optimizationofscanningtransmissionx-raymicroscopy ... · polyol-reinforcing particles in a...

Ultramicroscopy 88 (2001) 33–49

Optimization of scanning transmission X-ray microscopyfor the identification and quantitation of reinforcing particles

in polyurethanes

A.P. Hitchcocka,*, I. Koprinarova, T. Tyliszczaka, E.G. Rightorb, G.E. Mitchellb,M.T. Dineenb, F. Hayesb, W. Lidyc,1, R.D. Priesterc, S.G. Urquhartd,2,

A.P. Smithd, H. Aded

aBrockhouse Institute for Materials Research, McMaster University, 1180 Main Street West, Hamilton, ON, CanadabDow Chemical, 1897 Building, Midland, MI 48667, USA

cDow Chemical, Building B-1608, 2301 N. Brazosport Blvd. Freeport, TX 77541, USAdDepartment of Physics, North Carolina State University, Raleigh, NC 27695, USA

Received 2 May 2000; received in revised form 13 October 2000

Abstract

The morphology, size distributions, spatial distributions, and quantitative chemical compositions of co-polymerpolyol-reinforcing particles in a polyurethane have been investigated with scanning transmission X-ray microscopy(STXM). A detailed discussion of microscope operating procedures is presented and ways to avoid potential artifacts

are discussed. Images at selected photon energies in the C 1s, N 1s and O1s regions allow unambiguous identification ofstyrene–acrylonitrile-based (SAN) copolymer and polyisocyanate polyaddition product-based (PIPA) reinforcingparticles down to particle sizes at the limit of the spatial resolution (50 nm). Quantitative analysis of the chemical

composition of individual reinforcing particles is achieved by fitting C 1s spectra to linear combinations of referencespectra. Regression analyses of sequences of images recorded through the chemically sensitive ranges of the C 1s, N 1sand O1s spectra are used to generate quantitative compositional maps, which provide a fast and effective means ofinvestigating compositional distributions over a large number of reinforcing particles. The size distribution of all

particles determined by STXM is shown to be similar to that determined by TEM. The size distributions of each type ofreinforcing particle, which differ considerably, were obtained by analysis of STXM images at chemically selectiveenergies. # 2001 Elsevier Science B.V. All rights reserved.

Keywords: X-ray microscopy; Polymers; Quantitative chemical mapping; Particle size distributions

1. Introduction

Soft X-ray spectromicroscopy is a powerful toolfor chemical analysis of polymer microstructure.In this study, we demonstrate some of itscapabilities – in particular, qualitative identifica-tion, quantitative chemical analysis, and spatial

*Corresponding author. Tel.: +1-905-525-9140� 24749; fax:

+1-905-521-2273.

E-mail address: [email protected] (A.P. Hitchcock).1Dow Chemical, retired2Current Address: Department of Chemistry, University of

Saskatchewan, Saskatoon, SK, Canada S7N 5C9.

0304-3991/01/$ - see front matter # 2001 Elsevier Science B.V. All rights reserved.

PII: S 0 3 0 4 - 3 9 9 1 ( 0 0 ) 0 0 1 1 3 - 3

metrology of reinforcing particles. In manypractical applications the intrinsic chemical ormechanical properties of a polymer are modified invarious ways through the use of reinforcingparticles. In this case, the system investigated is apolyurethane containing two different types ofcopolymer polyol reinforcing particles.Polyurethane polymers, in the form of free-rise

or molded foams, are widely used in the auto-motive and furnishings industries. In order tomake foams with higher hardness several differentstrategies are presently employed [1]. Copolymerpolyol (CPP) particles, which are dispersions ofpolymer particles in polyether polyol, inorganicfibers, and low molecular weight cross-linkerpolyols, are preferred for stiffening slabstockfoams which are made in long, continuous blocksor buns. In order to understand how reinforcingparticles like co-polymer polyols affect mechanicalproperties such as elastic modulus, tear strengthand resiliency, and in order to develop improvedCPP substances, it is important to have analyticaltechniques which can probe the morphology andchemistry at the required spatial scale. Whiletraditional chemical spectroscopies such as infra-red or NMR are excellent at chemical character-ization, they do not have adequate spatialresolution to address questions relating to thesubmicron composition of polymer blends con-taining reinforcing particles. Analytical transmis-sion and scanning transmission electronmicroscopy (TEM/STEM) have superb spatialresolution and can provide useful polymer micro-analysis when performed with care [2] but in mostcases the high-energy electron beam causes toomuch radiation damage for detailed spectroscopicstudies of the composition of small radiation-sensitive polymer phases. Also, on account of thelower energy resolution of analytical TEM/STEMusing electron energy loss spectroscopy (EELS)[2,3], the ability to differentiate similar chemicalspecies is not as good as synchrotron-based X-raymicroscopy. Overall, NEXAFS microscopy per-formed on a high-resolution synchrotron beamlineprovides superior chemical contrast relative toTEM-EELS.Over the past decade analytical soft X-ray

microscopy has developed into a useful tool for

characterizing soft materials such as polymers[4–7], as well as environmental and biologicalsamples [8]. The variant used in this work,scanning transmission X-ray microscopy (STXM)[9–14] provides near-edge X-ray absorption spec-tra (NEXAFS) with �0.1 eV energy resolution,which provides quite good chemical analysis. Thespatial resolution is of the order of 50 nm,adequate for studies of many reinforcing particlesystems, including CPP particles in polyurethanes.While there is some radiation damage, particularlyto the dominant aliphatic polyether or softsegment component, the damage rate is muchlower than in electron microscopy studies of thesame materials. This is consistent with a recentstudy of polyethylene terephthalate in whichdamage rates of X-ray and electron microscopywere compared quantitatively [14]. Recent relatedstudies that describe complementary aspects of theapplication of NEXAFS microscopy to polymersinclude: spectroscopic investigations to developthe conceptual understanding of the origin ofspectral features [5–7,16–19]; a systematic ap-proach to quantitation through comparison tomodel polymer reference standards [19,20]; quan-titative studies of urea and urethane (carbamate)composition in model homogeneous [19–21], andtypical heterogeneous, phase segregated polyur-ethanes [22,23]; analysis of the orientation ofpolymer chains and functional groups using thepolarization dependence of images and spectra[4,7].In this work, scanning transmission X-ray

microscopy has been used to study a polyurethanein which two types of copolymer polyol (CPP)reinforcing particles were used simultaneously.The polyurethane matrix is a high molecularweight polypropylene oxide (PPO) cross-linkedwith toluene diisocyanate (TDI). One CPP ispoly(styrene-co-acrylonitrile) (symbolized here asSAN), and the other is an aromatic-carbamate richpoly-isocyanate poly-addition product (symbolizedhere as PIPA), derived from methylene diisocya-nate (MDI) [24–26]. Both species were formed anddispersed in the polyether polyol component ofthe polyurethane formulations. The two types ofparticles are indistinguishable when an unstainedsample is viewed in a transmission electron

A.P. Hitchcock et al. / Ultramicroscopy 88 (2001) 33–4934

microscope (TEM). In this study, STXM spectro-scopy, imaging, and analysis procedures aredeveloped which optimize the ability to distinguishand quantitatively analyze these two CPP particletypes. The distributions of the sizes of the PIPAand SAN particles have been measured andcompared to the average particle size distributiondetermined from transmission electron microscopy(TEM). The composition of the SAN and PIPACPP particles has also been analyzed, both on aparticle-by-particle basis, and through the use ofcompositional maps derived from image sequences[27].

2. Experimental

Thin sections of the specimens were prepared bycryo-microtoming plaques of the slabstock poly-urethane at �1208C using a Reichert-Jung FC4Emicrotome. The plaques were made according to aprocedure described earlier [1]. Sections weretransferred dry to uncoated copper grids usingan eyelash. Typical sections were judged to be100 nm thick based on the interference colors ofthe sections on the knife edge. A JEOL 2000FXATEM was used to examine unstained samples.When grids were studied by TEM to examine themorphology, radiation-induced spectroscopicchanges were noted by STXM. Thus, the sectionsanalyzed by STXM in this work were notpreviewed by TEM to avoid radiation damageartifacts.C 1s images and NEXAFS spectra were re-

corded with the Stony Brook STXM at the X-1Abeamline of the National Synchrotron LightSource (NSLS) [9–11]. C 1s, N 1s and O1s imagesand NEXAFS spectra were recorded with thebeamline 7.0 STXM at the Advanced Light Source(ALS) [12–14]. The principles and experimentaldetails of the X-ray optics used in scanningtransmission X-ray microscopy have been de-scribed in detail elsewhere [8,9]. Briefly, mono-chromated undulator radiation illuminates aFresnel zone plate, which has a first-order focalposition defined by f ¼ DdrN=ml , where D is thezone plate diameter, drN is the width of theoutermost zone, m is the order of the zone plate

employed, and l is the X-ray wavelength (seeFig. 1). The diffraction limited resolution is givenby 1:22drN=m. An image is generated by monitor-ing the X-ray signal transmitted through a thinsection of a specimen as it is raster-scanned at thefocus of the zone plate. For these measurementsthe spatial resolution of the NSLS microscope wasabout 60 nm (Raleigh criterion) and the energyresolution was �200meV (the NSLS-based resultsreported herein were obtained during the commis-sioning of a new monochromator which nowprovides �100meV resolution). Measurementsmade with the ALS microscope provided higherenergy resolution (�100meV) but significantlylower spatial resolution (�100 nm) on account ofmechanical vibration and limitations of the X-rayoptics.Point spectra of the CPP particles acquired with

dwell times at each energy of 50.1 s at NSLS didnot appear to be noticeably affected by radiationdamage, although longer dwells were observed todamage the polyether matrix. Mass loss in thepolyether-rich matrix occurs with a critical dose(defined as the dose which produces a 1=e decreasein the intensity of a specified spectral feature)which is �60 times smaller than that for poly(ethylene terephthalate) [15]. The carbonyl groupof the PIPA particles exhibits spectral changeswith a critical dose that is �15 times smaller thanthat for PET [28]. At the ALS the undulator gapmust be scanned synchronously with the photonenergy on account of the narrow undulator

Fig. 1. Schematic of Fresnel zone plate focusing used in

scanning transmission X-ray microscopy. drN is the width of

the outermost zone, and f is the focal length of the first-order

diffraction spot.

A.P. Hitchcock et al. / Ultramicroscopy 88 (2001) 33–49 35

spectral output, whereas full range NEXAFSspectra at the NSLS are acquired without chan-ging the undulator gap. A mechanical shutter isused to block unwanted radiation during thechange of monochromator energy and undulatorgap, but the shutter has an uncertainly in itsopening time of the order of 100ms. Thus, forpractical reasons, acquisition of point spectra atthe ALS microscope is relatively slow – theminimum sampling period is about 1 s – andtypically some X-ray flux hits the sample forbetween 0.3 and 0.5 s/energy point. In contrastpoint spectra at the NSLS are acquired with 20–60ms/energy point since the undulator gap is notchanged. The long sampling period combined withthe �10� higher flux at ALS relative to NSLSlead to significant beam damage over the course ofacquiring a NEXAFS spectrum in point spectralmode at the ALS (when the focused X-ray beam isused). In order to obtain meaningful spectra withnegligible radiation damage, the ALS microscoperoutinely uses sequences of images with short dwelltimes (0.2–0.4ms per pixel) and a fine energyspacing. Subsequent post-processing aligns theseimages, then extracts spectra at any subset ofpixels in the sampled region. Image alignment isnecessary to correct for drifts of the field of viewassociated with lateral shifts in the position of theFresnel zone plate focusing optical element as its z-position is scanned to retain focus with changingphoton energy (wavelength). The software forimage sequence acquisition at NSLS, and thesubsequent analysis of the data (sometimes called‘stacks’) was developed by Chris Jacobsen [27]. Inaddition to curve fit analysis of point spectra, wehave also used newly developed procedures [29] togenerate quantitative composition maps by alinear regression analysis of the spectrum at eachpixel in an image sequence.In order to obtain the most reliable quantitative

analysis using STXM it is important to under-stand, monitor, and minimize a number of sourcesof systematic error. On the instrumental side, thisincludes factors such as detector non-linearity,spectral purity of the incident radiation, and thespatial profile of the zone plate focused beam [14].Regarding detector linearity, the NSLS micro-scope uses a gas proportional counter operating in

pulse counting mode which is linear to �1MHz,with an efficiency of �15% and negligible darkcount. The ALS microscope uses a phosphor(polycrystalline GdO2S2 : Tb, recently convertedto a single-crystal yttrium aluminum phosphide) toconvert X-rays to visible photons which are thendetected in pulse mode with a high performancephotomultiplier tube (Hammamatsu 647P). Theoverall system in the ALS microscope has a linearresponse up to 25MHz, a measured efficiency of30% at the C 1s edge, and a dark count of less that200Hz, which is negligible relative to the typicaldetected incident flux signal of 5–15MHz.With regard to spectral purity, the NSLS

microscope is equipped with a two-mirror ordersuppression device which reduces higher orders toless than 1% [30]. Currently, ALS beamline 7.0does not have a tunable order-sorting device butNi coatings on mirrors provide an effective meansto damp higher-order photons above 850 eV.Photoemission studies on this line indicate thatthe light incident on the zone plate has �4.5%second order and �1% third order at 300 eV.These levels are reduced to 0.6% and 0.5%,respectively if a 160 nm Ti filter foil is placed inthe beam; however, the Ti filter was not incorpo-rated in the beamline when these measurementswere made. At the time of these experiments, theusable linear optical density range was limited toOD values of less than 3 at the ALS microscope,whereas, in CO2 gas, linear optical densityresponse up to 7 has been observed with the NSLSmicroscope when the order sorting mirrors havebeen tuned carefully.The profile of the X-rays that impinge on the

sample to generate the transmitted signal has anumber of components: the desired focused spotfrom first-order diffraction, the portion of thesecond-order diffraction that passes through theorder sorting aperture, un-diffracted beam in caseswhere the central stop is not sufficiently opaque,and a diffuse scattering halo due to imperfectionsin the placement and profile of the zone platezones. The superimposition of light from allsources other than the first-order focused spotgives rise to a broad, structured distribution manytimes the focal spot size – the so-called ‘‘halo’’.Under most favorable conditions, the integrated


intensity of all of the halo (i.e. everything outsidethe focal spot) is 51% of the focused beam. If theorder-sorting aperture (OSA) and central stop ofthe zone plate are not aligned, then un-diffractedbeam which passes around the central stop willincrease the halo intensity. Under certain circum-stances, the halo can give rise to signal artifactswhich can affect the accuracy of quantitativeanalysis. For example, when the sample understudy is a strongly absorbing species embedded ina more transparent matrix (e.g. carbon black in apolymer matrix), then 1–2% of transmitted light inthe halo region becomes a significant fraction ofthe observed signal. In such cases, the halo signallimits the maximum optical density one canobserve without effective absorption saturation,and the X-ray absorption signal reflects the matrixas well as the dense embedded species. Clearlycareful tuning of the microscope is an importantfactor in producing accurate results.On the operational side, one must use acquisition

techniques which provide meaningful incident flux(I0) signals, good energy calibration, and one mustmonitor the extent of radiation damage. At boththe NSLS and ALS microscopes the I0 signal usedto generate optical densities is recorded before orafter the actual measurement. It is necessary to usethis approach in order to have the same detectorsensitivity and to incorporate spectral effects of thezone plate, such as its contributions to filteringhigher-order radiation. In the ALS microscope, asecondary signal, the current at the front of theOSA, is collected. This signal, which is a convolu-tion of total electron yield and gas ionization yield,serves to correct for instantaneous variations inincident flux (due to beam motion in the storagering) as well as to correct for the decay in ringcurrent between sample and Io measurements.Energy scale calibration at the NSLS micro-

scope is routinely carried out by injecting a smallamount of CO2 gas in the He stream bathing thephoton path length and recording the sharpRydberg lines of CO2 [14,31]. At the ALS, gasand solid reference samples can be recorded forcalibration. For this work, the N1s and O1sspectra were calibrated with N2 and O2, respec-tively, while the C 1s spectra were calibrated usingthe known positions of the strong resonances.

Prolonged study of these polyurethane samplesresults in mass loss of the polyether-rich matrixand selective damage of the carbonyl groups of thePIPA. In both microscopes, after critical analyticalmeasurements such as image sequences, it isroutine procedure to monitor the extent of sampledamage by recording an image at an energy thatshows large spectral changes due to radiationdamage (e.g. 289 eV for C 1s studies). Measure-ments with unacceptably high damage rates arenot used in subsequent analysis. The data reportedin this work are those recorded before significantdamage had occurred (55% change).In the data analysis one must evaluate the

significance of the derived quantitative results byappropriate statistical and reproducibility tests.The accuracy of the results is limited by the degreeto which materials used as reference standardsaccurately match the chemical character of thecomponents in the material under study. The set ofreference compounds must encompass all of thechemical components in the material and thereference spectra should be unaffected by radiationdamage or spectroscopic saturation. A largeamount of supporting spectroscopy [16–20] hasbeen carried out to make sure spectral features areassigned correctly and that the model spectra arethose appropriate to the problem.

3. Results and discussion

3.1. C 1s imaging and spectroscopy: qualitativeidentification

Prior to the STXM investigation, a differentsection of the same sample was examined bytransmission electron microscopy (Fig. 2). Althoughthe morphology and size distribution of the CPPparticles are clearly revealed, TEM imaging ofunstained samples cannot distinguish the chemicalidentity of the individual particles. By contrast, asoutlined below, the chemical sensitivity of selectiveenergy imaging NEXAFS microscopy makes iden-tification of the CPP particles very straightforward.Differences in the X-ray absorption spectra of

the components provide the basis for distinguish-ing PIPA and SAN CPP particles. This is


illustrated in Fig. 3 which plots C 1s NEXAFSspectra of the polyether-rich matrix, the SAN andPIPA particles, along with the average spectrum ofthe sample. These spectra were acquired at theALS microscope from the mixed CPP sampleusing the image sequence mode of acquisition. TheSAN spectrum was recorded from an area of�2 mm2 while the PIPA spectrum was recordedfrom an area of �0.2 mm2. The spectra are similarto those recorded in point-mode from othersamples containing larger spatial regions of theseCPP materials. The spectra of both PIPA andSAN exhibit a strong low-energy transition at285 eV, which is associated with C 1s ! p�C�Ctransitions at the C–H carbons of the phenyl ringcommon to both structures. In addition, the SANparticle has a strong transition at 286.7 eV arisingfrom C 1s ! p�C�N transitions of the acrylonitrilecomponent (AN) of the styrene/acrylonitrile(SAN) copolymer. This feature is unique to theacrylonitrile component of the SAN particleswhich provides the basis both for distinguishingSAN from PIPA and for quantitative analysis of

the styrene/acrylonitrile composition of individualSAN particles. At 287.2 eV, on the high-energyside of the p�C��N peak, the optical density of thePIPA particles and the matrix is the same and verylow relative to that for SAN. Thus, an idealstrategy for selective imaging of SAN is to imageat 287.2 eV, where the SAN particles are in highcontrast but the PIPA particles are indistinguish-able from the matrix. Note that the quality of this‘sorting’ of the two types of CPP particles dependson making a careful choice of energy, accurateenergy calibration, and also on the spectral purityof the imaging photon beam. Any higher order orstray light contamination reduces the chemicalcontrast. At higher energy, around 288 eV, the

Fig. 2. Transmission electron microscope image of a polyur-

ethane plaque containing a mixture of two types of copolymer

polyol (CPP) reinforcing particles. One type is SAN, a

copolymer of polystyrene and polyacrylonitrile, and the other

type is PIPA, a carbamate-rich polyurethane addition product.

How does one know which particle is which type of CPP?Fig. 3. C 1s X-ray absorption spectra (NEXAFS) of SAN

particles, PIPA particles, the polyurethane matrix, and the

spatially averaged spectrum of the sample. Vertical offsets of

0.4, 0.8 and 1.2 OD units have been used in plotting the upper

three spectra. These spectra were extracted using multi-pixel

sums over relevant regions from an image sequence recorded

with the scanning transmission X-ray microscope (STXM) at

the Advanced Light Source (ALS). The chemical structures

indicated for the SAN and PIPA particles are the structural

motifs that give rise to the dominant spectral features, but they

do not depict all aspects of their structure, which includes poly-

ether components. What are the values of m and n?


matrix begins to have a very strong absorptionassociated with the C 1s ! s�

C2O transitions in thepolyether chains. At somewhat higher energy,289.8 eV, the PIPA particles have a sharp peakassociated with C 1sðC¼OÞ ! p�C¼O transitions.In pure PIPA this feature is stronger than theintense 285.1 eV p�C¼C peak and it occurs at anenergy where SAN does not have a sharp feature.On this basis one might expect 289.8 eV to be anideal energy for selective imaging of PIPA.However, in practice, while imaging at 289.8 eVgives some relative enhancement of contrast ofPIPA relative to SAN, it is an inefficient means tohighlight PIPA relative to SAN because of thestrong absorption by both the matrix and SAN atthis energy which results in poor overall contrast.This energy could be useful for distinguishingPIPA particles from segregated ureas, whichabsorb at 289.4 for a TDI-based system [23]. Ingeneral, it is much easier to see strong absorptionagainst a weak background (as in imaging at285.1 eV which dramatically highlights both typesof reinforcing particles against the weakly absorb-ing polyurethane matrix) than it is to detect strongabsorption against a background of similarabsorption strength. As a further example of therelative contrast afforded by p�C�N versus p�C�Otransitions, the spatially averaged spectrum(Fig. 3, obtained by summing the spectra fromall pixels in the �100 mm2 area imaged), gives someindication of the SAN-specific peak at 287 eV butessentially no sign of the PIPA-specific peak at290 eV. Comparison of the spatially resolvedspectra to the spatially averaged spectrum empha-sizes the value of good spatial resolution inNEXAFS chemical analysis of heterogeneousmaterials.When using variable spectroscopic sensitivity to

highlight a specific component, one must beconcerned about the effect of density and thicknessvariations between different phases in the sample,and among different sections of the same sample.These factors can complicate the otherwisestraightforward approach of imaging using chemi-cally specific energies derived from model NEX-AFS spectra. Density and thickness variations canbe compensated by normalization of chemicallysensitive images to images measured at photon

energies outside of the chemically sensitive near-edge regions.C 1s NEXAFS microscopy works very well in

differentiating the two CPP particles in thissample. Fig. 4 compares images recorded at285.1 eV (p�C�C) and 287.2 eV (� p�C�N) of a14 mm� 14 mm region of the sample. As discussedabove, the energy chosen for the second image wasnot at the top of the p�C�N resonance peak butrather slightly above, in order to minimize thedifference between the absorption by the poly-ether-rich matrix and the PIPA, so that the SANparticles have the highest contrast. The image at285.1 eV (Fig. 4a) highlights both types of CPPparticles with very strong contrast relative to thepolyurethane matrix. This is as expected from thestrong p�C¼C absorption by the phenyl groups ofthe aromatic CPP particles and much smallerabsorption by the polyurethane matrix, which ispolyether rich. It is very clear that imaging at287.2 eV (Fig. 4b) provides complete selectivity forthe SAN particles, based on their acrylonitrilecontent. Images at 289.8 eV (not shown) provideonly limited ability to distinguish the two types ofCPP particles. However, the PIPA particles can bereadily identified in the difference of the 285.1 and287.2 eV images, as illustrated in Fig. 4c. We notethat careful choice of an acquisition strategy, suchas that employed to record the SAN-selectiveimage in Fig. 4b, is preferable to image subtractionas this avoids possible artifacts associated withimage drift, image scale changes, etc. Such artifactsgive rise to ghost images at the large SAN particlesin Fig. 4c, particularly the lower-half of the image,even though every effort was made to match thespatial scales of the two images prior to subtrac-tion. As described in the following section,regression analysis of image sequences using fitsof the spectrum at each pixel to spectral standards[29], provides extremely powerful chemical analy-sis. If the spectral standards are suitable, thisapproach gives the quantitative chemical composi-tion at each point in the region studied.

3.2. Chemical contrast at the N 1s and O1s edges

While the C 1s edge clearly provides excellentability to distinguish these two types of reinforcing


particles, it is also of interest to investigate thechemical contrast at the N1s and O1s edges. All ofthe nitrogen in SAN is at the unsaturated nitrilegroup (C�N) whereas the nitrogen in PIPA and

the matrix is in essentially saturated environments(N–H). Thus, one might expect very strongselectivity for SAN at the energy of theN 1s ! p�C�N transition. Similarly, since theSAN particles nominally do not contain oxygen,one might expect strong contrast at the O 1s edge.In spite of the seemingly ‘‘obvious’’ potential forimage contrast at these energies, we must remem-ber that the total absorption cross-section at eachX-ray energy for a particular component isgoverned by a number of factors, including densitydifferences, thickness differences and the back-ground from lower energy core/valence absorptionedges, in addition to the chemically specifictransition.Fig. 5 compares N 1s spectra of the three

components, and images recorded with the ALSSTXM at several energies in the N 1s spectralregion. The contrast in the N1s region is lowerthan at the C 1s edge (on account of the large,underlying C 1s signal), except at 399.8 eV (imageb), the energy of the strong N 1s ! p�C�N transi-tion of the nitrile group. Imaging at 399.8 eVpreferentially highlights the SAN relative to thePIPA particles and is clearly as effective as C 1sselective energy imaging for particle type identifi-cation. The N1s spectrum of PIPA is very similarto that of the matrix, as expected since all of thematrix nitrogen is in a carbamate environment inthis particular polyurethane formulation [20].Although their spectral shapes are similar, thePIPA particles are somewhat more dense than thematrix. Therefore they have a higher opticaldensity at all energies and they remain visiblethroughout the whole N1s region.Fig. 6 compares O 1 s spectra of the PIPA, SAN

and matrix components, along with images re-corded with the ALS STXM at the indicated

3

Fig. 4. Comparison of X-ray microscopy images at (a) 285.1 eV

and (b) 287.2 eV of a polyurethane containing SAN and PIPA

CPP particles recorded with the Stony Brook STXM at the

National Synchrotron Light Source (NSLS). (c) Difference of

the 285.1 and 287.2 eV images which highlights the PIPA

particles only. The ghost signals at the positions of many of the

SAN particles are from imperfect registry of the two images. In

these transmission images dark regions correspond to higher

absorption.


energies. In the O 1s spectral region it is the PIPAparticle that has the unique spectral signature,namely the p�C¼O resonance at 532.5 eV. Whileimaging at this energy does highlight the PIPArelative to the SAN particles (PIPA are the darkersmall particles in Fig. 6b, see arrow), the contrastis relatively low since the O 1s ! p�C¼O transition isa weak feature relative to the remainder of the O 1ssignal, in part associated with the polyethercontent of the PIPA particles but mainly withunderlying N 1s and C 1s absorption. Since thechemical structures of both the matrix and PIPAcontain oxygen while the nominal structure ofSAN does not, we expected the SAN to appearwith strong negative contrast in the O 1s region.However, the SAN particles used in this CPPformulation do have some polyether component(as do all copolyol polymer particles, see below)and, therefore, they have an O 1s spectral signaturerather similar to the ether signal present in boththe PIPA particles and the matrix. The increase insignal across the O 1s edge is only 10% for SAN,much less than the increase for the PIPA and

matrix particles. This leads to somewhat lowercontrast of SAN in the O 1 s continuum. However,the optical density of the SAN CPP particlesremains larger than the surrounding matrix at allenergies and thus the expected contrast reversalwas not observed. This could be because the SANparticles are thicker than the rest of the section,they have a higher density, or they have a higherabsorption cross-section. A higher overall spectro-scopic cross section does exist because of thestrong N 1 s absorption in SAN. The SANparticles are also known to be more dense thatthe matrix. Overall, there is relatively weakspectral contrast at the O 1s edge since it issuperimposed on the strong C 1s and N1scontinua This limits the ability to differentiateSAN and PIPA particles by means of selectiveenergy imaging at the O 1s edge.

3.3. Particle size analysis

It is clear from Figs. 4–6 that the size distribu-tion of the SAN particles differs from that of the

Fig. 5. (left) N1s spectra of the matrix, PIPA and SAN particles recorded with the ALS STXM in linescan mode. No spectral offsets

were used. (right) Transmission mode images recorded at 397, 400 ðp�C��NÞ and 403 eV, respectively, in the N1s region. Note the clear

selective identification of the SAN particles in the 400 eV image.


PIPA particles. We have quantitatively analyzedthe apparent particle size distributions of each typeof CPP particle, based on chemically selective C 1simaging. (We denote these particle sizes as‘apparent’ since the section is much thinner thanthe diameter of most particles and thus there areartifacts associated with the random sampling ofthe levels of particles the section was cut through[32]). In addition we have determined the apparentparticle size distributions from TEM images, insome cases of exactly the same thin section of thesample that was analyzed by STXM. The particlesize analysis was carried out in the followingmanner. A threshold value corresponding to about10% greater than the minimum intensity in theparticles was applied to the full image to generate amask image in which each pixel is either black (0)or white (1). The area of each particle was thendetermined by summing the number of contiguous

white pixels. Finally, all particles are assumed tobe circular in order to compute the particlediameter from the measured area. Systematicerrors of this procedure include: reliability of thethresholding (a range of values was explored andthe distributions reported correspond to the bestthresholding achievable); occurrence of someparticles sectioned through the top or bottomregions rather than the middle; effects of particlecurvature which blur the boundaries of theparticles; errors associated with finite sampling;non-circular particle cross-sections. The impact ofthese types of systematic errors on particle sizedistributions is discussed elsewhere [32].Histograms of the distributions of the particle

sizes (�diameters) derived from the two STXMimages recorded at NSLS (Fig. 4) and frommultiple ALS STXM and TEM images are shownin Fig. 7. One hundred to several hundred particles

Fig. 6. (left) O 1s spectra of the matrix, PIPA and SAN particles recorded with the ALS STXM in linescan mode. The appropriate OD

scales for each species are indicated by arrows. (right) Transmission images recorded at 525, 532.5 ðp�C¼NÞ and 548 eV respectively, in

the O 1 s region. Note the enhanced contrast of the PIPA particles in the 532.5 eV image.


in each image were analyzed. Fig. 7a compares theparticle size distribution, independent of chemicalcomposition, as measured from STXM images at285.1 eV (e.g. Fig. 4) and from TEM images. The285.1 eV STXM and TEM distributions aresimilar, each exhibiting two maxima – one around50 nm, the other around 150–200 nm, as well as along tail to larger size. Fig. 7b is the sizedistribution of the SAN particles as determinedfrom analysis of 287.2 eV STXM images (Fig. 4).Note the change in vertical scale. The SANdistribution is clearly different. Relative to the

average particle size distribution (Fig. 7a), thereare more large particles, fewer particles in themiddle range, and a relatively large number of sub-100 nm SAN particles. Fig. 7c depicts the particlesize distribution of the PIPA particles derived fromthe difference in the size distributions derived from285.1 and 287.2 eV STXM images. The PIPAparticle distribution is very different from theaverage distribution, with the PIPA particleshaving a narrower range of sizes, clustered around200 nm. This analysis shows that there are very fewPIPA particles larger than 0.5 mm, and that thepeak of the PIPA distribution lies below 200 nm.In contrast, the SAN particle size distribution ismore complicated. The size-weighted average sizeis around 400 nm. Most SAN particles are smallerthan 1 mm and many SAN particles are smallerthan 0.1 mm, but there are also a number of SANparticles in the 2–3 mm range. In several cases thelarger SAN particles have a ‘cauliflower’ shape (seeFigs. 2 and 4), suggesting they may be agglomer-ates of smaller particles. Results from the two X-ray microscopes are in good agreement, althoughthe higher spatial resolution of the NSLS micro-scope allows better definition of the size distribu-tions at small sizes. The power of STXM topartition particle distributions according to che-mical composition is well illustrated by thisexample.

3.4. Composition maps from regression analysesof image sequences

Fig. 8 presents maps of the chemical compo-nents of this system – polystyrene (pS), polyacry-lonitrile (pAN), polyol, matrix, and PIPA –obtained by a 5-component regression analysis ofa sequence of 72 images (12� 8 mm) between 280and 291 eV recorded at the ALS. The indicatedgrey scales correspond to relative mol% since themodel spectra were on absolute oscillator strengthper repeat unit scales. The model spectra for thematrix and the PIPA are those plotted in Fig. 3,extracted from regions of this image sequence. Themodel spectra of polypropylene oxide (analog tothe polyol component of the CPP particles),polystyrene (pS) and polyacrylonitrile (pAN) wererecorded on un-oriented samples with the Stony

Fig. 7. (a) Histograms of particle sizes derived from STXM

images recorded at 285.1 eV at NSLS (image shown in Fig. 4)

and at ALS. The inset shows the particle size distribution

derived from analysis of four TEM images (total area of

�2000mm2, �400 particles). (b) SAN particle size distribution

derived from STXM images recorded at 287.2 eV at NSLS

(Fig. 4) and ALS. (c) The difference of the particle size

distributions of STXM images at 285.1 and 287.2 eV, which is

the PIPA particle size distribution.


Brook STXM at NSLS using a defocused beam toensure no radiation damage artifacts. The modelspectra from NSLS have been carefully calibratedand the energy resolution matched to that used forthe ALS measurements.

The individual component images clearly sup-port the conclusions drawn from selective energyimaging (Fig. 4). In particular, they show thatthere are both large and small SAN particles butonly small PIPA particles. The polyol map showssignal at the SAN and PIPA particles whereasthere is negligible signal at the correspondinglocations in the matrix map. While there is someoverlap with the matrix signal, the very largedifference between the matrix and polyol mapsindicates that C 1s NEXAFS microscopy is able todistinguish pure polyether–polyol in the CPPparticles from the partially aromatic polyurethanematrix despite their strong spectral similarity,which arises since the matrix is predominantlypolyether-polyol (see Fig. 3). This differentiationhas enabled measurement of the amount ofpolyether-polyol in the SAN and PIPA particles.The PIPA image identifies and quantifies the PIPAparticles. It also shows structure in the matrix,which may be evidence for the hard phasemicrosegregation intrinsic to aromatic polyur-ethanes [23]. Further study is needed to verifythat point. Finally, the pS map has a weak, diffusesignal in the matrix region which may help explaina discrepancy between the styrene/acrylonitrileratio measured by STXM and that expected fromthe formulation. This issue is discussed further inthe following section which presents the quantita-tive results for the CPP particle compositionsderived from analysis of the C 1s image sequence.Fig. 9 presents composition maps derived

from image sequences recorded in the N1s andO1s spectral regions. Since the N1s and O1sspectra of the pure materials were not available,the pS and pAN components were not analyzedseparately. Instead, the image sequence wasfit to linear combinations of spectra internallygenerated from selected regions known to bePIPA, SAN and the matrix. The N1s analysisclearly distinguishes the SAN particles andcould readily be used as the basis for quantitationof the acrylonitrile component of SAN. Regressionanalysis of O 1s image sequences gives reasonableidentification of the PIPA particles, but has limitedability to distinguish the SAN from the matrix.The O1s is the least favorable of the three coreedges, which is not surprising given the similarity

Fig. 8. Component maps for polystyrene, polyacrylonitrile,

polyol, polyurethane matrix and PIPA derived from linear

regression analysis of the spectrum at each pixel in a C 1s image

sequence recorded at the ALS. The portion of the aligned

sequence analyzed consisted of 171� 111 pixel images (80 nm/

pixel) recorded at 72 energies between 280 and 291 eV. See text

for details of the models used. The vertical scales indicate

relative mol%.


of the O 1s spectra of the three main components(Fig. 6).

3.5. Quantitative compositional analysis of theCPP-reinforcing particles

In addition to identification of different CPPmaterials, and the qualitative aspects of thecompositional mapping, it is possible togenerate quantitative chemical compositionsfrom either the component maps or from pointspectra [19,20,33–36]. In this section, wedescribe the use of signals extracted from imagesequences to (i) quantify the composition of thetwo types of reinforcing particles and (ii) todetermine the ratio of polystyrene (pS) andpolyacrylonitrile (pAN) in several of the largerSAN copolymer particles.We have used two approaches to the quantita-

tive analysis of the chemical composition ofindividual reinforcing particles. For both types ofparticles, the compositions were determined fromquantitative component maps derived from theC 1s image sequence (Fig. 8) as outlined below. Wehave also performed regression analyses of selectedPIPA and larger SAN particles by curve fitanalysis of spectra extracted by summation over

multi-pixel regions from the C 1s image sequence.The compositions determined by spectral curve fitanalysis are in good agreement with the results ofthe image sequence analysis, so only the latter isreported in detail here. In order to establishquantitative intensity scales, the model spectrawere background subtracted and converted to anoscillator strength per repeat unit scale, by settingthe far C 1s continuum to match the value foratomic carbon [37,38] and then multiplying by thenumber of carbon atoms per repeat unit –polyether (3), polystyrene (8), polyacrylonitrile(3), and PIPA modeled as MDI reacted with aC4 diol (19). The extracted PIPA and SAN C1sspectra were then curve fit to find the mixture ofthe spectra of pure components which bestreproduced the spectral shapes in the chemicallysensitive 284–291 eV region. An equivalent least-squares minimization procedure was accomplishedwith the regression analysis of the image sequencefrom 280 to 291 eV (Fig. 8). With either approach,the derived regression coefficients are the relativemol% values for each species so that normal-ization to the sum of these values gives absolutemol%. The absolute mol% values were thenconverted to wt% using the molecular masses ofthe repeat units – polyether (42), polystyrene (104),

Fig. 9. Component maps for PIPA, SAN and matrix derived from linear regression analysis of the spectrum at each pixel in N1s and

O1s image sequences recorded at the ALS. In these component maps brighter regions correspond to higher concentration or thickness.


polyacrylonitrile (53), and PIPA modeled as MDIreacted with a C4 diol (278).Table 1 reports the analysis results from the

component mapping for four SAN and four PIPAparticles. The quality of the fits to the extracted C1 s spectra are illustrated in Figs. 10 and 11 forSAN and PIPA, respectively. In both the SAN andPIPA particles there is an appreciable amount ofpolyether – 12(2) wt% in SAN and �15(4) wt% inPIPA. The four SAN particles analyzed are allmuch larger than the thickness of the microtomedthin film – the smallest particle from whichcompositional numbers were derived is about0.5 mm diameter, whereas the film was �0.1 mmthick. Thus the polyether component identified bythis quantitative analysis is unlikely to arise frommatrix above or below the particle, but rather tobe part of the chemical make-up of the particleitself. Since the size of the PIPA particlesapproaches that of the specimen thickness, thiscomment is less applicable to the PIPA particles,

and thus some of the polyether signal foundnominally in the PIPA may be from matrix, eitheron top or below the PIPA, or even at the edgeswhere it may be partly mixed in on account of thelimited spatial resolution. CPP particles aresynthesized in polyether polyol in a manner thatexplicitly forms covalent links to the CPP particles[26]. Thus it is very likely that polyol is incorpo-rated into the CPP particles. Infrared spectro-scopic analysis of extracted SAN solid gave anestimate of 10wt-% incorporated polyol, inreasonable agreement with our analysis. The fourlarge SAN particles in the image sequence whichwere analyzed were found to have relativelyuniform compositions, with a variation of lessthan 5% in the ratio of the styrene to acrylonitrilecontent. The average weight ratio of (pS) to (pAN)in these larger SAN particles is 0.86(2). Interest-ingly, this pS/pAN ratio is lower than thatpredicted from the relative mass of the monomerfeedstocks in the bulk formulation of the SAN

Table 1

Quantitative in situ analysis of the chemical composition of individual polyurethane copolymer polyol-reinforcing particles

Mol%a Wt%b

# Ec Sc ANc S/ANc Ec Sc ANc S/ANc

A. SAN particles (see Fig. 10)

1 21 25 54 0.45 14 40 46 0.88

2 18 25 57 0.45 12 40 48 0.85

3 16 25 59 0.43 11 40 49 0.82

4 22 24 54 0.45 14 40 46 0.88

B. PIPA particles (see Fig. 11)

Mol%a Wt%b

# Cc Ec Cc Ec

1 49 51 86 14

2 40 60 81 19

3 70 30 94 6

4 66 34 92 18

aBased on regression analysis of the C 1s image sequence (Fig. 8) using model spectra whose vertical scales had been established on a

per-repeat-unit oscillator strength basis. The estimated error based on the quality of the fits is 3mol% {this does not account for

systematic errors associated with mismatch of models and the material under study} The quality of the fits are illustrated in Figs. 10

and 11. Results for four particles are presented to indicate reproducibility of the analysis and uniformity of the composition of the

particles. While the fit was made to five components (polyurethane matrix, pure polyol, polystyrene, polyacrylonitrile and PIPA), the

table reports the compositions of the main components for each particle type, ignoring the minority components which contribute a

few % to the fit.bWt% (a)=mol% (a)*mw (a)/(Si mol%i *mwi) where mol% (a) is mol% of species a and mwi represents the molecular weight of

species i.cE=polyether, S=polystyrene, AN=polyacrylonitrile, C=carbamate (PIPA)


particles, which corresponds to a pS to pAN wt%ratio of 1.50. The reason for this discrepancy is notfully understood at present. One possibility isincomplete polymerization of the styrene duringformation of the CPP-polyol. Unreacted styrenemonomer or polystyrene oligimer could then bepreferentially located in the polyurethane matrixrelative to the amount in the SAN particles. A lowlevel of signal is observed in the polystyrenecomponent map (Fig. 8) throughout the matrixregion which provides qualitative support for thisinterpretation.

These results illustrate the ability of STXM-based NEXAFS microscopy to provide quantita-tive chemical analysis on submicron regions withbetter than 5% precision. Other examples ofquantitative polymer analysis with STXM include:urea/urethane composition in model polyur-ethanes [6,19] and high water content polyurethanes[23]; polystyrene/poly(methyl meth-acrylate) blends[35], PS, Br-PS thin film structures [36], andconfinement structures [39]. The level of precisiondemonstrated in this work should be achievablewhenever there are characteristic spectral signaturesof the individual components and the radiationdamage rates are low enough to allow acquisitionof representative images and spectra. Aided by

Fig. 10. Illustration of the quality of the C 1s image sequence

curve fit analysis summed over four SAN particles. The dots are

the spectral data summed over �1000 pixels, the solid line is thefit, and the thin solid line indicates the pS, pAN and polyol

component contributions. The circles superimposed on the four

SAN particles in the inset image indicate the regions of the

sample from which this spectrum was derived. This navigation

image was derived from the SAN and PIPA component maps

(Fig. 8) by thresholding to have 0/1 coded maps which were

summed with a 1 : 3 weighting.

Fig. 11. Illustration of the quality of the C 1s image sequence

curve fit analysis summed over four PIPA particles. The dots

are the spectral data summed over �200 pixels, the solid line is

the fit, and the thin solid lines indicates the carbamate and

polyol component contributions. The four PIPA particles

analyzed are indicated in the inset navigation image (see

caption to Fig. 10).


image sequence procedures, it is possible to carryout this level of detailed analysis on features downto and even slightly smaller than 0.1mm indiameter.

4. Summary

Based on its spectroscopic and imaging cap-abilities, scanning transmission X-ray microscopy(STXM) has been shown to be a suitable tool fordistinguishing polymer-reinforcing particles ofdifferent chemical composition, and for quantify-ing their chemical composition and size distribu-tions. Comparisons of images at selected energiesprovide a quick and reliable means to differentiateSAN and PIPA CPP particles. The C 1 s region isbetter than the N 1s or O 1s regions, partly becauseof better intrinsic spectral resolution (smallernatural linewidths), but also because the contrastat the C 1s edge is much higher since there isnegligible underlying background continuum. N 1simaging at 400 eV also provides high selectivity.Quantitative spectral analysis showed that bothtypes of CPP reinforcing particles contain�15wt% polyether polyol. The presence of thiscomponent in the SAN particles leads to relativelypoor ability to differentiate SAN and PIPAparticles at the oxygen edge. The polystyrene/polyacrylonitrile ratio in large SAN particles wasdetermined to be constant within the precision ofthe measurements. This study is a good illustrationof the power of soft X-ray spectromicroscopy forquantitative microanalysis of soft, radiation sensi-tive materials such as polymers and biologicalspecimens. It is one of many possible areas ofapplication of soft X-ray spectromicroscopy topolymer science.

Acknowledgements

The authors thank Gene Young for assistancewith sample preparation, and Dana Gier andMark Adams for helpful discussions. We thankDr. Chris Jacobsen (SUNY Stony Brook) forgenerously sharing his image (‘stack’) data acqui-sition and analysis software, and Eric Kneedler for

supplying the core of the image-spectra regressionanalysis code. Data was recorded using the StonyBrook STXM at the X-1A NSLS beamline and theBL7.0 STXM at ALS. The Stony Brook STXMwas developed by the groups of J. Kirz and C.Jacobsen, with support from the Office of Biolo-gical and Environmental Research, US DOEunder contract DE-FG02-89ER60858, and theNSF under grant DBI-9605045. The zone plateswere developed by S. Spector and C. Jacobsen ofStony Brook and Don Tennant of Lucent Tech-nologies Bell Labs, with support from the NSFunder grant ECS-9510499. The ALS STXM wasdeveloped by T. Warwick (ALS), B. Tonner(UWM) and collaborators, with support fromthe US DOE under contract DE-AC03-76SF00098. Zone plates at ALS were providedby Eric Anderson of CXRO, LBNL. H. Ade, S.G.Urquhart, and A.P. Smith are supported by NSFYoung Investigator Award (DMR-9458060) andby Dow Chemical. We thank Pierre Brassard foruse of his particle size analysis program in earlystudies. A.P. Hitchcock’s work is supported byNSERC research and strategic grants.

References

[1] D.R. Gier, R.E. O’Neill, M.R. Adams, R.D. Priester Jr.,

W.A. Lidy, E.G. Rightor, C.G. Barnes, B.L. Davis

(Eds.), Filler, Hard Phases and Copolymer Polyols –

Their Different Mechanisms to Reinforce Flexible Poly-

urethane Foams, Proceedings of the Society of Plastics

Institute 1998, Dallas, TX.

[2] L.M. Brown, C.A. Walsh, A. Dray, A.L. Bleloch,

Microsc. Microanal. Microstruct. 6 (1995) 121.

[3] R. Egerton, Electron Energy Loss Spectroscopy, Plenum

Press, New York, 1986.

[4] H. Ade, X. Zhang, S. Cameron, C. Costello, J. Kirz, S.

Williams, Science 258 (1992) 972.

[5] H. Ade, B. Hsiao, Science 262 (1992) 1427.

[6] H. Ade, A. Smith, S. Cameron, R. Cieslinski, C. Costello,

B. Hsiao, G.E. Mitchell, E.G. Rightor, Polymer 36 (1995)

1843.

[7] A.P. Smith, H. Ade, Appl. Phys. Lett. 69 (1996) 3833.

[8] J. Kirz, C. Jacobsen, M. Howells, Quart. Rev. Biophysics

33 (1995) 33.

[9] H. Ade, in: J.A.R. Samson, D.L. Ederer (Eds.), Experi-

mental Methods in the Physical Sciences, Vol. 32,

Academic Press, New York, 1998, p. 225.


[10] C. Jacobsen et al., Proceedings of the International

Conference on X-ray and Optical Microanalysis Institute

of Physics, Bristol, 1993.

[11] D. Sayre, H.N. Chapman, Acta Crystallogr. A 51

(1995) 237.

[12] T. Warwick, H. Padmore, H. Ade, A.P. Hitchcock, E.G.

Rightor, B. Tonner, J. Electron Spectrosc. 84 (1997) 85.

[13] T. Warwick, K. Franck, J.B. Kortwight, G. Meigs, M.

Moronne, S. Myneni, E. Rotenberg, S. Seal, W.F. Steele,

H. Ade, A. Garcia, S. Cerasari, J. Denlinger, S.

Hayakawa, A.P. Hitchcock, T. Tyliszczak, E.G. Rightor,

H.-J. Shin, B. Tonner, Rev. Sci. Inst. 69 (1998) 2964.

[14] H. Ade, A.P. Smith, H. Zhang, B. Winn, J. Kirz, E.

Rightor, A.P. Hitchcock, J. Electron Spectrosc. 84 (1997)

53.

[15] E.G. Rightor, A.P. Hitchcock, H. Ade, R.D. Leapman,

S.G. Urquhart, A.P. Smith, G. Mitchell, D. Fischer, H.J.

Shin, T. Warwick, J. Phys. Chem. B 101 (1997) 1950.

[16] S.G. Urquhart, A.P. Hitchcock, R.D. Leapman, R.D.

Priester, E.G. Rightor, J. Polym. Science B: Polym. Phys.

33 (1995) 1593.

[17] S.G. Urquhart, A.P. Hitchcock, R.D. Priester, E.G.

Rightor, J. Polym. Sci. B: Polym. Phys. 33 (1995) 1603.

[18] S.G. Urquhart, A.P. Hitchcock, A.P. Smith, H. Ade,

E.G. Rightor, J. Phys. Chem. B 101 (1997) 2267.

[19] S.G. Urquhart, A.P. Smith, H. Ade, A.P. Hitchcock,

E.G. Rightor, W. Lidy, J. Phys. Chem. B. 103 (1999)

4603.

[20] S.G. Urquhart, A.P. Hitchcock, A.P. Smith, H. Ade, W.

Lidy, E.G. Rightor, G.E. Mitchell, J. Electron Spectrosc.

100 (1999) 119.

[21] A.P. Hitchcock, S.G. Urquhart, E.G. Rightor, W. Lidy,

H. Ade, A.P. Smith, T. Warwick, Microsc. Microanal. 3

(S-2) (1997) 909.

[22] A.P. Hitchcock, S.G. Urquhart, E.G. Rightor, W. Lidy,

H. Ade, A.P. Smith, T. Warwick, Microsc. Microanal. 5

(S-2) (1998) 808.

[23] E.G. Rightor, G.E. Mitchell, S.G. Urquhart, A.P. Smith,

H. Ade, A.P. Hitchcock, in preparation.

[24] K. Picken, Urethanes Technology, June 1984 p. 23.

[25] T. Narayan, J.T. Paton, US Patent 4,326,043 to BASF

Wyandotte Corp. April 20, 1982. (ref 2.78 of Herrington,

cited p. 2.14).

[26] R. Herrington, Flexible Polyurethane Foams, 2nd edi-

tion, The Dow Chemical Company, Midland, MI, 1997.

[27] C. Jacobsen, S. Wirick, G. Flynn, C. Zimba, J. Microsc.

197 (2000) 173.

[28] T. Coffrey, M.Sc. Thesis, NCSU, 1998.

[29] A.P. Hitchcock, I. Koprinarov and R. Kneedler, in

preparation.

[30] H. Ade et al., Proceedings of VUV-10, World Scientific,

Singapore, 1993.

[31] A.P. Smith, T. Coffey, H. Ade, in: J. Thieme, G. Schmahl,

E. Umbach, D. Rudolph (Eds.), X-ray Microscopy and

Spectromicroscopy, Springer, Berlin, 1998.

[32] J.C. Russ, Computer-Assisted Microscopy: The Measure-

ment and Analysis of Images Plenum Press, New York,

1990.

[33] G.E. Mitchell, et al. 1998 Compendium of ALS

Abstracts, LBNL Pub 43431, 1999, p. 282.

[34] G.E. Mitchell, L.R. Wilson, M.T. Dineen, S.G. Urqu-

hart, F. Hayes, E.G. Rightor, A. Hitchcock, H. Ade,

Science (2000), submitted for publication.

[35] H. Ade, D.A. Winesett, A.P. Smith, S. Qu, S. Ge, S.

Rafailovich, J. Sokolov, Europhys. Lett. 45 (1999) 526.

[36] H. Ade, D.A. Winesett, A.P. Smith, S. Anders, T.

Stammler, C. Heske, D. Slep, M.H. Rafailovich, J.

Sokolov, J. Stohr, Appl Phys. Lett. 73 (1998) 3773.

[37] B.L. Henke, P. Lee, T.J. Tanaka, R.L. Shimabukuro,

B.K. Fujikawa, Atom. Nucl. Data Tables 27 (1982) 1.

[38] A.P. Hitchcock, D.C. Mancini, J. Electron. Spectrosc. 67

(1994) 1.

[39] S. Zhu, Y. Liu, M.H. Rafailovich, J. Sokolov, D.

Gersappe, D.A. Winestt, H. Ade, Nature 400 (1999) 49.


optimizationofscanningtransmissionx-raymicroscopy ... · polyol-reinforcing particles in a...

Documents