Balanced Detection in Femtosecond X-ray Absorption Spectroscopy to Reach the Ultimate Sensitivity Limit
W.F. Schlotter, M. Beye, S. Zohar, G. Coslovich, G.L. Dakovski, M.-F. Lin, Y.Liu, A. Reid, S. Stubbs, P. Walter, K. Nakahara, P. Hart, P. S. Miedema, L. LeGuyader, K. Hofhuis, Phu Tran Phong Le, Johan E. ten Elshof, H. Hilgenkamp, G. Koster, X.H. Verbeek, S. Smit, M.S. Golden, H.A. Durr, A. Sakdinawat
BBalanced XAS
Balanced Detection in Femtosecond X-ray Absorption Spectroscopy to Reach theUltimate Sensitivity Limit
W.F. Schlotter, a) M. Beye, S. Zohar, G. Coslovich, G. L. Dakovski, M.-F. Lin, Y.Liu, A. Reid, S. Stubbs, P. Walter, K. Nakahara, P. Hart, P. S. Miedema, L. LeGuyader,
3, 4
K. Hofhuis, Phu Tran Phong Le, Johan E. ten Elshof, H. Hilgenkamp, G. Koster, X.H. Verbeek, S. Smit, M.S. Golden, H.A. D¨urr, and A. Sakdinawat Linear Coherent Light Source, SLAC National Accelerator Lab, 2575 Sand HillRd., Menlo Park, CA 94025, USA FS-FLASH, Deutsches Elektronen-Synchrotron (DESY), Notkestr. 85 D-22607Hamburg, Germany SLAC National Accelerator Lab, 2575 Sand Hill Rd., Menlo Park, CA 94025, USA Spectroscopy & Coherent Scattering, European X-Ray Free-Electron Laser FacilityGmbH, Holzkoppel 4, 22869 Schenefeld Germany Faculty of Science and Technology and MESA+ Institute for Nanotechnology,University of Twente, P.O. Box 217, 7500 AE, Enschede, The Netherlands Van der Waals-Zeeman Institute, Institute of Physics, University of Amsterdam,Science Park 904, 1098 XH, Amsterdam, The Netherlands Uppsala Universitet, Dept of Physics and Astronomy, Box 516, 751 20 Uppsala,Sweden (Dated: 26 June 2020) a r X i v : . [ phy s i c s . i n s - d e t ] J un alanced XASX-ray absorption spectroscopy (XAS) is a powerful and well established techniquewith sensitivity to elemental and chemical composition. Despite these advantages,its implementation has not kept pace with the development of ultrafast pulsed x-raysources where XAS can capture femtosecond chemical processes. X-ray Free Elec-tron Lasers (XFELs) deliver femtosecond, narrow bandwidth ( ∆ EE < . ∼ photons. However, the energy contained in each pulse fluctuatesthus complicating pulse by pulse efforts to quantify the number of photons. Improve-ments in counting the photons in each pulse have defined the state of the art forXAS sensitivity. Here we demonstrate a final step in these improvements through abalanced detection method that approaches the photon counting shot noise limit. Indoing so, we obtain high quality absorption spectra from the insulator-metal tran-sition in VO and unlock a method to explore dilute systems, subtle processes andnonlinear phenomena with ultrafast x-rays. The method is especially beneficial forx-ray light sources where integration and averaging are not viable options to improvesensitivity. a) Electronic mail: [email protected]
Soft X-rayabsorption spectroscopy is sensitive to both elemental composition and chemical state. Itcan be used to probe bonds formed by specific elements such as carbon, nitrogen and oxy-gen as well as to understand the role that d-orbital electrons play in the many emergentproperties of transition metal compounds. The high sensitivity that enabled these advancesis provided by the stable and sizable x-ray photon flux (number of photons per second)generated by modern storage ring synchrotron light sources.The advent of ultrafast pulsed x-ray sources sparked excitement in the promise of trackingelectron dynamics with femtosecond resolution. Early demonstrations at low flux sourcesfurther fueled the enthusiasm.
However, the development of XAS at next generation x-ray free electron laser (XFEL) sources has not been limited by the number of availablephotons but instead by methods for normalization. The temporal structure of the x-raypulse FEL sources, which deliver millijoule-scale pulses with tens of femtosecond durations,has hindered the development of XAS. The deluge of photons from each pulse challenges thelinearity and dynamic range of established detectors such as photodiodes and multi-channelplates. Consequently the application of XAS at XFELs to time resolved measurementshas been restricted, often requiring extensive experimental beamtime for limited, albeitsignificant, results.
The scarcity of short pulses at FEL sources, which typically operate between 10-120 Hz,has favored spectroscopic methods that take advantage of the full pulse bandwidth. Im-proved normalization can be achieved using a transmission grating to generate two spectrallydispersed copies of the incident beam. Recently, off axis zone plate illumination was usedto fill an area detector (e.g. CCD) in order to resolve ultrafast pump probe signals in asingle shot. Because the spectrum for each pulse is different, improvements to signal qual-ity require improved detection or averaging and careful normalization over multiple pulses.However, the complete measurement of both energy spectrum and temporal evolution ina single pulse is not conducive to high sensitivity spectroscopy because of limitations inaveraging due to detector readout noise.By restricting each single shot measurement to a narrow spectral energy bandwidthrecorded with superior normalization afforded by high dynamic range detection, one can3alanced XASobtain high fidelity absorption measurements that improve with averaging. This is the stan-dard method for XAS measurement at synchrotron storage rings where advances in gratingmonochromators enabled high spectral resolution XAS using detection methods includingtotal electron yield, fluorescence and transmission.
This approach has been used to recordthe highest quality XAS spectra at x-ray FEL sources. Recently the use of a CCD to mea-sure transmission greatly enhanced the sensitivity for time-resolved XAS measurements ata FEL. By expanding the x-ray beam on the area detector the low noise properties of theCCD are exploited while improving dynamic range. Here we employ the ability of a grating to generate consistent copies of a beam com-bined with a zone plate to uniformly illuminate a high sensitivity CCD area detector, thusdemonstrating photon counting noise as the main limitation to XAS sensitivity.Photon counting noise or shot noise represents the ultimate limit to photon detectionsensitivity as described by statistical optics. Because each soft x-ray photon observed ona silicon detector generates hundreds of electrons, cooled, low noise electronics can easilyresolve a single photon. Therefore, the dynamic range in this regime is not limited byreadout electronics, but rather the total number of detected photons which scales with theilluminated detector area. A typical megapixel CCD area detector can linearly observe 10 photons ( λ =1 nm). The uncertainty in counting because of photon shot noise is √ N orSNR=10 for 10 detected photons. However, a single shot sensitivity of 10 is far from thecurrent single shot state-of-the-art for XAS at an FEL.Sensitivity at (or near) the photon counting limit will extend XAS to study dilute samplesand open the possibility to measure subtle changes associated with non-linear processes. Itrepresents the most efficient data collection method, therefore it is also ideally suited forradiation sensitive samples.Our experimental arrangement is illustrated in Figure 1 where monochromatic, soft x-rays ( λ =2.4 nm, λ ∆ λ =3500) illuminate a transmission diffraction grating. The transmissiondiffractive optical element includes a zone plate and grating combined on a single structure.In this way two identical, balanced, highly divergent beams are generated by amplitudedivision. Even if the two beams differ because of imperfection in the diffractive element,their photon number ratio will be consistent. Upon detection, the number of photons in eachof the first order beams differs based on a Poisson statistical parent distribution. For theXAS measurement one beam passes through the sample (signal, s ) while the other serves as4alanced XAS FIG. 1. The two diffracted beams form the reference, r , that does not pass through the sampleand the signal, s , that does. a) The grating zone plate structure is fabricated by electron beamlithography (see methods) and contains a 600 nm outer zone width. Because the average single pulsefluence illuminating the zone plate is 43 nJ/cm the zone plate is well below the damage threshold.b) A pnCCD destructively measures both beams from a single pulse as they are detected withinisolated regions of interest (ROI). c) The reference ROI containing 5.4 × photons spread over16,510 pixels is shown with a graininess characteristic of photon counting noise. a reference ( r ) beam. Because of the divergence of the zone plate focal length ( f =122 mmfor a photon energy of 525 eV), the beams expand rapidly to fill the large area detector.Following photon detection of the two beams on the CCD, the detector is read out andcorrected (see methods) and the two regions of interest (ROI) are selected for integration,see Figure 1. The transmission through the sample is calculated simply by computing theratio of the two integrated regions of interest.To demonstrate shot noise limited performance we validate the setup without a samplein place to form two identical beams.Figure 2 a) shows a correlation plot between the two ROIs. A line fit to the correlationgives the ratio of the number of photons in the two beams which is 0.97 ± FIG. 2. Data for the control case where no sample is inserted into the the signal beam. a)Correlation between signal and reference (integrated ROIs) for 8,739 single events (blue dots). Thered line is a fit to the data where the slope represents the transmission. b) Transmission vs. signalfor each shot are plotted. Solid lines indicate one standard error of the mean or the photon limitederror: T lim = T mean ± (cid:112) /N where N is the number of photons measured in the signal ROI. In thepurely photon noise limited case, 2 σ (68.2%) of the shots would fall between the error curves. Overthe full signal range shown 62.4% of the recorded shots are within this photon limited envelope. the systematic error to the ratio between the two beams. By analysing the ratio of the twoROIs vs. the reference ROI it becomes clear that the fidelity of the measurement increaseswith the number of photons in each pulse as illustrated in Figure 2b). The fraction ofrecorded events within the photon limited error envelope increases as the number of detectedphotons in the signal increases. 6alanced XAS FIG. 3. The SNR (left axis) data points with errorbars are plotted vs. the signal level in photons.The errorbars represent the standard error which is σ SNR √ N where N is the number of measurementshots per point. The photon limited SNR (solid line) is provided to illustrate the maximum possibleSNR. For shots collecting more than 6 × photons the SNR reaches the photon noise limit. Thehistogram (right axis) of shaded bars displays the number of shots used to calculate the SNR anderror. To quantify the sensitivity of this method we calculate the Signal to Noise Ratio (SNR)as the ratio of the mean transmission, T , to the standard deviation of the transmission σ T such that SN R = T /σ T .The SNR improves as the number of detected photons increases and Figure 3 shows theintensity dependence. For comparison the SNR in the photon counting noise limit, SN R = (cid:113) N , is plotted as well. The error bars on the SNR ( σ SNR ) are derived by propagating erroras follows: V ( σ T ) = ( n − n (cid:18) µ − n − n − σ T (cid:19) (1)where V ( σ T ) is the variance of the standard deviation of the transmission, µ q is the q th ordermoment of T and n is the number of data points per bin. Using error propagation we canarrive at the standard deviation of the SNR, σ SNR . σ SNR = (cid:112) V ( σ T )2 σ T (cid:12)(cid:12)(cid:12)(cid:12) < T >σ T (cid:12)(cid:12)(cid:12)(cid:12) (2)7alanced XASWe see that the SNR falls short of the shot noise limit by 35% at low signal and approachesit for the highest signal shots. The explanation for the closing of this gap may come fromdetector readout noise, like fixed pattern noise and errors in the common mode correction.To demonstrate the advantage of this spectroscopy method we recorded an XAS spec-trum of the often-studied transition metal oxide, VO . Vanadium dioxide is technologicallyinteresting because of its insulator-to-metal transition slightly above room temperature. Thevanadium L-edges and oxygen K-edge are within a 50 eV photon energy scan range. Whilethe ROI in the balanced beam detection method shift spatially during the 50 eV photonenergy scan, they do remain fully on the detector. To generate these spectra, the sampletransmission is calculated by taking the ratio of the sum of the detected photons in thesample ROI with respect to the reference ROI. The absorption length, α , corresponds to: α = − ln ( T ) t sample where T is the sample transmission and t sample is the thickness. The method affords asimple, yet absolute, calculation of the absorption length because the sample is studied intransmission and an identical reference is recorded. Figure 4 shows two XAS spectra forVO measured at temperatures above and below the metal-to-insulator transition. Each FIG. 4. Absorption spectra recorded on the 50 nm thick VO sample at temperatures above (394K) and below (301 K) the insulator-to-metal transition. The spectra span the vanadium L-edgesand the oxygen K-edge. The difference between the two spectra is plotted and depicts the changesin the unoccupied electronic states. spectra recorded using ultrafast x-ray pulses and are comparable to spectra recorded atsynchrotron storage rings for similar samples and conditions. The difference between thetwo spectra that is plotted in Figure 4 is also validated by storage ring lightsource spectra.VO is particularly sensitive to changes in temperature so it is important to minimizethe x-ray fluence incident on the sample. In this case the spot size was 80 µ m at 500 eVand 30 µ m at 550 eV resulting in an average per pulse fluence of 10 nJ/cm and 75 nJ/cm respectively. The pulse duration was 110 fs. Because the distance between the zone plateand the sample remained fixed during the scan we see this change in fluence as a result ofthe longitudinal dependence, of the zone plate focal length on wavelength. The spot size onthe sample can be increased simply by moving it toward the detector because of the largeangular divergence. However, because the total integrated energy for the measured spectrawas only 82 nJ, x-ray induced changes are, in fact, undetectable.Moving the sample position closer to the focus enables higher intensity and because thefocal spot can be less than 25 nm it is conceivable to reach a fluence of ∼ at anFEL source. Such intensities are unprecedented at a narrow bandwidth, thus opening newpossibilities to study x-ray induced non-linear absorption phenomena. The method presented scales with photon counting noise. Therefore an optimized setupcapable of illuminating a full area detector (2.5 cm ) could detect 10 soft x-ray photons(fewer at shorter wavelengths) and thus realize a per shot SNR of 20,000. Such sensitivitywould be suitable for measuring subtle differences in adsorption from dilute solutions orvery thin samples or interfaces. Because the position of the sample can be used to controlthe fluence and the method is in the photon noise regime this represents the most efficientXAS measurement possible. This is important at high intensity pulsed sources as well as lowintensity pulsed sources where the measurement of every photon is essential for obtaining thebest possible data quality. The experimental geometry affords sufficient space to introducea magnetic field or a pulsed laser to optically pump the sample.We have demonstrated x-ray absorption spectroscopy limited only by photon countingstatistics. This method is well suited for pulsed sources such as x-ray free electron laserslike the one used here. Moreover, thanks to this efficiency, samples prone to change ordamage upon x-ray illumination can be explored at the lowest possible exposure fluence.9alanced XASOur exemplary XAS spectra of the metal-insulator transition in VO validate that thebalanced beam detection is accurate, fast and robust. Application of balanced beam XASwill provide new opportunities for time resolved x-ray experiments. I. METHODSA. X-ray Parameters
Data were collected on the Soft X-ray Instrument for Materials Science (SXR) at theLinac Coherent Light Source (LCLS). For these spectroscopic energy scans the wavelengthwas continuously adjusted by scanning the SXR monochromator between 510 eV and 550eV.
The LCLS electron beam energy was concurrently scanned to maintain the maximumtransmission through the monochromator. To ensure the zone plate was illuminated with aclean wavefront, the exit slits on the monochromator were closed to 5 µ m and the KB mirrorsystem was set to focus in the horizontal 1 m downstream of the zone plate. Consequentlythe zone plate was illuminated and overfilled by a 1.3 x 1.9 mm (h x v) spot. B. Zone plate parameters and fabrication
The integrated beam splitting zone plate was fabricated with gold on a 100 nm thickSiN membrane using electron beam lithography and electroplating. See the micrograph inFigure 1 a). The diameter of the zone plate was 480 µ m with an outer zone width of 600nm while the grating period was 225 nm. The combination of the zone plate and gratinginto a single optic predicts a diffraction efficiency of 4.5% into the first order beams. Theparameters for the integrated zone plate grating were optimized to maximize the spot sizeon the detector as well as the distance between the two beams as they intersect the sampleplane.
C. Sample and Reference Geometry
Because 90% of the illuminating beam is transmitted by the zoneplate, care was takento fully attenuate it using a beamstop located 180 mm downstream of the zone plate optic.The horizontal focus of the KB mirror is downstream of the beamstop, thus preventing the10alanced XASdirect beam from illuminating and saturating the detector.For the VO measurements the sample and reference were positioned 130 mm downstreamof the zone plate at which point the spacing between the two diffraction orders was 3 mm.The reference was a commercially available 3 mm diameter 200 µ m thick Si substrate witha 200 nm thick Si N membrane to form a 250 x 250 µ m window at the center. The samplewas a 50 nm thick VO film grown on an identical Si N system. The Si substrate betweenthe two Si N windows serves as an order sorting aperture for the zone plate focal orders. D. Detection and Photon Calibration
To optically isolate the detector, a 200 nm thick Al film was introduced 650 mm down-stream of the zoneplate. The detector plane was located 2428 mm downstream of thezoneplate. The pnCCD consists of two halves, each consisting of 1024 x 512 pixels each witha size of 75 µ m. The two detector halves were separated by 1.4 mm. The pnCCD was oper-ated in high gain mode (1.1 ADU/eV) where the noise level is 0.12 ADU when the detectortemperature is -55 C, thus providing single photon sensitivity. The illumination levels werebelow the full well depth of 300 000 electrons, ensuring no distortion from saturation. The sensitivity of this measurement requires careful attention to background subtrac-tion, gain correction and common mode correction. The background was subtracted afteraveraging 3480 images collected from the detector under the same conditions at which thedata was recorded. To compensate for pixel by pixel variations in gain, a correction matrixis applied to each pixel as determined via flat field illumination. Finally, a common modecorrection was applied to compensate for the time-dependent variation in amplifier gain.For this it was important to mask the area of illumination by the signal and reference ROI.Determining the ADU per incident photon from the detector is crucial for correctlyevaluating the photon counting noise limit correctly. A histogram of the counts per pixel wasgenerated for various regions of interest using 1500 collected images. From these histogramsthe first photon peak coincides with 574 ±
61 ADU per photon.11alanced XAS
ACKNOWLEDGMENTS
Use of the Linac Coherent Light Source (LCLS), SLAC National Accelerator Laboratory,is supported by the U.S. Department of Energy, Office of Science, Office of Basic EnergySciences under Contract No. DE-AC02-76SF00515. L.L.G. would like to thank the Volk-swagenStiftung for the financial support through the Peter-Paul-Ewald Fellowship. P.M.and M.B. acknowledge funding the Helmholtz Association via grant VH-NG-1105. X.H.V.,S.S, M.G., K.H., H.H. and G.K. acknowledge the NWO/FOM programme DESCO (VP149),which is financed by the Netherlands Organisation for Scientific Research (NWO). P.T.P.L.,J.E.tE. and G.K. acknowledge the NWO/CW ECHO grant ECHO.15.CM2.043. P.T.P.Lacknowledges financial support from the Netherlands Organization for Scientific Research(NWO) in the framework of the Chemical Sciences ECHO programme. The authors kindlythank Daniel Higley for reviewing the manuscript.
AUTHOR CONTRIBUTIONS
W.F.S., A.S. and H.D. conceived the experiment. M.B., G.C., G.L.D., H.D., K.H., L.L.G.,M.F.L., Y.L., P.M., A.R., A.S., W.F.S., S.S., P.W., S.Z., S.S., K.N., P.A.H and X.H.V.planned and participated in the experiment. Y.L, A.S and W.F.S. designed and fabricatedthe grating zone plate structures. M.G., H.H., G.K., K.H., W.F.S., S.S., P.T.P.L. andJ.E.tE. designed, fabricated and characterized the VO sample systems. L.L.G, W.F.S, S.Z.and X.H.V. performed data analysis and model development. The manuscript preparationwas lead by W.F.S with the participation of all co-authors. COMPETING FINANCIAL INTERESTS
The authors declare no competing financial interests.
REFERENCES J. St¨ohr,
NEXAFS Spectroscopy (Springer Series in Surface Sciences) (Springer, 2003). C. Bostedt, S. Boutet, D. M. Fritz, Z. Huang, H. J. Lee, H. T. Lemke, A. Robert, W. F.Schlotter, J. J. Turner, and G. J. Williams, Reviews of Modern Physics (2016).12alanced XAS R. Schl¨ogl, Angewandte Chemie International Edition , 3465 (2015). M. Imada, A. Fujimori, and Y. Tokura, Reviews of Modern Physics , 1039 (1998). R. W. Schoenlein, S. Chattopadhyay, H. H. W. Chong, T. E. Glover, P. A. Heimann, C. V.Shank, A. A. Zholents, and M. S. Zolotorev, Science , 2237 (2000). A. Cavalleri, M. Rini, H. H. W. Chong, S. Fourmaux, T. E. Glover, P. A. Heimann, J. C.Kieffer, and R. W. Schoenlein, Phys. Rev. Lett. , 067405 (2005). C. Boeglin, E. Beaurepaire, V. Halt´e, V. L´opez-Flores, C. Stamm, N. Pontius, H. A. D¨urr,and J.-Y. Bigot, Nature , 458 (2010). M. Kubin, J. Kern, S. Gul, T. Kroll, R. Chatterjee, H. Loechel, F. D. Fuller, R. G. Sierra,W. Quevedo, C. Weniger, J. Rehanek, A. Firsov, H. Laksmono, C. Weninger, R. Alonso-Mori, D. L. Nordlund, B. Lassalle-Kaiser, J. M. Glownia, J. Krzywinski, S. Moeller, J. J.Turner, M. P. Minitti, G. L. Dakovski, S. Koroidov, A. Kawde, J. S. Kanady, E. Y. Tsui,S. Suseno, Z. Han, E. Hill, T. Taguchi, A. S. Borovik, T. Agapie, J. Messinger, A. Erko,A. Foehlisch, U. Bergmann, R. Mitzner, V. K. Yachandra, J. Yano, and P. Wernet,Structural Dynamics , 054307 (2017). D. J. Higley, K. Hirsch, G. L. Dakovski, E. Jal, E. Yuan, T. Liu, A. A. Lutman, J. P.MacArthur, E. Arenholz, Z. Chen, G. Coslovich, P. Denes, P. W. Granitzka, P. Hart,M. C. Hoffmann, J. Joseph, L. L. Guyader, A. Mitra, S. Moeller, H. Ohldag, M. Seaberg,P. Shafer, J. Stoehr, A. Tsukamoto, H.-D. Nuhn, A. H. Reid, H. A. Duerr, and W. F.Schlotter, Review of Scientific Instruments , 033110 (2016). T. Kroll, J. Kern, M. Kubin, D. Ratner, S. Gul, F. D. Fuller, H. Loechel, J. Krzywin-ski, A. Lutman, Y. Ding, G. L. Dakovski, S. Moeller, J. J. Turner, R. Alonso-Mori,D. L. Nordlund, J. Rehanek, C. Weniger, A. Firsov, M. Brzhezinskaya, R. Chatterjee,B. Lassalle-Kaiser, R. G. Sierra, H. Laksmono, E. Hill, A. Borovik, A. Erko, A. Foehlisch,R. Mitzner, V. K. Yachandra, J. Yano, P. Wernet, and U. Bergmann, Optics Express ,22469 (2016). D. P. Bernstein, Y. Acremann, A. Scherz, M. Burkhardt, J. Stoehr, M. Beye, W. F.Schlotter, T. Beeck, F. Sorgenfrei, A. Pietzsch, W. Wurth, and A. Foehlisch, AppliedPhysics Letters , 134102 (2009). T. Katayama, Y. Inubushi, Y. Obara, T. Sato, T. Togashi, K. Tono, T. Hatsui,T. Kameshima, A. Bhattacharya, Y. Ogi, N. Kurahashi, K. Misawa, T. Suzuki, andM. Yabashi, Applied Physics Letters , 131105 (2013).13alanced XAS M. Buzzi, M. Makita, L. Howald, A. Kleibert, B. Vodungbo, P. Maldonado, J. Raabe,N. Jaouen, H. Redlin, K. Tiedtke, P. M. Oppeneer, C. David, F. Nolting, and J. Luening,Scientific Reports (2017). V. N. Strocov, T. Schmitt, U. Flechsig, T. Schmidt, A. Imhof, Q. Chen, J. Raabe,R. Betemps, D. Zimoch, J. Krempasky, X. Wang, M. Grioni, A. Piazzalunga, andL. Patthey, Journal of Synchrotron Radiation , 631 (2010). L. Strueder, S. Epp, D. Rolles, R. Hartmann, P. Holl, G. Lutz, H. Soltau, R. Eckart, C. Re-ich, K. Heinzinger, C. Thamm, A. Rudenko, F. Krasniqi, K.-U. Kuehnel, C. Bauer, C.-D.Schroeter, R. Moshammer, S. Techert, D. Miessner, M. Porro, O. Hlker, N. Meidinger,N. Kimmel, R. Andritschke, F. Schopper, G. Weidenspointner, A. Ziegler, D. Pietschner,S. Herrmann, U. Pietsch, A. Walenta, W. Leitenberger, C. Bostedt, T. Moeller, D. Rupp,M. Adolph, H. Graafsma, H. Hirsemann, K. Grtner, R. Richter, L. Foucar, R. L. Shoe-man, I. Schlichting, and J. Ullrich, Nuclear Instruments and Methods in Physics ResearchSection A: Accelerators, Spectrometers, Detectors and Associated Equipment , 483(2010). C. R. Rao,
Linear Statistical Inference and its Applications (Wiley-Interscience, 2001). N. B. Aetukuri, A. X. Gray, M. Drouard, M. Cossale, L. Gao, A. H. Reid, R. Kukreja,H. Ohldag, C. A. Jenkins, E. Arenholz, K. P. Roche, H. A. Duerr, M. G. Samant, andS. S. P. Parkin, Nature Physics , 661 (2013). A. Gray, J. Jeong, N. Aetukuri, P. Granitzka, Z. Chen, R. Kukreja, D. Higley, T. Chase,A. Reid, H. Ohldag, M. Marcus, A. Scholl, A. Young, A. Doran, C. Jenkins, P. Shafer,E. Arenholz, M. Samant, S. Parkin, and H. Duerr, Physical Review Letters (2016). P. T. P. Le, K. Hofhuis, A. Rana, M. Huijben, H. Hilgenkamp, G. A. Rijnders, J. E.ten Elshof, G. Koster, N. Gauquelin, G. Lumbeeck, C. Schueler-Langeheine, H. Popescu,F. Fortuna, S. Smit, X. H. Verbeek, G. Araizi-Kanoutas, S. Mishra, I. Vaskivskyi, H. A.Duerr, and M. S. Golden, Advanced Functional Materials , 1900028 (2019). J. St¨ohr, Advances in Optics and Photonics , 215 (2019). W. F. Schlotter, J. J. Turner, M. Rowen, P. Heimann, M. Holmes, O. Krupin, M. Messer-schmidt, S. Moeller, J. Krzywinski, R. Soufli, M. Fern´andez-Perea, N. Kelez, S. Lee,R. Coffee, G. Hays, M. Beye, N. Gerken, F. Sorgenfrei, S. Hau-Riege, L. Juha, J. Chalup-sky, V. Hajkova, A. P. Mancuso, A. Singer, O. Yefanov, I. A. Vartanyants, G. Cadenazzi,B. Abbey, K. A. Nugent, H. Sinn, J. Luening, S. Schaffert, S. Eisebitt, W.-S. Lee, A. Scherz,14alanced XASA. R. Nilsson, and W. Wurth, Review of Scientific Instruments , 043107 (2012). P. Heimann, O. Krupin, W. F. Schlotter, J. Turner, J. Krzywinski, F. Sorgenfrei,M. Messerschmidt, D. Bernstein, J. Chalupsk´y, V. H´ajkov´a, S. Hau-Riege, M. Holmes,L. Juha, N. Kelez, J. Luening, D. Nordlund, M. F. Perea, A. Scherz, R. Soufli, W. Wurth,and M. Rowen, Review of Scientific Instruments , 093104 (2011). C. Chang, A. Sakdinawat, P. Fischer, E. Anderson, and D. Attwood, Optics Letters ,1564 (2006). G. Blaj, P. Caragiulo, G. Carini, S. Carron, A. Dragone, D. Freytag, G. Haller, P. Hart,J. Hasi, R. Herbst, et al. , Journal of synchrotron radiation22