Development of 3D microscopy of biological specimens using soft X-ray diffraction

1
Development of 3D microscopy of biological
specimens using soft X-ray diffraction

• David Sayre, Chris Jacobsen, Janos Kirz
• David Shapiro, Tobias Beetz, Enju Lima
• Stony Brook University
• Malcolm Howells
• Advanced Light Source

2
What is coherent X-ray diffraction imaging good
for?
3
Atomic resolution imaging electrons or photons?

• 100 keV electrons
• About 2.5 inelastic scatters per elastic scatter
• About 45 eV deposited per inelastic scatter
• Therefore about 102 eV deposited per elastic
  scatter
• A hundred scattered electrons 102102 eV into (2
  Å)3, or 2?1011 Gray

• 10 keV photons
• About 10 absorption events per elastic scatter
• About 10 keV deposited per absorption
• Therefore about 105 eV deposited per elastic
  scatter
• A hundred scattered photons 102 105 eV into (2
  Å)3, or 2?1014 Gray

• Electrons are better than photons for atomic
  resolution imaging
• J. Breedlove and G. Trammel, Science 170,
  1310 (1970) R. Henderson, Q. Rev. Biophys. 28,
  171 (1995).
• Crystallographys answer spread the dose out
  over many identical unit cells

4
Johnson Blundell, JSR 1999

• At DESY, a 500 GeV colliding-beam TESLA project
  plans to produce a free-electron laser (FEL) of
  breathtaking brilliance for 1 A X-rays. A
  totally new science has to be explored. What
  possibilities could this open for structural
  biology?
• With the extremely bright source it may be
  possible to escape the benevolent tyranny of the
  crystal and record molecular transforms from
  individual molecules. Studies by Miao,
  Charalambous, Kirz Sayre (private
  communication) have shown that the soft-X-ray
  molecular transform of a micrometre-size
  non-crystalline specimen can be inverted to form
  an image in which the phase problem is overcome
  by over-sampling the diffraction pattern and use
  of an iterative algorithm. In order to record a
  molecular transform, would the molecule need to
  be tethered in order to localize it sufficiently
  or would a spray technique prove possible?
  ....Finally, will the biological molecule
  withstand such a bright beam?

5
Thick samples, lower resolutionphotons come out
ahead
X-rays better for thicker specimens. Sayre et
al., Science 196, 1339 (1977) Schmahl Rudolph
(1990). Electrons better for thin!
This plot Jacobsen, Medenwaldt, and Williams, in
X-ray Microscopy Spectromicroscopy (Springer,
1998)
6
The role of coherent X-ray diffraction

• A technique for 3D imaging of 0.5 20 µm
  isolated objects
• Too thick for EM (0.5 µm is practical upper
  limit)
• Too thick for tomographic soft X-ray microscopy
  (depth of focus lt 1 µm at 10 nm resolution for
  soft X-rays even if lenses become available)

• Goals
• 10 nm resolution (3D) in 1 - 10µm size
  biological specimens
• (small frozen hydrated cell, organelle see
  macromolecular aggregates)
• Limitation radiation damage!
• lt4 nm resolution in less sensitive
  nanostructures
• (Inclusions, porosity, clusters, composite
  nanostructures, aerosols)
• eg molecular sieves, catalysts, crack
  propagation

7
Imaging beyond the lens limit?

• Working from diffraction patterns
• No optics-imposed resolution limits!
• Working from images
• The lens phases all the Fourier plane
  information.
• The lens limits resolution.

• Holography mix in a reference phase

8
Where we really want to be

• Collect a high resolution 3D data set in an hour
  or two
• Reconstruct reliably in a comparable amount of
  time

9
Challenges 1/ getting sufficient coherent
photons
Currently we run at NSLS X1B. Designed for
spectroscopy. Poor use of brightness This fall
we go to the ALS, where 9.0.1 delivers 1000 times
more coherent photons over restricted energy
range. After upgrade both ALS and NSLS will
deliver at least 1000 times more yet!
(assuming custom beamline)

2/Choice of energy - Reconstruction of
real objects is easier. - Electron density
becomes more nearly real as energy increases. -
Exposure time scales as E4 . (Coherent flux for
fixed brightness scales as E-2, and cross
section scales as E-2) - Best compromise
specimen dependent 0.5 4 KeV
10
ALS Brightness Now and After Upgrade
In-vacuum IDs
Superconducting ID
11
NSLS II Design (J. Murphy, 6/20/03)
12
Challenges 3/ recording the pattern

• Eliminate higher orders aperiodic undulator?
• Shielding detector from all but diffracted signal
• Aligning specimen with small beam-spot,
• Keeping it aligned as specimen is rotated
• Minimizing missing data
• (beam stop, large rotation angles, etc.)
• Dynamic range of detector
• Automation of data collection
• Mapping recorded data onto cubic array

13
(No Transcript)
14
(No Transcript)
15
Challenges 4/reconstruction

• How to avoid stagnation local minima?
• The enantiomorph problem
• How to tell whether algorithm converged?
• (easy when object known)
• Multiple random starts
• How to make best use of the data?
• Of prior knowledge?
• How to optimize use of computer resources?
• Want 10243 DFT in 0.5 sec
• Much work remains to be done!

3D easier than 2D highly overdetermined
16
Model calculations (E. Lima)

• Jacobsen model of biological cell (complex)
• Success in 3D using difference map
• (V. Elser), loose support
• Much easier for high contrast (especially binary)
  specimens

17
Challenges 5/ damage

• The ultimate limitation for radiation-sensitive
  materials only
• Dose fractionation
• (Hegerl and Hoppe 1976, McEwen 1995)
• M. Howells - tomorrow

18
Preparation of Frozen-hydrated Fibroblasts

• Grids with live cells are
• Taken from culture medium and blotted
• Plunged into liquid ethane (cooled by liquid
  nitrogen)
• Loaded into cryo holder

19
Radiation damage resistance in cryo
Maser et al., J. Microscopy 197, 68 (2000)
Left frozen hydrated image after exposing
several regions to 1010 Gray
Right after warmup in microscope (eventually
freeze-dried) holes indicate irradiated regions!
20
PMMA at room, LN2 temperature

• T. Beetz, Stony Brook
• Repeated sequence dose (small step size, long
  dwell time), spectrum (defocused beam)
• Images dose region (small square) at end of
  sequence

Room temperature mass loss immediately visible
LN2 temperature no mass loss immediately visible
After warm-up mass loss becomes visible
21
PMMA at LN2, room temperatureXANES spectra

• T. Beetz, C. Jacobsen JSR May 2003
• Peak at 531.4 eV C0 bond
• Plateau at 540 eV total mass (plus some emphasis
  on oxygen ? bonds)

C0 peak
Plateau
C0 peak
Plateau
Liquid nitrogen temperature
Room temperature
22
Cryo protects PMMA against mass loss, but not
against chemical change!
23
Near-term future

• Stony Brook/NSLS chamber
• Incorporates low resolution X-ray microscope
• Provision for frozen hydrated samples
• Precision positioning and rotation of specimen,
  optics
• Our Goals
• Develop technique (recording and reconstruction)
• Study dose vs resolution damage limits
• 3D reconstruction of frozen hydrated dwarf yeast
  cell

24
Freeze-dried yeast cells
25
Diffraction pattern from freeze-dried yeast cell

• ln v I
• Combination of multiple exposures
• Edge of figure 45 nm
• pattern extends half way

26
Azimuthally averaged power spectrum

• Slope -3.4
• Dose 106 Gy

27
Soft X-ray 3D imaging using holographic
techniques

• Tobias Beetz
• Chris Jacobsen

28
Why hologaphy?

• Encode phase for cases where it's difficult to
  reconstruct
• Hologram recorded interference between a
  reference wave and a object wave
• H (R S)(R S) RR SR RS SS

29
In-line holography

• Resolution given by detector resolution (need to
  record fine fringes)
• photographic film/ photoresists (Howells et al.,
  Science 238, 514 (1987) Jacobsen et al., JOSA
  A7, 1847 (1990) Lindaas et al., JOSA A13, 1788
  (1996))
• read out holograms, digitize gtgt not feasible for
  3D data sets

30
In-line holograms and reconstructions

• 1 µm diameter latex spheres

• 3 µm diameter gold dot

hologram
hologram
31
Fourier transform holography

• resolution limited by the reference source size,
  not by detector
• use zone plate (Reuter and Mahr, J. Phys. E9, 746
  (1976) McNulty et al., Science 256, 1009 (1992))
• better for radiation sensitive specimens (low
  efficiency optics are in front of the specimen)
• combine FTH with diffraction to help with phasing
  the diffraction pattern

Experiments and computer simulations in progress!
32
Conclusions

• Diffraction based imaging is the method of choice
  for micron-size specimens
• Damage will set limit on resolution for
  radiation-sensitive specimens
• Much progress on 2D problems, 3D just starting
• New, bright synchrotron sources will be superb
  for this!
• Surely an exciting prospect!

33

• Phase and amplitude reconstruction in STXM using
  a segmented detector
• M. Feser and C. Jacobsen

34
Detector developmentsegmented, current mode
M. Feser, B. Hornberger, C. Jacobsen (Stony
Brook) P. Rehak, G. de Geronimo (BNL
Instrumentation) L. Strüder, P. Holl (MPI
München)

• Silicon drift detector
• Simultaneous recording of bright field, dark
  field, differential phase and interference
  contrast (Polack Joyeux)
• No significant upper limit to signal rate.
  Acceptable dark noise (8 photons/msec
  equivalent room temperature)
• High quantum efficiency (gt90)

Assembly 40 mm across Active area 600 ?m
35
Segmented silicon drift detector

• Corner of silicon nitride window silicon at 45
  wall slope forms a prism
• Refraction of x-ray beam in opposite direction
  from visible light prisms

X-ray refractive index
All channels acquired simultaneously
36
Amplitude And Phase Contrast Transfer Functions

• Concept developed by electron microscopy
  community Waddel et al., Optik 54 (1979)
• Description for x-ray microscopy Morrison et al.
  1992, former XRM proceedings
• New transfer functions for multisegmented
  detectors with circular geometry (i.e., not CCD
  pixels)

37
Contrast Transfer Functions Example

• The transfer functions depend on the illumination
  pupil function and the detector element geometry
• Amplitude transfer has even symmetry
• Phase transfer has odd symmetry
• Difference images of opposing detector segments
  have no amplitude transfer, only phase transfer
  ? Differential phase contrast

38
Simultaneous Availability Of Contrast Modes

• Silica spheres 1 mm diameter or less
• Differential phase contrast filters out intensity
  fluctuations of the source!

39
Presenting DPC data

• Color code phase gradient in images

1mm
DPC
Incoherent BF
Nice, but can be difficult to interpret
40
Reconstruction Filters

• They phase the detector images with respect to
  each other

41
Experiment Germanium Test Pattern

• Fabricated by S. Spector
• Production process of zone plates
• Quite absorbing, but b/d 0.346 _at_ 520eV (Henke,
  Gullikson, et al.)

40 nmhalf-period
STEM image
42
Unfiltered Incoherent Image

• Determined thickness of spokes to be ? 100 nm
  from absorption profile
• Expected phase shift is ? 40o in spokes
• Imaging zone plate 80mm diameter30nm outermost
  zones

43
Test Pattern Specimen Estimate

• Phase image shows finer details polymer resist
  left from production
• X-ray beam intensity fluctuations only transfer
  to amplitude!
• Measured ratio of d/b 0.3-0.4 compares well with
  tabulated data

Phase
Amplitude
44
Power Spectral Density
45
Conclusions

• STXM amplitude and phase recovery with a
  configured detector with good quantitative
  agreement for high spatial frequencies
• Spatial resolution at the diffraction limit of
  the 30 nm zone plate (structures of 20 nm clearly
  visible)
• Coherence a prerequisite for full spatial
  resolution!
• Imaging without absorption by tickling
  electrons below the carbon edge?

46
Acknowledgements

• ALS Malcolm Howells
• DOE support
• NIH support
• David Sayre, Chris Jacobsen, Michael Feser,
  Chi-Chang Kao, David Shapiro, Tobias Beetz, Onur
  Mentez, Enju Lima