femtosecond serial imaging using fast integrating detectors · 2013. 5. 29. · j.mol.biol. 221,...

```

Anton Barty(and 85+ collaborators)

Center for Free Electron Laser Science (CFEL)DESY, Hamburg, Germany

Femtosecond serial imaging using

fast integrating detectors

This work was the product of a large international team

Henry N. Chapman1,2, Petra Fromme3, Anton Barty1, Thomas A. White1, Richard A. Kirian4, Andrew Aquila1, Mark S. Hunter3, Joachim Schulz1, Daniel P. DePonte1, Uwe Weierstall4, R. Bruce Doak4, Filipe R.N.C. Maia5, Andrew Martin1,

Ilme Schlichting6,7, Lukas Lomb7, Nicola Coppola1, Robert L. Shoeman7, Sascha Epp6,8, Robert Hartmann9, Daniel Rolles6,7, Artem Rudenko6,8, Lutz Foucar6,7, Nils Kimmel10, Georg Weidenspointner11,10, Peter Holl9, Mengning Liang1,

Miriam Barthelmess12, Carl Caleman1, Sébastien Boutet13, Michael J. Bogan14, Jacek Krzywinski13, Christoph Bostedt13, Sa!a Bajt12, Lars Gumprecht1, Benedikt Rudek6,8, Benjamin Erk6,8, Carlo Schmidt6,8, André Hömke6,8,

Christian Reich9, Daniel Pietschner10, Lothar Strüder6,10, Günther Hauser10, Hubert Gorke15, Joachim Ullrich6,8, Sven Herrmann10, Gerhard Schaller10, Florian Schopper10, Heike Soltau9, Kai-Uwe Kühnel8, Marc Messerschmidt13, John D. Bozek13, Stefan P. Hau-Riege16, Matthias Frank16, Christina Y. Hampton14, Raymond Sierra14, Dmitri Starodub14, Garth

J. Williams13, Janos Hajdu5, Nicusor Timneanu5, M. Marvin Seibert5, Jakob Andreasson5, Andrea Rocker5, Olof Jönsson5, Stephan Stern1, Karol Nass2, Robert Andritschke10, Claus-Dieter Schröter8, Faton Krasniqi6,7, Mario Bott7,

Kevin E. Schmidt4, Xiaoyu Wang4, Ingo Grotjohann3, James Holton17, Stefano Marchesini17, Sebastian Schorb18, Daniela Rupp18, Marcus Adolph18, Tais Gorkhover18, Martin Svenda5, Helmut Hirsemann12, Guillaume Potdevin12,

Heinz Graafsma12, Björn Nilsson12, and John C. H. Spence4

1. Center for Free-Electron Laser Science, DESY, Notkestrasse 85, 22607 Hamburg, Germany.2. University of Hamburg, Luruper Chaussee 149, 22761 Hamburg, Germany.3. Department of Chemistry and Biochemistry, Arizona State University, Tempe, Arizona 85287-1604 USA.4. Department of Physics, Arizona State University, Tempe, Arizona 85287 USA.5. Laboratory of Molecular Biophysics, Department of Cell and Molecular Biology, Uppsala University, Husargatan 3 (Box 596), SE-751 24 Uppsala, Sweden.6. Max Planck Advanced Study Group, Center for Free Electron Laser Science (CFEL), Notkestrasse 85, 22607 Hamburg, Germany.7. Max-Planck-Institut für medizinische Forschung, Jahnstr. 29, 69120 Heidelberg, Germany.8. Max-Planck-Institut für Kernphysik, Saupfercheckweg 1, 69117 Heidelberg, Germany.9. PNSensor GmbH, Otto-Hahn-Ring 6, 81739 München, Germany.10. Max-Planck-Institut Halbleiterlabor, Otto-Hahn-Ring 6, 81739 München, Germany.11. Max-Planck-Institut für extraterrestrische Physik, Giessenbachstrasse, 85741 Garching, Germany.12. Photon Science, DESY, Notkestrasse 85, 22607 Hamburg, Germany.

13. LCLS, SLAC National Accelerator Laboratory, 2575 Sand Hill Road. Menlo Park, CA 94025, USA.14. PULSE Institute and SLAC National Accelerator Laboratory, 2575 Sand Hill Road. Menlo Park, CA 94025, USA.15. Forschungszentrum Jülich, Institut ZEL, 52425 Jülich, Germany.16. Lawrence Livermore National Laboratory, 7000 East Avenue, Mail Stop L-211, Livermore, CA 94551, USA.17. Advanced Light Source, Lawrence Berkeley National Laboratory, Berkeley, California 94720, USA.18. Institut für Optik und Atomare Physik, Technische Universität Berlin, Hardenbergstrasse 36, 10623 Berlin, Germany.

X-ray sources have developed at a staggering pace since their discovery in 1895

1E+06

1E+10

1E+14

1E+18

1E+22

1E+26

1E+30

1E+34

1910 1940 1970 2000 2030

Peak

Bri

llian

ce

Year

X-ray tubes

1st generation

2nd generation

3rd generation

Free-electron lasers

Synchrotrons

Some great moments in X-ray science

1E+06

1E+10

1E+14

1E+18

1E+22

1E+26

1E+30

1E+34

1910 1940 1970 2000 2030

Peak

Bri

llian

ce

Year

Röntgen

Bragg & Braggreflections

von Lauecrystal diffraction

Hodgkinpenicillin,B12

Perutz & Kendrewmyoglobin

Franklin, Crick,

WatsonDNA

Holmes & Rosenbaum

Walker ATP

MacKinnonPotassium channel

KornbergRNA polymerase

Yonath, Steitz, Ramakrishnan, NollerRibosome

Hauptman, Karledirect methods

Sayre equation

Jacobsen HolographyKirz & Schmahl Microscopy

The number of solved protein structures is now increasing linearly with time

0

10,000

20,000

30,000

40,000

50,000

60,000

19721976

19801984

1988 1992 1996 2000 2004 2008

X-rayNMRElectron

Year

Str

uct

ure

sThe bulk of protein structures have been solved using X-ray crystallography

X-ray crystallography requires large, well ordered crystals to overcome radiation damage

http://en.wikipedia.org/wiki/Image:X_ray_diffraction.png

• Grand challenge: Can we revolutionise molecular biology by imaging isolated molecules ?

• >52,684 PDB entries• but only

~10,000 distinct structures114 integral membrane proteins

• The bottleneck is in growing good crystals

• Membrane proteins are especially important (eg: for drug delivery)

X-ray crystallography is powerful, but growing the crystals is often difficult

Consider the case of RNA polymerase II

1972: started to investigate structure1983: 2D ‘crystal’ obtained

Nature 301, 125 (1983)

1991: 3D crystal growth observedJ.Mol.Biol. 221, 347 (1991); Nat.Struct.Biol 1, 195 (1994)

1999: Crystal oxidation problem solvedCell, 98, 799 (1999)

2000: First detailed crystal structureCramer et.al, Science 228, 640 (2000)

RNA Polymerase II Extracted from Cramer et.al, Science 228, 640 (2000)

(Fig.3).Eachofthesmallsubunitsoccursinasinglecopy,arrayedaroundtheperiphery.Thestructureiscross-struttedbyelementsofRpb1andRpb2thattraversethecleft:AhelixofRpb1bridgesthecleft,andtheCOOH-terminalregionofRpb2extendstotheoppositeside.TheRpb1-Rpb2complexisanchoredatoneendbyasubassemblyofRpb3,Rpb10,Rpb11,andRpb12.

Theactivesitewaslocatedcrystallographi-callybyreplacementofthecatalyticMg

2!ion

withZn2!

,Mn2!

,orPb2!

(40).AnativezincanomalousFouriershoweda10-"peakthatlikelyresultsfrompartialreplacementofthe

activesiteMg2!

byZn2!

duringproteinpuri-fication(Fig.1),anddifferenceFouriersob-tainedfromcrystalssoakedwitheitherMn

2!or

Pb2!

showedasinglepeakatthesamelocation(41).Themetalionsiteoccurswithinaprom-inentloopofRpb1(Fig.3),which,onthebasisofpreliminarysequenceassignment,harborstheconservedaspartateresiduemotif(42).Onlyonecatalyticmetalionwasfound,andonlyonewasreportedforabacterialRNApolymerase(43),althoughatwo-metalionmechanism,asdescribedforsingle-subunitpolymerases(44),isnotruledout.

ThelocationofduplexDNAdownstreamof

theactivesite(aheadofthetranscribingpoly-merase)waspreviouslydeterminedbydiffer-ence2Dcrystallographyofanactivelytran-scribingcomplex(27).CanonicalB-formDNAplacedinthislocationliesintheRpb1-Rpb2cleft,andcanfollowastraightpathtotheactivesite(Fig.3).About20basepairsarereadilyaccommodatedbetweentheedgeofthepoly-meraseandtheactivesite,consistentwithnu-cleasedigestionstudiesshowingtheprotectionofaboutthislengthofdownstreamDNA(45).ThisproposalforthepolII–DNAcomplexisalsoconsistentwithresultsofprotein-DNAcross-linkingexperiments:Rpb1andRpb5

Fig.3.ArchitectureofyeastRNApolymeraseII.Backbonemodelsforthe10subunitsareshownasribbondiagrams.Secondarystructurehasbeenassignedbyinspection.Thethreeviewsarerelatedby90°rotationsasindicated.DownstreamDNA,thoughnotpresentinthecrystal,isplacedontotheribbonmodelsas20basepairsofcanonicalB-DNA(blue)inthelocationpreviouslyindicatedbyelectroncrys-

tallographicstudies(27).Eightzincatoms(bluespheres)andtheactivesitemagnesium(pinksphere)areshown(Table1).Thebox(upperright)containsakeytothesubunitcolorcodeandanin-teractiondiagram.Thesameviewsandcolorcodingareusedthrough-outthearticle.ThisandotherfigureshavebeenpreparedwithRIBBONS(87).

RESEARCHARTICLES

www.sciencemag.orgSCIENCEVOL28828APRIL2000643

The bottleneck in structural biology is frequently growing good crystals

Why crystals?

Signal-to-noise

Radiation damage

Combine 105-107 measurements

Classification Averaging Orientation Reconstruction

Diffraction pattern(low signal = noisy)

10 fs FEL pulse

Particle injection

One pulse per diffraction pattern

X-ray free-electron lasers may enable atomic-resolution imaging of macromolecules without the need to grow large crystals

Animation courtesy of Sébastien Boutet, CXI instrument scientist, SLAC

Single particle imaging at LCLS

Our diffraction camera can measure forward scattering close to the direct soft-X-ray FEL beam

Multilayer reflectivity is uniform across the 30° to 60° gradient

“Soft edge” prevents any scatter from the hole

Bajt et.al. Appl.Opt. 47, 1673 (2008)

GRADED MULTILAYER MIRROR: Si, Mo, and B4C layers, period graded laterally. Variation matches angle of incidence (30º to 60º) to maintain Bragg condition for λ = 32 nm. Reflectivity: 45% for 32 nm pulses.

The mirror protects the CCD and works as a(i) bandpass filter (bandwidth = 9 nm at 45º) (ii) filter for stray light (1% off-axis reflectivity)(iii) low-scatter beam-stop

multilayer mirror

CCD

filter

samplemultilayer mirror

CCD with transmission filtershield

15o

-15o

sample

60°d = 32 nm

30°d = 18 nm

Sasa Bajt, Eberhard Spiller, and Jennifer Alameda

The VUV-FEL diffraction experiment employs a unique camera to measure forward scattering with high SNR

We performed single particle imaging of viruses in the CAMP instrument at LCLS

!

Max-Planck-Institutfür Kernphysik

Sascha Epp1, Robert Hartmann1,2, Daniel Rolles1, Artem Rudenko1, Lutz Foucar1, Benedikt Rudek1, Benjamin Erk1, Carlo Schmidt1, André Hömke1, Nils Kimmel2, Christian Reich2, Günther Hauser2, Daniel Pietschner2, Peter Holl2, Hubert Gorke3, Helmut Hirsemann4, Guillaume Potdevin4, Tim Erke4, Jan-Henrik Mayer4, Heinz Graafsma4, Michael Matysek5, Sebastian Schorb6, Daniela Rupp6, Marcus Adolph6, Tais Gorkhover6, Christoph Bostedt7, John Bozek7, Marc Messerschmidt7, Joachim Schulz4, Lars Gumprecht4, Andrew Aquila4, Nicola Coppola4, Frank Filsinger8, Kai-Uwe Kühnel9, Christian Kaiser9, Claus-Dieter Schröter9, Robert Moshammer9, Faton Krasniqi1, Simone Techert1,10, Georg Weidenspointer2, Robert L. Shoeman11, Ilme Schlichting1,11, Lothar Strüder1, 2, Joachim Ullrich1,9

We have performed experiments in the CAMP instrument

Experiments were performed in the CAMP chamber on the AMO beamline at LCLS

Imaging experiments at LCLS generate large volumes of data

Data rate:4 MB/image430 GB/hr (30 Hz)1.7 TB/hr (120 Hz)288 hard drives

Feb 2011: 120 Hz200 TB data...>20,000,000 images

Max-Planck-Institutfür Kernphysik

High-speed area detectors are essential for LCLS imaging experiments

The pnCCD enables direct detection at X-ray wavelengths;central hole allows the direct beam to pass through

1024

pixe

l, 7.8

cm

512 pixel, 3.7 cm

Area

: 29.

6 cm

2

1024x1024 pixels

pixel size: 75 x 75 !m"

active area 60 cm"

frame rate 125-900 Hz

single-photon resolution

!E=50/80eV @ 800/2000eV

Q.E. # 90 % from 0.4 to 10 keV

operating range 0.1 < E < 24 keV

The LCLS beam easily cuts through stainless steel

An accidental direct hit on the pnCCD at full power drilled straight through the detector

~25 µm~100 µm

Georg Weidenspointer, HLL Proc. SPIE 8070 (May 2011)

Amazingly, most of the detector remained useable despite the direct hit

0 200 400 600 800 1000Pixel x Coordinate

600

700

800

900

1000

Pixe

l y C

oord

inat

e

0 200 400 600 800 1000Pixel x Coordinate

600

700

800

900

1000

Pixe

l y C

oord

inat

e

Figure 2. Illustration of the pixels that had to be masked in the upper pnCCD of the front detector after the focused FELhad burned a hole into it. The dashed line indicates the dividing line between the two halves of the pnCCD, which arenot physically, but electrically separated. As indicated by the arrows, in the left (right) half of the pnCCD signal chargeis transferred to left (right) to the readout anodes. These are located at the left and right edges of the pnCCD as orientedin the figure. The semi-circle at the bottom center indicates a manufactured hole in the pnCCD through which theundi!racted FEL beam is supposed to pass. The burned-in hole is approximately at pixel coordinates (493,617). Becauseof the extremely strong charge generation around the hole, about 13 rows (running vertically in the figure) intersectingthe vicinity of the hole were flooded with charge and had to be masked. For the same reason, about 6 channels (runninghorizontally in the figure) around the hole had to be masked. In addition, because the burned-in hole severed electricalcontacts, charge transfer did no longer work for pixels in the bottom right corner of the left half of the pnCCD.

registers resulted in short circuits. Therefore pixels around the hole, and some rows and columns intersectingthe vicinity of the hole, could no longer be used for analysis and had to be masked, as illustrated in Fig. 2.

In addition, pixels in a region in the bottom right corner of the left half of the top pnCCD (see Fig. 2) couldno longer be read out. This is due to the fact that the burned-in hole is located in a region of the pnCCD in whichthe registers are electrically connected only to one side of the sensor (the top in Fig. 2); no electrical contactsexist to the bottom side in the region of the pre-manufactured half-hole (pixel x coordiantes 482–541). Thereforethe burned-in hole servered the electrical contact to registers below it (pixel y coordinates less than 617), whichmeans that charge could no longer be transferred across these registers and hence pixels in the indicated regioncould no longer be read out. In total, a fraction of about 2% of the pixels in the hit pnCCD was lost for analysis.In addition, the very large charge generation around the burned-in hole resulted in a small overall increase ofthe detector noise and hence slightly worsened the spectral resolution by several percent.

2.2 Impact of high speed ??? hot ??? particles on CCDOn Jun. 10, 2010, in preparation of an experiment, the position of a newly built steel shroud around the waterjet from a hydrodynamic sample injector was being adjusted. The shroud is equipped with a pre-manufacturedhole which permits the focused FEL beam to intersect the samples in the water jet. When scanning for the holein the shroud with the focused FEL, over the course of about an hour the ??? unattenuated ??? X-ray beamwas repeatedly focused on the shroud, rapidly burning each time small holes into the steel as depicted in Fig. 3.This process resulted in the ejection of tiny steel particles, some of which hit the pnCCDs of the front detectorwith su!cient energy to penetrate their entrance window. ??? kinetic energy or rather hot particles ???.

The impact of these high speed ??? or hot ??? particles in the bulk of the pnCCDs resulted in the creation ofso-called bright pixels, i.e. pixels with a strongly elevated signal level due to internal charge generation. Usually,bright pixels also exhibit an elevated noise level. The damage from high speed particles appears as point likedefects in the sense that the damage is limited to individual pixels. The accidentially created bright pixels couldno longer be used for scientific analysis and consequently had to be masked. In increase in the number of brightpixels is illustrated in Fig. 4. Before the impact of high speed particles, only 6 bright pixels had been identifiedon the front detector. After about half an hour of nozzle adjustment, an additional 20 bright pixels had been

Georg Weidenspointer, Robert Hartmann, MPG-HLL Proc SPIE 8070 (May 2011)

LCLS pulses: !! 10 to 300 fs, <3 mJ! 1.8 keV (0.68 nm)Focused to 7 μm: ! >1016 W/cm2

! 900 J/cm2

Expect single-shot resolution of 2 nm with 1019"W/cm2.

Janos Hajdu, Uppsala UniversityCNRS Marseille

Single particles at 20 nm resolution

1/d (nm)

0

1/(19 nm)

1/(19 nm)

Clear diffraction is measured from individual mimivirus

25 µ

m Ø

PA

RTI

CLE

BE

AM

We achieved an unprecedented 43% hit rate using an aerosol injector

Sample consumption: ~ 2 µl /min Sample concentration: !1012 particles/ml

1.2 MILLION HITS on viruses in ~36 hours of beam time(7.7% assuming 100% up-time)

Janos HajduUppsala

Mimivirus diffraction from LCLS reconstructs in 2D

Diffraction from individual Mimivirus can easily saturate the detector at low resolution

Janos Hajdu, Filipe Maia, Thomas Ekberg - Uppsala

Saturation = missing data

1024

pixe

l, 7.8

cm

512 pixel, 3.7 cm

Area

: 29.

6 cm

2

pnCCD

Weak signal at edges

High intensity regions can overload readout channels

Strange drop in signal

Resolution is limited by dynamic range when objects must be imaged with a single shot

1 10 100 1000Resolution q(cycles/micron)

10-2

100

102

104

106

Aver

age

coun

ts (A

U)

0.1µm object

1µm object

33nm

Attenuate with graded filters?

Scattered signal falls off as q-4

We can assemble individual snapshots in 3D

200 µm

Rear pnCCDz=564 mm

Front pnCCDz=68 mm

LCLS beam

Interaction point

We performed protein nanocrystallography at room temperature in a flowing water microjet

Liquid jet

The cspad pixel array detector was almost completely populated

120 Hz1 week>20,000,000 images 200 TB data...

https://confluence.slac.stanford.edu/display/PCDS/CSPad+detector

Single photon events produce about 8 ADU counts (at 8 keV)

histo0Entries 574240

Mean 8.827

RMS 3.023

ADU-10 -5 0 5 10 15 20 25 300

10000

20000

30000

40000

50000

singlePhoton gain q1, run 259 histo0Entries 574240

Mean 8.827RMS 3.023

/ ndf 2! 1.338e+05 / 97

Prob 0Constant 1.028e+02± 3.772e+04

Mean 0.006± 9.141

Sigma 0.004± 1.812

singlePhoton gain q1, run 259

Design specification was 24 ADU counts per photon

Philip Hart, SLAChttps://confluence.slac.stanford.edu/display/PCDS/CSPad+detector

The offset on each pixel fluctuates over time

Mean = 1213 ADUStd dev = 4.2 ADU3-sigma = 12.7 ADU

Distribution of dark pixel values looks roughly like a normal distribution

Mean = 1213 ADUStd dev = 4.2 ADU3-sigma = 12.7 ADU

Offsets on individual ASICS vary with total signal (!)

The first hard X-ray nanocrystal experiments were performed in the CXI instrument in February 2011

Submicron water jets are produced using a gas dynamic virtual nozzle

Neutral drops do not disperseDroplets triggered by piezo Flow alignment possibleWater acts as a tamperSub-micron drops achievable

Dan DePonte (CFEL), Bruce Doak, Uwe Weierstall, John Spence (ASU)

! "! #! $! %!!!

&'!(!!!

")!(!!!

#*!(!!!

$+!(!!!

"!

"&

#!

,-./0123425067825012569901:

;2342569901:82<

=9>2.31029>6:2,

250678,-.

/012342569901:8

Average over entire run

Best 30 minutes of data

Figure 2

A

B

¯

¯

¯

¯

¯

¯¯

¯¯

¯

¯ ¯

¯ ¯

¯

¯ ¯

¯

¯

¯

¯

¯¯

¯

¯¯

¯

¯¯

¯

¯

¯

¯

¯¯

¯¯

¯

¯¯

¯¯

¯

¯

¯

¯

¯¯

¯ ¯¯

¯¯

¯

¯

¯¯¯¯

3 5 04 6 05 7 06 8 0

7 9 0

11 3 1

10 2 1

9 1 18 0 1

11 0 110 0 1

12 0 1

9 2 18 1 1

10 21 4

15 26 7

19 19 718 22 7

16 21 617 19 6

15 19 514 21 5

13 17 414 14 4

18 14 618 15 6

20 14 7

19 9 7

14 8 414 2 5

15 4 5

14 2 6

20 6 8

21 5 9

23 9 10

21 11 8

14 9 4

12 1 1

11 11 0 10 15 1

7 13 1

3 13 24 14 2

7 16 2

15 20 2 14 22 3

14 24 4

2 9 1

21 17 1

16 17 1

13 12 0

19 12 0

19 2 0

19 10 0

21 11 022 10 0

12 3 113 3 113 2 1

22 9 022 7 022 6 0

¯¯

¯¯¯¯¯

¯¯ ¯

¯

¯

¯

¯¯

¯¯

¯¯¯¯¯

¯¯

¯¯

¯¯

¯

¯¯

Upper front detector

Lower front detector

Peaks can saturate detectorMust resolve spots

(no spillover)

Nanocrystal diffraction gives rise to separated bright peaks, which must be distinguished and quantified

Sum of all frames is dominated by water ring background

LCLS pulses: 2,292,468

Acquisition time: 19,103 sec(5 hr 18 min)

Photon energy: 9.4 keV

Ice gives rise to strong diffraction peaks on the detector

FEL pulses: 4,293

Acquisition time: 35 seconds

Photon energy: 9.4 keV

35 sec of ice deliveredroughly the same local dose as30 minutes of data collection

Dead pixels accumulate during the course of the experiment

Strong diffraction from accidentally forming ice can be very damaging to the detector

1/20 actual speed

Death of an ASICFrame 1/4

Frame 1: Feb21_r0427_151008_c4a4


Frame 2:Feb21_r0427_151008_c4a7


Frame 3:Feb21_r0427_151008_c4aa


Frame 4:Feb21_r0427_151008_c4ad

Accidents can happen: radiation dose event whilst moving hardware in the chamber

Our processing pipeline is an exercise in data volume reduction

Detector Storage Offline reduction Analysis

> 20,000,000 frames

Store all data(no corrections)

120 Hz4 MB/event1.7 TB/hr200TB/expt(5 days)

< 1,000,000 frames

Retain only ‘hits’(detector corrected)

<100,000 frames

Automated high volume image processing is essential(reliable background correction, automatic identification of useful data)

Science analysis

My data must be somewhere here....

Data Direct Networks SFA1000060-bay HDD enclosures in 4U format ~1.4 PB formatted per rack (600 x 3 TB HDDs)

600x 3TB hard drives

Scalable to over 13,440 HDDs(over 10,000 TB formatted capacity)

Combine 105-107 measurements

Classification Averaging Orientation Reconstruction

Diffraction pattern(low signal = noisy)

10 fs FEL pulse

Particle injection

One pulse per diffraction pattern

X-ray free-electron lasers may enable atomic-resolution imaging of macromolecules without the need to grow large crystals

0 100 200 300 400 500Pixels on detector

10-4

10-2

100

102

104

Aver

age

phot

on c

ount

12 keV8.2 keV4.1 keV2.5 keV1.7 keV820 eV

Moving to longer wavelengths increases the number of detected photons at the expense of spatial resolution

12 keV(1 Å)

(1,500 photons total)

8.2 keV(1.5 Å)


4.1 keV(3 Å)


2.5 keV(5 Å)


1.7 keV(7 Å)


820 eV(15 Å)


Anthrax lethal factor (1YQY)12924 atoms, mol.wt. = 86,283 Da

LCLS configuration1012 X-rays in 100nm square760x760x110µm pixel detectorplaced 50mm from focus

Ultrafast coherent imaging requires integrating detectors that can read out a full frame on each pulse

Heterogeneousobjects

Single moleculesviruses, etc

Protein nanocrystals

Reconstruct unique objects Average weak signal Index Bragg peaks

Very weak:Must average many shots

Single photon discrimination

No averaging:All data in a single shot

High dynamic range

Bright, isolated peaksHigh dynamic range

6-12 keV2 - 8 keV500 eV - 2 keV

Our ideal detector must satisfy many constraints

Property Why?Pixel arrays 2k x 2k or more(1k x 1k minimum)

Need to resolve fine diffraction features

Actual pixel size not critical Detector must fit in facility or vacuum vessel

Readout at facility repetition rate(LCLS: 120 Hz, XFEL: ~3000/bunch)

Each pulse creates a unique event

Dynamic range >104 Highly varying signal intensity

Low noise, single photon detection Signals can be weak at high resolution

Stable pixel positions (<1/10 pixel) Location of peaks need to be well defined

Photon integrating (not photon counting) Multiple photons/pixel all come in <100 fs

Saturation is well controlled Need to separate adjacent strong peaks

Correctable and well characterised artifacts Robust background subtraction essential

Reliable, works when needed Beamtime is very expensive

Replaceable modules Radiation damage is a concern

This work was the product of a large international team

Henry N. Chapman1,2, Petra Fromme3, Anton Barty1, Thomas A. White1, Richard A. Kirian4, Andrew Aquila1, Mark S. Hunter3, Joachim Schulz1, Daniel P. DePonte1, Uwe Weierstall4, R. Bruce Doak4, Filipe R.N.C. Maia5, Andrew Martin1,

Ilme Schlichting6,7, Lukas Lomb7, Nicola Coppola1, Robert L. Shoeman7, Sascha Epp6,8, Robert Hartmann9, Daniel Rolles6,7, Artem Rudenko6,8, Lutz Foucar6,7, Nils Kimmel10, Georg Weidenspointner11,10, Peter Holl9, Mengning Liang1,

Miriam Barthelmess12, Carl Caleman1, Sébastien Boutet13, Michael J. Bogan14, Jacek Krzywinski13, Christoph Bostedt13, Sa!a Bajt12, Lars Gumprecht1, Benedikt Rudek6,8, Benjamin Erk6,8, Carlo Schmidt6,8, André Hömke6,8,

Christian Reich9, Daniel Pietschner10, Lothar Strüder6,10, Günther Hauser10, Hubert Gorke15, Joachim Ullrich6,8, Sven Herrmann10, Gerhard Schaller10, Florian Schopper10, Heike Soltau9, Kai-Uwe Kühnel8, Marc Messerschmidt13, John D. Bozek13, Stefan P. Hau-Riege16, Matthias Frank16, Christina Y. Hampton14, Raymond Sierra14, Dmitri Starodub14, Garth

J. Williams13, Janos Hajdu5, Nicusor Timneanu5, M. Marvin Seibert5, Jakob Andreasson5, Andrea Rocker5, Olof Jönsson5, Stephan Stern1, Karol Nass2, Robert Andritschke10, Claus-Dieter Schröter8, Faton Krasniqi6,7, Mario Bott7,

Kevin E. Schmidt4, Xiaoyu Wang4, Ingo Grotjohann3, James Holton17, Stefano Marchesini17, Sebastian Schorb18, Daniela Rupp18, Marcus Adolph18, Tais Gorkhover18, Martin Svenda5, Helmut Hirsemann12, Guillaume Potdevin12,

Heinz Graafsma12, Björn Nilsson12, and John C. H. Spence4

1. Center for Free-Electron Laser Science, DESY, Notkestrasse 85, 22607 Hamburg, Germany.2. University of Hamburg, Luruper Chaussee 149, 22761 Hamburg, Germany.3. Department of Chemistry and Biochemistry, Arizona State University, Tempe, Arizona 85287-1604 USA.4. Department of Physics, Arizona State University, Tempe, Arizona 85287 USA.5. Laboratory of Molecular Biophysics, Department of Cell and Molecular Biology, Uppsala University, Husargatan 3 (Box 596), SE-751 24 Uppsala, Sweden.6. Max Planck Advanced Study Group, Center for Free Electron Laser Science (CFEL), Notkestrasse 85, 22607 Hamburg, Germany.7. Max-Planck-Institut für medizinische Forschung, Jahnstr. 29, 69120 Heidelberg, Germany.8. Max-Planck-Institut für Kernphysik, Saupfercheckweg 1, 69117 Heidelberg, Germany.9. PNSensor GmbH, Otto-Hahn-Ring 6, 81739 München, Germany.10. Max-Planck-Institut Halbleiterlabor, Otto-Hahn-Ring 6, 81739 München, Germany.11. Max-Planck-Institut für extraterrestrische Physik, Giessenbachstrasse, 85741 Garching, Germany.12. Photon Science, DESY, Notkestrasse 85, 22607 Hamburg, Germany.

13. LCLS, SLAC National Accelerator Laboratory, 2575 Sand Hill Road. Menlo Park, CA 94025, USA.14. PULSE Institute and SLAC National Accelerator Laboratory, 2575 Sand Hill Road. Menlo Park, CA 94025, USA.15. Forschungszentrum Jülich, Institut ZEL, 52425 Jülich, Germany.16. Lawrence Livermore National Laboratory, 7000 East Avenue, Mail Stop L-211, Livermore, CA 94551, USA.17. Advanced Light Source, Lawrence Berkeley National Laboratory, Berkeley, California 94720, USA.18. Institut für Optik und Atomare Physik, Technische Universität Berlin, Hardenbergstrasse 36, 10623 Berlin, Germany.

femtosecond serial imaging using fast integrating detectors · 2013. 5. 29. · j.mol.biol. 221,...

Documents