The Velociprobe: An ultrafast hard X-ray nanoprobe for high-resolution ptychographic imaging

As a scanning version of coherent diffraction imaging (CDI), ptychography is able to image an extended sample at a high spatial resolution.^1–3 With the short wavelength and high penetration length of X-rays, ptychography has been a powerful tool for nondestructive characterization of sample’s structures with the state-of-the-art spatial resolution reaching sub-10 nm.^4,5 Similar to CDI, the resolution of ptychography is theoretically limited by the Q range of the scattered signals in the measured diffraction patterns rather than the spot size of the illumination X-ray beam and the scan step size. However, the achieved spatial resolution in practice is also limited by the scanning position accuracy even though some advanced position correction algorithms^6–8 can alleviate the effect of position errors at the cost of higher signal-to-noise ratio, more overlapping between scan points, and better coherence.

Therefore, high position accuracy is always being pursued in ptychography and other types of scanning microscopy. This pursuit of high position accuracy drives instrument designers to develop new types of stability- and accuracy-enhancing mechanics. A common approach is to operate in a high-vacuum environment for a more stable instrument and sample environment^9–12 although this introduces other difficulties with heat rejection and stage availability. Overall instrument performance can be improved using a granite base, which when well designed can both minimize amplification of floor vibrations and increase thermal stability when compared with a metal support. High-precision positioning is usually achieved by positioning the sample and optics with piezoactuated, flexure-guided stages that have high stiffness and correspondingly high resonant frequencies.

Even with careful consideration of the mechanical design, it is virtually impossible to engineer out all vibrations, thermal drifts, and other error sources to the nanometer level, while having an instrument that is practical. Therefore, precise sample and optics metrology can be used to increase the position accuracy. For example, laser interferometry is able to measure the relative position between the sample and optics at nanometer (nm) or even sub-nanometer resolution, which is much easier than controlling the positions to that same level. Because of this, X-ray microscopes integrate laser interferometry with motion control for closed-loop feedback on the sample and optics positions.^10,12,13 This can correct for thermal thrift and some vibration of the system.

Despite the continuous improvements on the instrumentation for high-resolution ptychography, the throughput of a ptychographic scan is still comparatively low. This is because the conventional implementation of ptychography relies on a move-settle-measure (also called “step scan”) mode. The time spent on moving and settling of the scanning stage is the scan overhead, which is not used for diffraction pattern collection. In a recent demonstration of high-throughput ptychography with step-scanned 10 μm beam,¹⁴ the overhead per scan point was about 150 ms which already contributed about 43% of the total scan time. State-of-the-art optics are able to focus the X-ray beam down to sub 100 nm; thus, there is motivation to perform ptychography with a small focused beam,¹⁵ especially when combined with other imaging modalities (such as X-ray fluorescence)^16–19 in which the resolution is limited by the focused spot size. With such a high-photon-density beam, the exposure time per scan point can be correspondingly reduced down to a few milliseconds or even sub-milliseconds. In this case, the step scan overhead dominates the measure time, making ptychography grossly inefficient. To reduce or eliminate the scanning overhead, a continuous motion scheme (called “fly-scan”) has been recently implemented in ptychography.^20–22 The decoherence effect in the diffraction patterns caused by the continuous motion of the sample is then solved by a mixed-state reconstruction method.²³ In this way, fly-scan ptychography significantly reduces the scan overhead and improves the throughput of high-resolution imaging.

With advances in accelerator technology, synchrotron facilities around the world, including the advanced photon source (APS) at the Argonne National Laboratory, are undergoing upgrades.²⁴ The upcoming advanced photon source upgrade (APS-U) will provide significantly increased brightness and at least two orders of magnitude increase in coherent flux. The Velociprobe²⁵ is a next generation, ultra-high-resolution, X-ray ptychographic microscope designed to demonstrate the technologies and techniques that are necessary to take advantage of the more than 100-fold increase in coherent flux from the APS-U. As the sample and/or optics are fast and continuously moved in fly-scan ptychography, new concepts and methods for the aforementioned high-instrument stability and high-precision positioning need to be developed correspondingly. On the Velociprobe like in many instruments, we have separated the coarse alignment motions from the fine scanning motions. However, unlike other X-ray microscopes, the coarse positioning axes on the Velociprobe are novel, bespoke air bearing-supported granite stages rather than rolling-element bearings (such as ball or cross-roller type). As such, these axes do not amplify the ambient vibration and provide maximum stability with the minimum number of degrees of freedom. The fine motion optics and sample stage stacks are mounted on these stable granite stages.

Likewise, a novel, high-precision, fast scanning scheme for the zone plate optics was developed and a custom, closed-loop, interferometric feedback system is implemented on a high-bandwidth FPGA (Field Programmable Gate Array) controller. High-efficiency scanning schemes with various trajectories or arbitrary scan paths²⁶ have been developed. Since the installation of the Velociprobe at beamline station 2-ID-D at the APS, we have demonstrated fly-scan ptychography at different scan speeds, with an interferometer position recording speed of 25 kHz. The obtained 2D reconstructions show a spatial resolution of 8.8 nm on a Au test sample with an acquisition frame rate of 200 Hz, and about 10 nm on an integrated circuit sample with a detector’s full continuous frame rate of 3000 Hz (Eiger 500K).

The overall instrument concept has an ultrastable coarse motion system supporting both a fast optics scanner and a set of sample positioning stages, including a rotation stage for tomography. A computer rendering of the design of the Velociprobe setup is shown in Fig. 1. A corresponding cutaway on the right highlights the drive mechanism for the granite axes.

The ultrastable coarse axes (3, 6, 10 in Fig. 1) use novel air bearing-guided stages that provide repeatable motion (<1 μm) and high-rigidity, through granite-to-granite contact, when the air bearings are vented (US patent application 20180058552). These air bearing axes all have tens of millimeters of travel allowing for convenient alignment operation to place the optics in the X-ray beam. This vented air bearing design is stiffer than a conventional linear, rolling element bearing design. It is known that for the size of rolling-element bearings found in typical tables, the stiffness of the bearings can play a significant role in the system dynamics.²⁷ The higher stiffness of the granite axes means that amplification of ground motion is reduced as compared to that of a conventional support system. Because of the vented air bearing design, there are few vibration-amplifying components between the floor and the optics or sample. The sample is positioned by four stiff cross-roller bearing guided stages, whereas the optics are positioned by a stiff flexure stage.

The high stiffness of the vented air bearing design contributes to high vibrational stability, while the low thermal expansion (compared to metal) contributes to good drift stability. Figure 2 shows the measured relative vibration between the granite gantry that holds optics and the granite X axis supporting the sample stack. Integrated over a 5–100 Hz bandwidth, the motion is 1.4 nm rms in the direction transverse to the beam (X) and 0.75 nm rms in the vertical direction (Y). This is an extremely low vibration for a multiaxis motion system with tens of mm of travel. For reference, we also show a similar measurement made on an X-ray microscope of conventional construction located in the same hutch as the Velociprobe. The Velociprobe exhibits almost 50 and over 100 times less vibration in the X and Y directions, respectively, when compared to the conventional instrument. Many new instruments at the APS and some at the Sirus Light Source in Brazil have adopted this Velociprobe-style granite stage system.²⁸ The total mass of the granite components is about 3612 kilograms. The granite coarse X and Z (6, 10) are encoded with an incremental encoder (Renishaw TONiC), while the granite coarse Y axis is encoded with laser displacement sensors (Keyence LK-G3000 series). In addition, the granite gantry provides a common reference for the six interferometer axes which will be discussed in Sec. II B.

Figure 3 shows the sample stage stack. The sample stage system sits on the coarse sample X axis and consists of a rotation stage carrying three translation axes. The coarse sample axis (8-1 and 8-2) also uses air bearing to move the rotation axis of sample rotation stage (8-3) into the beam. The sample rotation stage was selected for stiffness and large aperture. The stiffness ensures stability, while the aperture enables the sample stage stack to have a shorter height and enhanced stability. The coarse sample X and the rotation stage are both driven with conventional stepper motors, while three sample translation axes (8-5) use precision linear stages (Physikinstrumente, Q-545.240) with 26 mm travel range. The sample holder (8-6) with magnetic base is kinematically mounted on the top of the sample stages, allowing for easy changing of the sample. Three optical reflectors are installed on the sample holder as references, and the sample position relative to the gantry position is measured with three interferometers. The reflectors are arranged at a distance of 10 mm or less from the sample. This close position minimizes the out-of-loop drifts and errors. At this time, the reference reflector for tomographic operation is not installed.

The zone plate scanner is a parallel-kinematic, flexure-based, piezo-driven stage (Physikinstrumente, P-733.3DD). The scanner assembly is shown in Fig. 4. The stage (9-5) has 30 μm, 10 μm, 30 μm travel in the X, Y, and Z axes, respectively. Kinematic mounting features align the zone plate holder (9-4) to the stage assembly. The completed scan assembly has a mass less than 10 grams. The zone plate translations are measured with three laser interferometers indicated by green beams in Fig. 4. Downstream of the zone plate, there is a 30 μm-diameter order sorting aperture (OSA) integral to the scanner mounting plate. The scanner is rigidly mounted upside down to the granite gantry which then can be moved along the beam direction to adjust the focusing on the sample when different energies or/and different zone plates are used.

The translation stages for the optics and the sample, especially the piezo stage, can provide accurate positions on stage motion through the built-in position encoders; however, they are not able to measure the thermal drifts, dynamics, and error of the system as a whole. In addition, the bandwidth of position encoders is usually not enough for fast scanning that requires >kHz bandwidth. On the Velociprobe, we are using laser interferometers to precisely measure the relative position between the zone plate and the sample. Laser interformeters can provide very high measurement bandwidth, tens of MHz. While precision control is necessary for the microscope operation, it is inherently limited by actuator and system dynamics. The high-bandwidth position measurement not only allows for precise closed-loop motion control but also provides accurate knowledge of sample and zone plate positions for ptychography reconstruction. Six laser interferometer axes (Attocube FPS3010) are all installed on the granite gantry, with three axes (green beams in Fig. 4) for zone plate X, Y, and Z and the other three axes (red beams) for the sample. The granite gantry provides high stability and also a common reference for the relative position measurement between the sample and the zone plate. To reduce the amount of out-of-measurement-loop path, the reference mirrors are installed on the zone plate and sample holders to minimize the distance between the mirrors and the place of interest.

The scanning control is a key aspect of achieving high-bandwidth, high-stability, and high-resolution performance with the Velociprobe instrument. On the Velociprobe, the advanced control algorithm, H-infinity,²⁹ is implemented on an FPGA hardware for closed-loop measurements. This type of the control method has higher resolution and bandwidth than the conventional PID (proportional-integral-derivative) feedback design so that it provides better disturbance rejection, such as ambient noise and thermal drift. Figure 5 shows the closed-loop system layout for the zone plate scanning stages. The closed-loop control system is comprised of a piezo stage for the zone plate, Attocube laser interferometric sensors, and a National Instruments (NI) control hardware. The NI control hardware includes an embedded CompactRio controller (cRIO-9039, with a real-time processor and reconfigurable FPGA), NI-9402 digital input-output (DIO) modules, and NI-9263 analog voltage output modules. Although the FPGA has a 25 MHz clock, the running frequency of the main loop for the scanning controller is chosen at 25 kHz which is sufficient for the current use and can be changed if necessary. In a closed-loop scan, the scan trajectory which can be arbitrary scan pattern is sent to the scanning controller as a table. The controller sends voltage signals provided by the NI-9263 analog voltage output modules to the PI amplifier at 25 kHz sampling frequency. The amplified signals go to the piezo actuators and move the stage to the requested position. The stage displacements are then measured by a high resolution laser interferometric sensor and read back through the NI-9402 DIO modules directly into the FPGA in the NI cRIO chassis at a frequency of 16 MHz; however, logging of the laser interferometer positions and running of the feedback closed-loop controller are done on the FPGA at a 25 kHz rate. With the H-infinity control algorithm implemented in the closed-loop system, the positioning resolution (3σ) can be achieved at 1.9 nm and 1.4 nm for the X, Y direction, respectively. For data acquisition in the scan, the external detector trigger signals are generated on the same controller clock with 25 kHz sampling frequency. In this manner, the trigger and position logging are synchronized and can be updated every 40 μs.

To better compensate the drift of the sample during the ptychographic scan by the zone plate, the laser interferometer information of the sample stage can be fed into this closed loop system with H-infinity controllers. As a result, the relative position of the zone plate and the sample is accurately measured in this differential motion mode. Two new fly-scan trajectories including the snake-raster scan pattern and spiral-scan pattern have been implemented on the Velociprobe using this H-infinity closed-loop system, with other high-efficient scan schemes (such as combined motion of the sample and the optics³⁰) being developed. Figures 6(a) and 6(b) show the fly-scan trajectory on an area of 2 μm × 2 μm with the raster-scan pattern and spiral-scan pattern, respectively. A Dectris Eiger X 500K area detector was triggered with a frequency of 500 frames/second, while the piezo scanner was continuously moved with a fly-scan step size of 50 nm (the motion distance between two exposures). The X and Y translations of the piezo scanner were measured by laser interferometers and were recorded at 25 kHz so that each exposure contained redundant position points to show the beam trajectory within that scan step. That additional position information can be fed into the reconstruction, resulting in improved reconstruction quality in fly-scan ptychography especially with arbitrary trajectory.²⁶ The raster fly-scan and the spiral fly-scan were finished in 3.2 second and 2.5 second, respectively, which is 75 times faster than a step scan on the same piezo stage with 150 ms overhead. At such a high scan speed, the trajectories still show that the H-infinity closed-loop controller gives high performance: the zone plate positions (interferometer positions) were very close to the requested position except the first few exposures when the stage had large accelerations. In the raster fly-scan, the difference between the laser interferometer positions and requested positions was within 2 nm on the linear scan path and was slightly larger (∼5 nm) at the turning corners. In the spiral fly-scan, this difference is smaller than 3 nm as can be seen in Fig. 6(b). In addition, both continuous scan schemes with time-based triggering have advantages of eliminating the scan periodicity from the conventional mesh scan pattern and thus being able to remove “raster grid pathology”^22,31 in the ptychographic reconstruction.

A Fresnel zone plate with 50 nm outmost zone width and 180 μm diameter is installed on the zone plate scanner. The first-order diffracted beam from the zone plate is selected by the combined use of a 60 μm diameter tungsten central stop and a 30 μm diameter order-sorting aperture (OSA) placed ∼62 mm downstream of the zone plate. The instrument is aligned according to the following procedure: First, with each of the three axes of the zone plate piezo scanner centered in its travel range, the granite coarse X and Y stages are used to move the zone plate into the X-ray beam. Next, the gantry is used to position the beam focus at a given X-ray energy. Finally, the precision sample stages are used to position the region of interest (ROI) of the sample in the beam. During a typical X-ray ptychography measurement, shown schematically in Fig. 7, the sample and the OSA are kept static, while the X and Y axes of the zone plate scanner are continuously moved in a trajectory that is chosen from the scan trajectory library, which currently includes conventional one-direction fly-scan,²¹ snake-raster fly-scan, Archimedean spiral fly-scan, and rectangular spiral fly-scan. As previously described, the FPGA-based scanner minimizes the error between requested trajectory and the scanner output using the H-infinity controller. The Eiger detector with a 75 μm pixel size is placed 1.92 m downstream of the sample and is triggered with a frame rate specified by the controller. A helium-filled flight tube is installed between the sample and the detector to reduce air absorption and scattering. The maximum image area of a single ptychography fly-scan is limited both by the travel range of the piezo stage (30 μm and 10 μm for X and Y, respectively) and by the beam clearance envelope defined by the diameter difference and axial distance between the OSA and the central stop for the zone plate. The latter constraint imposes a limit of approximately 15 μm for both X and Y. Furthermore, it is generally better to scan the zone plate in an area as small as possible in order to minimize deleterious effects due to inhomogeneities in the beam and thereby provide a uniform probe for ptychography. Therefore, the scan area for ptychography is typically less than 9 μm × 9 μm. To image larger fields of view, tile scanning via stepwise translations of the sample stages is implemented.

To demonstrate the achievable spatial resolution of the Velociprobe instrument, a Au Siemens star test pattern was imaged by X-ray ptychography. Two ptychographic fly scans were collected with a snake-raster and spiral pattern shown in Figs. 6(a) and 6(b), respectively. In the measurements, an X-ray beam with a photon energy of 8.8 keV was filtered by a double crystal monochromator (DCM) with a bandwidth of ΔE/E ≈ 1.4 × 10⁻⁴ and then focused by the zone plate. The sample was placed around 70 μm downstream of the focus spot, resulting in a full-width at a half-maximum (FWHM) beam size of ∼150 nm on the sample. The snake-raster scan covered a square field of view of 3 μm × 3 μm as marked by a dash cyan box in Fig. 8(a), while the spiral scan covered a circular area with a diameter of 3 μm as indicated by a dash cyan circle in Fig. 8(b). Both scans used a fly-scan step size of 50 nm and a detector acquisition speed of 200 frames/second. In the reconstruction, diffraction patterns with 512 × 512 pixels were cropped, resulting in a real-space pixel size of 7 nm in our experimental setup. A maximum-likelihood (ML) algorithm⁸ with illumination wavefront refinements was used for ptychography reconstructions. Figures 8(a) and 8(b) show the reconstructed phase of the test sample for the snake scan and spiral scan, respectively. The features on the scan regions are highly consistent with each other. Fourier ring correlation (FRC)³² was carried out on a common center region [blue box in Fig. 8(b)] of each scan, and one bit threshold criterion was used to estimate the achieved image resolution from the FRC. Figure 8(c) shows the FRC vs spatial frequency of two images along with the one bit threshold. The intersection between the FRC and one bit threshold gives a spatial resolution of about 8.83 nm. The consistency of the image signal in this spatial frequency correlation function also shows that the spiral fly-scan ptychography is able to achieve the same quality as a snake-pattern scan which has the same illumination coverage.

As an example of tile scanning and stitching of ptychographic datasets on the Velociprobe, a sample of LaFe_0.3Co_0.7O₃ perovskite oxide particles was imaged at 10 keV with an illumination spot size of about 150 nm. In each tile covering 7.5 μm × 8.5 μm, the sample was fly-scanned in a snake-raster pattern with 70 nm step size and 200 Hz detector acquisition rate. A total of 3 × 3 scanning tiles were acquired, covering a sample area of about 21 μm × 24 μm. Scanning positions measured by the laser interferometers and the diffraction patterns were input into an ePIE (extended ptychographic iterative engine³³) based reconstruction code with multi-GPUs computation capability.³⁴ Figure 9 shows the reconstructed phase of the perovskite particle aggregate with a pixel size of 6.2 nm. The high resolution structure image reveals the presence of low-electron-density ultrastructures in the oxide particles. Those porous structures may affect the performance of electrocatalysts on oxygen-reduction activity.³⁵ Such information can inform sample synthesis and lead to improved catalytic performance.

In this tile scan, 0.5 μm overlap between tiles was originally set in both directions for image stitching purpose, which decreased the effective imaging speed by about 8% due to the duplicate imaging on the overlap regions. It turns out that 0.5 μm overlap is excessive because the relative positions between the sample and optics can be measured with an accuracy of a few nanometers which should allow nonduplicate imaging area between scan tiles. However, in the practical tile, scanning an overlap of one or two step sizes is still recommended to assure enough illumination coverage on the boundaries between tiles.

Currently, the imaging speed of the Velociprobe is limited by the maximum continuous frame rate of the detector (3000 Hz) rather than the optics scanning mechanics (5.7 mm/s). To demonstrate the performance of the instrument at high scan speed limit, a 16 nm-node integrated circuit (IC) sample with a 130 μm Si substrate was imaged with the Eiger detector operated at the maximum continuous frame rate of 3000 Hz with 12 bit counter depth. The zone plate was fly-scanned in a snake-raster pattern with a motion distance (step size) of 50 nm during each 0.33 ms exposure, resulting in a linear speed of 150 μm/s which was 15 times faster than the scan shown in Fig. 8(a) and was about 65 times faster than a previous study³⁶ on an IC with the same thickness. During each fly-scan, a total of 3600 diffraction patterns were acquired on a 3 μm × 3 μm area in 1.2 seconds. Figure 10(a) shows the reconstructed phase image of the IC sample imaged with 8.8 keV X-rays from a DCM source which provides a flux of ∼5 × 10⁸ ph/s. Figure 10(c) shows the stacked phase image calculated from chip design files on the same region displaying tungsten source/drain connections (high-electron-density horizontal bars with 40 nm width, labeled by “OD”) to the bottom oxide diffusion layer and the gate vias to the polysilicon layer (PO) highlighted by the red color. A comparison of the images in Figs. 10(a) and 10(c) shows high agreement on circuitry features between the ptychographic image and the image simulated from the design file. In addition to the 720 nm-sized metal blocks (with rounded corners likely resulting from the manufacturing process) in the topmost layers, the measured image in Fig. 10(a) also shows consistent 90 nm-spaced OD connections and some of the PO gate vias visible in Fig. 10(c). However, the spatial resolution of such a high-speed scan is dose-limited: the increase in the scan speed and 76% absorption by 130 μm silicon substrate yielded only 16 photons/nm² on the sample. To increase X-ray dose on the sample, we used a double-multilayer monochromator (DMM) X-ray source which provides a higher flux of ∼1 × 10¹⁰ ph/s with a broader bandwidth of about 1%. Broad-bandwidth radiation has been used previously for X-ray imaging including ptychography.³⁷ A ptychographic scan with the DMM source was carried out with the same scan parameters (150 μm/s scan speed and 3000 Hz detector frame rate) as the one with the DCM source, and Fig. 10(b) shows the ptychographic reconstruction of the same chip region with the mixed-state reconstruction method.^23,37 The ptychographic image with the DMM source shows much more improved quality compared with the DCM-source scan, with OD and PO connections clearly displayed at a high spatial resolution of about 10 nm.

In this work, we have developed a high-speed, high resolution scanning X-ray microscope—the Velociprobe. The novel hardware and mechanical designs provide high stability for nanometer-scale imaging. The use of the high bandwidth laser interferometry and advanced motion control algorithm in a closed-loop system allows one to do precise and fast measurements. Advanced ptychographic scanning techniques and methods have been developed through high-efficiently scanning the low-mass zone plate in a continuous trajectory and accurately recording the interferometer positions at high frequency for fly-scan ptychography reconstruction.

The performance of this instrument was demonstrated by 2D ptychographic scans. A spatial resolution of 8.8 nm was achieved in a fly-scan measurement on a gold test sample. Ptychographic tile scanning was demonstrated on a perovskite oxide sample to show the instrument’s capability of imaging a large sample at high resolution. The ultrafast ptychography scan was implemented with a detector’s maximum continuous frame rate of 3000 Hz on an integrated circuit, which was 450 times faster than a step scan on the Velociprobe using the same dwell time but with 150 ms overhead per point. The setup still showed high accuracy of positioning and interferometer reading at such a high scan speed. The result with a DCM source implies that the current scan speed of ptychography on the Velociprobe is limited by the X-ray flux instead of the hardware, which was proved by a same scan with a DMM source that provided at least 20 times more flux. The DMM reconstruction from the scan with 3 μm × 3 μm was obtained in 1.2 second, with a 7 nm pixel size and about 10 nm resolution, giving an imaging rate of 9 × 10⁴ resolution elements per second or an effective dwell time of 11 μs per resolution unit (a method used in Ref. 14 to evaluate the throughput of ptychography). Such high-speed high-resolution ptychography will open up the opportunities for in situ, operando measurements with a time resolution down to subsecond per μm². To continuously increase the image speed, high-efficient scan schemes will be explored by optimizing the scan parameters, scan trajectories, and even the combined scan of the optics and the sample.³⁰ As the scan speed is currently limited by the X-ray flux, sources with higher flux (such as pink beam) and high-efficiency optics (such as capillary) will be explored for high-speed ptychography. For high-flux X-ray sources, a semitransparent central stop will be placed in front of the detector to solve the counting-rate saturation issue and to improve the dynamic range³⁸ or signal-to-background ratio.³⁹ The coming upgrade of the APS will provide at least 100 times coherent flux and can theoretically speed up the scan by two orders of magnitude, which, however, requires the development of a faster detector.

The current setup is only operated for ptychography. As X-ray fluorescence microscopy shares the same scanning mechanism with ptychography, we plan in the future to integrate a fluorescence detector into the instrument for correlative X-ray fluorescence and ptychographic imaging.^17,18,40

While the presented datasets are in 2D, the 3D ptychography test has been started on the Velociprobe. However, the preliminary 3D scans show that the existing stepper rotation stage has a runout of ∼2 μm, which would cause difficulties in projection alignment and deteriorate the 3D reconstruction quality. The rotation stage is currently being replaced with a high precision air bearing stage to minimize runout and wobble (Professional Instruments Company 10R-606 with <0.1 μrad wobble) and thereby improve tomographic ptychography. Strategies to have a better interferometer setup for accurate relative positions of the optics and the sample at all rotation angles will be implemented with the new rotation stage. This requires installing well-calibrated reference mirrors, especially the one for rotation degree of freedom. Therefore, the accurate interferometer positions will not only allow for high-resolution ptychography reconstruction but also for better projection alignment for tomography.

The authors would like to thank Joseph Arko, Pavel Shevchenko, Bruce Hoster, and the staff of the Argonne Central Shops for their assistance with manufacturing and modifying various components. We thank Chris Jacobsen for many useful discussions. We would also like to acknowledge Dennis Ethen and his colleagues at Starett Tru-Stone for helpful advice and discussions. The Velociprobe was supported by Argonne LDRD 2015-153-N0. This research used resources of the Advanced Photon Source and the Center for Nanoscale Materials, U.S. Department of Energy (DOE) Office of Science User Facilities operated for the DOE Office of Science by the Argonne National Laboratory under Contract No. DE-AC02-06CH11357. This work is partially supported by the Office of the Director of National Intelligence (ODNI), Intelligence Advanced Research Projects Activity (IARPA). The views and conclusions contained herein are those of the authors and should not be interpreted as necessarily representing the official policies or endorsements, either expressed or implied, of the ODNI, IARPA, or the U.S. Government.