Dr Paul J. Ellis, Physicist, Engineering (Beamline Scientist), SSRL, Stanford University.
Brief summary of expertise and significant research achievements:
Dr Ellis has made important contributions in the fields of x-ray absorption
spectroscopy (XAFS) and macromolecular crystallography and been responsible
for significant advances in software, hardware and technique in addition to
conducting my own research.
He has focused particlularly on developing and disseminating
new techniques. This emphasis was a conscious choice arising out of a desire
to take advantage of my creativity and ability to see innovative solutions to
problems, my expertise in the areas of chemistry, education, mathematics,
software development, electronics, and mechanical design, and the reasoning
that while the determination of the geometry of a particular metal site by
x-ray absorption spectroscopy or of a particular protein structure by
x-ray crystallography could be very valuable, a new or improved technique
will facilitate many such investigations and make a much larger
contribution overall.
Through his association with SSRL, first as a visiting researcher during
graduate school and later as a member of staff, has allowed him to work
with many outstanding scientists, to take advantage of the most modern
technology, and interact with hundreds of visiting researchers each year,
discussing their needs, offering guidance from my own experience and
being guided by their insights in return.
Examples: Software development
Dr Ellis has been solely responsible for, or contributed to, the
development of many software packages used by scientific researchers
around the world. Two notable examples are the XFIT XAFS analysis program
and the CBFlib imgCIF/CBF library:
XFIT: XAFS analysis
The XFIT XAFS analysis package (Ellis and Freeman, 1995) was written as part
of my graduate research. XFIT incorporates the state of the art FEFF ab initio
single- and multiple-scattering calculations developed by Prof. John J. Rehr
at the University of Washington in a user-friendly interface providing:
(i) Fourier filtering of both experimental and calculated XAFS.
(ii) A generalised formalism for parameter constraints and restraints.
(iii) The ability to fit a model simultaneously to several (polarised
and/or unpolarised) XAFS data sets.
(iv) The ability to fit a model with more than one absorption site.
(v) Monte-Carlo error analysis.
Other programs in the package included AVERAGE for averaging multiple
absorption scans, SPLINE for extracting the XAFS from averaged absorption
data, and GRAPH for drawing publication graphs and removing monochromator
glitches.
After completing graduate work, the package was rewritten under contract to the Australian Nuclear Science and Technology Organization (ANSTO) and Sydney University and ownership was transferred to the Australian Synchrotron Research Program (ASRP). The package has been used by more than 40 groups in Australia, England, Germany, Holland, Italy, Japan, Korea, New Zealand, Poland, Spain and the United States of America and has been cited in more than 50 publications.
imgCIF/CBF library
In the late 1990's, the crystallographic community was encountering several
serious problems related to the large volumes of image data being generated.
In particular, there was no generally-accepted standard for image files,
leading to a proliferation of proprietary formats; such formats as were
in use were inflexible in the types of associated information that
could be accommodated, and often did not take advantage of compression
to make most efficient use of finite disk space and archive systems.
In October 1997, 23 scientists representing universities, research laboratories and private industry, including several detector manufacturers, in the United States and Europe met at the Brookhaven National Laboratory to address these problems. At this workshop, it was decided to define a model for efficient storage of area detector data and other large datasets based on the standard means of information interchange in crystallography, the crystallographic information file (CIF) format maintained by the International Union of Crystallography (IUCr).
Following three years of work by Dr Ellis and the other members of the working group, version 1.0 of the Image-supporting CIF (imgCIF) standard was approved by the IUCr Committee for the Maintenance of the CIF Standard (COMCIFS) and published in December 2000. Incremental development of this first version is continuing and a new standard is expected in 2003.
Although the creation of the standard was a large step, real success demands that it be actually implemented in software and incorporated into data collection and analysis. Given the complexity of imgCIF, without help in this area this would probably have proven an impossible burden for most software developers, particularly those not in a corporate environment.
The imgCIF/CBF library was developed by Prof. Herbert J. Bernstein (Dowling College) and Paul Ellis, concurrently with the development of the standard. The library consists of a set of approximately 100 ANSI C functions useful for creating and interpreting files in the imgCIF or the closely related crystallographic binary file (CBF) formats. It:
(i) is the only existing publicly available interface to the
imgCIF and CBF formats.
(ii) dramatically reduces the effort required to create and read
imgCIF/CBF format files.
(iii) has been incorporated into a variety of data collection and
processing programs used worldwide by macromolecular crystallographers,
and thus forms a vital component of the imgCIF initiative.
imgCIF/CBF library in MOSFLM and HKL2000
MOSFLM and HKL2000 are programs for processing crystallographic
image data to produce the integrated intensities used in subsequent
analysis. MOSFLM was developed by Andrew Leslie at Cambridge University
and is available free of charge. HKL2000 is a commercial package
developed by HKL Research Inc. Together, these are the programs of
choice for the great majority of macromolecular crystallographers
in the United States.
Harry Powell (Cambridge University) and Dr Ellis have demonstrated incorporation of imgCIF/CBF format into the version of MOSFLM in use at SSRL (and are working to make similar changes to the latest development version for distribution to researchers around the world) and Dr Ellis is working with HKL Research Inc. to facilitate similar enhancement of HKL2000.
These developments will remove a major impediment to the acceptance of imgCIF/CBF in the crystallographic community, and starting with the 2002/2003 SSRL experimental run visiting experimenters will be able to collect their image data in imgCIF/CBF format, benefiting not only from much more complete and accessible experimental information, but also from a reduction in storage requirements by a factor of 3.
Other software projects include: CENTER: beam-center determination
Indexing a set of diffraction images, the first step in reducing raw data to integrated intensities, generally relies on accurate direct-beam coordinates (the point at which the x-ray beam would pass through the plane of the detector if there were no beam-stop). Although these coordinates can be measured by removing the beam-stop, this can be hazardous for the detector. Of course, in situations where the beam does not intersect the active area, even this approach cannot be used.
The CENTER program was developed to safely, rapidly and reliably determine these coordinates from a powder diffraction pattern, typically obtained by collecting an image with a sample of polyethylene or powdered silicon. It has been used routinely at SSRL by both staff and users since 1998.
SSRL data reduction scripts
With Dr. Aina Cohen (SSRL), Dr Ellis developed a set of shell scripts
greatly facilitating data reduction using the programs MOSFLM, SCALA
and TRUNCATE.
These scripts have been available at the SSRL crystallography beamlines
since 1998 and have been successfully used by both staff and hundreds
of visiting researchers.
BLU-ICE/DCS: beamline control and data collection.
BLU-ICE/DCS is the beamline control and crystallographic data collection
software developed under the direction of Dr. Timothy McPhillips at SSRL.
It consists of two parts: (1) the Distributed Control System (DCS),
comprising a central server process communicating with one or more
low-level processes controlling individual beamline hardware components
and possibly executing on highly heterogeneous systems, and (2) one
or more instances of the graphical interface (BLU-ICE).
Dr Ellis hascontributed several components to this system, notably:
(i) An optimized CCD image transform process used to apply
geometric and non-uniformity corrections to images collected with
the ASDC Quantum4 detectors.
(ii) Various alignment procedures, including automated table
alignment and automated energy calibration.
(iii) Fluorescence energy selection in XAS scans for improved
signal/noise and a facility for measuring the x-ray emission spectra
of samples for the identification of potential heavy atom derivatives.
(iv) With Dr. Aina Cohen, functions for calculating foil attenuation as
a function of energy or (inversely) calculating the combination of foils
yielding a desired attenuation at a given energy. These functions are
used both for manual attenuation of the beam and as part of a simple
one?button fluorescence scanning procedure.
FIRST: data reduction beyond 3D profile fitting
With Dr. Aina Cohen (SSRL), Dr Ellis is currently developing a new data reduction
package based on global optimization and capable of fully exploiting both
fine- and coarse-sliced data as well as problematic images not readily
analyzed with current programs. Common examples include:
(i) Images with non-ideal spot shapes. One common situation resulting
in poor spot shapes is crystal cracking.
(ii) Images with overlapping spots. Also a very common problem, this
can be the result of high mosaicity, large unit cells, or positioning
the detector too close to the sample. Like the similar problems associated
with processing Laue data, this situation can often be addressed with
statistical methods.
(iii) Images with spots from multiple crystals. This can result from
twinning, severe cracking, or simply from inter-grown crystals. The ideal
program would be capable of interpreting the data as a combination of
multiple lattices. In many cases, this would also involve the problem of
overlapped spots mentioned earlier.
Current software packages generally refine parameters against data recorded on one or a few images, producing a set of integrated intensities, either one for each spot on each image or one for each individual reflection, possibly corresponding to spots on several successive images. After the parameters are refined, a scaling model is used to merge the data, producing a set of final intensities corresponding to the unique reflections.
The new approach is based on the idea of treating both the final merged intensities and the scaling factors as parameters in a single model, and all of the images as comprising a single data set from which the parameters of the model can be estimated. Because each unique intensity is represented just once, the number of real parameters is minimized and because all the parameters are estimated in a single global refinement, the bias introduced by the assumption of parameter independence is eliminated.
The single model/single data set scheme also greatly simplifies the process of estimating the model parameters. Conceptually, this would take the form of a single maximization problem, albeit in a very high-dimensional space, at least the size of the number of unique reflections. In practice, because most reflections are well separated from each other minimizing correlations between the parameters for one reflection and those for another, this process should converge rapidly.
Energy tracking on synchrotron sidestations:
As phasing methods relying on the anomalous signal (SAD, MAD, SIRAS,
MIRAS) have become more widely used, to the point now where SAD and
MAD are the methods of choice, the demand for beamlines allowing a
user to easily control the x-ray beam energy has increased dramatically.
Dr. Aina Cohen and Paul Ellis have implemented software on the formerly
monochromatic SSRL stations 9-1 and 11-1 permitting users to arbitrarily select
any x-ray photon energy in the range 12500 eV to 16500 eV on 9-1 and
10500 eV to 15000 eV on 11-1. With the addition of fluorescence detectors
early in 2002, this permitted experiments to be carried out on the
single-crystal monochromator sidestations 9-1 and 11-1 that were formerly
restricted to the heavily oversubscribed double-crystal monochromator
endstation 9-2.
3. Hardware development
In collaboration with Dr. Aina Cohen and others at SSRL, Ellis has been
responsible for the conceptual design of new beamlines and the detailed
design and fabrication of many of the components now in use both at
SSRL and other synchrotrons. Particular examples are:
Sample handling robot
With Prof. Paul Phizackerley (SSRL) and Dr. Aina Cohen (SSRL), Ellis was one
of the principal designers of the robotic sample handling systems now
operational at beamlines 1-5, 9-1, 9-2 and 11-1. This robot transfers
crystals mounted on standard Hampton copper pins between a cylindrical
cassette in a liquid N2 dewar and the magnetic post on the goniometer.
As the system is based on a small and inexpensive industrial robot and
a standard storage dewar, has no actuators or electronics under liquid
N2, and doesn't require either that the pin be rotated from vertical to
horizontal, or that the cold-stream or beam-conditioning system be
moved for mounting, it is relatively cheap and has proven very reliable.
The robot dramatically reduces the time required to mount and dismount prefrozen samples at a beamline and is a critical component of the Structure Determination Core program of the Joint Center for Structural Genomics (JCSG) and during the 2001/2002 SSRL experimental run this group made extensive use of the prototype system at beamline 11-1. During the 2002/2003 experimental run, refined systems at beamlines 1-5, 9-1, 9-2 and 11-1 will also be made available to visiting researchers.
PIN diode beam stop
With Dr. Aina Cohen (SSRL), Dr Ellis designed and fabricated a new beamstop
incorporating a miniature Si PIN diode x?ray detector. This beamstop
was used throughout on the 2000/2001 SSRL run on beamline 9?1 and in
combination with a motorized xyz beamstop positioner greatly
facilitated alignment. Based on this experience and with the aim of
reducing the diameter of the beamstop from 2.0 mm to 1.5 mm or
below, a simplified second-generation device was fabricated and
installed on beamline 11-1 late in the run where it proved invaluable
in monitoring beam stability during the tests of a new detector.
This new device incorporates several innovations, notably the use
of a PC board as the main structural element and the substitution
of a tungsten/plastic composite for the solid tungsten cup. These
changes have made it dramatically easier to fabricate, reduced the
minimum diameter to 1.2 mm, and reduced cost by a factor of 5. The
new device is now the standard beamstop at the SSRL macromolecular
crystallography beamlines and will also be used in the polarized XAFS
program. As several groups at other beamlines, both in the United
States and Europe have also expressed interest in this technology,
a paper describing its construction and use was prepared. This
paper has been accepted for publication by the Journal of
Synchrotron Radiation.
Energy-sensitive fluorescence detector system
Dr Ellis designed and implemented both the SCA-based energy-sensitive
fluorescence detector system used during the 1999/2000 and 2000/2001
experimental runs on beamline 9-2 and a new ethernet MCA-based
system first now in use at SSRL. As well as enhanced control of
the shaping and analysis parameters over the present arrangement,
the new system allows fluorescence emission spectra to be recorded
in only seconds rather than minutes and thus facilitates rapid
screening for heavy atom derivatives. This new system has now been
implemented on beamlines 1-5, 9-1, 9-2, 11-1 and 11-3.
In a related development, Dr. Aina Cohen (SSRL) and Dr Ellis also designed and fabricated a motorized fluorescence detector positioner able to move the fluorescence detector toward the sample position during a scan and away from the sample for safety when not in use. Together with the computer-controlled MCA, these were the final hardware components required for fully automated rapid XAS and x-ray emission scanning.
Next-generation modular beam definition system
The beam-definition system presently used on beamlines 1-5, 9-1,
9-2 and 11-1 consists of a series of foils, ion-chambers,
motorized slits, a shutter and final guard shield filled with He
and separated by O-ring seals. Unfortunately, this system has two
significant problems: (1) it is bulky, obstructing the space in
the vicinity of the sample and restricting the range of the
kappa goniometer; (2) it is very difficult to seal effectively
as some of the components, particularly the slits, were never
designed to prevent gas leakage.
Dr. Aina Cohen and I are presently working on the design of a new compact He-tight beam-definition/shutter system. By replacing the large 1.6" square stepping motors and 0.63" diameter DC motors presently used on the slits and guard shield with 0.39" diameter stepper and 0.20" diameter permanent-magnet DC motors, we have reduced the size of the components to the point where they can be enclosed in a single small He-tight box. As this change also substantially reduces the motor power requirements, we plan to locate the drive circuits inside the box also and control them using FPGA-based embedded processors. This will allow us to replace the 12 bulky cables presently joining the front-end to the equipment racks with a single thin ethernet link.
The design of most of the mechanical and electronic components is in an advanced stage and acquisition of parts and software for mechanical and electronics prototyping has begun. We anticipate that a working prototype will be available for testing with beam in March 2003. If successful, the system will be replicated on all beamlines during the long 2003/2004 SPEAR3 shutdown.
As a member of the crystallography group at SSRL since 1997, I have also assisted in the construction, alignment, testing and operation of the crystallographic beamlines 1?5, 7?1, 9?1, 9?2 and 11?1. Typical responsibilities have been:
Installation and upgrades of crystallographic software
As the coordinator of the crystallographic software group, I have
been responsible for the acquisition, installation and maintenance
of the crystallographic software used by both staff and visiting
experimenters. Packages presently in use include ARP/wARP, CCP4,
CNS, DENZO/HKL2000, MOSFLM, O, PHASES, SHARP, SHELX, SnB,
SOLVE/RESOLVE and XTALVIEW.
Acquisition, installation and commissioning of experimental hardware
As a member of the beamline development group, I have been responsible
for the procurement, installation and commissioning of a variety of
major beamline components acquired from outside vendors or developed
by other groups within SSRL. Examples include area detectors from
MAR Research and Area Detector Systems Corporation, fluorescence
detectors from Eurisys and Amptek and sample cooling systems from
Oxford Instruments and Oxford Cryosystems.
