each box.


84

clear that changing the box geometry has only a very slight effect on the relative

height of the peaks, and no effect on the positions of the peaks, which is the most

important information anyway. Therefore, the arbitrary choice of box geometry can

safely be ignored in the two-point analysis.

The final factor was of the nature of the random sources. If, by chance, a

random source should be placed extremely close to another, so close that in a sample

of a few hundred sources such a placement was unusual and significantly raised the

Figure 4.13: The correlation to a random set of sources with different numbers of iterative

averaging of the random field. Clockwise from top left: 1, 10, 100, 10000 iterations. Bottom, the

full scale random correlation.


85

correlation at short lengths; or if a number of sources managed to get into the far

corners of the box and raised the correlation at long lengths, then the final correlation

would be distorted. To avoid this, some sort of averaging of the random sources was

needed. The way this was done was to create a ‘random field’ of sources, whereby the

real-random correlation was performed many times with many random distributions

and an average built up in the bin counts. Figure 4.13 shows the correlation of a

random set of sources with the random field. Small differences are just about

noticeable between the 1, 10 and 100 iteration plots; but the 100 and 10,000 iteration

plots are indistinguishable. For the data analysis, all plots were run at 10,000

iterations, which took about 24 hours CPU time each. Also shown in figure 4.13 is a

large-scale plot of the random correlation, showing that the correlation does indeed lie

flat at ~1. Note the ‘noise’ at the high scale lengths end. This is because the only way

to reach these extreme lengths is by getting sources right in opposite corners of the

box, which because of the small amount of space there is not very likely. Because it is

not very likely, the ‘luck’ of the random sources is emphasised, even when correlated

with a random field. In order to get a perfectly flat correlation two random fields

would need to be correlated.

86

Chapter 5

Analysis of W3(OH) – 1665 MHz

The W3 complex lies in an obscured part of the Perseus Galactic arm, at a distance of

~2.2 kpc (Humphreys 1978), and is one of the most intensively studied SFRs in the

Galaxy. It is one of the many condensations on a ‘ridge’ of material swept along by

the expansion of the giant H II region W4, itself located just to the east. This scenario

illustrates the theory of triggered star-formation very well. At optical wavelengths,

this area is associated with the Cassiopeia OB6 cluster of massive stars, which lie in

the plane of the Galaxy at l = 134–138°. The total mass of the W3 GMC has been

estimated at 7×104 M (Lada 1978). The complex contains three major sites of star

formation: W3 Main (also called W3(continuum); it is itself broken into W3 East and

W3 West), W3 North and W3(OH). W3 Main is a cluster of numerous infrared

sources, compact, ultra-compact and hyper-compact H II regions ionised by a recently

formed cluster of O and B stars (Tieftrunk et al. 1997, 1998). Also present are OH

and H2O masers (e.g. Claussen et al. 1994), outflows (Hasegawa et al. 1994), and

several hundred low-mass stars detected in the near infrared (Mageath et al. 1996).

W3(OH) is situated 13' (~7 pc) southeast of W3 Main. In addition, 6'' (0.06

pc) to the east of W3(OH) is a strong H2O maser source with very faint OH maser

emission – often referred to as W3(H2O) – identified by Turner and Welch (1984).

Within 2' (~1 pc) of W3(OH) are over 100 near infrared sources, presumed to be low

mass stars. Because of the thickness of the dust in this area, this is thought likely to

only represent stars in the outer shell of the cloud (Zinnecker et al. 1993). The UCH II

region associated with W3(OH) is considered as prototypical of such regions, and is

one of the most luminous situated close to the sun. Its radio continuum spectrum is

consistent with a ZAMS O-type star (Dreher and Welch 1981, Baudry et al. 1993) still

embedded in an obscuring envelope of dust and ionised gas; the dust having a V-

Chapter 5: Kinematic Analysis of W3(OH) – 1665 MHz

87

magnitude extinction of ~50 (Wynn-Williams 1972). It is associated with a far-

infrared source with a total luminosity of 105 L (Campbell et al. 1989), which

corresponds to a main sequence O7 star. The high density and temperature, and small

size of the ionised region imply a short dynamical timescale of just a few thousand or

tens of thousand years, and as such, this represents an extremely young star. W3(OH)

and W3(H2O) are associated with gas condensations of ~50 M and ~70 M

respectively, of hydrogen densities ~107 cm–3 and kinetic temperature 100 to 140 K.

The two condensations are embedded in a larger, cooler, less dense cloud of mass

~2000 M , kinetic temperature ~20 K and density ~105 cm–3 (Wilson et al. 1991).

W3(OH) is unique in that it is the only source which exhibits emission and/or

absorption in all of the ground and excited-state hyperfine transitions of OH which

have been identified in space so far. The proximity of the region means that the

masers in W3(OH) are not significantly affected by interstellar scattering, unlike those

of e.g. W49. The maser emission is confined to the western half of the continuum

region, although very faint maser emission is found in the east outside the boundary of

the continuum emission. The UCH II region itself is a limb-brightened shell of

diameter ~1.5 " (~3000 AU) at 15 GHz (Dreher and Welch 1981), but has a larger size

of ~3 " at 1.6 GHz (Reid and Moran 1988). At 15 GHz, the region is expanding at 3

to 5 km s–1 (Kawamura and Masson 1998).

5.1 The Observational Data

The 1665 MHz total power spectra for RCP and LCP are shown in figure 5.1a. They

were taken at Dwingeloo in 1998 (Masheder, private communication), two years after

the interferometry observation. The spectrum is complex, which is typical of 1665

MHz in star-forming regions (Caswell 1998). It is immediately clear that the LCP and

RCP are not similar in form, and so are not a simple Zeeman split pattern.

Figure 5.1b shows the Stokes I spectrum of the maser fluxes calculated from

the channel maps. The 1665 MHz integrated flux contour map and the maser spot

map are shown in figures 5.2 and 5.3. Results of 2-dimensional Gaussian fitting of

maser emission in the maps are shown in table 5.1. Channel maps of the masers in

contour form can be found in appendix A.

As is immediately apparent, the 1665 MHz map is exceedingly complex

featuring 211 detected masers. There is no immediately obvious pattern to the

distribution (other than that the masers cover only the western half of the H II region),


88

although there is an indication of clustering which will be investigated in chapter 5.10.

The brightest masers (i.e. >10 Jy) appear in the north or in the south, with just one

bright maser in the centre.

1665 Stokes I Spectra

0

50

100

150

200

250

300

350

-50 -48 -46 -44 -42 -40

Velocity / (km/s)

Flux

/ Jy

Figure 5.1a) and b): a) Top, the 1665 MHz total power spectra for RCP (dotted) and LCP (solid).

b) Bottom, the Stokes I spectrum of the maser fluxes calculated from the channel maps.

1665 LCP & RCP Spectra

-20

0

20

40

60

80

100

120

140

160

180

200

-50 -49 -48 -47 -46 -45 -44 -43 -42 -41 -40

Velocity / (km/s)

Flux

/ Jy


89

ARC

SEC

ARC SEC0.5 0.0 -0.5 -1.0

0.5

0.0

-0.5

-1.0

-1.5

-2.0

Figure 5.2: The 1665 Stokes I integrated flux map. Channels 59 to 63 have been omitted because

of their high levels of dynamic range limited noise. This does not affect the masers significantly.

Contours are at 300 mJy/beam; beam at lower left. For reference centre, see chapter 4.3.6.


90

Figure 5.3: The 1665 MHz maser spot map. Contours are the 0.5, 7.1 and 16.6 mJy/beam levels of

a VLA 2 cm continuum map (Kawamura and Masson 1998)

1665 MHz Maser Spots

-2100

-1900

-1700

-1500

-1300

-1100

-900

-700

-500

-300

-100

100

300

500

-1500-1300-1100-900-700-500-300-100100300500700900

RA Offset / mas

DE

C O

ffset

/ m

as

Less than 0.400 Jy

0.400 -> 2.000 Jy

2.000 -> 10.000 Jy

Greater than 10.000 Jy

91

Tab

le 5

.1:

The

166

5 M

Hz

Mas

ers.

Dat

a is

for

pea

k ch

anne

l, ex

cept

last

fiv

e co

lum

ns, w

hich

are

for

the

who

lefe

atur

e. F

eatu

res

with

no

FWH

M m

easu

rem

ent e

ithe

r fa

iled

lines

hape

fitt

ing

or h

ad to

o fe

w c

hann

els

for

a fi

t. E

rror

s ar

e de

taile

d in

cha

pter

4.4

.1.

RA

and

DE

C o

ffse

ts a

re f

rom

the

refe

renc

e fe

atur

e.

P

eak

Flu

x B

right

ness

R

A

DE

C

Vel

ocity

D

econ

v.

Dec

onv.

D

econ

v.

Tot

al

Circ

ular

Li

near

C

ircul

ar

Fea

ture

Z

eem

an

FW

HM

C

hann

els

Inte

grat

ed

Den

sity

T

emp.

O

ffset

O

ffset

‘V

LSR

’ M

aj. A

xis

Min

. Axi

s P

os. A

ng.

Pol

. P

ol.

Pol

. D

irect

ion

Num

ber

Pai

r

Pre

sent

F

eatu

re

(Jy/

Bea

m)

(K)

(mas

) (m

as)

(km

/s)

(mas

) (m

as)

(deg

rees

) (p

erce

nt)

(per

cent

) (p

erce

nt)

Num

ber

(m/s

)

Flu

x (

Jy)

74.5

2 8.

51E

+11

-906

.7

98.3

-4

4.91

11

.3

4.9

41

90

90

12

R

600

5 37

0 28

61

5.1

68.8

7 8.

18E

+11

-905

.1

116.

0 -4

5.18

9.

5 5.

6 65

74

73

14

R

27

1 4

245

11

483.

6

51.6

4 3.

38E

+12

0.0

0.0

-47.

46

3.5

2.8

163

22

17

15

L 47

4 -

241

17

185.

8

49.1

3 7.

42E

+11

-909

.8

94.5

-4

4.82

7.

4 5.

7 42

93

92

11

R

61

0 -

- 6

192.

8

45.9

8 8.

39E

+11

-958

.1

-107

.6

-46.

41

8.0

4.3

27

101

101

1 L

383

20

423

21

403.

1

33.9

2 3.

16E

+11

-102

1.3

-127

.7

-46.

23

12.5

5.

5 81

98

98

0

L 60

1 -

317

16

325.

8

27.0

3 1.

86E

+12

-48.

6 -1

93.1

-4

5.18

3.

4 2.

7 16

0 96

96

0

L 11

1 22

30

1 14

12

1.6

21.5

7 9.

43E

+11

-138

.2

-119

1.3

-44.

65

4.7

3.1

155

102

102

0 L

91

38

265

14

89.1

19.3

9 1.

13E

+12

25.1

31

.5

-41.

13

3.4

3.3

154

96

96

1 R

8

8 30

2 12

89

.9

15.5

4 9.

76E

+11

-4.5

21

.0

-44.

03

3.5

2.8

156

90

90

2 R

64

9

- 14

88

.6

15.5

4 1.

32E

+12

25.7

31

.6

-44.

82

3.0

2.4

170

98

98

0 L

103

8 34

2 12

75

.7

11.4

3 2.

42E

+11

-846

.7

-173

1.8

-44.

91

8.3

3.7

104

102

102

1 L

97

- -

11

73.3

10.9

4 2.

81E

+11

-591

.4

-191

6.9

-45.

53

5.3

4.6

112

97

97

0 L

374

55

234

9 49

.6

10.5

2 2.

42E

+11

-519

.7

-177

8.2

-46.

14

9.3

3.1

119

102

102

3 L

354

49

403

10

90.4

10.0

8 1.

85E

+11

-691

.3

-187

2.7

-41.

13

6.4

5.3

131

98

98

1 R

9

54

316

11

64.1

9.28

1.

67E

+11

-477

.4

-177

6.5

-45.

26

8.5

4.0

120

95

95

3 L

147

50

495

13

101.

9

8.49

1.

36E

+11

-25.

3 -1

743.

7 -4

4.21

10

.4

3.8

51

97

97

0 L

62

61

- 13

94

.7

8.32

1.

04E

+11

-725

.0

-167

5.3

-44.

47

9.3

5.5

72

43

40

15

L 83

43

.1

247

11

47.1

8.31

2.

56E

+11

-706

.8

-167

0.5

-45.

62

5.1

4.1

23

97

97

4 L

367

- 23

9 8

32.9

7.26

1.

74E

+11

-220

.3

-547

.0

-45.

88

8.6

3.1

155

95

95

4 L

385

- 30

2 11

44

.6

7.24

2.

09E

+11

-490

.0

-178

8.6

-44.

21

6.5

3.4

114

73

72

12

L 73

51

-

12

62.9

7.12

1.

71E

+11

-591

.6

-191

6.9

-41.

57

5.7

4.7

97

100

100

1 R

18

55

24

1 8

32.8

6.68

7.

60E

+11

-168

.1

-48.

4 -4

6.32

3.

2 1.

8 14

2 97

97

4

L 40

1 15

30

9 10

29

.1

6.60

7.

21E

+10

-753

.4

-170

1.7

-45.

18

9.7

6.1

179

101

101

0 L

256

- 31

5 8

54.2

6.60

9.

89E

+10

-730

.2

-168

2.0

-42.

89

9.1

4.7

23

98

98

5 R

51

43

21

8 9

35.6

6.40

4.

04E

+10

-859

.1

-652

.2

-44.

21

16.3

6.

2 17

1 98

98

1

R

604

30

265

8 71

.2

6.27

1.

66E

+11

-138

.7

-117

9.3

-41.

66

7.5

3.2

173

101

101

1 R

22

37

33

9 10

40

.3

92

5.81

8.

27E

+10

24.2

39

2.9

-48.

95

13.4

3.

3 10

6 79

79

4

L 60

3 62

.1

229

8 29

.9

5.47

2.

88E

+11

-20.

4 -1

739.

3 -4

4.03

3.

7 3.

2 89

10

3 10

3 0

L 74

60

-

3 13

.3

4.97

1.

55E

+11

-138

.2

-119

2.2

-41.

22

5.9

3.5

153

104

104

0 R

15

38

30

0 7

24.4

4.95

1.

77E

+11

-302

.9

-107

8.6

-42.

63

6.4

2.8

1 97

97

1

R

42

33

- 7

22.7

4.66

1.

55E

+11

198.

5 -1

793.

5 -4

4.38

5.

5 3.

5 62

90

90

1

L 75

52

22

6 9

20.4

4.63

3.

64E

+11

7.6

-132

.5

-45.

44

3.5

2.3

6 98

98

1

L 35

6 -

255

6 16

.2

4.28

1.

80E

+11

27.5

16

.4

-45.

26

5.7

2.7

42

107

107

1 L

152

- 28

5 10

19

.3

4.25

2.

01E

+11

-271

.1

-178

3.8

-44.

82

4.0

3.4

42

66

66

3 L

107

- 20

1 7

14.4

4.21

2.

44E

+11

-489

.9

-177

0.1

-45.

18

5.0

2.2

144

99

99

3 L

151

- -

8 21

.6

3.94

1.

16E

+11

-491

.1

-176

6.0

-45.

79

7.2

3.0

111

96

96

4 L

380

- -

8 28

.0

3.80

1.

00E

+11

-513

.9

-177

4.7

-45.

44

7.8

3.1

130

102

102

3 L

355

70

- 7

26.4

3.73

7.

53E

+10

14.0

39

3.8

-48.

87

9.0

3.5

110

55

53

14

L 60

2 1

- 7

12.8

3.66

2.

96E

+11

-107

.5

-59.

1 -4

5.79

3.

1 2.

6 15

4 91

91

6

L 37

7 16

37

9 9

20.1

3.46

2.

28E

+11

25.0

-1

722.

0 -4

5.35

3.

4 2.

8 12

7 99

99

1

L 28

0 45

34

0 9

16.7

3.45

1.

54E

+11

-144

.7

-118

4.2

-44.

47

4.1

3.5

9 10

2 10

2 1

L 95

37

-

13

29.4

3.36

7.

62E

+10

-32.

0 -1

745.

4 -4

1.22

7.

8 3.

6 97

97

97

1

R

13

61

373

9 23

.1

3.22

1.

06E

+11

-210

.7

-576

.0

-45.

62

6.2

3.1

171

96

96

1 L

371

26

276

6 15

.0

3.07

5.

10E

+10

123.

4 -1

777.

1 -4

5.53

8.

3 4.

6 10

1 96

96

2

L 37

5 48

18

3 5

11.8

2.90

7.

64E

+11

-52.

2 -8

2.4

-45.

00

1.9

1.3

160

101

101

4 L

193

- 19

7 5

7.8

2.80

4.

92E

+10

43.1

37

3.9

-48.

52

11.6

3.

1 88

92

91

11

L

495

2 39

7 10

24

.0

2.74

6.

18E

+10

-850

.6

-616

.8

-45.

70

9.0

3.2

167

99

99

1 L

381

28

252

6 14

.0

2.67

4.

52E

+10

-104

8.6

-136

.7

-46.

05

10.5

3.

6 77

96

96

1

L 38

9 -

292

6 18

.5

2.58

8.

00E

+10

-418

.3

-181

5.5

-44.

47

6.7

3.1

123

95

95

2 L

94

- -

7 14

.7

2.51

6.

71E

+10

-699

.5

-162

9.7

-40.

52

6.0

4.0

72

99

99

2 R

3

- 18

4 5

8.1

2.46

6.

70E

+10

-162

.7

-112

6.0

-42.

89

6.9

3.3

22

98

97

3 R

50

34

-

7 12

.9

2.31

9.

94E

+10

26.4

-1

722.

4 -4

2.36

4.

7 3.

2 14

9 97

97

1

R

39

45

372

8 13

.1

2.29

5.

97E

+10

-960

.6

-64.

9 -4

5.35

9.

2 2.

7 14

4 93

93

2

L 36

5 -

314

9 13

.5

2.17

3.

91E

+10

34.1

-1

725.

2 -4

2.01

8.

5 4.

1 11

3 10

4 10

4 0

R

32

- 25

1 6

9.1

2.15

8.

23E

+10

-25.

3 -1

731.

6 -4

0.87

4.

4 3.

8 46

98

98

0

R

6 -

- 8

11.7

2.09

1.

44E

+11

8.2

-108

.8

-45.

53

3.9

2.4

7 98

98

1

L 37

0 -

262

6 7.

7

2.06

6.

34E

+10

20.6

10

.2

-45.

44

5.2

4.0

11

93

93

1 L

366

- -

7 10

.7

2.05

3.

18E

+10

-850

.5

-619

.5

-43.

86

9.3

4.4

165

100

100

1 R

65

28

24

7 6

11.0

1.82

3.

82E

+10

-745

.9

-168

3.1

-45.

26

6.7

4.5

30

94

94

4 L

276

- 27

2 6

9.3

93

1.81

1.

03E

+11

-407

.3

-182

1.0

-44.

21

4.8

2.3

140

76

76

5 L

77

- 23

8 5

6.1

1.78

9.

78E

+10

-48.

9 -1

93.2

-4

0.69

3.

9 3.

0 17

5 97

97

0

R

4 22

31

3 6

7.9

1.68

1.

34E

+10

-891

.6

-595

.2

-44.

38

9.5

8.5

115

99

99

2 R

82

27

31

3 7

14.6

1.66

9.

21E

+10

-76.

5 -1

407.

1 -4

2.01

3.

6 3.

2 86

99

99

2

R

33

41

236

5 5.

7

1.60

1.

14E

+11

34.0

37

4.0

-45.

26

3.0

3.0

0 22

18

10

L

314

- -

7 10

.1

1.56

5.

85E

+10

-902

.0

39.3

-4

7.11

4.

5 3.

7 12

3 98

98

3

L 47

8 7

- 12

9.

7

1.42

1.

01E

+11

-79.

6 -1

754.

8 -4

1.92

3.

1 2.

9 10

4 10

0 10

0 2

L 34

-

214

5 4.

4

1.31

2.

63E

+10

-349

.8

-117

9.2

-43.

33

7.3

4.5

26

100

100

4 R

56

36

32

1 8

8.1

1.30

5.

16E

+10

10.6

-9

3.5

-45.

62

6.9

2.4

25

95

95

3 L

361

- -

10

9.2

1.29

1.

92E

+10

-956

.3

-104

.0

-43.

33

9.6

4.3

15

98

98

3 R

57

20

35

9 9

10.0

1.29

6.

07E

+10

-148

.4

-116

9.8

-44.

38

3.9

3.4

119

102

102

2 L

93

- -

3 3.

6

1.29

6.

87E

+10

45.3

-1

730.

1 -4

1.66

4.

4 2.

6 1

100

100

2 R

27

46

-

6 5.

9

1.16

1.

99E

+10

-601

.5

-191

5.0

-45.

53

7.9

4.7

75

112

112

3 L

611

- -

3 3.

4

1.12

5.

41E

+10

-475

.7

-180

9.8

-44.

56

4.4

3.1

131

31

29

11

L 10

2 -

179

4 2.

9

1.08

9.

11E

+10

26.0

-2

3.4

-41.

13

3.3

2.3

48

95

95

0 R

16

-

338

6 4.

8

1.05

1.

69E

+11

16.2

-1

5.5

-41.

13

3.0

1.4

163

93

93

6 R

11

12

-

7 4.

3

1.05

6.

95E

+10

46.0

37

4.0

-45.

26

3.0

3.0

0 50

47

20

R

62

6 2

- 5

0.0

1.03

7.

25E

+10

-95.

9 -6

4.9

-46.

41

4.2

2.2

141

94

94

4 L

402

- -

9 5.

8

1.03

6.

29E

+10

-716

.7

-347

.7

-44.

38

3.4

3.2

72

100

100

3 R

89

23

25

8 5

3.9

1.02

3.

64E

+10

26.6

-3

2.4

-40.

78

5.0

3.5

177

98

98

2 R

7

14

248

5 4.

1

0.96

3.

41E

+10

-895

.4

66.5

-4

4.21

7.

0 2.

6 52

10

5 10

5 1

R

76

6 -

7 6.

0

0.95

4.

17E

+10

-99.

8 -1

380.

9 -4

5.18

4.

8 3.

1 15

3 91

91

5

L 30

1 -

194

4 2.

7

0.94

1.

11E

+10

-831

.1

-718

.3

-44.

21

11.7

4.

8 17

9 83

83

4

R

88

- 27

5 5

6.7

0.94

2.

50E

+10

24.5

39

3.3

-49.

66

12.5

2.

0 10

4 69

68

10

L

505

62

260

4 5.

5

0.93

5.

69E

+10

157.

4 32

2.6

-44.

65

3.7

2.7

104

85

84

2 R

10

1 3

- 9

5.5

0.92

4.

17E

+10

348.

6 -1

680.

0 -4

4.38

5.

2 2.

6 15

9 24

24

2

L 90

-

229

5 3.

2

0.92

5.

35E

+10

8.8

390.

6 -4

5.09

3.

5 3.

0 87

11

6

9 R

27

2 1

272

5 3.

4

0.91

2.

87E

+10

26.8

47

.6

-41.

48

6.1

3.2

4 92

91

4

R

24

- 21

2 4

3.0

0.90

1.

78E

+10

-258

.2

-688

.9

-42.

54

7.0

4.6

136

98

98

4 R

44

-

243

5 4.

1

0.90

1.

48E

+10

-101

7.6

-110

.8

-46.

32

7.4

5.2

172

96

96

10

L 60

9 -

376

4 5.

4

0.87

5.

51E

+10

-4.6

20

.9

-48.

52

3.4

3.0

30

91

91

4 L

497

9 36

7 8

4.8

0.84

2.

48E

+10

-275

.1

-749

.1

-42.

63

6.5

3.3

11

94

94

1 R

46

31

28

9 7

4.6

0.84

1.

21E

+10

-866

.3

-625

.2

-44.

12

8.7

5.1

168

92

92

1 R

60

8 -

234

4 3.

9

94

0.82

3.

15E

+10

-720

.6

-167

5.6

-45.

62

5.5

3.0

68

93

93

8 L

376

43

215

4 2.

5

0.82

2.

73E

+10

-826

.3

-174

2.1

-44.

91

4.9

3.9

65

75

75

6 L

255

- 25

3 4

3.0

0.79

1.

11E

+10

-737

.6

-170

0.6

-42.

80

8.7

5.2

173

94

94

7 R

52

-

263

5 5.

1

0.78

4.

28E

+10

-378

.4

-186

1.6

-44.

74

4.4

2.7

94

85

85

3 L

104

- 22

9 5

2.7

0.77

9.

96E

+09

-515

.5

-178

1.0

-42.

63

15.8

3.

1 11

8 92

92

4

R

49

49

334

8 4.

2

0.76

-

-167

.9

-49.

0 -4

1.05

0.

0 0.

0 0

99

99

2 R

14

15

20

3 4

2.0

0.76

2.

21E

+10

-9.1

-1

7.5

-41.

57

6.9

3.2

24

105

105

4 R

23

11

28

3 7

3.9

0.75

4.

47E

+10

-76.

0 -1

406.

7 -4

5.62

4.

2 2.

6 98

96

96

5

L 35

9 41

-

8 4.

7

0.74

4.

46E

+09

-104

8.4

127.

0 -4

5.79

15

.6

6.7

31

22

16

15

L 38

7 -

253

4 4.

3

0.74

1.

19E

+11

-56.

0 -1

426.

5 -4

4.91

3.

3 1.

2 15

3 10

7 10

6 7

L 19

8 -

- 3

1.5

0.73

1.

86E

+10

-301

.8

-143

4.0

-43.

95

6.9

3.6

22

97

97

2 R

69

42

24

9 6

4.1

0.73

2.

21E

+10

198.

1 -1

793.

9 -4

1.57

6.

4 3.

3 65

94

94

1

R

26

52

231

6 2.

3

0.71

3.

16E

+10

-876

.5

82.3

-4

6.85

4.

0 3.

6 11

5 93

93

7

L 47

0 -

227

5 2.

5

0.71

1.

32E

+10

-861

.3

-642

.1

-46.

05

9.4

3.6

179

96

96

5 L

607

30

205

5 3.

1

0.70

1.

93E

+10

16.3

-1

714.

8 -4

5.44

8.

2 2.

8 11

5 92

92

1

L 35

8 -

328

5 4.

8

0.69

3.

71E

+09

-831

.9

-706

.9

-44.

21

27.3

4.

3 17

1 10

8 10

8 3

R

87

- -

3 3.

9

0.68

2.

68E

+10

-488

.8

-178

8.9

-39.

99

5.2

3.1

112

98

98

5 R

1

51

252

5 3.

9

0.64

-

-137

.8

-120

5.6

-44.

91

0.0

0.0

0 74

74

6

L 50

8 -

- 2

1.3

0.61

2.

67E

+10

-302

.6

-108

0.1

-45.

44

5.5

2.7

172

99

99

3 L

372

33

- 3

1.8

0.61

5.

46E

+09

-835

.1

-680

.1

-44.

03

14.1

5.

0 15

7 72

72

3

R

605

- -

3 3.

3

0.60

5.

58E

+09

-838

.0

-670

.1

-43.

95

12.9

5.

3 15

6 10

2 10

0 5

R

606

- -

3 2.

9

0.59

1.

94E

+10

-476

.5

-177

9.5

-41.

84

5.5

3.5

67

97

97

2 R

31

50

21

7 4

2.1

0.56

-

-525

.0

-191

6.3

-45.

35

0.0

0.0

0 10

4 10

4 9

L 36

9 57

27

8 5

2.6

0.55

1.

39E

+10

-244

.6

-705

.2

-42.

19

8.2

3.1

169

92

92

6 R

38

-

256

8 3.

6

0.54

4.

62E

+09

-899

.5

-576

.4

-44.

12

10.9

6.

8 12

10

1 10

1 4

R

70

25

513

10

6.7

0.52

2.

62E

+10

-44.

9 -1

00.6

-4

1.40

6.

6 1.

9 10

3 99

99

2

R

19

19

269

5 2.

2

0.52

5.

77E

+10

-18.

4 -1

733.

5 -4

1.75

3.

2 1.

8 45

10

0 10

0 3

R

28

- 24

0 5

1.9

0.51

4.

31E

+10

-878

.8

62.9

-4

7.11

3.

1 2.

4 75

97

97

2

L 47

7 -

275

5 2.

2

0.49

-

-22.

6 -1

740.

5 -4

1.22

0.

0 0.

0 0

111

111

4 R

62

2 60

-

2 1.

4

0.49

9.

53E

+09

-622

.0

-191

8.0

-41.

40

6.9

4.7

99

97

97

5 R

20

56

25

4 4

1.9

0.49

1.

30E

+10

-893

.7

-599

.8

-46.

23

7.0

3.4

105

100

100

4 L

407

27

259

5 2.

3

0.48

1.

07E

+10

-510

.7

-178

3.1

-42.

54

7.1

4.0

108

98

97

5 R

47

70

-

3 1.

6

0.46

2.

56E

+10

-149

.1

-64.

9 -4

0.17

3.

6 3.

2 9

91

90

6 R

2

17

340

5 1.

9

95

0.44

5.

49E

+09

719.

3 -1

512.

8 -4

7.20

8.

4 6.

1 17

4 61

61

3

L 47

9 -

220

4 2.

4

0.42

2.

37E

+10

-100

.0

-137

0.9

-41.

84

4.3

2.6

99

103

103

7 R

35

40

-

3 1.

1

0.40

1.

54E

+10

-159

.4

-114

2.9

-45.

35

6.1

2.7

23

94

94

2 L

264

35

- 8

2.8

0.40

4.

53E

+10

-43.

3 -5

3.7

-43.

33

3.0

1.9

108

95

95

5 R

58

-

285

4 1.

3

0.39

3.

19E

+10

-99.

9 -1

370.

6 -4

5.26

4.

5 1.

8 15

4 95

94

14

L

324

40

240

4 1.

1

0.38

3.

84E

+10

-274

.9

-748

.1

-45.

97

5.3

1.2

36

93

92

9 L

399

31

213

4 1.

3

0.37

3.

73E

+09

-301

.9

-845

.2

-43.

51

14.5

4.

3 17

3 10

0 10

0 4

R

59

32

- 7

4.0

0.37

4.

34E

+09

-53.

2 -1

970.

5 -4

8.25

10

.5

5.1

8 27

17

21

R

49

6 10

0 26

6 4

2.1

0.36

4.

76E

+09

-112

2.3

-169

.7

-46.

23

10.7

4.

5 2

92

92

7 L

409

- 24

5 4

1.9

0.36

1.

84E

+10

-94.

0 25

.5

-47.

20

3.9

3.2

37

27

16

22

L 47

6 -

- 7

2.4

0.35

-

-111

.4

-133

3.1

-45.

18

0.0

0.0

0 12

5 12

4 13

L

266

39

230

5 1.

4

0.35

3.

92E

+09

-139

6.9

-432

.2

-45.

62

12.6

4.

5 24

10

5 10

5 9

L 37

8 -

302

5 2.

7

0.35

-

2.9

-82.

4 -4

1.05

0.

0 0.

0 0

102

101

11

R

17

18

- 3

0.9

0.34

1.

56E

+10

-44.

0 -1

750.

3 -4

3.59

4.

8 2.

9 70

10

3 10

3 4

L 61

-

- 6

1.9

0.34

9.

98E

+09

-164

.5

-112

8.5

-45.

79

7.8

2.8

26

97

97

7 L

386

34

283

4 1.

5

0.34

9.

85E

+08

-696

.6

-166

7.7

-45.

62

29.3

7.

5 47

11

0 10

9 12

L

612

- -

3 2.

2

0.33

-

-38.

2 -1

92.4

-4

5.44

0.

0 0.

0 0

98

97

11

L 61

8 -

- 2

0.7

0.31

2.

56E

+10

-13.

0 -1

3.4

-49.

04

3.2

2.5

174

87

86

11

L 50

3 11

32

0 5

1.3

0.31

5.

68E

+09

-903

.6

103.

4 -4

8.43

6.

2 5.

6 13

5 81

80

8

L 50

1 5

237

4 1.

3

0.30

5.

82E

+09

174.

2 -1

795.

9 -4

3.77

7.

5 4.

5 85

10

4 10

4 1

L 66

-

284

4 1.

2

0.30

2.

25E

+10

25.9

-4

5.7

-45.

88

4.3

2.0

8 10

4 10

3 12

L

394

- 30

6 4

1.2

0.30

3.

00E

+09

-946

.6

20.7

-4

4.38

14

.4

4.3

106

83

82

6 R

98

-

- 3

1.4

0.30

5.

52E

+09

-105

6.4

-148

.2

-46.

41

9.0

3.8

180

81

81

3 L

456

- -

3 1.

4

0.30

1.

38E

+10

-6.8

-2

2.9

-40.

87

4.2

3.3

27

89

89

5 R

10

-

- 3

1.0

0.29

1.

76E

+09

-100

3.8

191.

2 -4

5.97

14

.2

7.3

49

36

30

21

L 39

2 -

278

5 2.

8

0.29

-

-621

.9

-191

7.9

-45.

44

0.0

0.0

0 94

94

4

L 62

8 56

-

2 0.

7

0.29

-

17.7

-7

9.3

-45.

88

0.0

0.0

0 10

5 10

5 8

L 39

1 -

299

5 1.

0

0.29

4.

67E

+09

-526

.5

-191

9.0

-41.

22

9.8

4.0

58

79

78

5 R

63

1 57

26

9 4

1.6

0.28

5.

85E

+09

-7.5

-1

.8

-48.

43

7.5

4.0

72

34

32

9 L

637

- -

7 2.

6

0.28

-

26.9

-3

4.5

-45.

62

0.0

0.0

0 10

9 10

5 29

L

382

14

- 3

0.9

0.28

5.

93E

+09

-654

.8

-190

3.9

-41.

31

9.8

3.0

124

76

76

7 R

21

-

- 3

0.9

0.28

-

3.1

-82.

3 -4

4.38

0.

0 0.

0 0

105

105

2 L

99

18

238

4 0.

8

0.27

1.

66E

+10

-85.

9 -1

87.5

-4

0.96

5.

5 1.

9 64

10

9 10

9 6

R

12

21

334

5 1.

2

96

0.26

-

26.0

-5

5.5

-45.

88

0.0

0.0

0 11

0 10

9 11

L

395

- -

3 0.

8

0.26

1.

09E

+10

-107

.3

-58.

8 -3

9.73

4.

2 3.

6 62

93

92

9

R

615

16

- 2

0.5

0.26

6.

36E

+09

-143

.3

-177

7.3

-43.

33

6.4

4.1

93

94

94

4 L

60

- 25

6 4

1.3

0.26

-

-111

.3

-41.

0 -4

5.70

0.

0 0.

0 0

55

49

25

L 61

9 -

- 2

0.6

0.25

7.

18E

+09

-690

.5

-187

2.4

-44.

82

5.8

3.8

72

104

103

14

L 15

6 54

29

6 4

1.1

0.25

1.

13E

+10

-716

.8

-347

.7

-46.

76

4.0

3.6

36

103

103

8 L

471

23

- 3

0.7

0.25

-

-898

.5

41.2

-4

4.12

0.

0 0.

0 0

105

105

9 R

63

2 7

- 6

1.1

0.25

-

-108

.5

-197

.0

-45.

18

0.0

0.0

0 87

75

44

L

621

- 33

4 5

1.1

0.25

3.

28E

+09

700.

0 -1

524.

4 -4

7.64

10

.7

4.6

157

35

34

10

R

492

- -

3 1.

1

0.25

4.

14E

+09

120.

2 -1

776.

0 -4

2.54

8.

5 4.

5 11

0 81

80

10

R

48

48

28

4 5

1.5

0.24

8.

42E

+09

-200

.2

-177

8.1

-43.

77

4.6

3.9

164

107

106

12

L 68

-

- 3

0.8

0.24

1.

41E

+10

-890

.6

-167

5.1

-42.

45

4.0

2.7

174

97

96

14

R

41

- 33

6 5

1.0

0.23

5.

02E

+09

-110

.3

-133

2.6

-41.

75

5.8

5.1

96

97

96

12

R

30

39

- 5

1.1

0.23

8.

83E

+09

-904

.6

-584

.5

-45.

97

5.1

3.2

78

104

101

22

L 40

0 -

- 3

1.0

0.22

6.

27E

+09

-730

.7

-188

7.2

-44.

56

7.9

2.8

40

74

72

18

L 61

3 -

- 2

0.6

0.22

8.

31E

+09

354.

3 -1

867.

6 -4

1.84

5.

4 3.

0 15

6 74

74

8

R

29

53

- 6

1.2

0.22

-

-165

.9

-120

1.4

-45.

18

0.0

0.0

0 10

6 97

43

L

614

- -

2 0.

3

0.21

3.

47E

+09

-61.

0 -1

89.7

-4

0.61

9.

6 4.

1 11

5 64

64

4

R

5 -

- 3

0.9

0.21

4.

43E

+09

-53.

4 -1

972.

1 -4

6.93

8.

2 3.

6 22

66

66

9

R

475

100

- 3

0.9

0.21

5.

38E

+09

13.5

-2

0.8

-44.

30

8.2

3.0

169

101

101

7 L

96

12

- 3

0.9

0.20

-

-287

.5

-762

.6

-43.

15

0.0

0.0

0 96

96

4

R

55

- 26

3 4

0.9

0.20

1.

86E

+10

156.

4 32

5.8

-47.

46

4.9

1.4

172

88

74

49

L 61

7 3

- 2

0.4

0.19

2.

39E

+09

-160

.5

-114

3.7

-42.

45

12.9

3.

9 18

11

6 11

6 8

R

43

35

323

4 1.

3

0.19

3.

66E

+09

-165

.5

-115

7.9

-42.

19

7.8

4.2

18

107

106

10

R

37

- 35

5 5

1.0

0.18

1.

13E

+10

-301

.1

-185

7.3

-44.

91

4.5

2.2

40

101

99

19

R

259

- -

3 0.

7

0.18

7.

19E

+09

-210

.7

-576

.0

-41.

75

4.1

3.8

179

87

87

7 R

36

26

-

3 0.

6

0.18

5.

24E

+09

-896

.1

64.8

-4

7.90

6.

9 3.

2 77

88

88

5

L 49

4 6

- 3

0.6

0.18

-

-946

.3

-644

.6

-46.

05

0.0

0.0

0 82

82

2

L 63

6 29

35

3 4

0.8

0.17

2.

22E

+09

-910

.8

121.

3 -4

9.04

8.

1 6.

0 16

5 94

94

9

L 50

4 4

- 3

0.7

0.17

1.

10E

+10

39.9

36

7.7

-47.

64

4.6

2.1

64

43

41

13

L 49

3 -

287

4 0.

7

0.17

8.

02E

+09

-350

.5

-117

9.7

-45.

88

4.6

2.9

125

97

97

13

L 39

6 36

-

3 0.

5

0.16

-

-148

.0

-64.

6 -4

6.58

0.

0 0.

0 0

96

95

5 L

469

17

- 3

0.4

0.16

6.

05E

+09

-53.

2 -9

2.0

-47.

02

4.3

3.9

75

102

100

19

L 62

0 -

- 2

0.3

97

0.16

1.

62E

+09

-894

.0

-578

.9

-46.

05

19.7

3.

2 71

95

94

12

L

403

25

- 3

1.2

0.16

-

10.6

17

.8

-43.

77

0.0

0.0

0 77

77

10

R

67

-

301

4 0.

5

0.15

1.

85E

+09

-286

.0

-566

.4

-42.

80

12.4

4.

2 16

6 87

86

17

R

53

-

- 3

0.9

0.15

1.

68E

+09

-858

.6

-582

.0

-43.

86

11.8

4.

9 14

9 88

86

19

R

72

-

- 3

0.9

0.15

-

24.3

39

3.9

-46.

05

0.0

0.0

0 7

1 7

R

629

62

- 2

0.3

0.14

1.

67E

+09

-300

.8

-143

0.7

-47.

20

10.1

5.

4 48

10

8 10

4 30

L

480

42

- 3

0.7

0.14

1.

10E

+09

-867

.9

-533

.4

-43.

86

13.3

6.

2 14

3 10

5 10

5 7

R

616

24

- 2

0.7

0.14

4.

93E

+09

-40.

5 -1

00.8

-4

8.34

5.

2 3.

6 15

8 11

4 11

3 7

L 49

8 19

-

3 0.

5

0.13

5.

05E

+09

-287

.9

-101

3.4

-42.

28

6.8

2.5

176

101

100

14

R

40

- -

5 0.

7

0.13

-

-86.

3 -1

90.0

-4

4.91

0.

0 0.

0 0

97

90

35

L 63

3 21

-

3 0.

4

0.13

1.

37E

+09

-130

8.5

-356

.4

-45.

79

8.2

7.1

67

108

106

19

L 39

0 -

- 3

0.6

0.12

-

46.7

-1

730.

5 -4

4.82

0.

0 0.

0 0

39

38

5 L

634

46

- 3

0.3

0.12

3.

08E

+09

-901

.5

-571

.5

-45.

62

10.0

2.

4 43

12

0 12

0 7

L 38

4 -

- 3

0.7

0.11

3.

94E

+09

-167

.7

-140

3.7

-42.

89

9.2

1.9

62

96

94

18

R

54

- -

3 0.

5

0.10

-

-863

.6

-540

.5

-45.

70

0.0

0.0

0 90

86

27

L

635

24

- 3

0.6

0.10

8.

32E

+08

-943

.5

-637

.1

-44.

21

12.0

6.

3 8

64

62

15

R

85

29

- 3

0.6

0.09

-

353.

6 -1

867.

0 -4

4.47

0.

0 0.

0 0

36

34

11

L 62

5 53

-

2 0.

2

0.08

1.

56E

+09

-303

.0

-843

.1

-46.

76

10.7

3.

1 13

4 91

81

41

L

630

32

- 2

0.3

0.08

9.

95E

+08

130.

0 -1

595.

6 -4

5.35

15

.6

3.1

119

72

71

14

L 37

3 -

- 3

0.5


98

5.2 Comparison with Previous Work at 1665 MHz

The 1665 MHz main line is the most studied hydroxyl line in W3(OH), and generally

in SFRs. W3(OH) has been studied at 1665 MHz by Barrett and Rogers (1966),

Moran et al. (1968), Cooper et al. (1971), Harvey et al. (1974), Mader et al. (1978),

Reid et al. (1980), Norris and Booth (1981), Norris et al. (1982), Fouquet and Reid

(1982); but by far the most comprehensive interferometer studies of 1665 MHz until

this work were those of Garcia-Barretto et al. (1988) and Bloemhof et al. (1992).

Garcia-Barretto et al. observed W3(OH) with five US and one European antenna in

1978 in all Stokes parameters with a resolution of ~8 mas, or for small parts of the

map, ~4 mas. They detected 81 masers stronger than 1.5 Jy, and 5 Zeeman pairs.

Bloemhof et al. observed W3(OH) in 1986 with six US and one European antenna in

RCP and LCP. Their aim was to re-reduce the data of Garcia-Barretto in an identical

way to their new data, with the intention of making direct proper motion

measurements of the individual masers from two sets of maps. Their rms noise flux

density was ~65 mJy for their new data and ~0.2 Jy for the Garcia-Barretto data, with

~5 mas resolution for both. They detected 23 Zeeman pairs (using excessively

generous criteria for a successful pair: only 12 would have passed the present criteria).

Simply re-reducing the old data and position fitting in a different way typically

produced positional differences of ~2 mas and resulted in highly varying fluxes as a

result of breaking or merging complex maser features. This gives an indication of

some of the problems that might be expected when comparing the present work to

these older observations.

Comparing the total power spectra of 1978 (Garcia-Barretto), 1986

(Bloemhof) and 1996 (figure 5.1a) it is seen that the 1665 MHz emission has

remained fairly similar in form, with the notable exception that the very bright RCP

emission at –45 km s–1 appears to have faded in flux from ~250 Jy to ~180 Jy. The

distribution of masers from 1978 and 1986 matches excellently with that in figure 5.3,

indicating that the 1665 maser groups are relatively stable features. Care must be

taken when comparing fluxes to distinguish between the total flux of a maser feature

(in Jy) and the peak flux density of a maser feature (in Jy/beam). Table 5.2 shows the

fluxes of identified masers in all three epochs, and some identified in just two. All the

1978/1986 data is from Bloemhof et al. (1992), except the two asterisked fluxes which

are from Garcia-Barretto et al. (1988).


99

Table 5.2: 1665 MHz masers compared at epochs 1978, 1986 and 1996

1996 1978 1986 1996 1978 1986 1996 1996 1996 1978 1996 Feature Flux Density Flux Density Flux Density Pos. Ang. Pos. Ang. Pos. Ang. RA Offset DEC Offset Velocity Velocity

Number (Jy/Beam) (Jy/Beam) (Jy/Beam) (Degrees) (Degrees) (Degrees) (mas) (mas) (km/s) (km/s)

600 104.9 65.6 74.5 61 63 41 -906.7 98.3 -44.9 -44.91 271 86.3 68.8 68.9 64 38 65 -905.1 116.0 -44.9 -45.18

383 83.5 58.4 46.0 53 59 27 -958.1 -107.6 -46.5 -46.41

91 33.5 29.9 21.6 6 26 155 -138.2 -1191.3 -44.8 -44.65

64 54 31 15.5 142 134 156 -4.5 21.0 -43.8 -44.03

97 15.2 16.6 11.4 113 101 104 -846.7 -1731.8 -44.9 -44.91

374 10.9 11 10.9 107 78 112 -591.4 -1916.9 -45.5 -45.53

9 11 7.6 10.1 89 94 131 -691.3 -1872.7 -41.2 -41.13

147 18 6.5 9.3 73 82 120 -477.4 -1776.5 -45.4 -45.26

62 11.6 9.3 8.5 69 75 51 -25.3 -1743.7 -44.1 -44.21

83 12.5 12.9 8.3 58 76 72 -725.0 -1675.3 -44.5 -44.47

385 9 8.4 7.3 148 155 155 -220.3 -547.0 -45.9 -45.88

73 13.1 9.4 7.2 84 114 114 -490.0 -1788.6 -44.2 -44.21

18 10.9 6.3 7.1 93 88 97 -591.6 -1916.9 -41.5 -41.57

401 34 11.5 6.7 132 150 142 -168.1 -48.4 -46.2 -46.32

51 6.6 5.9 6.6 18 24 23 -730.2 -1682.0 -42.9 -42.89

22 5.4 5.4 6.3 171 75 173 -138.7 -1179.3 -41.7 -41.66

603 19.4 15.2 5.8 104 104 106 24.2 392.9 -48.9 -48.95

74 8.4 7.5 5.5 4 63 89 -20.4 -1739.3 -44.1 -44.03

15 13.4 4.4 5.0 66 88 153 -138.2 -1192.2 -41.3 -41.22

42 4.1 4.4 5.0 177 1 1 -302.9 -1078.6 -42.6 -42.63

152 4.6 5.6 4.3 156 24 11 27.5 16.4 -45.4 -45.26

107 10.2 12.9 4.3 23 83 42 -271.1 -1783.8 -44.8 -44.82

151 8.5 11.1 4.2 0 0 144 -489.9 -1770.1 -45.2 -45.18

280 11.3 9.3 3.5 122 148 127 25.0 -1722.0 -45.4 -45.35

193 4.1 3.6 2.9 137 138 160 -52.2 -82.4 -45 -45.00

495 2.7 3.1 2.8 98 82 88 43.1 373.9 -48.5 -48.52

94 12.6 11.4 2.6 162 113 123 -418.3 -1815.5 -44.6 -44.47

39 4.4 2 2.3 144 165 149 26.4 -1722.4 -42.2 -42.36

77 4.5 3.3 1.8 131 122 140 -407.3 -1821.0 -44.3 -44.21

82 2.1 1.6 1.7 75 82 115 -891.6 -595.2 -44.5 -44.38

478 1.9 1.4 1.6 137 9 123 -902.0 39.3 -47.4 -47.11

56 3.4 2.2 1.3 3 19 26 -349.8 -1179.2 -43.4 -43.33

57 2.7 1.8 1.3 36 28 15 -956.3 -104.0 -43.4 -43.33

27 4.1 3.1 1.3 140 92 1 45.3 -1730.1 -41.8 -41.66

301 14.6 9.6 1.0 88 117 153 -99.8 -1380.9 -45.2 -45.18

44 2.7 1.9 0.9 135 122 136 -258.2 -688.9 -42.6 -42.54

104 5.1 2.5 0.8 155 135 94 -378.4 -1861.6 -44.8 -44.74

38 4.4 1.5 0.5 161 142 169 -244.6 -705.2 -42.1 -42.19

19 4.7 4.1 0.5 124 87 103 -44.9 -100.6 -41.5 -41.40

N/A 19.4 2.6 - 15 107 - - - -41.2 -

N/A 9.2 15 - 177 86 - - - -45.4 -

N/A 4 3.9 - 64 92 - - - -44.5 -

N/A 1.9 2 - - - - - - -45.8 -

N/A 1.8 1.5 - 115 100 - - - -44.1 -

474 5* N/A 51.6 - - 163 0.0 0.0 - -47.46

601 5* N/A 33.9 - - 81 -1021.3 -127.7 - -46.23

111 N/A N/A 27.0 - - 160 -48.6 -193.1 - -45.18

8 N/A N/A 19.4 - - 154 25.1 31.5 - -41.13


100

Caution needs to be exercised when comparing the flux densities because of

the possible effects of different fitting procedures and different sized beams, but it is

immediately apparent that the masers are indeed noticeably variable over a decade

timescale. The matching of position angle at different epochs has not been attempted,

since this is very much dependent on how complex areas of maser emission have been

broken down during fitting. Only 5 of the 45 masers listed in Bloemhof et al. (1992)

could not be found in the current data. The brightest of these in 1986 (15 Jy/beam) is

probably hidden now underneath the much brighter masers in the complex north-

western group. As explained in chapter 4.4.1, exhaustive efforts were not made in the

present work to break bright extended masers, such as these, into numerous smaller

masers. Another of the missing masers has a maser of opposite polarization at the

same velocity just 1 mas away – these two are probably the same maser with the linear

polarization flux detected in both hands of circular polarization. The remaining three

missing masers have fluxes of 2.6 to 1.5 Jy/beam and could have all faded since 1986.

Bloemhof et al. (1992) detected a trend towards lower flux densities – peaks in 1986

were on average 80% down on masers in 1978. They suggested that this could be

explained by a flux density miscalibration of 20% at either epoch. This trend

continues however into the current data: Flux densities in 1997 were on average 75%

of those detected in 1986, and 60% of those in 1978. It is unlikely that this is an

accidental trend of miscalibrations; rather it is probably a reflection of the limited life

of the masers. Even if the lifespan of the masers is many decades, after enough

decades have passed we would expect all of the spots to have faded away.

It may be more accurate to say that it is maser groups that are stable features,

whilst individual masers within a group may come and go over a period of decades. A

similar effect was noted in Desmurs et al. (1998) (observing at the 6 GHz excited

state) that, in comparison with Moran et al. (1978), the general distribution of maser

groups had been preserved but there were many changes in the major components.

This would suggest the scenario where ‘maser-friendly’ conditions persist on the large

scale in certain locations for many decades, but individual masers have a lesser

lifespan of just a few decades.

Table 5.2 also contains the four brightest masers in the current work that were

not listed by Bloemhof, one of which is the reference feature of the current work

(feature 474). Feature 474 and feature 601 were listed in the older work of Garcia-

Barretto (both at 5 Jy/beam), where feature 474 is marked down as the Zeeman


101

partner of their reference feature (feature 64) – although this cannot be the case since

the two masers have moved apart ~10 mas in two decades! (See chapter 5.8 for

proper motions. In addition, the true Zeeman partner of this maser has been detected:

see chapter 5.6.) These two reference masers are particularly interesting, and will

crop up in the following sections due to their peculiar lineshape, polarization and

proper motions. The two masers which were not in the Garcia-Barretto data (features

8 and 111) may actually have been in the 1986 data of Bloemhof – Bloemhof et al.

only published data for which there was an older counterpart in order to measure

movement. In any case, these two masers have risen from negligible emission to 27

and 19 Jy/beam in two (or possibly one) decades, whilst the two that were detected by

Garcia-Barretto have risen by 29 and 47 Jy/beam in 2 decades – indicating that some

OH mainline masers can be highly variable over decade timescales.

Figure 5.4: Position angles of 2-Dimensional Gaussian fits to the 1665 MHz masers. Length of the

vector is proportional to the ratio of major to minor axes


102

5.3 Maser Morphology

From figure 5.2, some patterns in the morphology of the masers are quickly apparent.

Most notable is the extended nature of the masers in the centre-west of the map at

offset coordinates –0.8'' RA, –0.6'' DEC; these masers have relatively high average

major axis of about 12 mas, and relatively low brightness temperatures. They are

generally elongated in the north-west to south direction – of the masers above 0.5 Jy 9

out of 11 have position angles in the range 155° to 179°. This is more apparent in

figure 5.4, which shows the position angles of the masers brighter than 300 mJy. Also

apparent is that in the extreme north and south of the map, the masers tend to be

elongated east west; in the south, this east-west orientation follows tentative lines or

arcs of masers. This may indicate shock fronts – see chapter 6.4.1.

The masers near the reference feature are typically small and often unresolved;

these masers have an average deconvolved major axis of 4.5 mas, and have relatively

high brightness temperatures. The complexity of emission in 1665 MHz makes

Gaussian fitting very difficult in some areas because the masers are blended. Fitting

was extremely difficult in the crowded southern areas of emission, and in the far

northeast emission. In the northwest are some very intense and large masers where

140

120

100

80

53 54 55 56

140

120

100

80

57 58 59 60

Milli

ARC

SEC

140

120

100

80

61 62 63 64

-880 -900 -920 -940

140

120

100

80

65 66

MilliARC SEC-880 -900 -920 -940

67 68

Figure 5.5: The three most intense masers, found in the northwest of W3(OH). Their extended

nature means that they are blended at this resolution. Contours are at 250 mJy/beam.


103

two of the masers merge completely, resulting in a maser fragment that is feature 610

in table 5.1, and a complex lineshape for feature 600 since it has effectively

‘consumed’ another maser. This area is shown in figure 5.5.

5.4 Maser Lineshapes

Examination of the lineshapes of 1665 MHz is complicated by the number of masers

in close proximity to each other – which can often lead to distorted lineshapes where

outlying flux from one maser adds to the peak of its neighbour.

Figures 5.6 and 5.7 show the strongest 12 maser lineshapes (all those above

8.5 Jy) that were well fit by Gaussians. In figure 5.6 there are some deviations from

the Gaussian: feature 600 merges with the fragment feature 610 at ~–45.9 km s–1;

feature 601 has a close neighbour which is probably responsible for the excess

emission at the red end of its lineshape; feature 271 has a close neighbour which is

probably responsible for the excess emission at the blue end of its lineshape. Note the

Figure 5.6: The most intense 1665 MHz maser lineshapes showing slight deviations from a

Gaussian form. Observed flux errors are negligible at these scales. Clockwise from top-left:

Features 9, 601, 600, 271.


104

Figure 5.7: The most intense 1665 MHz maser lineshapes showing Gaussian form. Clockwise

from top-left: Features 147, 8, 103, 474, 374, 383, 111 and 91.


105

excess emission at the blue end of the lineshape of feature 9; feature 9 is a well-

resolved maser with no nearby emission. The unsymmetrical lineshape is probably a

genuine attribute of the maser. This type of behaviour could result from the growth of

emission in the line wing where the optical depth is greater. Normally the line

rebroadening following saturation would be symmetrical, and both wings would grow,

but if conditions are slightly less favourable for masing in different parts of the maser,

then rebroadening may be suppressed in one wing. Features 103, 111 and 147 show a

noticeable deficit of intensity in the line centre. This is probably a symptom of

saturation in the maser, and may indicate physical conditions tending towards NVR

(see chapter 2.4.2).

Of particular interest however are the maser lineshapes that are not

approximated by a Gaussian lineshape. Theoretical modelling of maser lineshapes has

been attempted by several authors (e.g. Goldreich and Kwan 1974, Bettwieser and

Kegel 1974, Bettwieser 1976, Nedoluha and Watson 1988, 1991, Elitzur 1990b, Wu

and Elitzur 1992, Wu 1992, Field et al. 1994); and theoretical work by Nedoluha and

Watson (1990b), and Field et al. (1994), has predicted that saturating masers may

evolve into complex or multi peaked lineshapes (see chapter 2.4.2). Theoretical

modelling by Gray et al. (1992) has shown that masers should start to saturate at

brightness temperatures greater than about 5×1010 K.

Of the 211 masers detected in 1665 MHz, 87 failed fitting with a Gaussian

lineshape. A minimum of 4 points was required in the lineshape for an attempt to be

successful. Table 5.3 indicates the numbers and reasons for failure of lineshape

fitting. The spots allocated as “probable unresolved double” have the attribute of

being morphologically elongated, and typically, the maser centre moves noticeably

Reason for Fit Failure Quantity

Close Doublet affects lineshape 10

Close Multiplet affects lineshape 6

Probable unresolved Double 3

Near Gaussian 6

Less than 4 line points 58

Genuine non-Gaussian lineshape 5

Table 5.3: The 1665 MHz masers that failed to be fit with Gaussian lineshapes


106

through the frequency channels, indicating that a second maser probably sits only

slightly offset in both position and velocity.

The five masers with genuine non-Gaussian lineshapes are shown in figure

5.8. Feature 97, like feature 9 has an excess of emission in one line wing. The blue

end of feature 476 is in some very noisy channels due to the rising peak of the

reference feature ~100 mas away which causes rings of beam artefacts. A coincident

artefact may be the cause of the jump in intensity. Feature 632 (when shown in figure

5.9 actually appears triple peaked) is in the area just red-ward of the very strong maser

emission in the far northwest of W3(OH). Around feature 632 are several patches of

emission (most of which failed fitting criteria) which also show multi-peaked

lineshapes – this is probably caused by beam artefacts. Beam artefacts exhibit

variation with frequency as well as with varying spatial position. Feature 359 (figure

5.9 also) appears to be a true double peaked maser. Such a lineshape is what would be

expected from the phenomenon of ‘hole burning’ (see chapter 2.4.2), which can occur

in a maser under NVR conditions such that the line centre may go into anti-inversion.

The fact that just one single maser out of 211 exhibits this kind of lineshape suggests

that NVR conditions are very much the exception in 1665 MHz SFR masers.

This leaves the interesting feature 64, also shown in figure 5.9: This is a very

dynamic maser, which as can be seen from table 5.2 has been rapidly declining in

intensity in the last 2 decades. It is by far the most intense and brightest maser with a

non-Gaussian lineshape – with Tb of 1012 K it is definitely of saturated intensity, and

its flux deviation from Gaussian is the largest of any maser. It is also unusual in that it

has changed velocity in the last decade; typically, masers may move spatially,

brighten or fade, but they stay constant in velocity. Indeed this is usually the best way

to identify masers between epochs; the average change in frequency of the peak in

table 5.2 is 0.08 km s–1, which is less than one spectral channel. The change in

velocity, position and intensity of this maser was first noted by Norris and Booth

(1981), who measured the changes relative to feature 271. In the 1978 data, it has a

peak velocity of –43.78 km s–1 (Garcia-Barretto) or –43.8 km s–1 (Reid). In 1997 its

velocity peak was –44.03 km s–1; a change of –0.24 km s–1 from the mean of the 1978

values. Only two other masers have a velocity change of this magnitude; both in the

far northwest cluster, which may be partly due to the difficulty of resolving the

complex emission there. For this reason measuring changes relative to feature 271


107

Figure 5.8: Distinctive non-Gaussian maser lineshapes. The Stokes parameters are shown for

feature 359 to emphasise that it is not an overlapping LCP and RCP maser. On these scales, noise is

only significant in features 632 and 476, where it is about 0.015 Jy/beam.

Lineshape for 1665 feature 64

0

2

4

6

8

10

12

14

16

18

-44.8 -44.6 -44.4 -44.2 -44.0 -43.8 -43.6 -43.4

Velocity / (km/s)

Flu

x D

ensi

ty /

(Jy

/bea

m)


0.00

0.05

0.10

0.15

0.20

0.25

0.30

-44.3 -44.2 -44.2 -44.1 -44.1 -44.0 -44.0 -43.9 -43.9 -43.8 -43.8 -43.7

Velocity / (km/s)

Flu

x D

ensi

ty /

(Jy

/bea

m)


0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

-47.4 -47.3 -47.2 -47.1 -47.0 -46.9 -46.8

Velocity / (km/s)

Flu

x D

ensi

ty /

(Jy

/bea

m)

Lineshape for 1665 feature97

0

2

4

6

8

10

12

-45.3 -45.2 -45.1 -45.0 -44.9 -44.8 -44.7 -44.6 -44.5 -44.4 -44.3 -44.2

Velocity / (km/s)

Flu

x D

ensi

ty /

(Jy

/bea

m)

Stokes Parameters for 1665 feature 359

-0.8

-0.6

-0.4

-0.2

0.0

0.2

0.4

0.6

0.8

-45.9 -45.8 -45.7 -45.6 -45.5 -45.4 -45.3 -45.2 -45.1

Velocity / (km/s)

Flu

x D

ensi

ty /

(Jy

/bea

m)

IQUV


108

(which is in the northwest cluster) is probably not reliable. Nevertheless, the trends

detected above are in agreement with those of Norris and Booth. Over a period of 19

years, this velocity change represents an acceleration towards us (or a deceleration

away from us) of 13 m s–1 year–1. This velocity shift could also be caused by a

decrease in the maser cloud magnetic field of 0.8 mG (about 10% for this region, see

chapter 5.6), or 0.04 mG year–1. However, given that magnetic fields in SFRs are

coupled to matter (e.g. Shu et al. 1987) this would require the maser cloud to be

expanding. Given the unusual nature of this maser’s lineshape and its acceleration, it

may be reasonable to assume that the phenomena are connected, possibly this masing

cloud is experiencing a collision.

-1390

-1395

-1400

-1405

-1410

-1415

-1420

-1425

-1430

62 63 64

Milli

ARC

SEC

-1390

-1395

-1400

-1405

-1410

-1415

-1420

-1425

-1430

65 66 67

MilliARC SEC-60 -70 -80 -90

-1390

-1395

-1400

-1405

-1410

-1415

-1420

-1425

-1430

68 69

-60 -70 -80 -90

70

140

120

100

80

60

40

43 44 45

140

120

100

80

60

40

46 47 48

Milli

ARC

SEC

140

120

100

80

60

40

49 50 51

MilliARC SEC-880-890-900-910-920-930

140

120

100

80

60

40

52 53

40

30

20

10

44 45 46 47

40

30

20

10

48 49 50 51

Milli

ARC

SEC

40

30

20

10

52 53 54 55

10 0 -10 -20

40

30

20

10

56 57

MilliARC SEC10 0 -10 -20

58 59

Figure 5.9: Contour plots of masers with non-Gaussian lineshapes. Clockwise from top left:

feature 359 (‘hole-burn’ is channel 67, contoured at 40 mJy/beam), feature 632 (contoured at 40

mJy/beam), feature 64 (extended line wing are channels 52+, contoured at 75 mJy/beam)


109

1665 MHz Maser Groups

-2100

-1900

-1700

-1500

-1300

-1100

-900

-700

-500

-300

-100

100

300

500

-1500-1300-1100-900-700-500-300-100100300500700900

RA Offset / mas

DE

C O

ffset

/ m

as

a RCP a LCP b RCP b LCP c RCP c LCP

d RCP d LCP e RCP e LCP f RCP f LCP

g RCP g LCP h RCP h LCP i RCP i LCP

Figure 5.10: W3(OH) maser groups.


110

5.5 Polarization

When analysing the properties of the masers, it is convenient to separate the W3(OH)

map into groups of masers which appear clustered, shown in figure 5.10. At this stage

the divisions are largely aesthetic, but it will be seen that the divisions may represent

underlying maser associations.

The full Stokes polarization of W3(OH) was studied by Garcia-Barretto et al.

(1988). They found the vast majority of masers to be ~100% circularly polarised,

except in the area to the far northwest (region c). Here, many of the masers showed

relatively high linear polarizations and unpolarised emission. Outside this area, just 6

of 71 masers were less than 100% circularly polarised. They found 41 (58%) LCP

and 30 RCP masers.

Of the 211 masers detected in 1665 MHz in the present data, 122 (58%) are

LEP and 89 REP. 154 (74%) have greater than 90% circular polarization. Of the

masers with peak flux-density over 1 Jy/beam the latter rises to 68 of 80 (85%).

Previous authors (Nedoluha and Watson 1990) have suggested that only bright masers

seem to have significant linear polarization, but this is partially contradicted by the

percentages above. No masers exhibit more than 50% linear polarization. In terms of

linear flux density, only 7 masers have fluxes higher than 0.35 Jy/beam. They are

shown in table 5.4. As was explained in chapter 4.4.1, the polarization figures

become less reliable as the flux decreases, such that the error is potentially ~10% by

0.2 Jy/beam.

Feature Peak Flux RA DEC VLSR Total Circular Linear Linear Polarisation Circular

Number Density Offset Offset Pol. Pol. Pol. Flux Density Pos. Ang. Direction

(Jy/Beam) (mas) (mas) (km/s) (percent) (percent) (percent) (Jy/Beam) (degrees)

271 68.87 -905.1 116.0 -45.18 74 73 14 9.92 166 R

600 74.52 -906.7 98.3 -44.91 90 90 12 8.94 156 R

474 51.64 0.0 0.0 -47.46 22 17 15 7.54 95 L

610 49.13 -909.8 94.5 -44.82 93 92 11 5.35 152 R

83 8.32 -725.0 -1675.3 -44.47 43 40 15 1.27 125 L

73 7.24 -490.0 -1788.6 -44.21 73 72 12 0.85 10 L

602 3.73 14.0 393.8 -48.87 55 53 14 0.53 103 L

Table 5.4: 1665 MHz masers with high linear polarisation. Information shown is for the peak

channel in the maser feature.


111

The first four masers in table 5.4 are also the first four in total flux density, one

of which is the reference feature. Values for feature 602 should be considered with

caution since it merges with feature 603, an effect, which as explained in chapter

4.4.1, will have an effect on the lineshape and polarization values. Feature 600 also

has a merger with feature 610. The Stokes lineshapes of the four unhindered features

are shown in figure 5.11. The top two and bottom two in the table have Zeeman

partners, all of which are much the fainter of the pair.

Feature 73 is probably an unresolved double, which leads to the shoulder on

the blue end of the spectrum in figure 5.11. Features 83 and 474 are striking because

most of their emission is unpolarised. Only 16 of the 211 masers are less than 50%

polarised, all in the north or south of the map. There are two masers (features 272 and

629) that are 10% or less circularly polarised, and both of these are in the interesting

region to the extreme northeast of the map (region a). The masers of this region have

typically low total polarizations and moderately high linear polarizations. Features

272 and 629, along with features 476 and 496, are the only four masers with more

Figure 5.11: Stokes lineshapes of simple 1665 MHz masers with high linear flux densities.


-10

0

10

20

30

40

50

60

70

80

-45.8 -45.7 -45.6 -45.5 -45.4 -45.3 -45.2 -45.1 -45.0 -44.9

Velocity / (km/s)

Flu

x D

ensi

ty /

(Jy

/bea

m)

IQU

V


-4

-2

0

2

4

6

8

10

-45.1 -45.0 -44.9 -44.8 -44.7 -44.6 -44.5 -44.4 -44.3 -44.2 -44.1 -44.0

Velocity / (km/s)

Flu

x D

ensi

ty /

(Jy

/bea

m)

IQUV

Stokes Parameters for 1665 feature73

-6

-4

-2

0

2

4

6

8

-44.8 -44.6 -44.4 -44.2 -44.0 -43.8 -43.6

Velocity / (km/s)

Flu

x D

ensi

ty /

(Jy

/bea

m)

IQU

V


-20

-10

0

10

20

30

40

50

60

-49.0 -48.5 -48.0 -47.5 -47.0 -46.5

Velocity / (km/s)

Flu

x D

ensi

ty /

(Jy

/bea

m)

I

QU

V


112

linear than circular polarization. It is still assumed that these are σ components,

though they are amongst the most likely candidates for any π components in the map.

The interpretation is made more difficult because many of the masers in group a are

close doublets, where the allocation of polarization flux to each maser can become

ambiguous.

Elliptically polarised σ components result from magnetic field angles

intermediate between parallel and perpendicular to the line of sight (see figure 2.5).

Elliptically polarised π components would require velocity gradients and possible

magnetic fields to conspire to overlap σ and π components created in different parts of

the maser (Garcia-Barretto et al. 1988). In two places, at offsets of ~(24, 394) mas

and ~(–725, –1679) mas, three masers were found within 12 mas of each other (the

criterion for possible Zeeman components – see next section). The two cases are

Figure 5.12: The two candidate 1665 MHz maser Zeeman triplets. Numbers are polarization angles

in degrees.

Figure 5.13: A possible 1665 MHz maser π component.

Stokes Parameters for 1665, features 496, 475 (left to right)

-0.10

-0.05

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

-50.0 -49.5 -49.0 -48.5 -48.0 -47.5 -47.0 -46.5 -46.0 -45.5 -45.0

Velocity / (km/s)

Flu

x D

ensi

ty /

(Jy

/bea

m)

IQUV

118

20

a) Stokes Parameters for 1665 features 376, 83, 51 (left to right)

-4

-2

0

2

4

6

8

10

-46.0 -45.5 -45.0 -44.5 -44.0 -43.5 -43.0 -42.5 -42.0

Velocity / (km/s)

Flu

x D

ensi

ty /

(Jy

/bea

m)

IQUV

159

125

21

b) Stokes Parameters for 1665, features 505, 603, 629 (left to right)

-6

-4

-2

0

2

4

6

8

-50.0 -49.5 -49.0 -48.5 -48.0 -47.5 -47.0 -46.5 -46.0 -45.5 -45.0

Velocity / (km/s)

Flu

x D

ensi

ty /

(Jy

/bea

m)

IQUV

88

69

92


113

shown as Stokes plots in figure 5.12. In case a) the central peak is not midway

between the other two, and so a Zeeman triplet is probably not what we are seeing;

this could simply be a Zeeman pair and a Zeeman singlet coincident. For the purposes

of magnetic field measurement features 51 and 376 were taken to be the Zeeman pair.

Notice however, that in feature 83 the linear polarization (Stokes Q and U) peak

nearer the centre of the Zeeman pair than the circular and unpolarised emission does.

In case b) the situation is more unusual. The RCP component (feature 629) has only a

tiny fraction of circular polarization (0.5%), and is perhaps the best candidate for a π

component on the data. If interpreted as a π component it gives a magnetic field

measurement well out of the ordinary for this maser group, and so it is still interpreted

as a RCP Zeeman component, paired with feature 505. If both the LCP emission

peaks are from the same individual maser, then the LCP component has been split,

possibly by a large velocity gradient within the maser.

The third case with slightly ambiguous Zeeman interpretation is that of the

southernmost maser emission detected, allocated to maser group h. The brighter of

the two masers, feature 496 is largely unpolarised, with less circular polarization than

linear (shown in figure 5.13). The separation between the peaks gives a sensible

magnetic field value if interpreted as a π component and one σ component. Feature

496 is therefore tentatively interpreted as a π component.

Figure 5.14 shows the variation of polarization with RA and DEC. Only

masers with flux density grater than 750 mJy/beam are included to remove those with

large uncertainties. A clear trend emerges in DEC – the curves are least-squares

quadratic fits, shown simply to illustrate the nature of the trend. There is no clear

trend in RA. In general, total polarization and circular polarization are closely tied.

This means that moving to the north and south, the amount of unpolarised and linear

emission in the masers increases. An explanation for this picture of changing

polarizations moving away from the centre of the region is that the angle of the

magnetic field with respect to the line of sight is increasing as we move away from the

centre of the map. Since the molecular gas surrounds the approximately spherical

UCH II region, a simple model where magnetic field lines in the gas point generally

towards the central star would satisfy this proposition – whether the gas was in a

shell/envelope or a torus.


114

The projection of the angle of the field on to the sky could be deduced from

the angle of linear polarization of the masers that contain a reliable linear percentage,

but as shown in figure 5.15 there are certainly no obvious trends in the data. On

closer inspection, it can be seen that on small scales the masers are often aligned with

their neighbours. This is most clear in regions a, b and c which have the highest

fraction of linear polarization. Also, of the 88 masers with linear intensities greater

than 30 mJy/beam, 30 have polarization angles of 0≤θ <90, while 58 have polarization

angles of 90≤θ <180 – which might indicate that the large scale cloud field has a NW

to SE alignment.

However, there are several ways that the polarization data can be distorted:

1. Some of the masers could be π components, which would be offset by 90°

– but given that perhaps only one such component is seen this should not

affect the polarization map significantly

2. The polarization data are more affected by noise. The Stokes Q and U

maps are more noisy than the I or V maps (see chapter 4.4.1), coupled with

the low level of linear flux that the measurements are being derived from

means that the position angles have a high degree of error.

% Linear Polarisation vs RA (>750mJy)

0

10

20

30

-1200-1000-800-600-400-2000200400600RA Offset / mas

Dem

ag.

Vel

/ (

km/s

)

% Linear Polarisation vs DEC (>750mJy)

0

10

20

30

-2500-2000-1500-1000-50005001000DEC Offset / mas

Dem

ag.

Vel

/ (

km/s

)

% Circular Polarisation vs DEC (>750mJy)

0

10

20

30

40

50

60

70

80

90

100

-2500-2000-1500-1000-50005001000DEC Offset / mas

Dem

ag. V

el /

(km

/s)

% Total Polarisation vs DEC (>750mJy)

0

10

20

30

40

50

60

70

80

90

100

-2500-2000-1500-1000-50005001000DEC Offset / mas

Dem

ag.

Vel

/ (

km/s

)

Figure 5.14: The variation of linear polarisation with RA and DEC, and the variation of circular

and total polarisation with DEC. Curves are least-squares fit quadratics.


115

3. The magnetic field angle projection could be highly variable on a very

small scale (~100 AU)

4. Most likely: Faraday effects are at work within the maser, effectively

scrambling the linear polarization direction

The symmetry of the variation in polarization percentages with DEC most

likely reflects a degree of north-south symmetry in the structure of W3(OH). The OH

masers have previously been predicted to lie in a north-south torus about the UCH II

region (Harvey et al. 1974), a shell around the UCH II region, or in matter in front of

the UCH II region (Reid et al. 1980, Norris and Booth 1981). In any case, the first

two of these cases would automatically exhibit north-south symmetry.

Current maser theories do not predict the large amount of unpolarised emission

coming from masers that are presumed to lie in gas where the magnetic field is at a

high angle to the line of sight. The probable cause is Faraday depolarization in the

presence of high electron densities, or long maser path lengths giving higher effective

electron densities. Longer path lengths might be expected at the edges of a spherically

symmetric region of gas. Nedoluha and Watson (1990b) show that the overlap (or

Figure 5.15: The Projection of the angle of linear polarisation for masers with over 30 mJy/beam of

linear flux intensity.


116

blending) of Zeeman components could cause the linear polarization to cancel, giving

unpolarised emission, but this does not explain how unpolarised emission could be

present in a separated Zeeman pair where the components are clearly not overlapping

(see next section).

5.6 Magnetic field Structure

From table 5.1, 58 Zeeman pairs were found, which are listed in table 5.5. This

represents 55% of the masers being members of a Zeeman pair. The distribution of

Zeeman pairs is illustrated in figure 5.16, which shows that the Zeeman pairs appear

to be a fairly random selection of masers over the map, with all maser groups well

represented.

1665 MHz Zeeman Pairs

-2100

-1900

-1700

-1500

-1300

-1100

-900

-700

-500

-300

-100

100

300

500

-1500-1300-1100-900-700-500-300-100100300500700900

RA Offset / mas

DE

C O

ffset

/ m

as

a b c

d e f

g h i

Figure 5.16: 1665 MHz Zeeman pairs in W3(OH), with maser groups labelled. The dotted lines are

at +18° and –30° to north.


117

Table 5.5: 1665 MHz Zeeman pairs. Information for peak flux, RA and DEC are from the peak

channel in the brighter of the two Zeeman components. Pair quality is explained in chapter 4.5.3

Peak Flux RA DEC Magnetic Magnetic Demagnetised Demagnetised Pair

Density Offset Offset Field Field error Velocity Velocity error Quality

(Jy/Beam) (mas) (mas) (mG) ±(mG) (km/s) ±(km/s)

3.73 14.0 393.8 6.35 0.11 -47.00 0.03 a

0.94 24.5 393.3 5.97 0.11 -47.82 0.03 b

2.80 43.1 373.9 5.58 0.11 -46.91 0.03 a

0.93 157.4 322.6 4.59 0.11 -46.11 0.03 b

68.87 -905.1 116.0 6.62 0.09 -47.09 0.02 b

45.98 -958.1 -107.6 5.09 0.10 -44.89 0.03 a

74.52 -906.7 98.3 5.99 0.09 -46.68 0.02 b

1.56 -902.0 39.3 5.15 0.11 -45.64 0.03 a

0.96 -895.4 66.5 6.26 0.11 -46.05 0.03 a

6.68 -168.1 -48.4 8.98 0.10 -43.65 0.03 a

0.46 -149.1 -64.9 10.82 0.11 -43.39 0.03 a

3.66 -107.5 -59.1 10.29 0.11 -42.76 0.03 b

0.27 -85.9 -187.5 6.67 0.11 -42.95 0.03 a

27.03 -48.6 -193.1 7.58 0.09 -42.93 0.03 a

0.52 -44.9 -100.6 11.79 0.11 -44.86 0.03 a

0.76 -9.1 -17.5 12.75 0.11 -45.30 0.03 a

15.54 -4.5 21.0 7.37 0.11 -46.25 0.03 a

0.35 2.9 -82.4 5.73 0.11 -42.73 0.03 a

1.05 16.2 -15.5 5.37 0.11 -42.72 0.03 a

19.39 25.1 31.5 6.45 0.09 -43.00 0.03 a

1.02 26.6 -32.4 8.26 0.11 -43.18 0.03 a

6.40 -859.1 -652.2 3.19 0.09 -45.11 0.03 b

0.18 -946.3 -644.6 3.14 0.11 -45.13 0.03 b

2.74 -850.6 -616.8 3.14 0.10 -44.80 0.03 a

1.68 -891.6 -595.2 3.10 0.11 -45.32 0.03 a

0.54 -899.5 -576.4 3.34 0.11 -45.07 0.03 b

0.14 -867.9 -533.4 3.13 0.11 -44.78 0.03 b

3.22 -210.7 -576.0 6.50 0.11 -43.67 0.03 a

0.84 -275.1 -749.1 5.78 0.11 -44.30 0.03 a

0.37 -301.9 -845.2 5.52 0.11 -45.13 0.03 b

4.95 -302.9 -1078.6 4.77 0.11 -44.03 0.03 a

2.46 -162.7 -1126.0 4.89 0.11 -44.33 0.03 a

0.40 -159.4 -1142.9 4.66 0.11 -43.81 0.03 a

1.31 -349.8 -1179.2 4.33 0.11 -44.60 0.03 a

6.27 -138.7 -1179.3 4.73 0.11 -43.08 0.03 b

21.57 -138.2 -1191.3 5.84 0.09 -42.96 0.03 a

0.35 -111.4 -1333.1 5.74 0.11 -43.21 0.03 a

0.42 -100.0 -1370.9 5.78 0.11 -43.54 0.03 a

1.66 -76.5 -1407.1 6.14 0.11 -43.81 0.03 a

0.73 -301.8 -1434.0 5.58 0.10 -45.55 0.03 a

3.46 25.0 -1722.0 5.11 0.11 -43.84 0.03 a


118

1.29 45.3 -1730.1 5.37 0.11 -43.24 0.03 a

5.47 -20.4 -1739.3 4.77 0.11 -42.63 0.03 b

8.49 -25.3 -1743.7 5.10 0.11 -42.71 0.03 b

4.66 198.5 -1793.5 4.78 0.10 -43.00 0.03 a

3.07 123.4 -1777.1 5.07 0.09 -44.04 0.03 a

0.22 354.3 -1867.6 4.47 0.11 -43.15 0.03 b

9.28 -477.4 -1776.5 5.86 0.11 -43.54 0.03 a

10.52 -519.7 -1778.2 5.83 0.10 -44.40 0.03 a

7.24 -490.0 -1788.6 7.10 0.11 -42.15 0.03 a

10.08 -691.3 -1872.7 6.29 0.10 -42.98 0.03 a

0.56 -525.0 -1916.3 6.90 0.11 -43.30 0.03 a

10.94 -591.4 -1916.9 6.69 0.08 -43.52 0.02 a

3.80 -513.9 -1774.7 4.92 0.11 -43.99 0.03 b

0.49 -622.0 -1918.0 6.87 0.11 -43.41 0.03 a

6.60 -730.2 -1682.0 4.66 0.09 -44.24 0.03 b

0.37 -53.2 -1970.5 4.42 0.11 -48.24 0.03 b

1.03 -716.7 -347.7 4.05 0.11 -45.57 0.03 a

Maser

Group

Average

Field /

mG

Number

of masers

Field

Range /

mG

a 5.6 4 1.8

b 5.8 5 1.5

c 8.5 12 7.4

d 3.2 6 0.2

e 5.4 13 2.2

f 5.0 7 0.9

g 6.1 9 2.4

h 4.4 1 -

i 4.0 1 -

Table 5.6: Magnetic field properties of the different maser groups. Overall average is 5.9 mG


119

The median ratio of the fluxes of the Zeeman pairs is 4.4:1. The mean is

heavily distorted by the Zeeman pairs containing the two brightest masers, which have

very weak Zeeman partners and so have flux ratios in the hundreds. The mean

including these two ‘outliers’ is 17.8:1; after dropping them, it is 6.9:1.

The variation of magnetic field strengths with RA and DEC is shown in figure

5.17. This shows in excellent clarity the exceptional nature of maser groups c and d:

Group c because it is the only group where there are large variations in magnetic field,

and also because of the generally high magnetic fields found in the group. Group d is

exceptional for essentially the opposite reasons: the remarkable constancy of the

magnetic fields, and the generally low magnetic fields found in the group. The field

characteristics in the maser groups are shown in table 5.6.

The tendency of most maser groups to have magnetic fields in the region of 4

to 7 mG is in good agreement with previous magnetic field measurements at 18 cm in

W3(OH), e.g. Moran et al. 1978, Reid et al. 1980, Garcia-Barretto et al. 1988 and

Bloemhof et al. 1992; and also with measurements at the 6 GHz (5 cm) excited state

in other SFRs e.g. Baudry et al. (1997), Desmurs and Baudry (1998). However, the

seven field measurements under 2.5 mG that Bloemhof et al. recorded were not

reproduced. Four of these definitely should not be called Zeeman pairs; having

unjustifiably large differences in spatial position (>20 mas). For the remainder,

Bloemhof et al. suggest that there is considerable uncertainly in the Zeeman

interpretation; in the light of the new results it is probable that these were not in fact

Zeeman pairs. Notably, the two highest field values recorded by Bloemhof et al. (8.2

mG and 9.1 mG) are located in region c. The relatively high magnetic field strengths

Field Strength vs RA

0

2

4

6

8

10

12

14

-1200-1000-800-600-400-2000200400600RA Offset / mas

Fiel

d S

treng

th /

mG

abc

defg

hi

Figure 5.17: Variation in magnetic field strength with RA and DEC in W3(OH). The arrowed

points are for comparison with figure 5.20 (see chapter 5.7).

Field Strength vs DEC

0

2

4

6

8

10

12

14

-2500-2000-1500-1000-50005001000

DEC Offset / mas

Fiel

d S

tren

gth

/ mG

abc

defgh

i


120

in group c are in agreement with excited state OH masers measured at 6 GHz in

Desmurs et al. (1998) and at 13 GHz in Baudry and Diamond (1998). In Desmurs et

al., their region 1 has the highest magnetic field strengths (average 7.6 mG at 6035

MHz, 9.7 at 6031 MHz) and corresponds in position to maser group c in this work. In

Baudry and Diamond, masers at 13 GHz are found only in the small area

corresponding to maser group c, and field strengths vary between 5.6 mG and 11.3

mG with an average of 9.6 mG; in excellent agreement with a range of 5.4 mG to 12.8

mG for maser group c.

It is worth noting also that the average field for group d (3.2 mG) is close to

that detected by Güsten et al. (1994) in thermally excited non-masing gas in

absorption against the radio continuum. They measured a line of sight component

field of +3.1(0.4) mG. However, from the polarization data in chapter 5.5 the

magnetic field direction is most likely near to the line of sight across much of the face

of the UCH II region, and so their figure may be approaching the total field strength of

the dense non-masing gas in front of the UCH II region. This would mean that group

d has conditions closest to the non-masing gas around the UCH II region. In this case,

group d may be just forming from the gas, or dissipating into the gas. In either case,

the conditions in the gas in group d are probably significantly different to those in

group c, which shows the greatest difference in conditions to that in the ambient gas.

Finally we note that the Zeeman pairs from maser groups h and i are unusual

in that they are isolated masers, separated by ~200 mas from the nearest clusters. The

relatively low magnetic field measured in i suggests that it may be the closest in

conditions to group d. Zeeman pair h should be considered tentatively: its Zeeman

nature is open to interpretation as discussed in chapter 5.5.

The strength of the magnetic field in maser group c has implications for the

phenomenon of Zeeman component overlap as suggested by Nedoluha and Watson

(1990b) in order to explain the suppression of one Zeeman component. The Zeeman

splitting of 12 mG at 1665 MHz is ~7 km s–1. According to their theory of component

overlap, this means that the velocity shift across the maser must be of this magnitude

or larger – which is unreasonable because this is approaching the velocity extent of the

entire source. Nedoluha and Watson themselves suggest that the 3 to 4 km s–1

splitting previously detected must be at the high end of what is plausible. It could be

argued that we will tend to see Zeeman pairs in regions of high magnetic field, and

single components in areas with low fields where overlap is plausible. With this in


121

mind, we might expect to see fairly equal fluxes (low Zeeman ratio) in the

components of Zeeman pairs in a high-strength field since it is unlikely that

sufficiently large velocity shifts will be found to cause enough overlap to even

diminish one component if not eliminate it, but this is not the case. Figure 5.18 shows

no such correlation.

5.7 Velocity Distribution

The velocity of the masers is plotted vs. RA and DEC in figure 5.19. The effect of the

Zeeman splitting is immediately clear – with RCP masers red-shifted and LCP blue-

shifted; indicating that the magnetic field points away from the observer. The Zeeman

effect is generally large enough compared to the intrinsic maser velocity distribution

in any region, such that RCP and LCP maser groups do not overlap in velocity space

when seen in the DEC plot. In just two instances does overlap occur: in maser groups

a and f, a single LCP maser appears in the midst of an RCP velocity range. Note also

that maser emission was detected in channel 1 of the data (–39.7 km s–1), so there may

be emission outside the red end of the observation band. In future, a slightly wider

bandwidth may be desirable. Group h (which consists of a solitary pair of masers in

the extreme south, and a pair of masers in the extreme east) is interesting. The masers

Figure 5.18: Magnetic field strength of Zeeman pairs versus their ‘Zeeman ratio’ (the ratio of their

peak flux densities).

Magnetic Field vs Zeeman Ratio

0

2

4

6

8

10

12

14

0 2 4 6 8 10 12 14 16 18 20

Zeeman Ratio

Mag

netic

Fie

ld /

mG


122

in the extreme east are not a Zeeman pair, but as described above, the masers are

almost always split into distinct velocity groups and so field direction can be inferred

from the sense of splitting of a maser group. In this case, the RCP maser is blue-

shifted with respect to the LCP maser – indicating that the magnetic field in this area

is probably pointing towards the observer. This is plausible given the trend in

polarization percentages when moving away from the centre of W3(OH) (as described

in chapter 5.5) since these masers inhabit the extreme edge of the UCH II region. If

the field is not in fact pointing towards the observer, then the LCP masers for the

eastern group h lie beyond the blue end of the bandwidth – at about –50 km s–1 for a 5

mG field.

The presence of Zeeman pairs allows a more accurate picture of the velocity

distribution of the masers to be constructed by ‘demagnetising’ the masers, as

explained in chapter 4.5.4. Every Zeeman pair can be reduced to a single maser at the

true gas velocity. This is shown in figure 5.20.

The distribution with DEC shows a clear trend, rarely mentioned in W3(OH)

literature (the trend is apparent in much previously published W3(OH) data but was

not commented on since Harvey et al. 1974, although Guilloteau et al. 1983 hinted at

it). The north of the cloud is moving towards us faster then the south. The UCH II

region is optically thick at cm wavelengths (e.g. Hollenbach et al. 1994), so the maser

emission certainly comes from the near side of the region. The dotted line on the

graph is a least squares fit plotted excluding maser groups c and h. The special nature

of group c was indicated in chapter 5.6 with respect to the high magnetic field

strengths there, and now this is reinforced by the fact that group c does not share the

same velocity characteristics as the rest of the masers either. The main body of group

Figure 5.19: Variation in maser velocity with RA and DEC in W3(OH).

Velocity vs DEC

-50

-48

-46

-44

-42

-40

-2500-2000-1500-1000-50005001000

DEC Offset / mas

Vel

ocity

/ (

km/s

)

a RCP

a LCP

b RCP

b LCP

c RCP

c LCP

d RCP

d LCP

e RCP

e LCP

f RCP

f LCP

g RCP

g LCP

h RCP

h LCP

i RCP

i LCP

pi?

Velocity vs RA

-50

-48

-46

-44

-42

-40

-2000-1500-1000-50005001000

RA Offset / mas

Dem

ag. V

el /

km

/s

a RCP

a LCP

b RCP

b LCP

c RCP

c LCP

d RCP

d LCP

e RCP

e LCP

f RCP

f LCP

g RCP

g LCP

h RCP

h LCP

i RCP

i LCP


123

c has a velocity dispersion in agreement with the 13 GHz masers which are a

coincident group; whilst at 6 GHz the large full spread is also observed in the region

corresponding to group c. The other groups of masers at 13 GHz have velocities in

good agreement with the 1665 MHz maser groups at coincident locations suggesting

that they share the bulk movement of the 1665 MHz maser groups, whilst the 6 GHz

masers share the bulk movement of the group c masers. The velocity gradient is also

clear in the 6 GHz methanol masers (Menten et al. 1992), which also had a clump of

masers over group c sharing the full velocity spread; and also in the 12 GHz methanol

masers (Menten et al. 1988, Moscadelli et al. 1999), which share a generally similar

position and velocity spread to the 6 GHz masers. If the two northernmost masers in

the 23 GHz methanol group (Menten et al. 1988) are assumed to be in the main body

of group c, with which they are coincident, then the velocity gradient is also apparent

in the arc of masers to the south of them.

The location of maser group c is between a ‘pinch’ in the continuum emission

– which could be caused by a denser clump in the outlying gas that is resisting

ionisation – and the northern peak in the continuum. Baudry and Diamond (1988)

suggest that this might be the location of the O-type star powering the H II region.

This means that the star would be away from the geometric centre of the continuum

emission. Moscadelli et al. (1999) show evidence for a velocity gradient in the

methanol masers that would lie within group c. The exceptional nature of the group h

Zeeman pair is also clear. The plot also shows that with the exception of group c,

turbulent motions within each maser group are spread across 2 km s–1.

Demag. Velocity vs RA

-49

-48

-47

-46

-45

-44

-43

-42

-41

-1200-1000-800-600-400-2000200400600RA Offset / mas

Dem

ag.

Vel

/ k

m/s

abcde

fghi

Figure 5.20: Variation in ‘demagnetised’ velocity with RA and DEC in W3(OH). The arrowed

points are indicated for comparison with figure 5.17. In velocity, the arrowed points appear

possibly part of the main velocity trend, but in figure 5.17 it is clear that their field strengths are

well above any others in the main trend; so they are definitely part of maser group c.

Demag. Vel vs DEC

-49

-48

-47

-46

-45

-44

-43

-42

-41

-2500-2000-1500-1000-50005001000

DEC Offset / mas

Dem

ag. V

el /

km/s


124

(5.3)

(5.1)

(5.2)

The gradient of the fit is 1.5 km s–1 / arcsecond, and the masers lie in the range

–42 to –48 km s–1, with an average of –44.3 km s–1. The maser centroid is in excellent

agreement with previous observations, even though this is the first time the velocity

gradient clearly identified. Collapsing the masers to discs at angles of +18° and –30°

to north (angles derived by plotting least squares linear fits though maser groups a, e,

g; and c, e, f respectively – see figure 5.16) did not produce a significantly different

velocity gradient. At this point, the instinctive reaction is to fit a rotation curve to

such a velocity gradient. It is justifiable to assume that the masers are orbiting in a

thin ring because the masers keep close to the linear fit at all velocities. In this

circumstance the line of sight velocity of the masers in an edge-on ring are a linear

function of distance from the location of the star they orbit. The gradient of this line is

a function of the mass of the star M, and the radius of the ring rmax. Equating the

centripetal force of circular motion with the gravitational force of attraction gives the

total velocity of a body in Keplerian orbit:

maxtot

Gr

Mv =

However, we are interested in the measured component of the velocity in the direction

of observation, which for an edge-on ring is just vtot cosθ (assuming that θ = 0 at

maximum elongation). Cosθ is itself equal to a/rmax, where a is the projected

elongation. So now,

maxmax

Gr

a

r

Mv ⋅=

Finally, differentiating gives the gradient:

3max

Gdd

r

M

a

v =

Unfortunately, we can only know a lower limit to rmax because we don’t know how

much of the ring is populated with masers. The maximum extent of the masers in


125

DEC is 2300 mas, which is 4600 AU at 2.2 kpc. Using half this value for rmax gives a

lower limit of the stellar mass as ~8 M . However, if we assume that the star is under

maser group c, then by using symmetry arguments we can say that the disc or torus

could be up to 3500 mas (7000 AU) in size. This gives a possible estimate of the

stellar mass of ~27 M . The range 8 to 27 M is in excellent agreement with an OB

star that is thought to power the region. This raises the additional question: why are

so few OH masers seen in the northern part of the disc? The answer could be that

there is a requirement for UV radiation to produce OH molecules, and this UV

radiation does not penetrate very far from the UCH II region.

Circumstellar discs are believed to be common around stars (Hollenbach et al.

1994) and are known to contain large amounts of molecular gas (e.g. Beckwith and

Sargent 1993). The idea of a rotating torus in W3(OH) has been suggested before

(e.g. Harvey et al. 1974, and Guilloteau et al. 1983 who observed a velocity gradient

in NH3 absorption), but has fallen from favour recently due to conclusive evidence of

the expansion (or growth) of the UCH II region (Kawamura et al. 1998, which does

not actually rule out a disc), and the conclusive evidence of divergent motion

(Bloemhof et al. 1992, which also does not rule out a disc, as demonstrated in chapter

5.8). However, circumstellar discs received a boost from the work of Norris et al.

(1998) who showed that half of a selection of methanol maser sources showed

evidence for circumstellar discs. Many of these have OH and/or H2O masers present

also, but have not been studied at high resolution with these species – mainly because

they lie in the southern hemisphere where the VLBA and EVN have poor visibility.

Only W3(OH) has been studied in detail in all the common molecular species, but

Norris et al. (1998) allocated W3(OH) as ‘complex’ in their scheme because no

simple linear structure could be found. However, in the light of the data above

concerning both the OH and the methanol masers in the area of maser group c it is

likely that W3(OH) is actually a combination of both ‘compact’ and ‘line’ in their

scheme (the masers under group c being ‘compact’, the rest being ‘line’).

The mass of material around the UCH II region has been estimated at 1 M by

Guilloteau et al. (1983), and an upper limit of 3 M was set by Wink et al. (1994)

using CO observations. Wilson et al. (1991) proposed a ‘cloudlet’ of 4 M containing

the OH masers, which could itself be the disc. Hollenbach et al. (1994) concluded

that discs of mass ~0.3 *M exist around massive stars early in their history, which are


126

then evaporated away by the star over about 105 years. A north-south disc is attractive

in W3(OH) because it can also explain why only the western half of the UCH II region

is populated with OH and methanol masers; excited state OH absorption (Baudry and

Menten 1995); and also formaldehyde (H2CO) and methyl formate (HCOOCH3) (all

in Wyrowski et al. 1999). Interestingly SO2 was detected only over the eastern half of

W3(OH). A north-south disc would also explain the east-west elongation of the UCH

II region, since this would be the easiest path of escape for the ionised gas and

outflows. No significant CO emission is detected over the UCH II region (Wyrowski

1997). An indication of a slight inclination of the disc in the plane of the sky can be

seen in the arc-like nature of the easternmost OH masers at 1665 MHz, and in the

methanol masers at 23 GHz. After noting that the variance of the distribution of

maser positions is maximised along the DEC axis, Bloemhof et al. (1992) suggested

that the 10° angle of the arc represented the star moving into the plane of the sky at

that angle for their cometry bow-shock model. Equally, this 10° angle could be the

inclination of the disc to the line of sight, and this would explain how the western half

of the UCH II region is covered with molecular material. The incline would also

explain why the continuum emission is more extended to the east, because the eastern

emission would be slightly pointed towards us, with the western pointed away from

us. Recent very sensitive VLA observations of the region show an extended

‘champagne flow’ of continuum emission to the east-northeast (Wilner et al. 1999)

that could be driven by outflows extending the H II region in this direction. This

slight inclination also has the effect of reducing the apparent rotation speed of the

disc. A 10° inclination would reduce the measured velocity by just 1.5% for simple

Keplerian rotation, with the corresponding effect of increasing the mass estimates by

3%.

At this point, the underlying systemic velocity of the star and UCH II region

becomes important; unfortunately, this is far from certain. Reid et al. (1980) took a

value of –50 km s–1 from 22 GHz recombination lines (Hughes et al. 1976) and using

this deduced from the velocity centroid of the masers that they were in an in-falling

shell or accreting envelope. Using this value, all the maser groups would be in-

falling; those in the south falling the fastest, those in the north much slower. Infall on

this scale is effectively ruled out by the divergent motion demonstrated in Bloemhof et

al. (1992). However, the hydrogen lines are very broad (e.g. Sams et al. 1996) and

observations at higher frequencies showed that recombination lines at lower quantum


127

numbers were redder. Berulis and Ershov (1983) showed that this effect could be

caused by a rapidly expanding shell of ionised material moving at 18 km s–1, away

from a central star at –46 km s–1. Welch and Marr (1987) observed that at lower

quantum numbers the hydrogen recombination lines approached the maser centroid,

and calculated a value of –45.6 km s–1 for the star. Keto et al. (1987) supported the

infall model with evidence for both a rotating disc of NH3 and a collapsing shell of

NH3, but poor resolution means that the emission cannot be definitely pinned down in

the UCH II region. Keto et al. (1995) derived a similar figure of –46 km s–1 by

modelling the UCH II region as a supersonic champagne flow of ionised gas to the

northeast. Using this last value, the southern masers are still in falling, but the rate

decreases moving north and the northern masers are moving out towards us (all the

while excepting maser group c which appears to be in different circumstances). This

is typical of ordered rotation. Recall that Diamond et al. (1998) had suggested that the

star lies under what is effectively maser group c, at a DEC offset of about –200 mas

with respect to the current reference point. Kawamura and Masson (1998) detected

continuum variability in the central region of the UCH II at a DEC of about –400 mas

with respect to the current reference point, which they suggested might be related to

the central star. It is interesting to note that the linear fit to the velocity trend passes

between DEC offsets –200 and –400 between velocities of –45.7 km s–1 and –45.4

km s–1 respectively. In this case, masers at the observational stellar velocity occur

right over the suggested location of the star – exactly as would be expected for

rotating material.

Bloemhof et al. (1992) use arguments based on models fitting: distance to

W3(OH), an unreliable hydrogen recombination line (Schraml and Mezger 1969); two

component models for the hydrogen recombination (Wilson et al. 1991); and HCN

emission (Turner 1984) – to justify a stellar velocity of –41 km s–1. This figure was

attractive to them, because having just shown conclusively divergent motion in the

proper motion of the masers (see chapter 5.8) it allows all the masers to be in an

expanding shell or cometry bow shock (Van Buren et al. 1992) around the UCH II

region. Applying this figure to the current data would mean that the southern masers

are expanding towards us more slowly than the northern masers. If the same stellar

location is taken as was suggested above, then the expansion directly in the line of

sight of the star is ~5 km s–1 – in excellent agreement with the general expansion

speed measured by Kawamura and Masson (1998) using VLA difference maps in the


128

continuum. The figure is also in agreement with the non-parametric expansion rate

first derived by Bloemhof et al. (1992) from proper motions (see next section).

However, this figure means also that the north-eastern maser group a is moving

towards us at about 5 to 8 km s–1, which is faster than the northwards expansion speed

of the region of ~3 km s–1 measured by Kawamura and Masson.

5.8 Proper Motion

Bloemhof et al. (1992) first produced relative proper motions of selected 1665 MHz

masers on W3(OH) by re-reducing the data of Garcia-Barretto et al. (1988) and

comparing it with their own observational data. They took care to reduce both sets of

data in the same way where possible to minimise differences between the data sets.

The relative proper motions they derived are shown in figure 5.21a (their figure 6),

whilst the motions derived using the present observations are shown in figure 5.21b.

The two sets are in excellent agreement; the only significant differences are in

the west-centre (maser group d) and the northeast (maser group a). This is good

because while Bloemhof et al. took exhaustive measures to ensure that both datasets

were treated identically, figure 5.21b was produced after qualitatively different

reduction, analysis and fitting methods. This indicates that – over sufficient

timescales – simple comparisons of Gaussian fitted maser locations should be enough

to produce very reliable proper motion results.

Bloemhof et al. 1996 showed by tracking maser morphology that the motions

are due to actual physical movement of discrete clumps of masing gas, and not non-

kinematic effects such as travelling excitation phenomena (e.g. density waves) or

chance realignments.

Figure 5.22 shows the change in separation of the masers from 1978 to 1996

measured as a relative velocity between them. The masers clearly show an overall

trend of divergent motion. Averaged over a whole edge-on disc there should be no

more divergence than convergence, but if we are seeing just part of the disc then there

will be a net amount of convergence or divergent depending on which part of the disc

is in view. This is demonstrated next.


129

Figure 5.21a) and b): a) Top, the relative proper motions measured by Bloemhof et al. (1992). b)

Bottom, the relative proper motions measured in the current data. The reference points are not the

same maser, but are offset by ~20 mas. This was accounted for in the calculations.


130

Figure 5.23: Apparent change in separation of two masers on the sky as they rotate first towards

then away from the observer. In the left chart, the masers are separated by 100 mas at maximum

(an angular separation of ~4° in the ring); in the chart on the right the masers are separated by a

rotation angle of 45°.

Figure 5.22: Histogram of the pairwise change in separation of the 1665 MHz masers converted

into a velocity (using W3(OH) at 2.2 kpc).


131

Figure 5.23 shows the effects of rotation on the separation of two masers in a

thin disc of radius 3000 AU where the rotation plane is at 90° to the plane of the sky.

One maser starts in the plane of the sky (at 0°), the other starts ahead of it. As the

masers rotate towards the observer, they generally show divergent motion in the plane

of the disc. This slows as the masers approach 90° (the line of sight) and turns over to

become convergence. The exact turnover depends on the angular separation between

the masers. This shows that all that is needed for an excess of divergence is that we

are seeing more of the northern half (which is rotating out of the plane of the sky) of

the disc than the southern half (which is rotating into the plane of the sky).

Turbulence within the disc will broaden the distribution of relative velocities.

It is important to remember that figures 5.21 do not represent the absolute

proper motion of the masers, but rather: movement relative to the reference maser.

There is therefore an unknown vector of the true motion of the reference maser added

to every other maser on the map. This is excellently demonstrated by Bloemhof et al.

(1992) in figure 5.24 (their figure 10).

Bloemhof et al. argued in favour of a cometry bow-shock flow for the masers,

and suggested that the underlying reference motion was ~7 km s–1 to the northeast.

They added to this with work in Bloemhof (1993), where a technique of

‘diagonalising the variance/covariance matrix’ was used to derive this motion (this is

an eigenvalue method, which puts the matrix into its ‘natural’ coordinates, one axis of

which will point along the direction of the dominant bulk motion). The model of a

cometry H II region fails however to explain the clear velocity gradient observed in

the masers.

If the masers were in a rotating disc, they would also shift in velocity with

time. Some masers were mentioned in this respect in chapter 5.4. If the masers are in

a bound ring of gas 3'' (~6000 AU) in diameter and rotating at 5 km s–1, then the

centripetal acceleration of a maser directly over the star is v2/r, which evaluates to a

velocity change of ~1.75 m s–1 year–1. Even over 20 years this is just 35 m s–1, which

is about equal to the error on the demagnetised velocity after fitting Gaussians to the

Zeeman components, and so could not be reliably measured. Measuring velocity

changes from a single Zeeman component would certainly not be reliable since 35 m

s–1 in 20 years also represents a change of just 1% in the magnetic field strength for a

maser at the typical 6 mG field, which is entirely plausible in a region where gas is

likely experiencing collisions.

13

2

Fig

ure

5.24

: Pr

oper

mot

ions

of

1665

MH

z m

aser

s in

W3(

OH

) w

ith v

aryi

ng r

efer

ence

vel

ociti

es a

dded

. Z

ero

refe

renc

e ve

loci

ty i

s ou

tline

d, a

s is

the

‘co

met

ry’

choi

ce o

f B

loem

hof

et a

l. (1

992)

. T

aken

fro

m B

loem

hof

et a

l.(1

992)

.


133

Figure 5.25a shows the W3(OH) measure proper motions assuming that the

reference maser has a movement of (4,–1) km s–1 (RA, DEC). Figure 5.25b shows an

idealised model torus, exhibiting no turbulence, that has radius 6000 AU. The slightly

disordered motions of the masers suggests that we could be seeing the final

destruction of the stellar accretion disc, as the increasing radiation and energetic

outflow finally take their toll on the integrity of the rotating gas. This could itself

introduce an element of expansion into the rotation as well. Remember that the

masers in group c (which includes the reference feature) are assumed not to be in the

disc, and so need not share its motion. This scenario of the proper motion requires

that the star be further south in order to satisfy the divergent motion of the masers;

perhaps under the southern continuum peak. This raises the question of what is

happening under maser group c? This could be the site of a molecular outflow, or

perhaps a second smaller star, which does not significantly gravitationally affect the

disc rotation. It could also be an instability, forming an eddy; or a developed turbulent

eddy in the circumstellar gas.

Figure 5.25 a) and b): a) Left, showing the 1665 MHz proper motions with a reference motion of

(4,–1) km s–1 (RA, DEC) added. b) Model proper motions in a rotating disc diameter 6000 AU,

inclined to the line of sight by 10°.


134

5.9 Statistical Relationships

The large numbers of masers at 1665 MHz make it the best frequency for studying

statistical relationships of properties of OH masers. The maser properties were tested

for a variety of correlations, shown in figures 5.26 and 5.27.

There were no correlations found between maser FWHM and: flux density,

maser area, field strength or DEC offset (position); 92% of the masers for which

FWHM was measured had widths between 200 and 400 m s–1. Note that there may be

a selection effect at work at the lower end of the line widths: at least 4 channels were

required for a Gaussian fit to be attempted. Extremely narrow lines therefore may not

show emission in enough channels to allow a Gaussian fit – thus the sharp cut-off of

masers with FWHM under 200 m s–1 should be regarded with caution. Neither maser

area nor magnetic field strength was correlated with maser flux density. Most of the

masers with area over 190 mas2 come from maser group d, which has already been

noted for its extended masers; the rest come from the positional centre of group e.

There was no correlation found between maser area and total polarization, nor

between maser flux density and total polarization, though the small number of low

polarization masers makes this less clear.

Three correlations were found:

1. FWHM with total polarization. Masers with total polarization below 80%

(and, typically, correspondingly low circular polarization values) tended to

have below average line widths. No maser theories currently predict this.

2. Magnetic field strength with total polarization. There were no masers with

high magnetic field strengths and low (<80%) total polarizations. This is

probably because the higher gas density presumed to accompany high

magnetic field strengths enhances polarization.

3. Magnetic field strength with maser area. Masers with high magnetic field

strengths tend to be smaller. The work of Gray and Field (1994, 1995) has

shown that maser amplification is enhanced along magnetic field lines,

and that the maser beam angle is narrowed by this effect – a phenomenon

they term ‘magnetic beaming’. Stronger magnetic fields could enhance

this effect. The correlation is also probably related to the higher gas

density presumed to accompany high magnetic field strengths; higher

densities may allow higher amplification for a maser of any given size.


135

Flux Density vs Field Strength

0

10

20

30

40

50

60

70

80

0 2 4 6 8 10 12 14

Field Strength / mG

Flux

Den

sity

/ (

Jy/b

eam

)

Flux Density vs Maser area

0

10

20

30

40

50

60

70

80

0 50 100 150 200 250 300 350 400

Maser Area / mas^2

Flux

Den

sity

/ (

Jy/b

eam

)

FWHM vs Flux Density

0

100

200

300

400

500

600

0 10 20 30 40 50 60 70 80

Flux Density / (Jy/beam)

FWH

M /

(m/s

)

FWHM vs Maser Area

0

100

200

300

400

500

600

0 50 100 150 200 250 300 350 400

Area / mas 2

FWH

M /

(m/s

)

Figure 5.26: Charts showing no relationship between various maser properties.

FWHM vs DEC

0

100

200

300

400

500

600

-2500-2000-1500-1000-50005001000

DEC Offset / mas

FW

HM

/ (

m/s

)

FWHM vs Field Strength

0

100

200

300

400

500

600

0 2 4 6 8 10 12 14

Field / mG

FWH

M /

(m

/s)


136

Flux Density vs Total Polarisation %

0

10

20

30

40

50

60

70

80

020406080100

Total Polarisation / %

Flu

x D

ensi

ty /

(Jy/

beam

)

FWHM vs Total Polarisation %

0

100

200

300

400

500

600

020406080100


FWH

M /

(m/s

)

Maser Area vs Total Polarisation %

0

50

100

150

200

250

300

350

400

020406080100


Mas

er A

rea

/ mas

^2

Magnetic Field Strength vs Total Polarisation %

0

2

4

6

8

10

12

14

020406080100


Fiel

d S

treng

th /

mG

Magnetic Field Strength vs Group ave. maser size

0

1

2

3

4

5

6

7

8

9

0 2 4 6 8 10 12

Maj axis / mas

Mag

fiel

d / m

G

Figure 5.27: Charts showing likely relationships between various maser properties. In the maser

area versus magnetic field strength chart, large markers represent the brighter Zeeman partner; small

markers represent the fainter partner.

Magnetic Field Strength vs Maser area

0

2

4

6

8

10

12

14

0 50 100 150 200 250 300 350

Area / mas^2

Fie

ld S

tren

gth

/ m

G


137

5.10 Two-Point Correlation

The large number of masers at 1665 MHz makes for a good two-point correlation,

which is shown in figure 5.28.

It has long been known that masers tend to cluster; Reid et al. (1980)

suggested that the scale size in W3(OH) was 60 mas (~120 AU), but this data shows

that in order to include faint outliers (which they did not detect) the true maser

clustering scale begins at about 300 mas (~600 AU). The correlation shows a

significant peak at just less than 2'' (~4000 AU) in size. This is likely related to the

size of the underlying UCH II region. The size of this region depends on the

observation frequency, being ~1.5'' at 15 GHz, but ~3'' at 1.6 GHz. Figure 5.29 shows

a zoom in on the first 300 mas of the correlation with smaller bins and higher

resolution. It shows that the distribution within a maser cluster concentrates the

masers ever more at closer distances, right down to between 10 and 20 mas (~20 to

~40 AU). At around 8 mas the correlation breaks down and suddenly drops to zero at

4 mas. This is because the beam size of the observation was about 8 mas, and so

Figure 5.28 a) and b): a) Top, the 1665 MHz masers and a random selection of points in the same

space. b) Bottom, the two-point correlation of the 1665 MHz masers.


138

Gaussian fitting routines run into difficulties with masers which are less than 8 mas

apart, and fail completely for masers less than 4 mas apart. This shows that the

masers may be clustering on a scale that is comparable to their physical size.

Figure 5.29: The two-point correlation of the 1665 MHz masers, with smaller bins and hence

‘higher resolution’.

139

Chapter 6

Analysis of W3(OH) – 1667, 1612 and 1720 MHz

This chapter contains the results for the three remaining ground-state frequencies.

6.1 1667 MHz

6.1.1 The Observational Data

The 1667 MHz total power spectra for RCP and LCP are shown in figure 6.1-1a.

They are from 1998 (Masheder, private communication), two years after the

interferometry observation. The spectra show a very much simpler source than the

1665 MHz data, immediately offering the prospect that the main peaks are Zeeman

components. Figure 6.1-1b shows the Stokes I spectrum of the maser fluxes

calculated from the channel maps.

The 1667 MHz integrated flux contour map and the maser spot map are shown

in figures 6.1-2 and 6.1-3. Results of 2-dimensional Gaussian fitting of maser

emission in the maps are shown in table 6.1-1. Channel maps of the masers in contour

form can be found in appendix A.

The 1667 MHz masers cover only the south-western sector of the continuum

region. There are many fewer masers than at 1665 MHz – just 41 are present, and the

most intense masers are only 10% as intense as the masers that are seen at 1665 MHz.

Immediately noticeable is the striking ‘arc’ of maser emission in the south of the map.

The majority of the maser flux comes from this area, and it will be examined in detail.

Chapter 6: Kinematic Analysis of W3(OH) – 1667, 1612 and 1720 MHz

140


0

5

10

15

20

25

-50 -49 -48 -47 -46 -45 -44 -43 -42 -41 -40

Velocity / (km/s)

Flux

/ Jy

Figure 6.1-1a) and b): a) Top, the 1667 MHz total power spectra for RCP (dotted) and LCP (solid).



-5

0

5

10

15

20

25

-50 -49 -48 -47 -46 -45 -44 -43 -42 -41 -40

Velocity / (km/s)

Flu

x /

Jy


141

Figure 6.1-2: The 1667 Stokes I integrated flux map. Contours are at 120 mJy/beam; beam at

lower left. For reference centre, see chapter 4.3.6.

AR

C S

EC

ARC SEC0.5 0.0 -0.5 -1.0

0.5

0.0

-0.5

-1.0

-1.5

-2.0


142


-2300

-2100

-1900

-1700

-1500

-1300

-1100

-900

-700

-500

-300

-100

100

300

500

-1500-1300-1100-900-700-500-300-100100300500700900

RA Offset / mas

DE

C O

ffset

/ m

as

Less than 0.400 Jy

0.400 -> 2.000 Jy

2.000 -> 10.000 Jy

Figure 6.1-3: The 1667 MHz maser spot map. Contours are the 0.5, 7.1 and 16.6 mJy/beam levels

of a VLA 2 cm continuum map (Kawamura and Masson 1998).

14

3

Tab

le 6

.1-1

: T

he 1

667

MH

z M

aser

s. D

ata

is f

or p

eak

chan

nel,

exce

pt la

st f

ive

colu

mns

, whi

ch a

re f

or th

e w

hole

fea

ture

. Fe

atur

es w

ith n

o FW

HM

mea

sure

men

t eit

her

faile

d

lines

hape

fitt

ing

or h

ad to

o fe

w c

hann

els

for

a fi

t. E

rror

s ar

e de

taile

d in

cha

pter

4.4

.1.

RA

and

DE

C o

ffse

ts a

re f

rom

the

refe

renc

e fe

atur

e.

Pea

k F

lux

Brig

htne

ss

RA

D

EC

V

LSR

D

econ

v.

Dec

onv.

D

econ

v.

Tot

al

Circ

ular

Li

near

C

ircul

ar

Fea

ture

Z

eem

an

FW

HM

C

hann

els

Inte

grat

ed

Den

sity

T

emp.

O

ffset

O

ffset

Maj

. Axi

s M

in. A

xis

Pos

. Ang

. P

ol.

Pol

. P

ol.

Dire

ctio

n N

umbe

r P

air

P

rese

nt

Fea

ture

(Jy/

Bea

m)

(K)

(mas

) (m

as)

(km

/s)

(mas

) (m

as)

(deg

rees

) (p

erce

nt)

(per

cent

) (p

erce

nt)

Num

ber

(m/s

)

Flu

x

(Jy

)

9.29

1.

00E

+11

-386

.7

-187

9.7

-44.

21

8.2

4.6

141

72

72

10

L 13

-

422

14

87.3

4.92

5.

40E

+10

-671

.4

-189

6.6

-42.

10

7.3

5.4

117

101

101

2 R

1

4 20

1 7

21.7

3.98

4.

40E

+10

-672

.9

-189

5.7

-44.

39

7.7

4.6

118

99

99

1 L

23

4 20

8 5

16.6

3.69

4.

10E

+10

-592

.8

-191

7.7

-44.

82

4.5

3.2

72

112

112

1 L

39

12

187

5 13

.1

2.16

2.

40E

+10

-651

.4

-191

3.0

-44.

56

7.2

5.2

113

97

97

2 L

28

9 22

5 6

9.8

2.02

2.

20E

+10

-603

.9

-191

4.9

-44.

82

7.0

4.6

62

103

103

2 L

32

11

218

5 11

.6

1.98

2.

20E

+10

-694

.5

-188

2.4

-42.

28

6.4

4.9

1 10

0 10

0 2

R

6 3

212

6 9.

2

1.90

2.

10E

+10

-592

.7

-192

5.9

-44.

82

11.0

7.

9 15

10

7 10

7 2

L 31

8

281

7 16

.5

1.55

1.

70E

+10

-656

.8

-190

2.4

-42.

10

3.7

3.2

40

99

99

0 R

2

5 20

2 4

4.6

1.37

1.

50E

+10

-592

.8

-192

6.2

-42.

45

13.5

7.

0 2

93

93

3 R

7

8 23

8 5

9.4

1.37

1.

50E

+10

-594

.0

-191

7.7

-42.

45

9.5

4.7

109

80

80

2 R

8

12

182

5 5.

9

1.35

1.

50E

+10

-656

.8

-190

2.5

-44.

39

3.4

2.4

23

95

95

0 L

24

5 20

4 4

4.0

1.30

1.

40E

+10

-43.

3 -1

440.

2 -4

2.89

5.

5 3.

5 12

9 10

0 10

0 2

R

10

- -

3 3.

2

1.26

1.

40E

+10

-651

.1

-191

3.9

-42.

19

7.0

5.4

116

101

101

5 R

5

9 20

7 5

6.3

1.20

1.

30E

+10

-514

.7

-190

8.7

-44.

56

17.2

3.

7 57

10

0 10

0 1

L 29

6

221

5 7.

9

1.18

1.

30E

+10

-695

.1

-188

2.6

-44.

65

5.4

3.7

11

94

94

3 L

25

3 30

5 6

6.5

0.88

9.

70E

+09

-514

.1

-190

7.6

-42.

19

21.7

4.

3 60

10

7 10

7 2

R

40

6 24

0 5

7.0

0.74

8.

20E

+09

-622

.4

-191

8.9

-44.

65

7.8

4.2

79

102

102

4 L

26

10

185

6 3.

7

0.71

7.

80E

+09

-564

.2

-193

2.8

-44.

47

7.5

2.2

74

115

114

14

L 27

7

- 3

2.1

0.59

6.

50E

+09

-425

.1

-180

9.0

-44.

03

6.7

4.0

131

106

105

14

R

15

- 19

2 4

2.2

0.48

5.

30E

+09

-65.

6 -2

021.

9 -4

7.81

15

.0

8.8

67

65

61

23

L 37

-

288

5 4.

4

14

4

0.41

4.

50E

+09

-863

.8

-652

.5

-44.

65

20.0

5.

9 7

103

103

4 R

30

1

263

4 3.

3

0.38

4.

10E

+09

-604

.3

-191

4.9

-42.

54

7.8

3.7

36

106

106

2 R

54

11

-

3 1.

3

0.37

4.

10E

+09

-860

.6

-667

.5

-44.

65

10.4

3.

6 3

113

113

4 R

41

-

- 2

1.3

0.34

3.

80E

+09

-623

.4

-191

9.4

-42.

28

7.2

5.4

66

102

102

9 R

4

10

319

5 1.

8

0.33

3.

60E

+09

-522

.1

-191

1.6

-42.

28

14.9

3.

7 52

10

0 10

0 7

R

51

- -

2 1.

0

0.31

3.

50E

+09

-536

.6

-174

9.1

-45.

88

8.5

4.1

127

98

97

13

L 35

-

227

4 1.

4

0.31

3.

40E

+09

-80.

8 -2

035.

3 -4

7.55

17

.1

5.3

41

52

47

22

L 36

-

323

6 3.

1

0.28

3.

10E

+09

-564

.3

-193

2.8

-42.

02

10.3

4.

4 96

10

5 10

5 2

R

50

7 -

2 1.

0

0.26

2.

90E

+09

-58.

6 -1

418.

1 -4

2.81

6.

7 2.

5 16

8 98

98

5

R

9 13

-

4 1.

2

0.24

2.

70E

+09

-395

.0

-189

9.3

-43.

95

7.2

6.2

32

65

63

17

L 53

-

- 3

1.1

0.23

2.

50E

+09

-155

.2

-690

.3

-44.

03

5.4

2.9

36

97

96

11

R

42

2 -

2 0.

5

0.22

2.

40E

+09

-242

.4

-199

9.2

-43.

24

8.1

4.0

49

97

96

10

L 11

-

- 3

0.9

0.21

2.

30E

+09

-736

.9

-181

3.0

-44.

03

3.4

3.0

179

103

103

8 L

16

- -

3 0.

5

0.19

2.

10E

+09

-339

.9

-187

9.9

-42.

02

5.3

3.5

87

92

91

11

R

3 -

- 3

0.6

0.18

2.

00E

+09

-862

.5

-651

.7

-45.

53

21.6

5.

2 2

121

119

24

L 33

1

- 3

1.5

0.17

1.

80E

+09

-154

.5

-691

.1

-45.

53

4.3

2.6

4 89

88

13

L

34

2 -

3 0.

5

0.16

1.

80E

+09

-735

.3

-189

3.0

-43.

68

7.7

5.1

121

90

89

12

L 12

-

- 3

0.8

0.13

1.

50E

+09

-544

.5

-191

9.1

-44.

74

5.9

0.9

94

110

109

10

L 55

14

-

2 0.

3

0.13

1.

40E

+09

-543

.1

-191

6.8

-42.

37

0.0

0.0

0 76

76

4

R

52

14

- 2

0.5

0.09

1.

00E

+09

-60.

9 -1

420.

4 -4

4.30

14

.0

5.2

157

121

113

43

L 56

13

-

2 0.

5


145

6.1.2 Comparison with Previous Observations at 1667 MHz

Mader et al. (1978) observed W3(OH) at 1667 MHz, and confirmed that 1667 MHz

emission existed near 1665 MHz emission, but little more could be stated. The first

good interferometer study of 1667 MHz in W3(OH) was in 1981, when Norris and

Booth (1981) published multi-frequency observations of both 1667 and 1665 MHz

with a single baseline interferometer in the UK. They detected 10 masers, with

positions relative to their 1665 MHz masers that match well with the brighter masers

in figure 6.1-3; but they did not list fluxes. The observation by Fouquet and Reid

(1982) suffered from very poor sensitivity with three US antennas, and they also did

not list fluxes; but the location, with respect to their 1665 MHz masers, of the single

maser detected at 1667 MHz is in agreement with the most intense masers in the

southwest of the map. Better observations were published by Norris et al. (1982),

using the MERLIN array to observe at 1612, 1665 and 1667 MHz, detecting 15

masers at 1667 MHz. Norris et al. had much lower resolution than the current data,

and made a mistake in assuming that masers coincident in position and velocity in

1665 and 1667 MHz were truly physically coincident. Since the Zeeman splitting of

1667 MHz transitions is only 60% of that of 1665 MHz transitions this is not a valid

assumption to make. As a result of this, their 1667 MHz masers are positioned about

200 mas too far northwest with respect to the 1665 MHz masers. Nevertheless, this

was still the best published 1667 MHz data until now. Finally, Gaume and Mutel

(1987) observed W3(OH) with the VLA in January 1985, though only for an

observational period of ~5 minutes. They detected 14 masers with a much lower

positional accuracy than Norris et al. (1982), and their positional agreement is

correspondingly poorer.

Comparing the individual maser fluxes measured by Norris et al. (1982) and

Gaume and Mutel (1987), with the current work and single dish flux measurements

leads to some puzzling results:

1. In Barrett and Rogers (1966) the spectra indicate equal peaks in RCP and

LCP of ~12 Jy.

2. In Norris and Booth (1981) the single dish spectra from 1977 indicate

similar shaped peaks in LCP and RCP of ~45 Jy.

3. In Norris et al. (1982) the single dish spectra from 1980 show a very

similar shape to 1977, but peak at a flux of ~20 Jy. However, they detect


146

individual masers with peak flux densities of 25.8, 25.0 and 20.0 Jy/beam

– indicating a probable miscalibration in one or other experiment. The

current data is compared in more detail with Norris et al. (1982) in the

next section.

4. In Fouquet and Reid (1982) the single dish spectra from 1980 show a

similar shape as before at a peak of ~20 Jy.

5. In Gaume and Mutel (1987), their 1985 observation gave only two masers

with peak flux above 1.8 Jy: 4.0 and 2.4 Jy! Their fluxes at 1665 MHz

however were in agreement with other data.

6. Figures 6.1-1a and b indicate the LCP at ~25 Jy and the RCP at ~15 Jy,

with peak individual maser fluxes from table 6.1-1 of 9, 5, 4 and 4

Jy/beam.

The data at face value suggest that the 1667 MHz masers have flared up

around 1977, then died right down by 1985, then flared back up by 1996, with peaks

all the while at the same velocities and probably the same locations. However, it

seems much more likely that flux miscalibrations in the previous work have

contributed some or most of this dramatic variation. The only concrete conclusion

that can be drawn is that the RCP emission has halved in intensity with respect to the

LCP, whilst the emission has stayed simple in velocity structure.

6.1.3 Comparison with Norris et al. (1982)

Although the alignment of 1667 MHz emission to 1665 MHz was incorrect,

Norris et al. (1982) correctly aligned the RCP and LCP of the 1667 MHz data. They

detected three areas of emission where emission in LCP and RCP would be coincident

if the maps were aligned thus: in one area, the RCP and LCP were separated by 2.2

km s–1, which they assumed was a Zeeman pair (see chapter 6.1.7). In the other two

areas, the emission was at roughly the same velocity (within their very poor velocity

resolution), but cannot be assumed to be a Zeeman pair in a near-zero magnetic field.

Their observation was in RCP and LCP only, but with the benefit of observational

data in all four Stokes parameters, it is now possible to say that these two areas

represent masers with significant linear polarization. An elliptically polarised maser

will show up as a brighter maser of the dominant hand and a weaker maser of the

other hand when measured with only circular feeds. This gives an indication of the


147

1667 MHz Old and New Data

-2300

-2100

-1900

-1700

-1500

-1300

-1100

-900

-700

-500

-300

-100

100

300

500

-1500-1300-1100-900-700-500-300-100100300500700900

RA Offset / mas

DE

C O

ffse

t / m

as

New Data Norris Data

Figure 6.1-4: 1667 MHz masers from the current new data overlaid with data from Norris et al.

1982.

14

8

Tab

le 6

.1-2

: M

aser

s th

at h

ave

been

iden

tifie

d in

bot

h 19

77 (

Nor

ris

et a

l. 19

82)

and

1997

epo

chs.

The

ast

eris

ked

mas

ers

have

bee

n as

sum

ed c

oinc

iden

t

1977

1997

V

LSR

P

eak

Flu

x R

A

DE

C

Circ

ular

F

eatu

re

VLS

R

Pea

k F

lux

RA

D

EC

C

ircul

ar

D

ensi

ty

Offs

et

Offs

et

Dire

ctio

n La

bel

D

ensi

ty

Offs

et

Offs

et

Dire

ctio

n

(km

/s)

(Jy/

Bea

m)

(mas

) (m

as)

(km

/s)

(Jy/

Bea

m)

(mas

) (m

as)

-4

4.8

3.3

-870

-6

62

R

a -4

4.6

0.4

-864

-6

52

R

a

-44.

6 0.

4 -8

61

-668

R

-44.

2 0.

5 -1

92

-672

R

b

-44.

0 0.

2 -1

55

-690

R

-43.

1 4.

6 -5

8 -1

440

R

c -4

2.9

1.3

-43

-144

0 R

-44.

3 7.

5 -3

87 *

-1

880

* L

d -4

4.2

9.3

-386

.7 *

-1

879.

66 *

L

-44.

5 1.

4 -3

87 *

-1

880

* R

d

-44.

0 0.

6 -4

48

-178

0 R

e

-44.

0 0.

6 -4

25

-180

9 R

-48.

0 2.

7 -6

4 -2

042

L f

-47.

8 0.

5 -6

6 -2

022

L

-48.

0 0.

6 -6

7 -2

024

R

f -4

7.5

0.3

-81

-203

5 L

-42.

3 25

.8

-645

-1

913

R

g -4

2.1

4.9

-671

-1

897

R

g

-42.

3 2.

0 -6

95

-188

2 R

g

-42.

1 1.

6 -6

57

-190

2 R

g

-42.

2 1.

3 -6

51

-191

4 R

g

-42.

3 0.

3 -6

23

-191

9 R

-45.

0 25

.0

-591

-1

920

L h

-44.

8 3.

7 -5

93

-191

8 L

h

-44.

8 2.

0 -6

04

-191

5 L

h

-44.

8 1.

9 -5

93

-192

6 L

-44.

5 20

.0

-645

-1

902

L i

-44.

4 4.

0 -6

73

-189

6 L

i

-44.

6 2.

2 -6

51

-191

3 L

i

-44.

4 1.

3 -6

57

-190

2 L

i

-44.

6 1.

2 -6

95

-188

3 L


149

absolute alignment between the present data and that of Norris et al. because only two

areas of maser emission show significant linear polarization in the current data. One

of the areas is currently the brightest maser at 1667 MHz; feature 13. This also

happens to be an isolated maser, and so even at the lower resolution of Norris et al. its

position should be accurate and unambiguous. By aligning feature 13 exactly with

their –44.5 km s–1 RCP feature and their –44.3 km s–1 LCP feature, an excellent

positional match is achieved which is shown in figure 6.1-4.

Figure 6.1-4 allows a comparison to be made between the two epochs. It

should be remembered from the previous section that there appear to be

inconsistencies in the fluxes measured in previously published data, and that

comparing fluxes from data of differing resolution may be unreliable because of the

effects of resolving out diffuse emission at higher resolutions. Table 6.1-2 shows

masers that have been identified in both epochs. Note that masers that show up as

LCP and RCP in the data of Norris (i.e. features d and f in table 6.1-2) are elliptically

polarised masers under one hand in the new data. Summing the flux for feature 4 in

the data of Norris et al. (feature d in table 6.1-2) indicates that this maser is at

approximately the same flux after nearly 2 decades. Features g, h, and i are more

complex to interpret, probably because they were not resolved by Norris et al., so their

fluxes would be consolidated. Their much higher fluxes may be a result of extended

emission or because the flux has in fact declined in intensity.

6.1.4 Maser Morphology

Immediately striking from figure 6.1-2 is the southern arc feature, where over half of

the masers are concentrated. The arc is just under 200 mas in extent, or ~400 AU at

2.2 kpc. The arc is shown in individual channels in figure 6.1-5. Channels 26-34 are

essentially RCP emission, 53-61 are LCP emission. Arcs and filaments are extremely

rare in ground state OH masers, although filamentary emission was seen in excited

state masers, at 13 GHz, that overlie the 1665 MHz maser group c (Baudry and

Diamond 1998).

The arc does not contain a velocity gradient so it is not an inclined disc. The

arc is most likely a shock front propagating to the south. Support for this can be seen

in the orientation of the masers’ position angles as shown in figure 6.1-6. Masers A

and B are in complex areas of overlapping maser emission where position angles from

Gaussian fitting are unreliable.


150

Figure 6.1-5: The 1667 MHz arc feature (rotated 90° anti clockwise). Channels 26-34 are

essentially RCP and channels 53-61 are LCP. Contours are at 75 mJy/beam.

-186

0-1

880

-190

0-1

920

-194

0-1

960

2627

28

MilliARC SEC

-186

0-1

880

-190

0-1

920

-194

0-1

960

2930

31

Mill

iAR

C S

EC

-500

-550

-600

-650

-700

-750

-186

0-1

880

-190

0-1

920

-194

0-1

960

3233

-500

-550

-600

-650

-700

-750

34

-186

0-1

880

-190

0-1

920

-194

0-1

960

5354

55

MilliARC SEC

-186

0-1

880

-190

0-1

920

-194

0-1

960

5657

58

Mill

iAR

C S

EC

-500

-550

-600

-650

-700

-750

-186

0-1

880

-190

0-1

920

-194

0-1

960

5960

-500

-550

-600

-650

-700

-750

61


151

Work by Elitzur et al. (1992) has shown that masers in shocks should be thin and

planar, e.g. discs, elongated in the plane of the shock. They also suggest however that

the true geometry of these masers may not be visible because of the beaming effects

of saturation. The above figure is evidence that the true geometry may in fact be

visible. Figure 6.1-7 shows a minimised best-fit circle plotted though the points in the

arc. The centre of the circle is at offset –580, –1746, and it has radius 177 mas (~350

AU). An arc such this might be expected if a spherically symmetric shock was

propagating outwards from the centre of the circle. The 1665 MHz masers in this area

(which actually overlap the 1667 MHz masers, see chapter 6.4.1) are indeed moving

south as shown in figure 5.25. The evolution and dissipation of this shock could be

responsible for the apparent decline in intensity of the masers in this area

demonstrated in table 6.1-2.

Figure 6.1-6: Position angles of the masers in the 1667 MHz arc feature. Note the occurrence of

Zeeman pairs, which gives the appearance of ‘crosses’.

A

B


152

1667 MHz Arc Feature

-2000

-1800

-1600

-800-600-400

RA Offset / mas

DE

C O

ffset

/ m

as

Less than 0.400 Jy

0.400 -> 2.000 Jy

2.000 -> 10.000 Jy

Figure 6.1-7: The 1667 MHz arc feature with minimised circular fit. The arrowed masers were not

used in the fitting.


153

6.1.5 Maser Lineshapes

Figure 6.1-8 shows the lineshapes of the six most intense masers at 1667 MHz (all the

masers above 2 Jy/beam). The 1667 MHz maser lineshapes are generally very well

approximated by Gaussians. 23 of the 41 masers were fit successfully; 17 of the

remaining 18 could not be fitted adequately because the maser consisted of less than

four channels of emission. The one maser that failed and had enough channels was

still approximately Gaussian.

Figure 6.1-8: Lineshapes of masers over 2 Jy/beam at 1667 MHz. Observed flux errors are

negligible at these scales. Clockwise from top left: Features 1, 13, 23, 28, 32, 39.


154

6.1.6 Polarization

This is the first VLBI dataset of 1667 MHz in W3(OH) which contains all polarization

data. From table 6.1-1 it is clear that all of the masers are highly circularly polarised.

Of the 41 masers detected, 22 (54%) were LCP, and 19 (46%) RCP. 80% of the

masers have over 90% circular polarization. Indeed, only four masers have less than

75% circular or total polarization, and these four masers are notable. Two of them

(features 36 and 37) are faint and are highly blue-shifted at the extreme south of the

map. Of the other two, one is the most intense maser at 1667 MHz (feature 13) and

the other is a close companion (feature 53). Feature 13 is twice as intense as the next

most intense maser, and is the only maser with a significant intensity of linear

polarization: it has 0.92 Jy of linear flux, compared to the next highest of 0.11 Jy. All

the rest have linear fluxes under 0.1 Jy. Unlike 1665 MHz, at 1667 MHz there is no

strong variation in the amount of polarization across the map. This may well be

because the 1667 MHz masers do not cover enough of W3(OH) for very significant

variations to become apparent.

Figure 6.1-9 shows the Stokes parameters for feature 13. Notice how the

Stokes Q and U fluxes vary across the maser lineshape, indicating a gradual change in

the linear polarization plane across the maser. Feature 13 is the only 1667 MHz maser

for which the linear flux is strong enough to give reliable polarization angle

measurements, but comparing it to masers at 1665 MHz (e.g. Feature 474), we see that

this angle change is unusual behaviour – polarization angles where they are reliable

are generally stable across the maser lineshape. This could be caused by a change in

Figure 6.1-9: Stokes parameters for 1667 MHz feature 13.

Stokes Parameters for 1667 Feature 13

-8.0

-6.0

-4.0

-2.0

0.0

2.0

4.0

6.0

8.0

10.0

12.0

-45.0 -44.8 -44.6 -44.4 -44.2 -44.0 -43.8 -43.6

Velocity / (km/s)

Flu

x D

ensi

ty /

(Jy

/bea

m)

IQ

U

V


155

the magnetic field direction within the propagation path of the maser. This is more

likely than Faraday rotation because of the high free electron density needed for

Faraday effects within a maser.

6.1.7 Magnetic Field and Velocity Structure

From table 6.1-1, 14 Zeeman pairs were found which are listed in table 6.1-3. This

represents a 68% chance of a maser being in a Zeeman pair. However, in the arc

feature 22 of the 23 masers are in a Zeeman pair, or 96%! This much higher incidence

of Zeeman pairs may indicate a difference in the conditions of the arc, which are more

favourable for the formation of Zeeman pairs – possibly that there are very much

lower velocity shifts in the line of sight, which is what might be expected in a shock

expanding to the south. The locations of the Zeeman pairs are shown in figure 6.1-10.

Figure 6.1-10: 1665 MHz Zeeman pairs in W3(OH). The masers have been divided up into three

groups for analysis.

1667 MHz Zeeman Pairs

-2300

-2100

-1900

-1700

-1500

-1300

-1100

-900

-700

-500

-300

-100

100

300

500

-1500-1300-1100-900-700-500-300-100100300500700900RA Offset / mas

DE

C O

ffset

/ m

as

East Arc Center


156

The median ratio of the fluxes of the Zeeman pairs is 1.7:1, and the mean is

2.1:1. Thus, the Zeeman pairs are much more even in intensity than at 1665 MHz.

Figure 6.1-11 shows the distribution of magnetic field strengths and demagnetised

velocities with RA and DEC. The field strength vs. RA map shows a striking

uniformity of the magnetic field in the arc feature, and that the isolated maser to the

east has a noticeably low field strength. The uniformity of the field in the arc

indicates a constant density on a scale of 400 AU in the arc. For the maser at a field

strength of 2.4 mG, this value is much lower than any maser at 1665 MHz, and

surprisingly is lower than the value of the field in the surrounding gas given in Güsten

et al. (1994) of 3.2 mG. This maser then may lie in gas that is actually less dense than

much of the cloud in this region. Such low fields were proposed by Bloemhof et al.

(1992) at 1665 MHz but not reproduced in this work at that frequency; they were also

proposed by Desmurs et al. (1998) at 6 GHz in W3(OH). The range of field strengths

is however still in good agreement with other 1667 MHz measurements in Gaume and

Mutel (1987), and in Norris et al. (1982) whose proposed Zeeman pair would have a

field of about 6 mG.

In the demagnetised velocity vs. RA graph, note that there is no possibility of a

velocity gradient along the arc, within the errors on the velocities. Note also in the

Peak Flux RA DEC Magnetic Magnetic Demagnetised Demagnetised Pair Location


(Jy/Beam) (mas) (mas) (mG) ± (mG) (km/s) ± (km/s)

0.41 -863.8 -652.5 2.43 0.18 -45.10 0.03 a East

1.98 -694.5 -1882.4 6.58 0.17 -43.42 0.03 a Arc

4.92 -671.4 -1896.6 6.45 0.14 -43.28 0.03 a Arc

1.55 -656.8 -1902.4 6.59 0.17 -43.25 0.03 a Arc

2.16 -651.4 -1913.0 6.62 0.16 -43.39 0.03 a Arc

0.74 -622.4 -1918.9 6.77 0.16 -43.43 0.03 a Arc

2.02 -603.9 -1914.9 6.36 0.17 -43.67 0.03 a Arc

3.69 -592.8 -1917.7 6.57 0.14 -43.63 0.02 a Arc

1.90 -592.7 -1925.9 6.67 0.18 -43.59 0.03 a Arc

0.71 -564.2 -1932.8 6.95 0.18 -43.24 0.03 b Arc

0.13 -544.5 -1919.1 6.70 0.18 -43.55 0.03 b Arc

1.20 -514.7 -1908.7 6.74 0.17 -43.37 0.03 a Arc

0.23 -155.2 -690.3 4.22 0.18 -44.78 0.03 b Center

0.26 -58.6 -1418.1 4.22 0.18 -43.55 0.03 b Center

Table 6.1-3: 1667 MHz Zeeman pairs. Information for peak flux, RA and DEC are from the peak

channel in the brighter of the two Zeeman components.


157

demagnetised velocity vs. DEC graph that there is a velocity gradient that closely

follows the gradient in 1665 MHz, suggesting that the two main lines are in the same

moving body of gas. The blue-shifted masers in the extreme south do not have a

Zeeman partner to deduce their demagnetised velocity, but if their magnetic field

strength is in the typical range then they are probably not part of the same bulk

movement as the rest of the 1667 MHz masers. This will be examined more in

chapter 6.4.3, when the different frequencies are compared.

6.1.8 Statistical Relationships

Figure 6.1-12 shows the relationship between maser size and field strength. The

effect of magnetic beaming is slightly masked because the majority of the masers are

concentrated in the arc where the field strength is very consistent. The average

FWHM of the Gaussian fit 1667 MHz masers is ~240 m s–1, compared to that of the

1665 MHz masers of ~280 m s–1. This is the lowest of all the ground state lines. This

is as would be expected from the modelling of Gray et al. 1992 (see chapter 3.3.3),

which suggested that 1667 MHz masers require lower velocity shifts over their

Figure 6.1-11: Variation in magnetic field strength and demagnetised velocity with RA and DEC at

1667 MHz. The points are plotted as error bars; errors were calculated as per chapter 4.5.4


0

1

2

3

4

5

6

7

8

-2500-2000-1500-1000-5000

DEC Offset / mas

Fiel

d S

tren

gth

/ m

G


0

1

2

3

4

5

6

7

8

-1000-900-800-700-600-500-400-300-200-1000

RA Offset / mas

Fiel

d S

tren

gth

/ m

G

Demagnetised Velocity vs DEC

-45.5

-45.0

-44.5

-44.0

-43.5

-43.0

-2500-2000-1500-1000-5000

DEC Offset / mas

Dem

ag. V

el /

(km

/s)

Demagnetised Velocity vs RA

-45.5

-45.0

-44.5

-44.0

-43.5

-43.0

-1000-900-800-700-600-500-400-300-200-1000

RA Offset / mas

Dem

ag.

Vel

/ (k

m/s

)


158

propagation length. It might also indicate that 1667 MHz masers inhabit cooler gas

than 1665 MHz masers.


Although there are a smaller number of masers at 1667 MHz than 1665 MHz, there

are still enough for a good two-point correlation, shown in figure 6.1-13a and b. The

much smaller extent of the 1667 MHz emission is shown by the plot, as is the strong

tendency to cluster at the scale of 300 mas (600 AU) and under. This clustering scale

is the same as in 1665 MHz, and lends weight to the idea that the same gas structures

may be responsible for both masing transitions.

Figure 6.1.14 shows a zoom in on the first 300 mas of the correlation with

smaller bins and higher resolution. It shows that the distribution within a maser

cluster concentrates the masers ever more at closer distances, right down to between

10 and 20 mas (20–40 AU), just as in 1665 MHz. At around 8 mas the correlation

breaks down and suddenly drops to zero at 4 mas. Again, this is due to the resolution

limit of the observation.

Maser size vs Field Strength

0

1

2

3

4

5

6

7

8

0 50 100 150 200 250 300 350 400

Maser Area / (mas)^2

Fiel

d S

tren

gth

/ m

G

Figure 6.1-12: 1667 MHz maser area versus magnetic field strength.


159

Figure 6.1-13 a) and b): a) Top, the 1667 MHz masers and a random selection of points in the

same space. b) Bottom, the two-point correlation of the 1667 MHz masers.

Figure 6.1-14: The two-point correlation of the 1667 MHz masers, with smaller bins and hence

higher resolution.


160

6.2 1612 MHz


The 1612 MHz total power spectra for RCP and LCP are shown in fig 6.2-1a. They

are from 1998 (Masheder, private communication), two years later than the

interferometry observation. Immediately it is clear that 1612 MHz emission comes

from a relatively small velocity spread that is towards the red-shifted end of the band.

The similarity of the LCP and RCP lineshapes suggests that the different velocities of

the peaks may be due the Zeeman effect, but the large difference in the ratio of fluxes

would suggest caution in any assumption – the RCP emission is clearly much brighter

than the LCP emission. As will be seen later, the channel maps themselves do indeed

show that the velocity differences are due to the Zeeman effect. Relatively narrow

and uncomplicated lineshapes are typical of SFR 1612 MHz maser regions, e.g.

Caswell (1999).

Figure 6.2-1b shows the Stokes I spectrum of the maser fluxes calculated from

the channel maps. The maser spectrum also shows a very weak but notable blue-

shifted maser feature at –47.5 km s–1.

Results of 2-dimensional Gaussian fitting of maser emission in the maps are

shown in table 6.2-1. The 1612 MHz integrated flux contour map and the maser map

are shown in figs 6.2-2 and 6.2-3. Channel maps of the masers in contour form can be

found in appendix A. As can be readily seen, the 1612 MHz masers are divided into

two groups; the brighter masers in the north and a line of weaker masers strung out in

a line in the south. Figure 6.2-3 shows that the northern masers lie just within the

contours of continuum emission which defines the UCH II region, whilst the southern

masers lie well outside the contours.

The low number of 1612 MHz masers detected means that the sample size is

too small for a statistical analysis to be performed on the data, certainly not to the

level which was possible for the 1665 MHz data in chapter 5.9.


161

Figure 6.2-1a) and b): a) Top, the 1612 MHz total power spectra for RCP (dotted) and LCP (solid).



-2

0

2

4

6

8

10

12

14

16

-50 -49 -48 -47 -46 -45 -44 -43 -42 -41 -40

Velocity / (km/s)

Flu

x /

Jy


0

2

4

6

8

10

12

14

16

18

-50 -48 -46 -44 -42 -40

Velocity / (km/s)

Flux

/ J

y


162

Figure 6.2-2: The 1612 Stokes I integrated flux map. Channels 32 to 34 have been omitted because

of their high levels of dynamic range limited noise. This does not affect the masers significantly.


ARC

SEC

ARC SEC0.5 0.0 -0.5 -1.0

0.5

0.0

-0.5

-1.0

-1.5

-2.0


163


-2300

-2100

-1900

-1700

-1500

-1300

-1100

-900

-700

-500

-300

-100

100

300

500

-1300-1100-900-700-500-300-100100300500700900

RA Offset / mas

DE

C O

ffset

/ m

as

Less than 0.400 Jy

0.400 -> 2.000 Jy

2.000 -> 10.000 Jy



of a VLA 2 cm continuum map (Kawamura and Masson 1998)

16

4

Tab

le 6

.2-1

: T

he 1

612

MH

z M

aser

s. D

ata

is f

or p

eak

chan

nel,

exce

pt la

st f

ive

colu

mns

, whi

ch a

re f

or th

e w

hole

fea

ture

. Fe

atur

es w

ith n

o FW

HM

mea

sure

men

t eit

her

faile

d

lines

hape

fitt

ing

or h

ad to

o fe

w c

hann

els

for

a fi

t. E

rror

s ar

e de

taile

d in

cha

pter

4.4

.1.

RA

and

DE

C o

ffse

ts a

re f

rom

the

refe

renc

e fe

atur

e.

Pea

k F

lux

Brig

htne

ss

RA

D

EC

V

LSR

D

econ

v.

Dec

onv.

D

econ

v.

Tot

al

Circ

ular

Li

near

C

ircul

ar

Fea

ture

Z

eem

an

FW

HM

C

hann

els

Inte

grat

ed

Den

sity

T

emp.

O

ffset

O

ffset

Maj

. Axi

s M

in. A

xis

Pos

. Ang

. P

ol.

Pol

. P

ol.

Dire

ctio

n N

umbe

r P

air

P

rese

nt

Fea

ture

(Jy/

Bea

m)

(K)

(mas

) (m

as)

(km

/s)

(mas

) (m

as)

(deg

rees

) (p

erce

nt)

(per

cent

) (p

erce

nt)

(Num

ber)

(m

/s)

F

lux

(Jy

)

11.3

4 5.

52E

+11

-50.

3 -2

14.2

-4

2.55

4.

6 3.

0 16

1 10

0 10

0 0

R

4 1

221

11

42.1

3.53

1.

94E

+11

-50.

3 -2

14.2

-4

3.46

4.

7 2.

6 16

8 10

0 10

0 1

L 25

1

215

6 11

.5

1.97

4.

11E

+10

172.

6 -1

817.

7 -4

2.09

6.

4 5.

0 83

10

1 10

1 1

R

3 4

357

8 13

.9

1.53

2.

80E

+10

112.

8 -1

831.

9 -4

2.91

7.

8 4.

8 79

94

94

2

R

22

3 26

8 7

8.2

1.15

2.

43E

+10

195.

6 -1

816.

4 -4

2.64

8.

1 3.

8 79

10

3 10

3 3

R

5 5

403

6 8.

7

0.99

4.

98E

+10

-38.

7 -2

13.4

-4

3.64

4.

6 3.

1 17

4 10

1 10

1 2

L 28

2

209

4 2.

9

0.98

4.

13E

+10

-38.

7 -2

13.5

-4

2.73

4.

1 3.

7 16

98

98

3

R

21

2 19

2 4

2.7

0.78

3.

90E

+10

9.3

-113

.2

-43.

00

5.3

2.6

8 96

95

1

L 24

6

207

4 2.

6

0.51

4.

73E

+09

184.

0 -1

817.

2 -4

3.09

14

.3

5.2

78

97

97

1 L

23

5 36

6 7

5.3

0.36

1.

58E

+10

175.

0 -1

818.

0 -4

3.18

6.

6 2.

3 78

89

89

4

L 26

4

- 11

2.

0

0.32

4.

93E

+09

118.

0 -1

830.

7 -4

3.91

9.

0 4.

9 70

96

96

5

L 27

3

- 9

3.1

0.28

7.

08E

+09

-156

.1

-179

9.1

-41.

46

7.3

3.6

72

96

96

5 R

1

- 34

1 5

1.5

0.27

3.

92E

+09

645.

2 -1

633.

4 -4

7.54

10

.1

4.5

18

90

90

12

L 29

-0

24

8 4

1.5

0.25

1.

02E

+10

264.

5 -1

710.

4 -4

1.46

5.

9 2.

9 95

74

74

5

R

2 -

- 3

0.9

0.19

2.

60E

+10

9.7

-113

.7

-41.

73

5.2

0.9

34

80

79

9 R

30

6

- 2

0.4


165


The first interferometer study of 1612 MHz in W3(OH) was in 1982, when both

Norris et al. (1982) and Fouquet and Reid (1982) published multi-frequency

observations. Fouquet and Reid had very poor sensitivity in their observation with

three US antennas, and did not list their fluxes, though their 1720 MHz and 1612 MHz

positions match well with subsequent data. Norris et al. observed at 1612, 1665 and

1667 MHz with the MERLIN observatory in 1980-81. The 1612 MHz transition was

observed only in RCP. They detected only two masers at 1612 MHz, which they

positioned on a map by assuming that the maser which was coincident in velocity with

a maser in their 1667 MHz map was also coincident in position. For a variety of

reasons including Zeeman effects and the results we find in chapter 6.4.1 this was not

a valid assumption to make. 1612 MHz in W3(OH) was also studied by Gaume and

Mutel (1987) with the VLA in January 1985, though only for an observational period

of ~5 minutes. There has been no newer data at 1612 MHz recently until this work

(maser interest at 1612 MHz has been mostly confined to late-type stars).

Direct comparison with other work at differing resolutions is difficult because

of the issues of resolving out more diffuse emission, and of the error in absolute

position. Even so, it ought to be possible to compare the relative positions of the

masers in earlier maps using velocity information that is generally more reliable.

Comparing fig 6.2-1a with the single dish spectrum of Norris et al. (1982) indicates

that the emission has remained qualitatively the same in the 16 years since their

observation, with the main peak at –42.5 km s–1.

Turning now to the maps, Norris et al. detected only two masers, separated by

just under 100 mas. This means that they can only have detected either the northern

or the southern cluster. A quick inspection of velocities indicates that, surprisingly,

they only detected the southern emission – and have placed it ~800 mas too far west

with respect to the continuum by aligning the peak with the brightest 1665 MHz

maser. Their two maser features correspond well to feature 22 and an unresolved

amalgamation of features 3 and 5 (and possibly 22) from table 6.2-1. The lower

intensity of masers in the current work (by a factor of three) is likely to be the result of

resolving out slightly diffuse emission – the channel maps of these features (Appendix

A, figure A2) show that they are well resolved – although an additional decline in

brightness cannot be ruled out. The intensity of the northern emission now would


166

certainly make it clearly visible then; and their single dish spectrum at the time shows

a peak at the –42.5 km s–1 velocity of the RCP northern feature. It is not clear what

the presence of the –42.5 km s–1 peak in the total power spectrum represents if it is not

the northern feature.

The VLA observations by Gaume and Mutel (1987) had much poorer spatial

resolution and velocity resolution (only 1.1 km s–1) than the present work or that of

Norris et al., but their sensitivity was better than that of Norris, having a 1σ noise of

~30 mJy/beam. Gaume and Mutel did detect some weak northern emission: they

detected emission of 0.3 Jy/beam from two LCP spots just 5 mas apart (although their

resolution gives them position errors of ~60 mas) which would appear to be features

28 and 25 in table 6.2-1. They also detected the weak south-eastern highly blue-

shifted emission of feature 29 at an unchanged flux, and a central-eastern RCP maser

at 0.1 Jy/beam, the only maser emission of theirs that was not detected. Notably

absent again is any indication of northern RCP maser emission. Taken with the Norris

et al. data this does seem to indicate that the 11 Jy RCP maser in the present data has

intensified from a negligible flux in 11 years, and as such demonstrates a high degree

of variability in the northern 1612 MHz masers, whilst at the same time showing a

quite constant level of emission from the southern masers.

Agreement with observations in December 1993 (M. D. Gray, unpublished

work) is excellent. Again, emission from the southern conglomerate of masers around

features 3 and 5 are the brightest emission with peak of 5.3 Jy/beam, but Gray did

detect the northern RCP maser (presumably unresolved with feature 21) with a flux of

2.7 Jy/beam. This gives a definitive indication of the appearance of this maser over

11 years, as shown in table 6.2-2.

Year RCP Flux

(Jy/beam)

LCP Flux

(Jy/beam)

1982 Undetected Undetected

1987 Undetected 0.3

1993 2.7 Undetected

1996 11.3 3.5

Table 6.2-2: Variation in intensity of the northern masers over recent years.


167


The small number of 1612 MHz masers show simple morphology, especially in the

north. The northern masers are very small, with an average semi-major axis of 4.7

mas, while the southern masers have an average semi-major axis of 8.4 mas. The

southern masers (excepting the highly blue-shifted far south-east maser) share a

common orientation also: they all have position angles in the range of 70-95°, with a

mean of 79°. The rough line the masers make on the sky also has an angle of ~80°. It

could be that whatever is causing these masers to lie in a rough line is also responsible

for position angles; possibly they line up with a weak advanced shock off the southern

edge of the UCH II region.


Figure 6.2-4 shows the six most intense maser lineshapes from the 1612 MHz data.

As can be seen, the maser lineshapes are in general very well approximated by

Gaussians. The two masers with four or more channels (the minimum to attempt

fitting) that failed lineshape fitting were both in the south of the map. Feature 26

merges with feature 23, which is the likely cause of its non-Gaussian lineshape.

Feature 27 has a non-Gaussian unsymmetrical lineshape, even though it is an isolated

spot. Its brighter RCP Zeeman counterpart however fits well with a Gaussian, the pair

is shown in figure 6.2-5.

6.2.5 Polarization

This is the first VLBI dataset of 1612 MHz in W3(OH) which contains all polarization

data. From table 6.2-1 it is clear that all of the masers are highly circularly polarised.

For the most intense masers, the circular polarization is very close to 100%, with

negligible linear and unpolarised emission. The only masers to have less than 90%

circular polarization have low fluxes, and are subject to larger errors. The only maser

to have significant linear polarization is interesting – it is the highly blue-shifted

maser in the extreme southeast of the map. The statistical significance of this is not

apparent with a sample of one maser, but has more significance when taken in

conjunction with the much more numerous 1665 MHz data (see chapter 6.4.3). Even

as a single feature, it is clear that this maser is not associated with the main southern

group of masers.


168

Figure 6.2-4: The 6 most intense maser lineshapes at 1612 MHz. Observed flux errors are

negligible at these scales. Clockwise from top-left: Features 3, 4, 5, 22, 25, 28.


169


From table 6.2-1, 6 Zeeman pairs were found – the first to be found in 1612 MHz in

W3(OH). They are shown in table 6.2-3. With only three of the masers not in a

Zeeman pair this means that 80% of the masers were in a Zeeman pair. All the masers

not in Zeeman pairs were in the southern group; and from the map, all of them were

solitary masers not part of any cluster. It is not possible to say whether their solitary

nature is responsible for the lack of a Zeeman partner (there is no obvious mechanism

to suggest why this might be so), but rather it is more likely that as they are weak

masers then the chances of observing an even weaker partner is slim.


3.5:1. From table 2.2 it can been seen that for 1612 MHz there are potentially a

number of possibilities for the magnetic field strength to produce a certain splitting

since three separate Zeeman pair separations can theoretically result. Clearly, this

result is never seen – only one pair of lines has ever been observed. The question is

then: which pair is being observed? From figure 2.5 it might be expected that the

strongest emitting Zeeman pair (labelled with splitting 1/4) is the one being observed;

and indeed comparing the various magnetic field strengths possible in the satellite

lines with unambiguous field strengths from e.g. 1665 MHz, one does indeed find the

best agreement for the inner splitting 1/4 Zeeman pair. It is conceivable that more

Figures 6.2-5: Zeeman pair 3 (features 22 and 27). Feature 27 has a non-Gaussian lineshape and is

an isolated spot.

Stokes Parameters for 1612 Zeeman Pair 3

-0.5

0.0

0.5

1.0

1.5

2.0

-44.2 -44.0 -43.8 -43.6 -43.4 -43.2 -43.0 -42.8 -42.6 -42.4

Velocity / (km/s)

Flu

x D

ensi

ty /

(Jy

/bea

m)

IQU

V


170

ambiguity could arise if the brighter component of a Zeeman pair were a splitting 1/4

line and a weaker component were of another splitting, e.g. 3/4. However, bearing in

mind the mechanisms proposed for suppression of Zeeman partners above it is

unlikely that conditions could be contrived so as to allow suppression of one Zeeman

partner on either side of the splitting. There is still the possibility of the case where

some Zeeman pairs could be of one splitting type, whilst others could be of another.

Fortunately, this scenario is eased by the next splitting up from the 1/4 pair being three

times larger in magnitude. This therefore ought to only be of cause for worry when

unexpectedly large magnetic-field strengths are encountered, such that a field strength

of one-third of this is also reasonable. Zeeman pair 6 in table 6.2-3 has the highest

field strength of 10.7 mG – alternate values are 3.5 mG and 2.1 mG – whilst Zeeman

pair 5 has the lowest field strength of 4.8 mG. It is not inconceivable therefore, that

Zeeman pair 6 might be exhibiting the weaker splitting 3/4 Zeeman lines.

Figure 6.2-6 shows the distribution of magnetic field strengths and

demagnetised velocities with RA and DEC. The range of magnetic field strengths is

~5 to ~11 mG, which is in agreement with magnetic field measurements at 1665 MHz,

and also with the previous 1612 MHz Zeeman pair measured in W49 by Gaume and

Mutel (1987) which gave a field of 9.1 mG. As can be seen the demagnetised velocity

spread of the masers is very narrow, especially in relation to the other ground state

frequencies. Unfortunately the blue-shifted feature at –47.5 km s–1 does not have a

Zeeman partner, so its unusual velocity cannot be probed further (but see chapter 6.4.3

for comparisons with other ground state masers in the vicinity).




11.34 -50.3 -214.2 7.56 0.39 -42.97 0.02 a North

0.99 -38.7 -213.4 7.37 0.51 -43.19 0.03 a North

0.78 9.3 -113.2 10.68 0.53 -42.39 0.03 b North

1.97 172.6 -1817.7 8.41 0.54 -42.67 0.03 a South

1.53 112.8 -1831.9 7.92 0.53 -43.42 0.03 a South

1.15 195.6 -1816.4 4.65 0.54 -42.84 0.03 b South

Table 5.5: 1612 MHz Zeeman pairs (assuming smallest magnetic splitting). Information for peak

flux, RA and DEC are from the peak channel in the brighter of the two Zeeman components.


171


Even with the small number of masers at 1612 MHz, a relationship is apparent linking

the size (or area) of the masers with the magnetic field. This is shown in fig 6.2-7.

This is an effect of magnetic beaming.

The average FWHM of the 1612 MHz masers which were fit with a Gaussian

is ~275 m s–1, which is in good agreement with the figure of ~280 for 1665 MHz.


The small number of masers makes for a poor graph of the two-point correlation.

Still, figure 6.2-8a does clearly show evidence for clustering of the masers on a 200

mas scale, and additional structure on a 1.6'' scale.


0

2

4

6

8

10

12

-1000-800-600-400-2000200400

RA Offset / mas

Fiel

d S

tren

gth

/ m

G


0

2

4

6

8

10

12

-2000-1800-1600-1400-1200-1000-800-600-400-2000

DEC Offset / mas

Fiel

d S

tren

gth

/ m

G


-43.6

-43.4

-43.2

-43.0

-42.8

-42.6

-42.4

-42.2

-1000-800-600-400-2000200400

RA Offset / mas

Dem

ag.

Vel

/ (

km/s

)


-43.6

-43.4

-43.2

-43.0

-42.8

-42.6

-42.4

-42.2

-2000-1800-1600-1400-1200-1000-800-600-400-2000

DEC Offset / mas

Dem

ag.

Vel

/ (

km/s

)


1612 MHz. Error bars are shown; errors were calculated as per chapter 4.5.4


172


Maser area vs Field Strength

0

2

4

6

8

10

12

0 50 100 150 200 250

Maser area / mas 2

Fie

ld S

tren

gth

/ m

G

Figure 6.2-8 a) and b): a) Top, the 1612 MHz masers and a random selection of points in the same

space. b) Bottom, the two-point correlation of the 1612 MHz masers.


173

6.3 1720 MHz


The total power spectra for RCP and LCP of the 1720 MHz emission are shown in fig

6.3-1a. They are from 1998 (Masheder, private communication), two years after the

interferometry observation. The 1720 MHz spectrum immediately suggests that the

Zeeman effect is at work, splitting the LCP and RCP fairly evenly. As will be

confirmed later, the 1720 MHz emission in W3(OH) is indeed very evenly Zeeman

split – by far the most even of all the ground-state lines. 1720 MHz maser emission,

like 1612 MHz emission, is often quite simple and often confined to a relatively

narrow velocity band, e.g. Caswell (1999); in this respect W3(OH) might be

considered quite a complex source at this frequency. Figure 6.3-1b shows the Stokes I

spectrum of the maser fluxes calculated from the channel maps.

Results of 2-dimensional Gaussian fitting of maser emission in the maps are

shown in table 6.3-1. The 1720 MHz integrated flux contour map and the maser spot

map are shown in figs 6.3-2 and 6.3-3. Channel maps of the masers in contour form

can be found in appendix A. All the 1720 MHz emission is concentrated in three very

small areas of emission – in the north, south and west – making it the simplest of all

the ground-state OH lines.


W3(OH) was observed by Lo et al.(1975) at 1720 MHz, who deduced using early

VLBI that the LCP/RCP doublet at that frequency originated from less than 3 mas

away from each other, providing some of the first strong evidence of the Zeeman

effect in these masers. The next work was Fouquet and Reid (1982), who published

multi-frequency ground-state observations of W3(OH). As was mentioned with

respect to 1612 MHz, Fouquet and Reid suffered from very poor sensitivity in their

observation with three US antennas, and they detected just one 1720 MHz maser. The

position of this maser matches well with the brightest 1720 MHz maser in this work,

in the north of the map. Gaume and Mutel (1987) observed W3(OH) at 1720 MHz,

along with the other ground-state lines, but were only able to observe in RCP at 1720

MHz. They detected 5 masers in two clusters, which lie roughly over the northern and


174

Figure 5.1a) and b): a) Top, the 1665 MHz total power spectra for RCP (dotted) and LCP (solid).



-2

0

2

4

6

8

10

12

-50 -49 -48 -47 -46 -45 -44 -43 -42 -41 -40

Velocity / (km/s)

Flux

/ J

y


0

1

2

3

4

5

6

7

8

9

10

-50 -49 -48 -47 -46 -45 -44 -43 -42 -41 -40

Velocity / (km/s)

Flu

x /

Jy


175

Figure 6.3-1: The 1720 Stokes I integrated flux map. Channels 34, 58, 59, 66-69 to 63 have been

omitted because of their high levels of noise. This does not affect the masers significantly.


AR

C S

EC

ARC SEC0.5 0.0 -0.5 -1.0

0.5

0.0

-0.5

-1.0

-1.5

-2.0


176


-2300

-2100

-1900

-1700

-1500

-1300

-1100

-900

-700

-500

-300

-100

100

300

500

-1500-1300-1100-900-700-500-300-100100300500700900RA Offset / mas

DE

C O

ffset

/ m

as

Less than 0.400 Jy

0.400 -> 2.000 Jy

2.000 -> 10.000 Jy


of a VLA 2 cm continuum map (Kawamura and Masson 1998).

17

7

Tab

le 6

.3-1

: T

he 1

720

MH

z M

aser

s. D

ata

is f

or p

eak

chan

nel,

exce

pt la

st f

ive

colu

mns

, whi

ch a

re f

or th

e w

hole

fea

ture

. Fe

atur

es w

ith n

o FW

HM

mea

sure

men

t eit

her

faile

d

lines

hape

fitt

ing

or h

ad to

o fe

w c

hann

els

for

a fi

t. E

rror

s ar

e de

taile

d in

cha

pter

4.4

.1.

RA

and

DE

C o

ffse

ts a

re f

rom

the

refe

renc

e fe

atur

e.

Pea

k F

lux

Brig

htne

ss

RA

D

EC

V

LSR

D

econ

v.

Dec

onv.

D

econ

v.

Tot

al

Circ

ular

Li

near

C

ircul

ar

Fea

ture

Z

eem

an

FW

HM

C

hann

els

Inte

grat

ed

Den

sity

T

emp.

O

ffset

O

ffset

Maj

. Axi

s M

in. A

xis

Pos

. Ang

. P

ol.

Pol

. P

ol.

Dire

ctio

n N

umbe

r P

air

P

rese

nt

Fea

ture

(Jy/

Bea

m)

(K)

(mas

) (m

as)

(km

/s)

(mas

) (m

as)

(deg

rees

) (p

erce

nt)

(per

cent

) (p

erce

nt)

(Num

ber)

(m

/s)

F

lux

(Jy

)

7.13

3.

40E

+11

-2.4

27

.5

-45.

60

4.8

2.6

7 98

98

2

L 1

12

509

16

60.7

5.08

1.

89E

+11

-2.4

27

.7

-44.

74

5.1

3.2

8 10

0 10

0 1

R

1 5

0 15

54

.6

5.08

2.

49E

+11

25.8

50

.6

-42.

70

4.1

2.9

172

100

100

1 R

2

1 24

3 10

19

.7

2.76

1.

48E

+11

25.7

50

.3

-43.

47

4.3

2.6

172

100

100

3 L

2 21

23

7 8

10.5

2.26

5.

33E

+10

-147

.0

-112

2.1

-43.

13

7.3

3.4

150

99

99

2 R

3

2 24

9 7

11.0

2.24

5.

93E

+10

-147

.0

-112

2.2

-43.

72

6.7

3.3

154

98

98

2 L

3 4

235

8 11

.5

0.40

6.

67E

+09

-156

.0

-111

3.7

-43.

38

9.6

3.7

145

97

97

6 R

5

3 32

2 5

2.2

0.12

8.

87E

+08

-853

.0

-600

.2

-45.

17

21.4

3.

8 17

9 92

91

7

L 4

11

- 3

1.0

0.12

5.

28E

+08

-853

.3

-600

.9

-44.

66

24.1

5.

7 7

93

91

20

R

4 23

-

2 0.

7

0.11

3.

90E

+09

-156

.6

-111

2.3

-44.

06

6.8

2.3

177

94

94

1 L

5 22

-

2 0.

3


178

southern emission shown in figure 6.2-3. Before the present work, the best data at

1720 MHz was the extremely high-resolution observation by Masheder et al. (1994).

Masheder et al. observed for 12 hours in LCP in November 1989 with two antennas in

the US, one in China, one in Sweden, and the 100 m antenna in Germany. This gave

them long baselines giving a beam size of 2.4 mas and a 1σ noise of ~24 mJy/beam –

a higher resolution but lower sensitivity than the current data, and in only LCP. They

also detected only the northern and southern emission: detecting just one maser in the

south and probably three masers in the north within about 15 mas (beam artefacts in

their maps give some level of ambiguity to the masers). The three masers in their

northern area are so close that they are not individually resolved in the current data; all

three are incorporated into feature 12. The effects of this can be seen in the lineshape

(see chapter 6.3.3) and in the motion of the Gaussian fit position of feature 12, which

is shown in figure 6.3-4.

First impressions of a velocity gradient such as shown in figure 6.3-4 might be

that this indicates ordered motion, perhaps rotational. But looking at the higher

resolution data of Masheder et al. it is clear that the direction of movement (with

increasingly negative velocity) – first moving southwest then west – is in good

agreement with the relative positions and velocities of their three masers. Therefore,

what is observed in figure 6.2-4 is the superposition of three unresolved masers at

slightly different velocities, indicating that this area of emission has remained

unchanged over 7 years. Motion of Gaussian fit positions of maser components like

that seen in figure 6.3-4 may be a useful diagnostic of close, unresolved masers.

Just 35 mas to the northeast is a maser that was not detected by Masheder et al.

(1994), and which in LCP has a flux density of 2.76 Jy/beam, and so clearly would

have been detected if present in 1989. This is therefore a new maser, which has

grown from undetectable in the 7 years between the two epochs. It is extremely

unlikely that this is the northeast maser in the close northern group of Masheder et al.

(which could conceivably have moved to this location) because the new maser is at a

velocity of –43.47 km s–1 – over 1.5 km s–1 from the old northern group. Figure 4 in

Masheder et al. (1994) shows that the 1720 MHz masers have been evolving slowly,

in particular a feature has been intensifying at a velocity of about –45.5 km s–1 (LCP)

over the last 3 decades. This velocity corresponds to the currently most intense maser,

feature 12.


179

The detection of the western source adds evidence to the alignment of 1720

MHz emission and 4765 MHz emission as suggested by Baudry et al. (1988), Baudry

and Diamond (1991), Gray et al. (1992), Masheder et al. (1994). More recent

comparisons between 1720 MHz and 4765 MHz clarify the association between the

two frequencies (Macleod 1997, Gray et al. 2001).

W3(OH) has also been observed at 1720 MHz with MERLIN in December

1993 (Gray et al. 2001), in an attempt to detect co-propagation (see chapter 6.4.1)

with the 4765 MHz line. With the much lower resolution of MERLIN, the results

agree very well in terms of maser positions and velocities with the current data. The

two close southern masers were not resolved in the MERLIN data, but the extra

northern maser not detected in Masheder et al. (1994) was detected. Fluxes as

determined by Gaussian fitting were typically 2 to 3 times higher in the MERLIN

data, suggesting that the 1720 MHz masers may be surrounded by slightly diffuse

emission that is resolved out on longer baselines. The exception here is the northern

maser that wasn’t detected by Masheder et al. (1994), which has approximately the

same flux in both more recent observations. There are two possible explanations to

this:

1. The maser does indeed have extended emission associated with it, but has been

rapidly intensifying such that the core emission in the current data is roughly

equal to the total emission in the earlier MERLIN data.

Figure 6.3-4: Movement of the Gaussian fit positions of the component channels of feature 12.

Error bars show formal position errors, calculated as per chapter 4.4.1. Velocities are labelled next

to the relevant spot. The circled spot is the peak channel.

1720 MHz Feature 12

23

24

25

26

27

28

29

30

31

32

33

34

35

-10-9-8-7-6-5-4-3-2-1012345

RA Offset / mas

DE

C O

ffset

/ m

as

-45.09

-46.36

-46.28

-46.19

-46.11-46.02

-45.94-45.85

-45.77-45.68

-45.60-45.51

-45.43-45.34

-45.26

-45.17


180

2. The maser has no extended emission – which may be related to its very recent

appearance.


The three groups of masers are distinct morphologically. The northern masers are the

smallest and most circular, having an average major axis of 4.6 mas. The southern

masers are slightly larger, having an average major axis of 7.6 mas; while the western

masers are much more elongated and well resolved, having an average major axis of

22.7 mas and a north-south position angle. It should be noted though that both

northern and southern masers were unresolved in these observations, and so circular

morphology would be expected; in the higher resolution observations of Masheder et

al. they were also unresolved, with a maximum size of 1.2 mas. The extended nature

of the western masers suggests that they may be more diffuse than the others, which

would explain why the flux measured by MERLIN is so much higher. It is likely that

the 1720 MHz masers consist of a bright core containing ~50% of the flux in <1.5

mas, and an extended halo (>10 mas) containing the remaining 50% of flux.


Figure 6.3-5 shows all the lineshapes that were successfully fit with Gaussians from

the 1720 MHz data. One of the lineshapes that failed fitting is shown at the bottom of

figure 6.3-5: feature 5. Of the three maser lineshapes not shown, all failed fitting

because they had less than four channels in them; these lineshapes also have fluxes

less than 120 mJy. The shoulder on the emission is one of the underlying unresolved

spots as described in chapter 6.3.2. The combination of systematic movement of the

Gaussian-fit centre of emission of a maser with ‘shoulders’ is indicative of unresolved

pairs or multiples of masers. 1720 MHz masers are in general well approximated by

Gaussian lineshapes; no evidence was found for any true deviations from the

Gaussian.


181


0.00

1.00

2.00

3.00

4.00

5.00

6.00

-45.4 -45.2 -45.0 -44.8 -44.6 -44.4 -44.2 -44.0 -43.8

Velocity / (km/s)

Flu

x D

ensi

ty /

(Jy

/bea

m)

Figure 6.2-5: Lineshapes of 1720 MHz masers. Observed flux errors are negligible at these scales.

Clockwise from top left: Features 1, 2, 3, 4, 12, 21. Feature 5 is at the bottom.


182

6.3.5 Polarization

This is the first VLBI dataset of 1720 MHz in W3(OH) that contains all polarization

data. From table 6.3-1 it is clear that all of the masers are highly circularly polarised,

no maser has less than 90% total or circular polarization. For the brightest masers, the

circular polarization is very close to 100%, with negligible linear and unpolarised

emission. Only the western masers, which are both very weak and therefore subject to

larger polarization errors (see chapter 4.4.1) have any significant linear polarization.

This means that the 1720 MHz masers show the least deviation from total circular

polarization of any line in the ground state.


From table 6.3-1, 5 Zeeman pairs can be recognised – representing 100% of the

masers detected. From the single-dish spectra, it was clear that the 1720 MHz

emission exhibits a relatively clear Zeeman pattern, and this is reflected in the above

percentage. The Zeeman pairs are shown in table 6.3-2. Magnetic field

measurements are in good agreement with Gray et al. (2001).


1.8:1. This represents the most even Zeeman pairs of the ground-state lines – the

conditions in which 1720 MHz masers form are clearly also favourable for near

perfect Zeeman patterns. Zeeman pair 3, consisting of features 2 and 4, is the most

perfect Zeeman pair detected in the whole dataset – the peaks are within 1% of each

other in peak flux. Its Stokes lineshape is shown in figure 6.3-6

As with 1612 MHz, there are potentially a number of possibilities for the

magnetic field strength to produce a certain splitting since three separate Zeeman pair




7.13 -2.4 27.5 7.47 0.51 -45.17 0.03 a North

5.08 25.8 50.6 6.80 0.41 -43.10 0.02 a North

2.26 -147.0 -1122.1 5.70 0.46 -43.44 0.03 a South

0.40 -156.0 -1113.7 6.02 0.51 -43.72 0.03 b South

0.12 -853.0 -600.2 4.51 0.51 -44.91 0.03 b East

Table 6.3-2: 1720 MHz Zeeman pairs (assuming smallest magnetic splitting). Information for peak

flux, RA and DEC are from the peak channel in the brighter of the two Zeeman components.


183

separations can theoretically result (see table 2.2). This is never seen in 1720 MHz –

only one pair of lines is ever observed. Therefore, as with 1612 MHz, the question is

which line is being observed. The field values in table 6.3-2 assume that the observed

emission is from the strongest Zeeman lines – the other possibilities would produce

field strengths of one-third and one-fifth of the magnitude. Even the highest field

strength would have a magnitude of just 2.49 mG from the next Zeeman splitting up –

a field that is considerably lower than any than any other masers except the single

exceptional 1667 MHz maser. It is therefore most likely that, as with 1612 MHz, all

the masers observed are of the intrinsically strongest (splitting 1/4) Zeeman lines.

Using this the range of the magnetic field strengths is ~4.5 to ~7.5 mG, which is in

good agreement with the masers at the other ground-state frequencies, and with Lo et

al. (1975) who measured a field of 6 mG at 1720 MHz.

Figure 6.3-7 shows the distribution of magnetic field strengths and

demagnetised velocities with RA and DEC. The demagnetised velocities of the

southern and western groups suggest that they could be part of the bulk rotational

motion mapped out in 1665 MHz and 1667 MHz. The northern masers though,

despite having similar field strengths and being so close to each other are widely

separated in velocity. The northern masers may be in the same environment as the

1665 MHz group c masers. This will be discussed further in chapter 6.4.3

Figure 6.3-6: Zeeman pair 3 – the most perfect Zeeman pair in W3(OH).

Stokes Parameters for 1720 Zeeman Pair 3

-2.5

-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

2.0

2.5

-44.2 -44.0 -43.8 -43.6 -43.4 -43.2 -43.0 -42.8 -42.6

Velocity / (km/s)

Flu

x D

ensi

ty /

(Jy

/bea

m)

I

QU

V


184

6.3.7 Proper Motion

In the same way as for 1665 MHz, it was possible to measure the motion of the 1720

MHz masers relative to a ‘reference’ maser from earlier VLBI observations. Since

only two masers could be identified from the observations of Masheder et al. (1994),

only the motion of one maser relative to the other would be measured. The result is

shown in figure 6.3-8.

The results are dramatic – the southern maser has moved east by 12 mas in 6.8

years, corresponding to a velocity of ~18 km s–1. This velocity is much higher than

any measured in 1665 MHz, and is 3 times the magnitude of the measured expansion

rate of the UCH II region. This means that the maser could be in fast moving material,

perhaps an outflow associated with a star in the south of the region, as proposed in

chapter 5.8. Possible sources of error in the measurement come from the slight

ambiguity in the true location of the northern maser, which was unresolved in the

most recent observation, as shown in figure 6.3-4. However, the position of the

northern maser in the recent observation is very unlikely to be more than 1.5 mas in


0

1

2

3

4

5

6

7

8

9

-1200-1000-800-600-400-2000200

DEC Offset / mas

Fie

ld S

tren

gth

/ m

G


0

1

2

3

4

5

6

7

8

9

-900-800-700-600-500-400-300-200-1000100

RA Offset / mas

Fie

ld S

tren

gth

/ m

G


-45.5

-45.0

-44.5

-44.0

-43.5

-43.0

-42.5

-1200-1000-800-600-400-2000200

DEC Offset / mas

Dem

agne

tised

Vel

ocity

/ (

km/s

)


-45.5

-45.0

-44.5

-44.0

-43.5

-43.0

-42.5

-900-800-700-600-500-400-300-200-1000100

RA Offset / mas

Dem

agne

tised

Vel

ocity

/ (

km/s

)


1720 MHz. The points are plotted as error bars; errors were calculated as per chapter 4.5.4


185

error in RA, and this has little effect on the magnitude of the velocity. It must be

remembered, as for 1665 MHz that the motion of the reference feature is unknown

and must be added onto both masers to find the true proper motion. If the northern

maser was in the same body of gas as 1665 MHz maser group c – which is proposed

to be moving east – then an additional eastward motion would be added to the

southern maser, probably taking its eastward velocity to over 20 km s–1. This velocity

is typical of an outflow and could indicate outflows in this area, possibly from a

southern star.


Even with the small number of masers at 1720 MHz, a relationship is apparent linking

the size (or area) of the masers with the magnetic field. This is shown in figure 6.3-9.

This is an effect of magnetic beaming.

The mean FWHM of the 1720 MHz masers which were fit with a Gaussian is

~299 m s–1, which is in fair agreement with the figure of ~280 for 1665 MHz. It

should be noted that the FWHM of the most intense maser (feature 12) is 509 m s–1,

which is one of the highest FWHM of any maser fit in the ground state. This is

possibly because of the unresolved multiple nature of this maser, as shown in chapter

6.3.2. Without this maser the mean FWHM is ~257 m s–1.

Figure 6.3-8: Relative proper motion of the southern maser to the northern maser at 1720 MHz.


186


The small number of masers at 1720 MHz makes for a poor graph of the two-point

correlation (see figure 6.3-10). The plot still shows the small scale clustering of the

masers. The clustering scale of <150 mas perhaps indicates that 1720 MHz maser

clusters are smaller than at the other ground frequencies; but the very small sample

means that this is only a weak proposition.

The largest scale structure at 1720 MHz is at around 1000 mas, and this is the

lowest value of maximum extent of all the ground state frequencies.

6.4 Comparisons between the Ground State Lines

6.4.1 Co-Propagating Masers

The complete mapping with phase referencing of the 4 ground state lines in W3(OH)

allows for the first time searches for co-propagating masers from different lines at

VLBI resolutions. As explained in chapter 4.3.6, the observation allows for the

alignment of images at different ground state frequencies with an accuracy of ~1 mas.

This is shown in figure 6.4-1.

From figure 6.4-1, it is apparent that many of the masers at 1612, 1667 and

1720 MHz lie near or amongst 1665 MHz masers. Only two 1612 MHz, three 1667

MHz masers and no 1720 MHz masers lie 100 mas or more from a 1665 MHz maser.

Maser size vs Field Strength

0

1

2

3

4

5

6

7

8

0 50 100 150 200 250 300 350 400 450 500

area / mas^2

Fie

ld S

tren

gth

/ mG



187

No other masers lie more than 200 mas from a 1665 MHz maser. This information

will be of importance in observations at lower resolutions looking for coincident

masers at other frequencies or other sources. Immediately striking is the number of

close coincident masers of 1665 and 1667 MHz in the southwest at the location of the

arc feature, which will be examined in detail. Also notable is the lack of masers from

any frequency other than 1665 MHz in the northwest and far northeast of W3(OH)

(1665 MHz maser groups a and b).

The much larger number of 1665 MHz masers means that even in unrelated

random distributions we would expect to see the most spatial coincidences between

Figure 6.3-10 a) and b): a) Top, the 1720 MHz masers and a random selection of points in the

same space. b) Bottom, the two-point correlation of the 1720 MHz masers.


188

All Maser Spots

-2300

-2100

-1900

-1700

-1500

-1300

-1100

-900

-700

-500

-300

-100

100

300

500

-1500-1300-1100-900-700-500-300-100100300500700900

RA Offset / mas

DE

C O

ffset

/ m

as

1612 MHz

1665 MHz

1667 MHz

1720 MHz

3

7

9

8

5 6

4

10

1

2

Figure 6.4-1: The ground state masers in W3(OH). Regions of distinct velocity profile are bounded

by the thin lines. Dashed lines indicate strings of masers that may outline shocks.


189

1665 MHz masers and those of other frequencies, and this is indeed seen. The

criterion for masers to be considered coincident is important here. Given the accuracy

of alignment of the maps, and the criterion applied when considering coincident

masers for Zeeman pair candidature (12 mas), it was decided that the same criterion

would be applied to the different frequencies, with separations of 0 to 6 mas having

the highest confidence. In addition, coincident masers with separations 12 to 24 mas

would be examined with caution. Maser coincidences and an assessment of the

likelihood of co-propagation are shown in table 6.4-1. The masers were checked for

agreement within error of their demagnetised velocity and magnetic field strengths.

Where there was only one maser at one frequency and a Zeeman pair at the other

frequency, the velocity of the maser was used under the assumption that its

demagnetised velocity was equal to that of the Zeeman pair at the other frequency.

From this, the magnetic field that would be required to give the correct Zeeman

splitting was calculated, and compared to that of the Zeeman pair. Where there was

only one maser at either frequency, a magnetic field can always be found which will

give the required separation – the test then becomes whether this magnetic field is in

agreement with other Zeeman pairs in the area. Clearly, this test is less rigorous than

direct comparison of demagnetised velocities and field strengths. The overlaps are

categorised into classes A, B, C and D, where: A is probable co-propagation, B is

possible association, C is indeterminate, and D is probably not co-propagation.

There are no 3-frequency or 4-frequency overlaps, not even up to 25 mas

separation (~50 AU). All overlaps are with one ground state frequency and 1665

MHz. This will be a useful constraint to future modelling of SFR masers.

There are no likely instances of co-propagation between 1720 MHz and 1665

MHz above the detection threshold of 75 mJy, so previous map alignments of strong

sources at these frequencies by authors are thus invalid. However, all 1720 MHz

emission comes from regions of significant and numerous 1665 MHz maser activity,

and it is likely that the activity at both frequencies has a common cause.

There is just one class A overlap between 1612 MHz and 1665 MHz, labelled

3 on figure 6.4-1. The calculated field is ~0.5 mG out, but this is within the error of

the field at 1612 MHz. Also of interest are the class B overlaps labelled 1, 2 and 4.

From figure 6.4.1 it can be seen that 1612 MHz masers lie ~20 mas from lines or

‘chains’ of 1665 MHz masers of the same demagnetised velocity and magnetic field

19

0

Tab

le 6

.4-1

: M

aser

ove

rlap

s am

ong

the

grou

nd s

tate

lin

es.

Ove

rlap

s be

twee

n fr

eque

ncie

s ar

e cl

asse

d A

to

D o

n ag

reem

ent

in d

emag

neti

sed

velo

city

and

mag

neti

c fi

eld

stre

ngth

bet

wee

n th

e fr

eque

ncie

s. N

ote

the

feat

ure

num

bers

ref

er to

eac

h in

divi

dual

fre

quen

cy, s

o it

is p

ossi

ble

to h

ave

the

sam

e fe

atur

e nu

mbe

r in

dif

fere

nt f

requ

enci

es.

Tra

nsiti

on

Fea

ture

Int

egra

ted

Pea

k F

lux

RA

D

EC

V

LSR

C

ircul

ar

Zee

man

M

agne

tic

Mag

netic

D

emag

netis

edD

emag

netis

edO

verla

p C

alcu

late

dO

verla

p

Fre

quen

cy

num

ber

Fea

ture

D

ensi

ty

Offs

et

Offs

et

d

irect

ion

P

air

Fie

ld

Fie

ld e

rror

V

eloc

ity

Vel

ocity

err

or S

epar

atio

nfie

ld

Cla

ss

(MH

z)

F

lux

(Jy

) (J

y/B

eam

) (m

as)

(mas

) (k

m/s

)

Num

ber

(mG

) ±(

mG

) (k

m/s

) ±(

km/s

) (m

as)

(mG

)

1612

25

11

.54

3.53

-5

0.3

-214

.2

-43.

46

L 1

7.82

0.

39

-42.

97

0.02

1612

4

42.1

2 11

.34

-50.

3 -2

14.2

-4

2.55

R

1

7.82

0.

39

-42.

97

0.02

21

.0

B

1665

4

7.88

1.

78

-48.

9 -1

93.2

-4

0.69

R

22

7.

58

0.09

-4

2.93

0.

03

1665

11

1 12

1.56

27

.03

-48.

6 -1

93.1

-4

5.18

L

22

7.58

0.

09

-42.

93

0.03

1612

21

2.

66

0.98

-3

8.7

-213

.5

-42.

73

R

2 7.

63

0.51

-4

3.19

0.

03

1612

28

2.

91

0.99

-3

8.7

-213

.4

-43.

64

L 2

7.63

0.

51

-43.

19

0.03

21

.0

B

1665

61

8 0.

73

0.33

-3

8.2

-192

.4

-45.

44

L -

- -

- -

7.

61

1612

30

0.

42

0.19

9.

7 -1

13.7

-4

1.73

R

6

11.0

6 0.

53

-42.

39

0.03

1612

24

2.

60

0.78

9.

3 -1

13.2

-4

3.00

L

6 11

.06

0.53

-4

2.39

0.

03

4.6

A

1665

37

0 7.

70

2.09

8.

2 -1

08.8

-4

5.53

L

- -

- -

-

10.6

5

1612

26

2.

02

0.36

17

5.0

-181

8.0

-43.

18

L 4

8.71

0.

54

-42.

67

0.03

1612

3

13.8

7 1.

97

172.

6 -1

817.

7 -4

2.09

R

4

8.71

0.

54

-42.

67

0.03

21

.9

D

1665

66

1.

23

0.30

17

4.2

-179

5.9

-43.

77

L -

- -

- -

3.

74

1612

23

5.

28

0.51

18

4.0

-181

7.2

-43.

09

L 5

4.81

0.

54

-42.

84

0.03

1612

5

8.67

1.

15

195.

6 -1

816.

4 -4

2.64

R

5

4.81

0.

54

-42.

84

0.03

22

.6

B

1665

26

2.

27

0.73

19

8.1

-179

3.9

-41.

57

R

52

4.78

0.

09

-43.

00

0.03

1665

75

20

.35

4.66

19

8.5

-179

3.5

-44.

38

L 52

4.

78

0.09

-4

3.00

0.

03

1665

15

6 1.

09

0.25

-6

90.5

-1

872.

4 -4

4.82

L

54

6.29

0.

10

-42.

98

0.03

1665

9

64.0

6 10

.08

-691

.3

-187

2.7

-41.

13

R

54

6.29

0.

10

-42.

98

0.03

10

.2

C

1667

6

9.23

1.

98

-694

.5

-188

2.4

-42.

28

R

3 6.

58

0.17

-4

3.42

0.

03

1667

25

6.

50

1.18

-6

95.1

-1

882.

6 -4

4.65

L

3 6.

58

0.17

-4

3.42

0.

03

19

1

1665

37

4 49

.62

10.9

4 -5

91.4

-1

916.

9 -4

5.53

L

55

6.69

0.

08

-43.

52

0.02

1665

18

32

.83

7.12

-5

91.6

-1

916.

9 -4

1.57

R

55

6.

69

0.08

-4

3.52

0.

02

1667

39

13

.14

3.69

-5

92.8

-1

917.

7 -4

4.82

L

12

6.57

0.

14

-43.

63

0.02

1.

2

A

1667

8

5.93

1.

37

-594

.0

-191

7.7

-42.

45

R

12

6.57

0.

14

-43.

63

0.02

1665

62

8 0.

67

0.29

-6

21.9

-1

917.

9 -4

5.44

L

56

6.87

0.

11

-43.

41

0.03

1665

20

1.

87

0.49

-6

22.0

-1

918.

0 -4

1.40

R

56

6.

87

0.11

-4

3.41

0.

03

1.0

A

1667

26

3.

71

0.74

-6

22.4

-1

918.

9 -4

4.65

L

10

6.77

0.

16

-43.

43

0.03

1667

4

1.77

0.

34

-623

.4

-191

9.4

-42.

28

R

10

6.77

0.

16

-43.

43

0.03

1665

21

0.

91

0.28

-6

54.8

-1

903.

9 -4

1.31

R

-

- -

- -

2.5

7.60

C

1667

2

4.59

1.

55

-656

.8

-190

2.4

-42.

10

R

5 6.

59

0.17

-4

3.25

0.

03

1667

24

3.

99

1.35

-6

56.8

-1

902.

5 -4

4.39

L

5 6.

59

0.17

-4

3.25

0.

03

1665

63

1 1.

57

0.29

-5

26.5

-1

919.

0 -4

1.22

R

57

6.

90

0.11

-4

3.30

0.

03

1665

36

9 2.

56

0.56

-5

25.0

-1

916.

3 -4

5.35

L

57

6.90

0.

11

-43.

30

0.03

5.

5 5.

74

C

1667

51

0.

99

0.33

-5

22.1

-1

911.

6 -4

2.28

R

-

- -

- -

1665

61

1 3.

42

1.16

-6

01.5

-1

915.

0 -4

5.53

L

- -

- -

- 2.

4 6.

30

A

1667

32

11

.55

2.02

-6

03.9

-1

914.

9 -4

4.82

L

11

6.36

0.

17

-43.

67

0.03

1667

54

1.

35

0.38

-6

04.3

-1

914.

9 -4

2.54

R

11

6.

36

0.17

-4

3.67

0.

03

1665

61

3 0.

59

0.22

-7

30.7

-1

887.

2 -4

4.56

L

- -

- -

- 7.

4 7.

44

B

1667

12

0.

75

0.16

-7

35.3

-1

893.

0 -4

3.68

L

- -

- -

-

1665

94

14

.67

2.58

-4

18.3

-1

815.

5 -4

4.47

L

- -

- -

- 9.

4 ~0

D

1667

15

2.

15

0.59

-4

25.1

-1

809.

0 -4

4.03

R

-

- -

- -

1665

10

4 2.

65

0.78

-3

78.4

-1

861.

6 -4

4.74

L

- -

- -

- 19

.9

4.46

B

1667

13

87

.32

9.29

-3

86.7

-1

879.

7 -4

4.21

L

- -

- -

-

1665

19

8 1.

46

0.74

-5

6.0

-142

6.5

-44.

91

L -

- -

- -

7.9

B

1667

56

0.

48

0.09

-6

0.9

-142

0.4

-44.

30

L 13

4.

22

0.18

-4

3.55

0.

03

4.

61

1667

9

1.23

0.

26

-58.

6 -1

418.

1 -4

2.81

R

13

4.

22

0.18

-4

3.55

0.

03

19

2

1665

60

7 3.

06

0.71

-8

61.3

-6

42.1

-4

6.05

L

30

3.19

0.

09

-45.

11

0.03

1665

60

4 71

.20

6.40

-8

59.1

-6

52.2

-4

4.21

R

30

3.

19

0.09

-4

5.11

0.

03

3.4

C

1667

33

1.

54

0.18

-8

62.5

-6

51.7

-4

5.53

L

1 2.

43

0.18

-4

5.10

0.

03

1667

30

3.

29

0.41

-8

63.8

-6

52.5

-4

4.65

R

1

2.43

0.

18

-45.

10

0.03

1665

60

6 2.

95

0.60

-8

38.0

-6

70.1

-4

3.95

R

-

- -

- -

22.7

5.

93

D

1667

41

1.

25

0.37

-8

60.6

-6

67.5

-4

4.65

R

-

- -

- -

1665

24

3.

02

0.91

26

.8

47.6

-4

1.48

R

-

- -

- -

2.9

5.48

C

1720

21

10

.49

2.76

25

.7

50.3

-4

3.47

L

2 6.

57

0.41

-4

3.10

0.

02

1720

1

19.7

3 5.

08

25.8

50

.6

-42.

70

R

2 6.

57

0.41

-4

3.10

0.

02

1665

38

6 1.

47

0.34

-1

64.5

-1

128.

5 -4

5.79

L

34

4.89

0.

11

-44.

33

0.03

1665

50

12

.93

2.46

-1

62.7

-1

126.

0 -4

2.89

R

34

4.

89

0.11

-4

4.33

0.

03

1720

3

2.17

0.

40

-156

.0

-111

3.7

-43.

38

R

5 5.

82

0.51

-4

3.72

0.

03

14.0

D

1720

22

0.

31

0.11

-1

56.6

-1

112.

3 -4

4.06

L

5 6.

02

0.53

-4

3.72

0.

03

1720

4

11.4

7 2.

24

-147

.0

-112

2.2

-43.

72

L 3

5.51

0.

46

-43.

44

0.03

16

.2

D

1720

2

11.0

1 2.

26

-147

.0

-112

2.1

-43.

13

R

3 5.

70

0.47

-4

3.44

0.

03

1665

49

7 4.

80

0.87

-4

.6

20.9

-4

8.52

L

9 7.

37

0.11

-4

6.29

0.

03

1665

64

88

.59

15.5

4 -4

.5

21.0

-4

4.03

R

9

7.37

0.

11

-46.

29

0.03

6.

9

D

1720

12

60

.67

7.13

-2

.4

27.5

-4

5.60

L

1 7.

21

0.51

-4

5.17

0.

03

1720

5

54.5

7 5.

08

-2.4

27

.7

-44.

74

R

1 7.

21

0.51

-4

5.17

0.

03

1665

72

0.

93

0.15

-8

58.6

-5

82.0

-4

3.86

R

-

- -

- -

19.6

3.

58

D

1665

38

1 13

.98

2.74

-8

50.6

-6

16.8

-4

5.70

L

28

3.14

0.

10

-44.

80

0.03

1665

65

11

.05

2.05

-8

50.5

-6

19.5

-4

3.86

R

28

3.

14

0.10

-4

4.80

0.

03

18.8

D

1720

23

0.

73

0.12

-8

53.3

-6

00.9

-4

4.66

R

4

4.36

0.

51

-44.

91

0.03

1720

11

1.

01

0.12

-8

53.0

-6

00.2

-4

5.17

L

4 4.

36

0.51

-4

4.91

0.

03


193

strength. The 1612 MHz masers could be in the same body of gas as the 1665 MHz

masers, but separated by ~40 AU. If the chains of 1665 MHz masers delineate shocks

in the gas (see chapter 5.3), then the 1612 MHz masers are likely physically just

before or just after the 1665 MHz in the shocks. In the north, in 1665 MHz maser

region c, the direction of any shock is unclear, but in the south, in maser group f

(overlap 4), any shock would most likely be propagating to the south. In this case, the

1612 MHz masers lie physically ahead of the 1665 MHz masers in the shock; at an

earlier stage in the shocking process. Applying this to the northern overlaps this

suggests that the shocks are also propagating south from the centre of maser group c.

The remaining six class A and B overlaps are all between 1665 MHz and 1667

MHz. Statistically the most overlaps might be expected between these two

frequencies since they are the most numerous maser sources. The two class B

overlaps 8 and 9 are classed as such because each consists of a single Zeeman

component from each frequency. This means that a single velocity and magnetic field

can always be calculated for the two masers. The overlaps may actually be class A,

but there is not enough evidence to support this. There is however no evidence

against them being co-propagating masers. Overlap 10 has a calculated magnetic field

value outside the error range, and is therefore classed as B. From figure 6.4-1, it is

clear that the arc feature in the south of W3(OH) is the source of several good

overlaps. Figure 6.4-2 shows contour maps of the arc feature in both 1667 MHz and

1665 MHz, where it is also visible.

While all the 1665 MHz and 1667 MHz masers in the arc feature have

magnetic field and velocity values that are close in range – and are hence almost

certainly in the body of gas – only those in the centre of the arc (between RA offsets

of –580 and –630) have the same magnetic field and velocity values and can therefore

be justified as co-propagating. The excellent positional agreement between the co-

propagating masers, and indeed the positional agreement in each Zeeman pair, is

testimony to the high accuracy of alignment between the maser maps. There are also

three class C overlaps in the arc features, and it is possible that the difficult nature of

the Gaussian fitting in this complex area has resulted in the ‘loss’ of some co-

propagating masers. From figure 6.4-2 it appears that there could be co-propagation

at ~–690 mas where a strong 1665 MHz maser has swamped weaker maser emission a

few mas to the west.


194

The fluxes of all class A and B overlaps are shown in table 6.4-2. Also shown

for the Class A overlap with 1612 MHz and all the overlaps for 1665 MHz is an

overall ratio of emission at each frequency, calculated by totalling the Stokes I

emission in the case of Zeeman pairs. Emission ratios such as this are extremely

useful for modelling of masers, and will provide constraints for the modelling of

masers in SFRs. In three of the four class A overlaps, the same hand of polarization is

strongest at each frequency. Having the same hand strongest would be expected

because both frequencies are experiencing the same velocity overlap – so we would

expect all of the masers to exhibit this. One possible reason why one overlap doesn’t

is that it isn’t a true overlap, but rather one which by coincidence occupies the same

projected location, magnetic field strength and velocity, but not the same physical

position – an unlikely combination, but still possible. In this case the two coincident

masers could be experiencing velocity shifts in different directions.

The total flux from all of the overlapping 1667 MHz masers is 127 Jy, while

the total flux from the 1665 MHz masers they overlap with is 93 Jy. This slightly

higher value for 1667 MHz is in stark contrast with the values for the W3(OH) area as

a whole, where the total flux from 1665 MHz is about 15 times larger. This indicates

Figure 6.4-2: The southern arc feature in both 1665 MHz and 1667 MHz emission. The contours

here indicate maximum channel intensity, not integrated intensity. Contours are at 100 mJy/beam.

Mill

iAR

C S

EC

MilliARC SEC-500 -550 -600 -650 -700 -750

-1840

-1860

-1880

-1900

-1920

-1940

-1960

Mill

iAR

C S

EC

MilliARC SEC-500 -550 -600 -650 -700 -750

-1840

-1860

-1880

-1900

-1920

-1940

-1960


195

that there is no justification in taking the ratios of emission at different frequencies

from any large-scale region to be indicative of the true ratios resulting from co-

propagation. This may be of relevance to megamasers, where there is usually a ~3:1

excess of 1667 MHz emission, but this may not represent the true co-propagation

ratios if megamasers are actually unresolved smaller masers. Conversely, if

megamasers are truly co-propagating, then this result has another significance in that

it too shows an excess of 1667 MHz emission. In the light of these results concerning

the strength of masers co-propagating at 1665 and 1667 MHz, future observers

looking for co-propagation should consider looking to 1667 MHz dominant maser

regions as a priority. The conditions in the arc may therefore be related to the grand-

scale conditions which drive megamasers; quite possibly the passage of shocks of the

same nature as the arc. It is also worth noting at this point that the work of Thissen et

al. (1999) predicts 1667 MHz dominant masing to arise from photodissociation of

OH. Criticisms of their chemical-pump model have centred on the lack of a

replenishable source of OH, since each molecule can mase only once after

Ratio Overlap Polarisation Overlap Polarisation Ratio

1 (B) Left (Jy) Right (Jy) 5 (A) Left (Jy) Right (Jy)

1612 3.529 11.339 1665 10.939 7.122 3.6

1665 27.027 1.779 1667 3.694 1.37 1

2 (B) Left Right 6 (A) Left Right

1612 0.987 0.98 1665 0.286 0.487 1

1665 0.333 <0.075 1667 0.744 0.344 1.4

3 (A) Left Right 7 (A) Left Right

1 1612 0.775 0.189 1665 1.158 <0.075 1

2.2 1665 2.094 <0.075 1667 2.017 0.375 2.1

4 (B) Left Right 8 (B) Left Right

1612 0.509 1.146 1665 0.219 <0.075 1.3

1665 4.656 0.734 1667 0.164 <0.075 1

9 (B) Left Right

1665 0.784 <0.075 1

1667 9.294 <0.075 11.9

10 (B) Left Right

1665 0.736 <0.075 2.1

1667 0.093 0.26 1

Table 6.4-2: The fluxes of all class A and B overlaps. Flux ratios are in the outer columns.


196

dissociation. If the arc is indicative of a shock, then this is just the environment where

locally produced UV radiation could produce large amounts of freshly dissociated

OH, and result in a transient burst of masing. As such, the arc may represent an

environment where the dominant maser pump is chemical.

What is surprising is the very small number of co-propagation cases that arise.

Using just the class A overlap for 1612 MHz and the A and B overlaps for 1667 MHz

shows co-propagation accounts for just 7% of 1612 MHz masers, 3% of 1665 MHz

masers, 15% of 1667 MHz masers and ~0% of 1720 MHz masers. This may be an

indication of the effectiveness of competitive gain in maser clouds.

6.4.2 Polarization

All of the ground state lines show the occurrence of Zeeman pairs. The occurrence

rates of Zeeman pairs should constrain the development of theories that explain why

Zeeman components are often suppressed. A comparison of the Zeeman properties of

the four lines are shown in table 6.4-3.

1665 MHz has the largest sample size and the largest deviation from equality

in the number of masers of each hand of polarization. This makes a significance test

valid. If LCP masers are assumed equally likely as RCP masers then the significance

of this deviation (continuity corrected) is 2.24 sigma. The ‘two-tailed’ probability

associated with values as or more extreme than this in the normal distribution is just

over 0.02. This deficit of RCP is unlikely therefore to be random, and is probably

rooted in the conditions of the masing: the most likely candidate is the velocity shift

present, and therefore this surplus of one hand could indicate a prevailing velocity

shift over the region.

Frequency

(MHz)

Number

of

Masers

Integrated

Flux

(Jy)

Zeeman

Pair

(%)

Median

Zeeman

Ratio

Mean

Zeeman

Ratio

RCP

(%)

LCP

(%)

1612 15 107 80 3.7 3.5 53 47

1665 211 4250 55 4.4 6.9* 42 58

1667 40 276 68 1.7 2.1 46 54

1720 10 172 100 1.4 1.8 50 50

Table 6.4-3: Polarisation and Zeeman pair properties of the ground state OH maser lines. The

‘Zeeman ratio’ is the ratio of the intensity of the stronger peak to the weaker peak. The asterisked

value has had 2 outliers dropped from the sample (see chapter 5.6).


197

Figure 6.4-3 shows the variation of ‘Zeeman probability’ (percentage of

masers in a Zeeman pair) at each frequency, with the Zeeman component separation at

that frequency. The component separation is determined by the Landé factor for the

transition (see chapter 2.2). As suggested in chapter 5.6, this appears to have major

implications for theories that explain how Zeeman pairs are destroyed, such as that of

Nedoluha and Watson (1990b). Nedoluha and Watson suggested that single Zeeman

components would tend to be formed in areas of low magnetic field, but Figure 6.4-3

suggests the opposite. A note of caution here: it is not possible to calculate the

magnetic field of single Zeeman components, so it is not possible to know whether

these singlets sample a particular range of magnetic fields. Therefore there is no way

of knowing whether masers at e.g. 1667 MHz would show the same Zeeman

probability as those at 1665 MHz if they were in a magnetic field that was on average

66% stronger (which would give both frequencies the same average Zeeman splitting).

Clearly, this raises difficult questions of the theory of Nedoluha and Watson, and may

require that the earlier theory of Cook (1966) be reinvestigated. Cook noted that a

mechanism that introduces some asymmetry into the two σ-transitions (and thereby

favours one component over the other) could be obtained by appropriately matching

the gradients of the magnetic field and velocity within the masing region. This does

not appear to be contradicted by figure 6.4-3.

The Zeeman ratio also follows the general trend that the ratio is higher for

larger Landé factors, but not with such clarity as shown in figure 6.4-3. It is likely

that the small sample size of 1612 MHz and 1720 MHz masers affects this. The

Zeeman ratio is likely to be an indicator of the effectiveness of Zeeman overlap in the

Figure 6.4-3: Variation of Zeeman probability with Zeeman component separation at the four

ground-state frequencies

Zeeman Probability vs Component Separation

0

10

20

30

40

5060

70

80

90

100

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

Component Separation / (km/s/mG)

Zeem

an P

roba

bilit

y /

%

1720

16651667

1612


198

same way as the Zeeman probability. It is certain that larger Zeeman component

separation leads to fewer Zeeman pairs. This is unfortunate because 1665 MHz is the

most common maser and has the largest component separation, and so gives relatively

the least magnetic field information. Observations at 6 GHz and 13 GHz can give

much more favourable results where magnetic field strengths or demagnetised

velocities are concerned.

6.4.3 Velocity Structure and Magnetic Field

The demagnetised velocity and magnetic field strengths of all the ground state lines

are shown in figure 6.4-4. All of the Zeeman pairs lie in the velocity range covered by

1665 MHz masers. All of the northern 1612 MHz masers appear to be in the same

body of gas as 1665 MHz maser group c (group c is highlighted by its large range of

velocities and high magnetic field strengths); as do at least one and possibly both of

the northern 1720 MHz masers. The rest of the 1612 MHz, 1720 MHz and all of the

1667 MHz masers lie within the velocity gradient delineated by the 1665 MHz

masers, and therefore they probably share the motion of this gas – argued in chapter

5.7 to be rotational motion with a possible expansion component. This picture is

supported by the magnetic field values also. In both 1667 MHz and 1720 MHz there

are low field strengths in the area of 1665 maser group d. In the south, two of the

1612 MHz masers have field strengths slightly above those of their 1665 MHz

neighbours.

Figure 6.4-5 shows the raw velocity data for all four lines, in order to examine

the velocity of those masers not in Zeeman pairs. Most notable are the small group of

masers that are very blue-shifted at the south of the map, between offsets of –1500 and

–2000 mas. The masers in this group are in the extreme southeast and extreme south

of the map, except the leftmost of this group (arrowed in figure 6.4-5), which is nearer

the centre of the map and is not associated with the rest. The remaining 7 masers are

all therefore on the fringes of the W3(OH) region, and do not share the velocity profile

of either maser group c or the large scale rotation. In the 1665 MHz masers, there is

evidence that the field may have swapped direction in this outlying area and so be

pointing towards the observer (see chapter 5.5). This could indicate that these masers

lie just on the far side of W3(OH), since this far from the centre of the UCH II region

the gas between us and the masers might not be sufficiently ionised that it is optically


199

thick. In any case, these 7 masers lie in a third distinct body of gas in the W3(OH)

region.

6.4.4 Morphology

In general, the ground state lines do not show different morphology at the individual

maser level. 1612 MHz, 1665 MHz and 1667 MHz all show lines or arcs of emission.

1720 MHz emission may be limited in its features by the small sample size, although

1720 MHz masers may be significantly smaller than the other frequencies, as shown

by Masheder et al. (1994). In the west of W3(OH) region, 1665 MHz, 1667 MHz and

1720 MHz masers in the area of 1665 MHz maser group d all show elongated

resolved north-south geometry, shown in figure 6.4-6. This region is unique because it

is the only place where masers of three transitions lie close to each other. The 1665

Figure 6.4-4: Demagnetised velocities and magnetic field strengths at all four ground state OH

frequencies in W3(OH).


-50

-48

-46

-44

-42

-40

-38

-2500-2000-1500-1000-50005001000

DEC Offset / mas

Dem

ag.

Vel

. /

(km

/s)

1665

16121667

1720


0

2

4

6

8

10

12

14

-2500-2000-1500-1000-50005001000

DEC Offset / mas

Fie

ld S

tren

gth

/ m

G

16651612

1667

1720

Velocity vs DEC

-50

-48

-46

-44

-42

-40

-38

-2500-2000-1500-1000-50005001000

DEC Offset / mas

Vel

ocity

/ (k

m/s

) 1665 RCP

1665 LCP

1612 RCP

1612 LCP

1667 RCP

1667 LCP

1720 RCP

1720 LCP

Figure 6.4-5: Velocities of all the masers in the ground state. The outlying blue-shifted group of

masers are separated by the dashed line.


200

MHz, 1667 MHz and 1720 MHz masers lie within 50 mas (~100 AU) of each other,

and are almost certainly in the same body of gas (in the area of maser group c, 1612

MHz, 1665 MHz and 1720 MHz masers come to within 140 mas of each other).

Maser region d is typified by the lowest magnetic field strengths at all three

frequencies.

The extended nature of the masers in this region is a consequence of the lower

field strengths, and therefore the lessened effect of magnetic beaming. The north-

south geometry of the masers probably reflects real physical condensations in the gas.

Mill

iAR

C S

EC

MilliARC SEC-820 -830 -840 -850 -860 -870

-600

-620

-640

-660

-680

-700

-720

Mill

iAR

C S

EC

MilliARC SEC-820 -830 -840 -850 -860 -870

-600

-620

-640

-660

-680

-700

-720

Mill

iAR

C S

EC

MilliARC SEC-820 -830 -840 -850 -860 -870

-580

-600

-620

-640

-660

-680

-700

-720

Figure 6.4-6: The elongated north-south nature of masers at all three frequencies present in the east

of W3(OH). Contours are at 75 mJy/beam.

201

Chapter 7

Concluding Remarks

7.1 Data Acquisition

7.1.1 The Observation

This was an observational dataset of outstanding detail – vastly improving on previous

ground-state observations of W3(OH) in almost every way. The VLBA has shown

itself to be an enormously efficient ‘value for money’ telescope, which makes up in its

contribution to understanding that which it lacks in public prestige. W3(OH) is itself

one of the most – if not the most – intensively studied massive-star forming region,

and the depth of information presented here greatly expand the understanding of the

kinematics and conditions in the molecular gas on the near side of the UCH II region.

As an archetypal UCH II region, the results are of great relevance to the understanding

of all UCH II regions, especially in terms of gas dynamics and magnetic fields.

7.1.2 Data Reduction

Spectral line polarimetry observations are renowned amongst radio astronomers for

being the most difficult to reduce and calibrate. The reduction sequence in AIPS is

complex and there are possible pitfalls at every step. The help of experienced staff at

the NRAO in person, and at the end of an email was crucial to the end success of the

project. Since the data reduction at the beginning of this project a great deal of ‘user

friendly’ progress has been made at the NRAO by the construction of routines and

‘run-files’ in AIPS that automate standard parts of the reduction sequence.

Unfortunately, these arrived too late for the present work, but future similar projects

should benefit greatly from such tools. Potentially, the whole field of VLBI data

reduction could be revolutionised by the ever-imminent release of AIPS++ (now

Chapter 7: Concluding Remarks

202

available in beta form), which will finally bring AIPS into the world of modern

computing by introducing a graphical user interface and intuitive help and

configuration sections.

It is hoped that the extensive commenting and code additions made to HAPPY

will allow continued use and development into a general tool for the collation of

maser emission Gaussian fits into maser features. Likewise, although considerably

simpler and smaller than HAPPY, the IDL code for the fitting and plotting of maser

lineshapes and the code for the two-point correlation have been commented enough

and written in a generalised format so that they might find future use as tools or a

basis for further development.

7.2 Analysis

The data, at 1665 MHz in particular, has given up a wealth of information. The

positional distribution, velocity distribution and polarization information together

paint a dynamic picture of the molecular environment around an UCH II region. A

total of 276 masers were detected; 15 at 1612 MHz, 211 at 1665 MHz, 40 at 1667

MHz and 10 at 1720 MHz.

7.2.1 Comparisons with Previous Work

The lack of comparable resolution/sensitivity observations at any frequency other than

1665 MHz limited the comparisons possible. Generally, it is seen that OH ground

state masers are slightly variable over decade timescales, but maser clusters persist.

Masers were identified with previous observations at all frequencies, most especially

at 1665 MHz with the work of Bloemhof et al. (1992). New masers were discovered

at all frequencies, indicating the ongoing activity of the W3(OH) region.

7.2.2 Morphology

The maser emission was generally well fit by 2-dimensional elliptical Gaussians, only

rarely were there cases of significant resolved extended emission. The deconvolved

major axis of the masers tended for all four ground state lines to be lie in the range of

~3 to 10 mas (6 to 20 AU), and this therefore is the likely size of the standard ground-

state OH maser cloud. At 1665 MHz, there are several groups of masers lying in

‘strings’ or arcs, and these may trace the passage of shocks – postulated as a source of


203

the OH required for masing by distruction of H2O. The position angle of masers in

these lines is usually parallel with the arc, a feature that would be expected if the

masers are flattened and elongated in the plane perpendicular to the propagation of the

shock.

In the west of the map, in maser group d, at 1665 MHz, 1667 MHz and 1720

MHz, masers in close proximity all showed extended, resolved, north-south

elongations, which are likely to reflect the shape of condensations in that area.

Aligned condensations could be caused by the effects of magnetic fields preferentially

supporting the gas in some directions. Maser group c displayed a tendency for

extremely small and unresolved masers – an attribute caused by the high magnetic

fields in that area and has shown statistically the effect of magnetic beaming at higher

magnetic field strengths. A spectacular arc of emission in the far south of the map

was discovered co-propagating in both 1665 MHz and 1667 MHz. This arc was well

modelled by an expanding shock of circular geometry, indicating a source 350 AU to

the north.

A complete catalogue of all maser emission above 75 mJy/beam is in

Appendix A, in contour plot form, for future reference.

7.2.3 Lineshapes

Study of the maser lineshapes shows that deviations from Gaussian form are

extremely rare. Minor deviations in the line wings – presumed to be from

unsymmetrical rebroadening – are detected for a small proportion of the masers, but

major deviations exist for just two cases, representing ~1% of the masers. One of

these cases showed a true double peak and may be indicative of the only instance of

negligible velocity redistribution conditions in the gas, since this is required for such

‘hole burning’. Future modelling of maser lineshapes in SFRs will be constrained by

the need to produce Gaussian form in the vast majority of cases. Predictions of

peculiar multi-peaked and multi-shouldered lineshapes that have appeared in the

literature were not borne out by the observations. The FWHM of the lines generally

lay in the range of 200 to 400 m s–1, although caution is required interpreting the

lower end of the range because a minimum of 4 channels were required of a maser for

a fit attempt. A statistical relationship was detected between the FWHM of a maser

and the level of total polarization, whereby low levels of polarization were


204

accompanied by narrow linewidths. As yet, no maser theories predict this

relationship.

7.2.4 Polarization and Magnetic Field

From the general splitting of the LCP and RCP emission, it is clear that the magnetic

field generally points away from the observer. Only in the far southeast is there

evidence that the field may be reversed. The polarization of the masers shows a clear

trend: the further from the centre of the W3(OH) region the masers are, the more

linear polarization and unpolarised emission they exhibit and the less circular and total

polarization they contain. This is interpreted as indicating a general change in

orientation of the ambient magnetic field, such that moving north or south of the

centre results in the angle between the magnetic field axis and the line of sight

increases. This description might indicate that the magnetic field lines point towards

the central star. It is tempting to suggest that the field lines in the disc lie radially, but

given the differential rotation in a disc the lines would quickly wind up into a tight

spiral. What is more likely is that with the 10° inclination of the disc we are seeing

the field entering the disc from ‘above’ or ‘below’.

Interpreting coincident RCP and LCP maser components as Zeeman pairs not

only clarifies the otherwise messy velocity distribution of the masers, but also gives

valuable information about the magnetic field strengths. The magnetic field strength

is found to be fairly constant within each maser cluster, with typical values of ~5 to 7

mG. The exception was maser group c, which exhibits a wide range of magnetic field

strengths from 5 to 13 mG. Relatively low field strengths are found at 1665 MHz,

1667 MHz and 1720 MHz in maser group d in the west of the region. It is generally

assumed that the magnetic field is coupled to the gas in SFRs. In this case, models

predicting different densities for emission from different lines must explain why

magnetic fields are all so similar, since all the frequencies exhibit similar field

distributions. While the overall density cannot vary much because of magnetic field

implications, the OH density could still vary significantly.

A correlation was found between the magnetic field strength and total

polarization, where masers of high magnetic-field strength always displayed high total

polarization. This is probably because the higher gas density presumed to accompany

higher magnetic field strength enhances polarization. A correlation was also found at

all four frequencies between the magnetic field strength and the size of the masers.


205

This is an effect of ‘magnetic beaming’ – which could be enhanced by stronger fields.

Since stronger fields would also indicate higher densities, a maser cloud of smaller

size can become strong enough to be detected. An unexplained correlation was found

between the degree of polarization and FWHM of a maser, where masers with low

total polarizations tended to have narrow linewidths.

A relationship was discovered between the Landé Zeeman line splitting factor

and the likelihood that masers will be in a Zeeman pair, showing that larger Landé

factors made it easier to destroy Zeeman pairs. This has major implications for

theories that intend to explain the occurrence of uneven Zeeman pairs, and isolated

Zeeman components. The general excess of LCP masers at 1665 MHz was shown

unlikely to be a chance phenomenon, and so is probably indicative of a prevailing

condition in the maser clouds that can affect Zeeman pair formation. The most likely

candidate for this is velocity shifts in the masing gas.

7.2.5 Velocity Structure

Before ‘demagnetisation’ the velocity distribution of the masers appears chaotic.

After demagnetisation, a clear pattern emerges: maser group c that contained a large

spread of magnetic field strengths also contains a wide velocity dispersion and so

appears much more chaotic. The masers in the far southeast are shown not to be part

of this group, nor do they share the motion of the rest of the masers, but instead appear

to be an outlying group. The rest of the masers lie in a tight velocity trend from north

to south. The evidence suggests a thick rotating disc or torus of minimum diameter of

4600 AU, aligned north-south around the central massive star – itself determined to

have a lower mass limit of 8 M . Using symmetry arguments for the dimensions of

the disc the stellar mass could be as high as 27 M and the disc diameter 7000 AU.

The special nature of maser group c was noted in reference to observations of

OH excited states and methanol lines, which are both strong in this area. The

phenomenon that is powering this cluster of masers sustains many lines and leads to a

wide velocity spread, and as shown by magnetic field measurements, this group

contains denser clumps of gas than the rest of the region.

7.2.6 Proper Motion

The proper motion of the masers measured between 1978 and 1997 agrees very well

with the measurement of Bloemhof et al. between 1978 and 1986. In the light of the


206

rotational motion showed by the velocity gradient, a different model is proposed to

that proposed by Bloemhof et al. in the light of their results. The proper motions are

consistent with a north-south disc of material at an angle of inclination to the line of

sight of 10°, with the exception of the group of northern masers already noted as

exceptional. A value is adopted for the unknown proper motion of the reference

maser that both emphasises the rotation of the disc but also places the exceptional

maser group in motion away from the disc in an orthogonal direction – indicative of

motion in an outflow. The present work reproduces the generally divergent motion of

the masers, but shows that this need not necessarily indicate expansion; in fact, a

partially sampled rotation can show such a divergent signature. An element of

divergence could also arise within the disc, indicating perhaps that we are witnessing

the destruction of a disc.


The two-point correlation on such a large sample of masers has revealed the true

clustering scale of the 1665 MHz masers. Previous observations estimated the cluster

size at ~60 mas (120 AU), but it is shown that inclusion of all faint outliers increases

the cluster size is ~300 mas (600 AU). The density of masers increases at smaller

scale lengths until the correlation breaks down at the resolution limit – below which it

becomes difficult to separate merged emission. The value here of not making

exhaustive efforts to break slightly extended maser emission down into numerous

Gaussians becomes apparent, since such attempts would simply create a large surplus

of dubious weak masers at scale lengths just above the resolution limit.

7.2.8 Co-Propagation

All the ground state masers show similar morphology, polarization, magnetic field and

velocity distributions. However, instances of co-propagation of two maser

frequencies are extremely rare, with instances of more than two lines not detected at

all. A large portion of the total number of maser overlaps occur in the arc of maser

emission in the south of the region, speculated to be tracing a shock. A propagating

shock scenario is also proposed to account for near-overlaps between 1612 MHz and

1665 MHz masers, where the 1612 MHz emission physically lies ~40 AU ahead of

1665 MHz emission in these shocks. The relative strengths of the lines co-

propagating is of great importance to theoretical modellers, and the results obtained in


207

this work show that – in strong contrast to the strength of emission from the region as

a whole – the 1665 MHz emission is slightly weaker than the 1667 MHz emission.

This will be a major constraint on future models.

7.3 Future Work

7.3.1 Observations

Future observations at the ground state are now less useful unless they have

significantly greater sensitivity, which may now be possible with the European VLBI

Network (EVN) since it has the large antennas at Bonn and Jodrell Bank in it. Greater

resolution is unlikely without orbiting satellite antennas. However, observations

where just one ground state line is observed will now be spatially aligned with all the

ground state lines because of the excellent spatial alignment between the ground state

lines achieved in this work. The exceptional nature of maser group c is clear in

observations of excited OH states and methanol masers. Ideally, positional alignment

between the ground states, excited states and methanol masers in this area is required,

and observations that make this possible would be most useful.

7.3.2 Code Enhancements

Computer programs can always be enhanced in some way or another – whether by

adding extra help documentation or versatile graphical user interfaces. Specifically,

there are useful additions that could be made to the codes developed for this project.

Currently, HAPPY does not propagate polarization measurement errors through to the

results files. This is partly because the AIPS task MFQUV does not measure the rms

noise in the polarization maps and pass it on. The AIPS task MFPRT, which is used

to output the results of SAD is also deficient because it cannot output any polarization

measurements or errors – they must be extracted separately and rather more

laboriously by printing out the AIPS tables with a task like PRTAB. Altering AIPS

tasks is no simple feat, and requires the blessing of the AIPS development crew if it is

to be included in general release, but the tasks MFQUV and MFPRT badly need

enhancing so that they tie together and output the required data in one pass.


208

7.3.3 Modelling

The clustering identified in the two-point correlation function of the 1665 MHz

masers could give useful information about the nature of the medium that holds the

clusters if properly modelled. The increase in clustering at shorter distances

demonstrated in chapter 5.10 (figure 5.29) could simply be a result of a spherical

distribution of uniform density (which itself has peak clustering at the centre) or of a

spherical distribution where density increases towards the centre, e.g. the Single

Isothermal Sphere theory. Modelling the two-point correlation function of clusters of

masers with uniform density and 1/r2 density could identify which of these situations

is present in the maser clusters, and provide evidence for or against the ailing SIS

model of SFRs.

Modelling of the proper motion of masers in a disc, along with masers in

orthogonal outflows, should allow more definite predictions of the unknown motion of

the reference-maser to be made in the proper-motion measurements. The disc model

shown in chapter 5.8 (figure 5.25) does not include any attempt to model masers in

outflows; and these masers could have an important effect on the demonstrated

generally divergent motion of the masers in W3(OH).

Finally, advanced computational models of maser emission are in development

by several groups. Such models aim to reproduce the maser emission observed given

inputs of density, temperature, [OH], [H2], etc. With the constraints laid down by this

work on magnetic field strength (and therefore density), velocity shifts and co-

propagation, the parameter space necessary to search with such codes is now reduced.

Modelling can therefore allow even more complete probing of such areas by

calculating directly the densities and temperatures present in masing star-forming

regions.

209

Appendix A

Channel Maps

The following pages contain individual channel maps of all areas containing maser

emission. The regions selected are indicated on the maser spot map of each

frequency, which comes before each set of contour maps. The contour maps were all

contoured at flux levels that gave the best visual representation of the masers, rather

than contouring all regions at the same flux level. The contour levels of each map are

indicated in brackets after the region label in each caption. All maps are of Stokes I

intensity. Some maps have been rotated by 90° anti-clockwise in order to better fit

them on the pages. Such maps have DEC on the bottom and RA on the right-hand

side. All positions are in mas offsets from the reference feature, which has an

absolute position of: RA 02h 27m 03.s825 ± 0.s001, DEC +61° 52' 25.''089 ± 0.''01

(J2000) (see chapter 4.3.6).

Appendix A: Channel Maps

210

Figure A.1: The 1612 MHz contour mapped maser regions.

1612 MHz Maser Regions

-2300

-2100

-1900

-1700

-1500

-1300

-1100

-900

-700

-500

-300

-100

100

300

500

-1300-1100-900-700-500-300-100100300500700900

RA Offset / mas

DE

C O

ffset

/ m

as

Less than 0.400 Jy

0.400 -> 2.000 Jy

2.000 -> 10.000 Jy


1

3

2


211

Figure A.2: Clockwise from top left: regions 3 (100 mJy/beam), 1 (100 mJy/beam), 2 (100

mJy/beam).

-100

-150

-200

-250

25 26 27 28 29

-100

-150

-200

-250

30 31 32 33 34

Milli

ARC

SEC

-100

-150

-200

-250

35 36 37 38 39

-100

-150

-200

-250

40 41 42 43 44

0 -50

-100

-150

-200

-250

45 46

MilliARC SEC0 -50

47 48

-170

0-1

750

-180

0-1

850

2021

2223

2425

-170

0-1

750

-180

0-1

850

2627

2829

3031

-170

0-1

750

-180

0-1

850

3233

3435

3637

MilliARC SEC-1

700

-175

0-1

800

-185

0

3839

4041

4243

-170

0-1

750

-180

0-1

850

4445

4647

4849

Milli

ARC

SEC

200

100

0-100

-170

0-1

750

-180

0-1

850

50

-1580

-1590

-1600

-1610

-1620

-1630

-1640

-1650

-1660

-1670

-1680

87 88

Milli

ARC

SEC

MilliARC SEC690 680 670 660 650 640 630 620 610

-1580

-1590

-1600

-1610

-1620

-1630

-1640

-1650

-1660

-1670

-1680

89 90


212



-2100

-1900

-1700

-1500

-1300

-1100

-900

-700

-500

-300

-100

100

300

500

-1500-1300-1100-900-700-500-300-100100300500700900

RA Offset / mas

DE

C O

ffset

/ m

as

Less than 0.400 Jy

0.400 -> 2.000 Jy

2.000 -> 10.000 Jy


2a,b

16a,b

10a,b

9a,b,c,d,

8a,b,c,

7

6

5a,b

4a,b3a,b

1a,b

1918

17a,b

15

14

13a,b

11a,b

12


213

Figure A.4: Clockwise from top left: regions 1a (75 mJy/beam), 1b (100 mJy/beam), 2b (80

mJy/beam), 2a (80 mJy/beam).

-185

0

-190

0

-195

0

1213

1415

-185

0

-190

0

-195

0

1617

1819

MilliARC SEC

-185

0

-190

0

-195

0

2021

2223

Milli

ARC

SEC

-500

-600

-700

-185

0

-190

0

-195

0

2425

-185

0

-190

0

-195

0

5556

5758

-185

0

-190

0

-195

0

5960

6162

MilliARC SEC-1

850

-190

0

-195

0

6364

6566

-500

-600

-700

-185

0

-190

0

-195

0

6768

Milli

ARC

SEC

-500

-600

-700

6970

-165

0

-170

0

-175

0

2930

3132

-165

0

-170

0

-175

0

3334

3536

MilliARC SEC

-165

0

-170

0

-175

0

3738

3940

Milli

ARC

SEC

-700

-800

-900

-165

0

-170

0

-175

0

4142

-165

0

-170

0

-175

0

5051

5253

54

-165

0

-170

0

-175

0

5556

5758

59

MilliARC SEC

-165

0

-170

0

-175

0

6061

6263

64

-165

0

-170

0

-175

0

6566

6768

69

-700

-800

-900

-165

0

-170

0

-175

0

7071

Milli

ARC

SEC

-700

-800

-90072

73


214

-175

0

-180

0

-185

0

2324

2526

-175

0

-180

0

-185

0

2728

2930

MilliARC SEC

-175

0

-180

0

-185

0

3132

3334

-175

0

-180

0

-185

0

3536

3738

Mill

iAR

C S

EC

-400

-500

-175

0

-180

0

-185

0

3940

-175

0

-180

0

-185

0

4849

5051

5253

-175

0

-180

0

-185

0

5455

5657

5859

-175

0

-180

0

-185

0

6061

6263

6465

MilliARC SEC

-175

0

-180

0

-185

0

6667

6869

7071

-175

0

-180

0

-185

0

7273

7475

7677

Milli

ARC

SEC

-400

-500

-175

0

-180

0

-185

0

7879

-170

0

-175

0-1

800

89

1011

12

-170

0

-175

0

-180

0

1314

1516

17

-170

0-1

750

-180

0

1819

2021

22

MilliARC SEC

-170

0

-175

0-1

800

2324

2526

27

-170

0

-175

0

-180

0

2829

3031

32

200

0

-170

0-1

750

-180

0

3334

Milli

ARC

SEC

200

0

35

-170

0

-175

0-1

800

4243

4445

46

-170

0

-175

0

-180

0

4748

4950

51

-170

0-1

750

-180

0

5253

5455

56

MilliARC SEC

-170

0

-175

0-1

800

5758

5960

61

-170

0

-175

0

-180

0

6263

6465

66

200

0

-170

0-1

750

-180

0

6768

Milli

ARC

SEC

200

0

6970

Figure A.5: Clockwise from top left: regions 3a (80 mJy/beam), 3b (100 mJy/beam), 4b (100



215

Figure A.5: Clockwise from top left: regions 5a (80 mJy/beam), 5b (110 mJy/beam), 6 (100

mJy/beam).

-1100

-1150

-1200

-1250

13 14 15 16 17

-1100

-1150

-1200

-1250

18 19 20 21 22

-1100

-1150

-1200

-1250

23 24 25 26 27

Milli

ARC

SEC

-1100

-1150

-1200

-1250

28 29 30 31 32

-1100

-1150

-1200

-1250

33 34 35 36 37

-100 -150

-1100

-1150

-1200

-1250

38 39

MilliARC SEC-100 -150

40 41

-1100

-1150

-1200

-1250

52 53 54 55

-1100

-1150

-1200

-1250

56 57 58 59

Milli

ARC

SEC

-1100

-1150

-1200

-1250

60 61 62 63

-1100

-1150

-1200

-1250

64 65 66 67

-100 -150

-1100

-1150

-1200

-1250

68 69


70 71

-550

-600

-650

-700

45 46 47 48 49 50

-550

-600

-650

-700

51 52 53 54 55 56

-550

-600

-650

-700

57 58 59 60 61 62

Milli

ARC

SEC

-550

-600

-650

-700

63 64 65 66 67 68

-550

-600

-650

-700

69 70 71 72 73 74

-850 -900 -950

-550

-600

-650

-700

75 76

MilliARC SEC-850 -900 -950

77


216

Figure A.6: Region 7 (75 mJy/beam).

400

350

300

5455

5657

5859

6061

400

350

300

6263

6465

6667

6869

400

350

300

7071

7273

7475

7677

400

350

300

7879

8081

8283

8485

MilliARC SEC

400

350

300

8687

8889

9091

9293

400

350

300

9495

9697

9899

100

101

400

350

300

102

103

104

105

106

107

108

109

150

100

500

400

350

300

110

111

150

100

500

112

113

Milli

ARC

SEC

150

100

500

114

115

150

100

500

116


217

0

-100

-200

1 2 3 4 5

0

-100

-200

6 7 8 9 10

Milli

ARC

SEC

0

-100

-200

11 12 13 14 15

0

-100

-200

16 17 18 19 20

0 -100

0

-100

-200

21 22

MilliARC SEC0 -100

23 24

0 -100

25

0

-100

-200

81 82 83 84 85

0

-100

-200

86 87 88 89 90

0

-100

-200

91 92 93 94 95

Milli

ARC

SEC

0

-100

-200

96 97 98 99 100

0

-100

-200

101 102 103 104 105

0 -100

0

-100

-200

106 107

MilliARC SEC0 -100

108 109

0

-100

-200

45 46 47 48 49 50

0

-100

-200

51 52 53 54 55 56

0

-100

-200

57 58 59 60 61 62

Milli

ARC

SEC 0

-100

-200

63 64 65 66 67 68

0

-100

-200

69 70 71 72 73 74

0 -100

0

-100

-200

75 76

MilliARC SEC0 -100

77 78

0 -100

79 80

Figure A.7: Clockwise from top left: regions 8a (110 mJy/beam), 8c (130 mJy/beam), 8b (110

mJy/beam).


218

Figure A.8: Clockwise from top left: regions 9a (80 mJy/beam), 9b (180 mJy/beam), 9d (80

mJy/beam), 9c (180 mJy/beam).

200

100

0

-100

39 40 41 42

200

100

0

-100

43 44 45 46

Milli

ARC

SEC

200

100

0

-100

47 48 49 50

200

100

0

-100

51 52 53 54


200

100

0

-100

55 56

200

150

100

50

0

-50

-100

-150

57 58 59

Milli

ARC

SEC

200

150

100

50

0

-50

-100

-150

60 61 62

MilliARC SEC-800 -900 -1000 -1100

200

150

100

50

0

-50

-100

-150

63 64

200

150

100

50

0

-50

-100

-150

65 66 67

Milli

ARC

SEC

200

150

100

50

0

-50

-100

-150

68 69 70

MilliARC SEC-800 -900 -1000 -1100

200

150

100

50

0

-50

-100

-150

71 72

-800 -900 -1000 -1100

73

200

150

100

50

0

-50

-100

-150

74 75

Milli

ARC

SEC

200

150

100

50

0

-50

-100

-150

76 77

MilliARC SEC-800 -850 -900 -950-1000-1050-1100-1150

200

150

100

50

0

-50

-100

-150

78 79


219

200

100

0

-100

80 81 82 83 84

200

100

0

-100

85 86 87 88 89

200

100

0

-100

90 91 92 93 94M

illiAR

C S

EC

200

100

0

-100

95 96 97 98 99

200

100

0

-100

100 101 102 103 104

-800 -1000

200

100

0

-100

105 106


107 108

-600

-700

-800

27 28 29 30 31

-600

-700

-800

32 33 34 35 36

Milli

ARC

SEC

-600

-700

-800

37 38 39 40 41

-600

-700

-800

42 43 44 45 46

MilliARC SEC-200-250-300

-600

-700

-800

47

-600

-700

-800

64 65 66 67

-600

-700

-800

68 69 70 71

Milli

ARC

SEC

-600

-700

-800

72 73 74 75

-600

-700

-800

76 77 78 79

-200-250-300

-600

-700

-800

80 81

MilliARC SEC-200-250-300

82 83

Figure A.9: Clockwise from top left: regions 9e (80 mJy/beam), 10b (75 mJy/beam), 10a (75

mJy/beam).


220

-1000

-1050

-1100

-1150

-1200

28 29 30 31

-1000

-1050

-1100

-1150

-1200

32 33 34 35

Milli

ARC

SEC

-1000

-1050

-1100

-1150

-1200

36 37 38 39

-1000

-1050

-1100

-1150

-1200

40 41 42 43

-300 -350

-1000

-1050

-1100

-1150

-1200

44 45


46 47

-1000

-1050

-1100

-1150

-1200

64 65 66

-1000

-1050

-1100

-1150

-1200

67 68 69

Milli

ARC

SEC

-1000

-1050

-1100

-1150

-1200

70 71 72

MilliARC SEC-260-280-300-320-340-360

-1000

-1050

-1100

-1150

-1200

73

-1750

-1800-1850

-1900

40 41 42 43 44

-1750-1800

-1850

-1900

45 46 47 48 49

Milli

ARC

SEC -1750

-1800

-1850

-1900

50 51 52 53 54

-1750

-1800

-1850-1900

55 56 57 58 59

-100 -200 -300

-1750

-1800-1850

-1900

60 61


62 63

Figure A.10: Clockwise from top left: regions 11a (mJy/beam), 11b (mJy/beam), 13 (mJy/beam).


221

Figure A.11: Region 12 (75 mJy/beam).

-135

0-1

400

-145

0

2324

2526

2728

29

-135

0-1

400

-145

0

3031

3233

3435

36

-135

0-1

400

-145

0

3738

3940

4142

43

MilliARC SEC

-135

0-1

400

-145

0

4445

4647

4849

50

-135

0-1

400

-145

0

5152

5354

5556

57

-135

0-1

400

-145

0

5859

6061

6263

64

-100

-200

-300

-135

0-1

400

-145

0

6566

Milli

ARC

SEC

-100

-200

-30067

68

-100

-200

-30069

70

-100

-200

-30071


222

Figure A.12: Clockwise from top left: regions 14 (75 mJy/beam), 15 (50 mJy/beam), 16b (75


-340

-360

-380

-400

-420

-440

-460

66 67

Milli

ARC

SEC

-340

-360

-380

-400

-420

-440

-460

68 69

MilliARC SEC-1300-1320-1340-1360-1380-1400-1420

-340

-360

-380

-400

-420

-440

-460

70 71

-1900

-1950

-2000

-2050

80 81 82 83 84

-1900

-1950

-2000

-2050

85 86 87 88 89

Milli

ARC

SEC

-1900

-1950

-2000

-2050

90 91 92 93 94

-1900

-1950

-2000

-2050

95 96 97 98 99

MilliARC SEC0 -50 -100

-1900

-1950

-2000

-2050

100 101

-320

-330

-340

-350

-360

-370

51 52 53

Milli

ARC

SEC

-320

-330

-340

-350

-360

-370

54 55 56

MilliARC SEC-700 -710 -720 -730

-320

-330

-340

-350

-360

-370

57

-320

-330

-340

-350

-360

-370

79 80M

illiAR

C S

EC

-320

-330

-340

-350

-360

-370

81 82

MilliARC SEC-695-700-705-710-715-720-725-730

-320

-330

-340

-350

-360

-370

83


223


mJy/beam), 19 (75 mJy/beam).

-1700

-1750

-1800

-1850

-1900

21 22 23

Milli

ARC

SEC

-1700

-1750

-1800

-1850

-1900

24 25 26

MilliARC SEC380360340

-1700

-1750

-1800

-1850

-1900

27 28

-1650

-1700

-1750

-1800

-1850

-1900

51 52 53

Milli

ARC

SEC

-1650

-1700

-1750

-1800

-1850

-1900

54 55 56

MilliARC SEC380360340320

-1650

-1700

-1750

-1800

-1850

-1900

57 58

-1550

-1560

-1570

-1580

-1590

-1600

64 65

Milli

ARC

SEC

MilliARC SEC140 130 120 110 100 90

-1550

-1560

-1570

-1580

-1590

-1600

66 67

-1490

-1500

-1510

-1520

-1530

-1540

81 82 83 84

-1490

-1500

-1510

-1520

-1530

-1540

85 86 87 88

Milli

ARC

SEC

-1490

-1500

-1510

-1520

-1530

-1540

89 90 91 92

740 720 700 680

-1490

-1500

-1510

-1520

-1530

-1540

93 94

MilliARC SEC740 720 700 680

95


224


-2300

-2100

-1900

-1700

-1500

-1300

-1100

-900

-700

-500

-300

-100

100

300

500

-1500-1300-1100-900-700-500-300-100100300500700900

RA Offset / mas

DE

C O

ffset

/ m

as

Less than 0.400 Jy

0.400 -> 2.000 Jy

2.000 -> 10.000 Jy

6

1a,b

5

4

3

72



225


mJy/beam).

-180

0-1

820

-184

0-1

860

-188

0-1

900

-192

0-1

940

-196

0

2627

28

MilliARC SEC

-180

0-1

820

-184

0-1

860

-188

0-1

900

-192

0-1

940

-196

0

2930

31

Milli

ARC

SEC

-400

-500

-600

-700

-180

0-1

820

-184

0-1

860

-188

0-1

900

-192

0-1

940

-196

0

3233

-400

-500

-600

-700

34

-180

0-1

850

-190

0-1

950

4546

4748

-180

0-1

850

-190

0-1

950

4950

5152

MilliARC SEC

-180

0-1

850

-190

0-1

950

5354

5556

-180

0-1

850

-190

0-1

950

5758

5960

Milli

ARC

SEC

-400

-500

-600

-700

-180

0-1

850

-190

0-1

950

6162

-660

-670

-680

-690

-700

-710

-720

66 67

Milli

ARC

SEC

MilliARC SEC-130 -140 -150 -160 -170 -180

-660

-670

-680

-690

-700

-710

-720

68


226


mJy/beam), 5 (75 mJy/beam).

-1400

-1420

-1440

-1460

-1480

35 36M

illiAR

C S

EC

-1400

-1420

-1440

-1460

-1480

37 38

MilliARC SEC-20 -30 -40 -50 -60 -70 -80 -90

-1400

-1420

-1440

-1460

-1480

39

-1390

-1400

-1410

-1420

-1430

-1440

-1450

-1460

-1470

-1480

52 53

Milli

ARC

SEC

MilliARC SEC-20 -30 -40 -50 -60 -70 -80 -90

-1390

-1400

-1410

-1420

-1430

-1440

-1450

-1460

-1470

-1480

54

-2000

-2020

-2040

-2060

-2080

87 88 89

Milli

ARC

SEC

-2000

-2020

-2040

-2060

-2080

90 91 92

MilliARC SEC-40 -60 -80 -100

-2000

-2020

-2040

-2060

-2080

93 94

-40 -60 -80 -100

95

-1970

-1980

-1990

-2000

-2010

-2020

-2030

-2040

40 41

Milli

ARC

SEC

MilliARC SEC-210 -220 -230 -240 -250 -260 -270 -280

-1970

-1980

-1990

-2000

-2010

-2020

-2030

-2040

42


227

Figure A.17: Top to bottom: regions 6 (75 mJy/beam), 7 (60 mJy/beam).

-1720

-1730

-1740

-1750

-1760

-1770

-1780

69 70

Milli

ARC

SEC

MilliARC SEC-520 -530 -540 -550 -560 -570

-1720

-1730

-1740

-1750

-1760

-1770

-1780

71 72

-620

-640

-660

-680

-700

55 56 57 58

-620

-640

-660

-680

-700

59 60 61 62

Milli

ARC

SEC

-620

-640

-660

-680

-700

63 64 65 66

MilliARC SEC-840 -860 -880 -900

-620

-640

-660

-680

-700

67 68


228


-2300

-2100

-1900

-1700

-1500

-1300

-1100

-900

-700

-500

-300

-100

100

300

500

-1500-1300-1100-900-700-500-300-100100300500700900

RA Offset / mas

DE

C O

ffset

/ m

as

Less than 0.400 Jy

0.400 -> 2.000 Jy

2.000 -> 10.000 Jy

3

2

1a,b



229

Figure A.19: Clockwise from top left: regions 1a (100 mJy/beam), 2 (75 mJy/beam), 3 (50

mJy/beam).

100

50

0

-50

30 31 32 33

100

50

0

-50

34 35 36 37

Milli

ARC

SEC

100

50

0

-50

38 39 40 41

100

50

0

-50

42 43 44 45

MilliARC SEC50 0 -50

100

50

0

-50

46 47

-1100

-1150

35 36 37 38

-1100

-1150

39 40 41 42

Milli

ARC

SEC -1100

-1150

43 44 45 46

-1100

-1150

47 48 49 50


-1100

-1150

51

-570

-580

-590

-600

-610

-620

-630

-640

56 57 58

Mill

iAR

C S

EC

-570

-580

-590

-600

-610

-620

-630

-640

59 60 61

MilliARC SEC-830-840-850-860-870-880

-570

-580

-590

-600

-610

-620

-630

-640

62 63

-830-840-850-860-870-880

64


230

100

50

0

-50

48 49 50 51 52

100

50

0

-50

53 54 55 56 57

100

50

0

-50

58 59 60 61 62

Milli

ARC

SEC

100

50

0

-50

63 64 65 66 67

100

50

0

-50

68 69 70 71 72

50 0 -50

100

50

0

-50

73 74

MilliARC SEC50 0 -50

75 76

50 0 -50

77

Figure A.20: Region 1b (150 mJy/beam).

231

References

Adams F. C., Lada C. J., Shu F. H., 1987, ApJ 312, 788

Adams F. C., Shu F. H., 1986, ApJ 308, 836

Alcock C., Ross R. R., 1985a, ApJ 290, 433-44

Alcock C., Ross R. R., 1985b, ApJ 299, 763-8

Alcock C., Ross R. R., 1986a, ApJ 305, 837-51

Alcock C., Ross R. R., 1986b, ApJ 306, 649-54

Allen A., Shu F. H., 2000, ApJ 536, 368-379

Allen R. J., Atherton P. D., Tilanus R. P. J., 1986, Nature 319, 296

Andresen P., 1985, A&A 154, 42-54

Andresen P., Ondrey G. S., Titze B., Rothe E. W., 1984, J. Chem. Phys. 80, 2548-69

Avery L. W., 1980, in "Interstellar Molecules", ed. B. Andrew, p47

Baan W. A., Wood P. A. D., Haschick A. D., 1982, ApJ 260, L49-52

Bakes E. L. O., 1997, "The Astrochemical Evolution of the Interstellar Medium",

Vledder: Twin 106

Bakes E. L. O., Tielens A. G. G. M., 1994, ApJ 427, 822-38

Ball J. A., Gottlieb C. A., Lilley A. E., Radford H. E., 1970, ApJ 162, 1203-10

Bally J., Stark A. A., Wilson R. W., Langer W. D., 1987, ApJ 312, L45

Barret A. H., Rogers A. E. E., 1966, Nature 210, 188

Bash F. M., Peters W. L., 1976, ApJ 205, 786

Baudry A., Desmurs J. F., Wilson T. L., Cohen R. J., 1997, A&A 325, 255

Baudry A., Diamond P. J., 1991, A&A 247, 551-555

Baudry A., Diamond P. J., 1998, A&A 331, 697-708

Baudry A., Diamond P. J., Booth R. S., Graham D., Walmsley C. M., 1988, A&A 201,

102-122

References

232

Baudry A., Menten K. M., 1995, A&A 298, 905-912

Baudry A., Menten K. M., Walmsley C. M., Wilson T. L., 1993, A&A 271, 552

Beckwith S. V. W., Sargent A. I., 1993, ApJ 402, 280

Berulis I. I., Ershov A. A., 1983, Soviet Astron. Lett. 9, 6

Bettweiser E. V., 1976, A&A 50, 231

Bettweiser E. V., Kegel W. H., 1974, A&A 37, 291

Bettweiser E., Misselbeck G., 1977, A&A 61, 567-74

Bevington P. R., Robinson D. K. 1992, "Data Reduction and Error Analysis for the

Physical Sciences", McGraw-Hill 2nd edition

Blitz L., 1978, PhD Thesis, Columbia University

Blitz L., 1993, in "Protostars and Planets III", eds E. H. Levy, J. I. Lunine

Blitz L., Shu F. H., 1980, ApJ 238, 148

Blitz L., Thaddeus P., 1980, ApJ 241, 676

Bloemhof E. E., 1993, ApJ 406, L75-78

Bloemhof E. E., Reid M. J., Moran J. M., 1992, ApJ 397, 500-519

Bodenheimer P., Sweigart A., 1968, ApJ 152, 515

Born M., Oppenheimer J. R., 1927, Ann. Physik 4-84, 457

Buhl D., Snyder L. E., Lovas F. J., Johnson D. R., 1974, ApJ 192, L97-100

Bujarrabal V., Guibert J., Nguyen-Q-Rieu, Ormont A., 1980b, A&A 84, 311

Burdyuzha V. V., Varshalovic D. A., 1973, Soviet Astronomy A. J. 16, 980-2

Burton W. B., 1976, Annu. Rev. Astron. Astrophys. 14, 275

Cameron A. G. W., 1985, ApJ 299, L83

Campbell M. F., Lester D. F., Harvey P. M., Joy M., 1989, ApJ 345, 298

Carrington A. P., 1974, "Microwave Spectroscopy of Free Radicals", Academic Press

London and New York, 181

Casassus S., Bronfman L., May J., Nyman L.-Å., 2000, A&A 358, 514-520

Caselli P., Hasegawa T. I., Herbst E., 1993, ApJ 408, 548-58

Caswell J. L., 1998, MNRAS 297, 215-235

Caswell J. L., 1999, MNRAS 308, 683-690

Caswell J. L., Hayes R. F., 1987, Aust. J. Phys. 40, 215

Cecchi-Pestellini C., Aiello S., 1992, MNRAS 258, 125-33

Cernicharo J., González-Alfonso E., Alcolea J., Bachiller R., John D., 1994, ApJ 432,

L59–62

Cesaroni R., Walmsley C. M., 1991, A&A 241, 537

References

233

Chandrasekhar S., Fermi E., 1953, ApJ 118, 116

Charnley S. B., 1997, ApJ 481, 396-405

Cheung A. C., Rank D. M., Townes C. H., Thornton D. D., Welch W. J., Crowther J.

H., 1969, Nature 221, 626-8

Cimerman M., Scoville N., 1980, ApJ 239, 526

Claussen M. J., Gaume R. A., Johnston K. J., Wilson T. L., 1994, ApJ 424, L41

Cohen R.J., 1989, Rep. Prog. Phys. 52, 881-943

Cook A. H., 1966a, Nature 210, 611

Cook A. H., 1966b, Nature 211, 502

Cook A.H., 1977, “Celestial masers”, Cambridge University Press

Cooper A. J., Davies R. D., Booth R. S., 1971, MNRAS 152, 383

Crawford J., Rowan-Robinson M., 1986, MNRAS 221, 923

Dalgarno A., 1985, in "Molecular Astrophysics", eds G. H. F. Diercksen, W. F.

Huebner, P. W. Langhof, p281-94

Davis R. D., Rowson B., Booth R. S., Cooper A. J., Gent H., Adgie R. L., Crowther J.

H., 1967, Nature 213, 1109-10

Desmurs J. E., Baudry A., Wilson T. L., Cohen R. J., Tofani G., 1998, A&A 334,

1085-1094

Desmurs J. F., Baudry A., 1998, A&A 340, 521

Desmurs J. F., Baudry A., Wilson T. L., Cohen R. J., Tofani G., 1998, A&A 334, 1085

Destombes J. L., Marliere C., Baudry A., Brillet J., 1977, Astr. Ap. 99, 320

Diamond P. J., Martinson A., Dennison B., Booth R. S., Winnberg A., 1988, in

"Radio Wave Scattering in the Interstellar Medium", NY: American Inst. Physics,

p195

Doel R. C., 1990, PhD Thesis, University of Bristol

Draine B. T., McKee C. F., 1993, Annu. Rev. Astron. Astrophys. 31, 373–432

Draine B. T., Roberge W. G., Dalgarno A.. 1983. ApJ 264, 485–507

Dreher J. W., Welch Wm. J., 1981, ApJ 245, 857

Elitzur M., 1979, ApJ 229, 560

Elitzur M., 1982, Rev. Mod. Phys. 54, 1225

Elitzur M., 1990a, ApJ 363, 638

Elitzur M., 1990b, ApJ 350, L17

Elitzur M., 1991b, ApJ 370, 407

Elitzur M., 1992, "Astronomical Masers", Dordrecht: Kluwer

References

234

Elitzur M., 1992, Annu. Rev. Astron. Astrophys. 30, 75-112

Elitzur M., de Jong T., 1978, Astr. Ap. 67, 323

Elitzur M., Hollenbach D. J., McKee C. F., 1992, ApJ 394, 221-227

Elmegreen B. G., Lada C. J., 1977, ApJ 214, 725

Evans J. II, 1999, Annu. Rev. Astron. Astrophys. 37, 311-362

Evans N. J., 1978, in "Protostars and Protoplanets", ed. T. Gehrels, p152

Evans N. J., 1999, Annu. Rev. Astron. Astrophys 37, 311

Evans N. J., Kutner M. L., 1976, ApJ 204, L131

Falgarone E., 1987, in "NATO/ASI Physical Processes in Interstellar Clouds", ed. M.

Scholer

Falgarone E., Panis J.-F., Heithausen A., Pérault M., Stutski J., Puget J.-F., Bensch F.,

1998, A&A 331, 669

Falgarone E., Pineau des Forêts G., Roueff E., 1995, A&A 300, 870-80

Field D., 1985, MNRAS 217, 1

Field D., Gray M. D., 1988, MNRAS 234, 253-72

Field D., Gray M. D., 1994, A&A 292, 271-275

Field D., Gray M. D., St. Paer P. de, 1995, A&A 282, 213-224

Field D., Richardson I. M., 1984, MNRAS 282, 213

Field G. B., 1978, in "Protostars and Protoplanets", ed. T. Gehrels p243, Tucson:

Univ. Ariz. Press

Flower D. R., Guilloteau S., 1982, A&A 114, 238

Forster J. R., Caswell J. L., 1989, A&A 213, 339

Fouquet J. E., Reid M. J., 1982, AJ 87, 691

Franco J., Tenorio-Tagle G., Bodenheimer P., 1990, ApJ 349, 126

Fuller G. A., Meyers P. C., 1987, in "NATO/ASI Physical Processes in Interstellar

Clouds", ed. M. Scholer

Garcia-Barretto J. A., Burke B. F., Reid M. J., Moran J. M., Haschick A. D., Schilizzi

R. T., 1988, ApJ 326, 954-966

Gardiner W. C., 1977, Acc. Chem. Res. 10, 326-31

Gaume R. A., Mutel R. L., 1987, ApJ Supplement 65, 193-253

Goldreich P., Keeley D. A., 1972, ApJ 174, 517-25

Goldreich P., Keeley D. A., Kwan J. Y., 1973a, ApJ 179, 111-34

Goldreich P., Keeley D. A., Kwan J. Y., 1973b, ApJ 182, 55-66

Goldreich P., Kwan J., 1974, ApJ 190, 27-34

References

235

Goldreich P., Scoville N., 1976, ApJ 205, 144

Gordy W., Cook R. L., 1984, "Microwave Molecular Spectra", Interscience

Publishers, chapter 4

Gray M. D., Cohen R. J., Richards A. M. S., Yates J. A., Field D., 2001, MNRAS in

press

Gray M. D., Doel R. C., Field D., 1991, MNRAS 252, 30

Gray M. D., Field D., 1994, A&A 292, 693-698

Gray M. D., Field D., 1995, A&A 298, 243-259

Gray M. D., Field D., 1995, A&A 298, 243-259

Gray M. D., Field D., Doel R. C., 1992, A&A 262, 555-569

Gredel R., Lepp S., Dalgarno A., Herbst E., 1989, ApJ 347, 289-93

Guilloteau S., Lucas R., Ormont A., 1981, A&A 153, 179

Guilloteau S., Stier M. T., Downes D., 1983, A&A 126, 10-15

Güsten R., Fiebig D., Uchida K. I., 1994, A&A 286, L51

Habing H. J., Israel F. P., 1979, Annu. Rev. Astr. Astrophys. 17, 345

Habing H. J., Olnon F. M., Winnberg A., Matthews H. E., Baud B., 1983, Astr. Ap.

128, 230

Hartmann L., 1998, "Accretion Processes in Star Formation", Cambridge UK,

Cambridge University Press, 237

Hartquist T. W., Williams D. A., eds 1998, "The Molecular Astrophysics of Stars and

Galaxies", Oxford: Oxford Univ. Press

Harvey P. J., Booth R. S., Davies R. D., Whittet D. C. B., McLaughlin W., 1974,

MNRAS 169, 545

Hasegawa T. I., Mitchell G. F., Matthews H. E., Tacconi L., 1994, ApJ 426, 215.

Heiles C. H., 1987, in "NATO/ASI Physical Processes in Interstellar Clouds", ed. M.

Scholer, Dordrecht: Reidel

Herbig G. H., 1962, Adv. Astron. Astrophys. 1, 47

Herbst E., Assousa G. E., 1978, in "Protostars and Protoplanets", ed T. Gehrels, p152

Herzberg G., 1950, "Spectra of Diatomic Molecules", (New York: van Nostrand

Reinhold) 658

Ho P. T. P., Klein R. I., Haschick A. D., 1986, ApJ 305, 714

Ho P. T. P., Townes C. H., 1983, Annu. Rev. Astron. Astrophys. 21, 239

Hollenbach D. J., 1997, in "Herbig-Haro Objects and the Birth of Low Mass Stars",

ed. B. Reipurth, IAU Symp. 182, p. 181–98

References

236

Hollenbach D. J., McKee C. F., 1989, ApJ 342, 306–36

Hollenbach D. J., Tielens A. G. G. M., 1997, Annu. Rev. Astron. Astrophys. 35, 179-

215

Hollenbach D., Johnstone D., Lizano S., Shu F., 1994, ApJ 428, 654

Hughes V. A., Viner M. R., 1976, ApJ 204, 55

Humphreys E. M. L., 1999, PhD Thesis, University of Bristol

Jones M. E., Barlow S. E., Ellison G. B., Ferguson E. E., 1986, Chem. Phys. Lett. 130,

218-23

Kahn F. D., 1974, A&A 37, 149

Kawamura J. H., Masson C. R., 1998, ApJ 509, 270-282

Keto E. R., Ho P. T. P., Reid M. J., 1987, ApJ 323, L117

Keto E. R., Welch W. J., Reid M. J., Ho P. T. P., 1995, ApJ 444, 765-769

Knowles S. H., Mayer C. H., Cheung A. E., Rank D. M., Townes C. H., 1969, Science

163, 1055-7

Kwan J., Valdez F., 1983, ApJ 271, 604

Kwan J., Valdez F., 1986, ApJ 309, 783

Lada C. J., 1985, Annu. Rev. Astron. Astrophys. 23, 267

Lada C. J., Elmegreen B. G., Cong H., Thaddeus P., 1978, ApJ 411, 708

Lada C. J., Kylafis N. D., eds 1999, "The Origin of Stars and Planetary Systems",

Dordrecht: Kluwer

Lada E. A., 1992, ApJ 393, L25–28

Lada E. A., Strom K. M., Myers P. C., 1993, in "Protostars and Planets III", ed. E. H.

Levy, J. I. Lunine, pp. 245–77. Tucson: Univ. Ariz.

Larson R. B., 1981, MNRAS 194, 809

Léger A., Jura M., Omont A., 1985, A&A 144, 147-60

Light J. C., 1978, J. Chem. Phys. 68, 2831-43

Linke R. A., Goldsmith P. F., 1980, ApJ 235, 437

Litvak M. M., 1973, ApJ 182, 711-30

Litvak M. M., McWhorter A. L., Meeks M. L., Zeiger H. J., 1966, Phys. Rev. Lett. 17,

821

Litvak M., 1969, ApJ 156, 471-92

Lo K. Y., Ball R., Masson C. R., Phillips T. G., Scott S. L., Woody D. P., 1987, ApJ

317, L63

References

237

Lo K. Y., Walker R. C., Burke B. F., Moran J. M., Johnston K. J., Ewing M. S., 1975,

ApJ 202, 650-654

Loren R. B., Sandquist A., Wootten A., 1983, ApJ 270, 620

Lucas R., 1980, A&A 84, 36-9

Macleod G. C., 1997, MNRAS 285, 635-639

Mader G. L., Johnston K. J., Moran J. M., 1978, ApJ 224, 115-124

Mannings V., Boss A., Russell S., eds, 2000, "Protostars and Planets IV", Tucson:

Univ. Ariz

Masheder M. R. W., Field D., Gray M. D., Migenes V., Cohen R. J., Booth R. S.,

1994, A&A 281, 871-881

Mathis J. S., Rumpl W., Nordsieck K. H., 1977, ApJ 217, 425-33

McKee C. F., 1989, ApJ 345, 782–801

McKee C. F., Zweibel E. G., Goodman A. A., Heiles C., 1993, in "Protostars and

Planets III", eds E. H. Levy, J. I. Lunine

Megeath S. T., Herter T., Beichmann C., et al., 1996, A&A 307, 775

Menten K. M., Johnston K. J., Wadiak E. J., Walmsley C. M., Wilson T. L., 1988, ApJ

331, L41-45

Menten K. M., Reid M. J., Moran J. M., Wilson T. L., Johnston K. J., Batrla W., 1988,

ApJ 333, L83-86

Menten K. M., Reid M. J., Pratap P., Moran J. M., 1992, ApJ 401, L39-42

Mestel L., 1965, Q. J. R. Astron. Soc. 6, 161

Mestel L., 1985, in "Protostars and Protoplanets II", eds D. C. Black, M. S.

Matthews, p320

Mezger P. G., Smith L. F., 1977, in "Star Formation, IAU Symp. No. 75", ed. T. de

Jong, A. Maeder, p133

Mizuno A., Onishi T., Yonekura Y., Nagahama T., Ogawa H., Fukui Y., 1995, ApJ

445, L161

Moore T. J. T., Cohen R. J., Mountain C. M., 1988, MNRAS 231, 887

Moran J. M., Burke B. F., Barrett A. H., Rogers A. E. E., Carter J. C., Ball J. A.,

Cudaback D. D., 1968, ApJ 152, L97

Moran J. M., Reid M. J., Lada C. J., Yen J., Johnston K. J., Spencer J. H., 1978, ApJ

224, L67

Moscadelli L., Menten K. M., Walmsley C. M., Reid M. J., 1999, ApJ 519, 244-256

Mouschovias T. Ch., 1976, ApJ 207, 141

References

238

Myers P. C., 1986, in "Star Forming Regions, IAU Symp. No. 115", ed. M. Peimbert,

J. Jugaku

Myers P. C., 1987, in "Interstellar Processes", ed D. Hollenbach, H. Thronson

Myers P. C., Benson P. J., 1983, ApJ 266, 309

Myers P. C., Fuller G. A., Mathieu R. D., Beichman C. A., Benson P. J., Schild R. E.,

Emerson J. P., 1987a, ApJ 319, 340

Nakahama T., Mizuno A., Ogawa H., Fukui Y., 1998, AJ 1165, 336

Nedoluha G. E., Watson W. D., 1988, ApJ 335, L19

Nedoluha G. E., Watson W. D., 1990a, ApJ 354, 660-675

Nedoluha G. E., Watson W. D., 1990b, ApJ 361, 653

Nedoluha G. E., Watson W. D., 1991, ApJ 376, L63

Neufeld D. A., Dalgarno A., 1989a, ApJ 340, 869–93

Neufeld D. A., Dalgarno A., 1989b, ApJ 344, 251–64

Norman C., Silk J., 1980, ApJ 238, 158

Norris P. R., Byleveld S. E., Diamond P. J., Ellingsen S. P., Ferris R. H., Gough R. G.,

Kesteven M. J., McCulloch P. M., Phillips C. J., Reynolds J. E., Tzioumis A. K.,

Takahashi Y., Troup E. R., Wellington K. J., 1998, ApJ 508, 275-285

Norris R. P., Booth R. S., Diamond P. J., 1982, MNRAS 201, 209-222

Norris R. P., Booth R. S., 1981, MNRAS 195, 213-226

Ohishi M., 1997, in "Molecules in Astrophysics: Probes and Processes", IAU Symp

178, Dordrecht: Kluwer, p61-74

Palumbo M. E., Geballe T. R., Tielens A. G. G. M., 1997, ApJ 479, 839-44

Pavlakis K. G., Kylafis N. G., 1996a, ApJ 467, 292

Pavlakis K. G., Kylafis N. G., 1996b, ApJ 467, 300

Pavlakis K. G., Kylafis N. G., 1996c, ApJ 467, 309

Pelletier G., Pudritz R. E., 1992, ApJ 394, 117–38

Pudritz R. E., 1986, Publ. Astron. Soc. Pac. 98, 709

Radford H. E., 1961, Phys. Rev. 122, 114-30

Reid M. J., Moran J. M., 1988, in "Galactic and Extragalactic Radio Astronomy", ed.

G. L., Verschuur, K. I. Kellermann, New York: Springer, p255

Reid M. J., Silverstein E. M., 1990, ApJ 361, 483

Reid M. J., Haschick A. D., Burke B. F., Moran J. M., Johnston K. J., Swenson G. W.,

1980, ApJ 239, 89-111

Roberts W. W., Stewart G. L., 1987, ApJ 314, 10

References

239

Rowan-Robinson M., 1979, ApJ 234, 111

Sams B. J. III, Moran, J. M., Reid M. J., 1996, ApJ 459, 632-637

Sandford S. A., Allamandola L. J., 1993, ApJ 417, 815-25

Sargent A. I., 1977, ApJ 218, 736

Scalo J., 1986, Fundam. Cosmic. Phys. 11, 1

Schraml J., Mezger P. G., 1969, ApJ 156, 269

Schutte W. A., 1996, in "Cosmic Dust Connection" ed. J. M. Greenberg, p1-42,

Dordrecht: Kluwer

Shu F. H., Adams F. C., Lizano S., 1987, Annu. Rev. Astron. Astrophys. 25, 23-81

Shu F. H., Lizano S., Adams F. C., 1986, in "Star Forming Regions, IAU Symp. No.

115", ed. M. Peimbert, J. Jugaku

Shu F. H., Najita J., Ostriker E. C., Shang H., 1995, ApJ 455, L155–58

Skinner C. J., Smith H. A., Sturm E., Barlow M. J., Cohen R. J., Stacey G. J., 1997,

Nature 386, 472

Snell R. L., Mundy L. G., Goldsmith P. F., Evans N. J., Erikson N. R., 1984, ApJ 276,

625

Snyder L. E., Buhl D., 1974, ApJ 189, L31-3

Solomon P. M., Sanders D. B., 1985, in "Protostars and Protoplanets II", eds D. C.

Black, M. S. Matthews, p81

Solomon P. M., Sanders D. B., Scoville N. Z., 1979, in "The Large Scale

Characteristics of Galaxies", ed W. B. Burton, p35

Spaans M., Hogerheijde M. R., Mundy L. G., van Dishoeck E. F., 1995, ApJ 455,

L167-70

Spitzer L., 1968, in "Nebulae and Interstellar Matter, Stars and Stellar Systems", ed B

Middlehurst, L. H., Aller, 7:1. Chicago: Univ. Press

Spitzer L., 1978, in "Physical Processes in the Interstellar Medium", p162, New

York: Wiley-Interscience

Stark A. A., 1985, in "The Milky Way, IAU Symp. No. 106", eds H. van Woerden, R. J.

Allen, W. B. Burton, p445

Sternberg A., Dalgarno A., 1995, ApJ Suppl. 99, 565-607

Strong A. W., et al, 1988, A&A 207, 1

Thissen T., Spiecker H., Andresen P., 1999, A&A Suppl. Ser. 137, 323-336

Thronson A. H., Harper D. A., 1979, ApJ 230, 133

References

240

Tieftrunk A. R., Gaume R. A., Claussen M. J., Wilson T. L., Johnston K. J., 1997,

A&A 318, 931

Tieftrunk A. R., Megeath S. T., Wilson T. L., Rayner J. T., 1998, A&A 336, 991

Tielens A. G. G. M., Hollenbach D. J., 1985a, ApJ 291, 722-46

Tielens A. G. G. M., Whittet D. C. B., 1997, in "Molecules in Astrophysics: Probes

and Processes", IAU Symp 178, Dordrecht: Kluwer, p45-60

Timmermann R., Poglitsch A., Nikola T., Geis N., 1996, ApJ 460, L65–68

Townes C. H., Schawlow A. L., 1975, "Microwave Spectroscopy", Dover

publications, chapter 7

Turner J. L., 1984, Ph. D. thesis, University of California, Berkeley

Turner J. L., Welch W. J., 1984, ApJ 317, L21

Van Buren D., Mac Low M.-M., Wood D., Churchwell E., 1990, ApJ 353, 570

van Dishoeck E. F., 1988, in "Millimeter and Submillimeter Astronomy", ed. R. D.

Wolstencroft ,W. B. Burton, p117-64, Dordrecht: Reidel

van Dishoeck E. F., 1997, ed. "Molecules in Astrophysics: Probes and Processes",

IAU Symp 178, Dordrecht: Kluwer, p588

van Dishoeck E. F., 1998a, in "The Molecular Astrophysics of Stars and Galaxies",

Oxford: Oxford Univ. Press

van Dishoeck E. F., 1998b, in "Formation and Evolution of Solids in Space", ed. J.

M., Greenberg, Dordrecht: Kluwer

van Dishoeck E. F., Black J. H., 1987, in "NATO/ASI Physical Processes in

Interstellar Clouds", ed M. Scholer

van Dishoeck E. F., Blake G. A., 1998, Annu. Rev. Astron. Astrophys. 36, 317-368

van Vleck J. H., 1929, Phys. Rev. 33, 467-506

Vázquez-Semadini E., Ostriker E. C., Passot T., Gammie C. F., Stone J. M., 2000, in

"Protostars and Planets IV", ed. V. Mannings, A. Boss, S. Russell, Tucson: Univ.

Ariz.

Vrba F. J., Strom K. M., Strom S. E., 1976, Astron. J. 81, 958

Wagner A. F., Graff M. M., 1987, ApJ 317, 423-31

Waller W. H., Clemens D. P., Sanders D. B., Scoville N. Z., 1987, ApJ 314, 397

Ward-Thompson D., 1999, in "The Universe as Seen by ISO", eds. P. Cox & M. F.

Kessler. ESA-SP 427

Watson W. D., 1993, in "The Structure and Content of Molecular Clouds", Springer-

Verlag, Germany

References

241

Weaver H., Dieter N. H., Williams D. R. W., Lum W. T., 1965, Nature 208, 29-31

Welch W. J., Marr J., 1987, ApJ 317, L21

Weliachew L. N., Lucas R., 1988, Rep. Prog. Phys 51, 605

Whiteoak J. B., Gardner F. F., 1973, Astrophys. Lett. 15, 211-5

Willacy K., Williams D. A., Duley W. W., 1994b, MNRAS 267, 949-56

Williams D. A., 1993, in "Dust and Chemistry in Astronomy", Bristol: Inst. Phys.

(IOP), p143-70

Williams D. A., Taylor S. D., 1996, Q. J. R. Astron. Soc. 37, 565-92

Wilson T. L., Johnston K. J., Mauersberger R., 1991, A&A 251, 220

Wilson T. L., Walmsley C. M., Baudry A., 1990, A&A 234, 159

Wilson W. J., Darrett A. H., 1968, Science 161, 778-9

Wink J. E., Duvert G. Guilloteau S., Güsten R., Walmsley C. M., Wilson T. L., 1994,

A&A 281, 505-516

Wiseman J., Ho P. T. P., 1998, ApJ 502, 676

Wu Y. C., 1992, Phys. Lett. A. 167, 243

Wu Y. C., Elitzur M., 1992, Phys. Lett. A. 162, 137

Wynn-Williams C. G., Becklin E. E., Neugebauer G., 1972, MNRAS 160, 1

Wyrowski F., Hofner P., Schilke P., Walmsley C. M., Wilner D. J., Wink J. E., 1997,

A&A 320, L17-20

Wyrowski F., Schilke P., Walmsley C. M., Menten K. M., 1999, ApJ 514, L43-46

Yorke H. W., Krugel E., 1977, A&A 54, 183

Zmuidzinas J., Blake G. A., Carlstrom J., Keene J., Miller D., et al., 1995, Proc.

Airborne Astron. Symp., ed. M. R. Haas, J. A. Davidson, E. F. Erickson, p. 33–40.

San Francisco: Astron. Soc. Pac.

Zuckerman B., Palmer P., 1974, Annu. Rev. Astron. Astrophys. 12, 279

probing the star-forming region w3(oh) with ground …

Documents