Polarization control with an X-ray phase retarder for high-time-resolution pump–probe experiments at SACLA

Polarization control using an X-ray phase retarder in combination with an arrival timing diagnostic on BL3 of SACLA is reported.

Polarization is a fundamental property of light. In the X-ray region, anisotropic and magnetic properties of matter have been widely investigated using various polarization states. A diamond X-ray phase retarder (XPR) (Hirano et al., 1991;Giles et al., 1994;Lang & Srajer, 1995) has been installed on BL3 of SACLA to control the polarization of the hard XFEL beam (Suzuki et al., 2014) so as to study the ultrafast dynamics of chemical bonding and magnetic states. In particular, pumpprobe X-ray magnetic circular dichroism (XMCD) spectroscopy is a powerful method to investigate spin dynamics such as a demagnetization and magnetization reversal observed on ISSN 1600-5775 a picosecond to femtosecond time scale (Kirilyuk et al., 2010;Takubo et al., 2017). Since the diamond XPR crystal has its diffraction plane rotated by 45 from the horizontal, the horizontally polarized incident XFEL beam is decomposed into equal partsand -polarized radiation relative to the diffraction plane. According to the dynamical theory of X-ray diffraction (Batterman & Cole, 1964), the XPR crystal produces a phase retardation between these two components near the Bragg condition, and consequently one is able to control the polarization states. In a conventional configuration [which we call a 'mono-XPR' (or scheme A) configuration], a monochromatic beam with a bandwidth of ÁE/E ' 1 Â 10 À4 produced with a perfect crystal monochromator is employed as the incident beam on the XPR crystal, because an energy bandwidth and/or angular divergence can cause degradation of the polarization states converted with the XPR crystal (Suzuki et al., 2014). However, one may consider locating the XPR crystal upstream of the monochromator [i.e. an 'XPRmono' (or scheme B) configuration] so as to expand the applicable range of experiments with various polarization states. For example, scheme B enables us to combine the XPR system with the arrival timing monitor (TM) on BL3 of SACLA, because the TM system requires the transport of a pink beam with a bandwidth of ÁE/E ' 5 Â 10 À3 through the beamline optics (including the XPR crystal) located in the optics hutch (OH) (Katayama et al., 2016).
In this study, we evaluated the degree of circular polarization (P C ) of the XFEL beam in scheme B that combines polarization control with the TM on BL3 of SACLA. We experimentally confirmed a high degree of P C , which is consistent with the calculated value and almost the same as that for scheme A (Suzuki et al., 2014).  (Tono et al., 2013). In the OH, the pink XFEL beam reflected with mirrors was incident on the XPR crystal (Suzuki et al., 2014). In this study, a diamond (100) crystal 1.5 mm thick was used in a 220 symmetric Laue geometry. A branch beam of the XFEL, separated with a grating in the OH, was used for the arrival TM in experimental hutch EH1 (Katayama et al., 2016). After passing through the XPR crystal, the main XFEL beam was monochromated at the Pt L 3 edge with a pair of channel-cut crystals (CCs) in a (+, À, À, +) geometry, which maintains the height of the beam, with Si(111) reflections installed in EH1.

Experimental
For comparison, we set up the conventional mono-XPR (scheme A) configuration as shown in Fig. 1(b). A doublecrystal monochromator (DCM) with Si(111) reflections was used to produce the monochromatic beam. The monochromatic XFEL pulse was incident on the XPR crystal. The grating, TM and CCs monochromator were not used in this configuration.
In EH2, we installed two Kapton scattering beam monitors (BMs) to measure the intensities of the horizontal and vertical linear polarized components as shown in Fig. 1. An FePtPd film (50 nm thick) grown on an MgO(100) substrate by magnetron sputtering at 773 K was used to evaluate the P C of the XFEL by measuring the XMCD. The sample forms an L1 0 structure with an out-of-plane easy magnetization direction (Seki et al., 2011;Takubo et al., 2017). Using an Nd-Fe-B permanent magnet, an external magnetic field of 0 H = AE0.59 T was applied along the surface normal of the sample film to saturate the magnetization. The sample was maintained at room temperature during measurements. We used two pairs of the sample and the permanent magnet, with opposite field directions for each pair, to measure the XMCD signal. The two samples were cut from the same parent film, and the magnetic field strengths were prepared to be nearly the same for the two configurations. The remaining systematic errors in the magnetic asymmetry were corrected. The XFEL beam was incident on the sample in the surface normal direction. To measure the absorbed intensity at the Pt L 3 edge, we detected the Pt L (9.443 keV) fluorescence with a multiport chargecoupled device (MPCCD) detector (Kameshima et al., 2014), as shown in Fig. 1. Fig. 2(a) shows the intensities of the horizontal (I x ) and vertical (I y ) polarization components as a function of the offset angle of the XPR crystal from the 220 symmetric Laue geometry at 11.567 keV detected with the Kapton scattering BMs in scheme B. This photon energy was selected so as to maximize the value of the XMCD. From the measured values of I x and I y , the degree of linear polarization, P L , was determined using the equation

Results and discussion
where Q is the correction factor of the BMs used in this experiment. The factor was estimated to be Q = 0.5990 AE 0.0005 by a measurement of the intensity of the horizontally polarized component without the XPR crystal (Suzuki et al., 2014). Fig. 2(b) shows the XPR crystal angle dependence of P L . We also operated the same measurement in scheme A. One finds good agreement between the two configurations, which shows that polarization control with the XPR crystal works properly even for scheme B. At the points of P L = 0, the XPR generates AE/2 phase retardation, and we obtain circular polarization with right-and left-helicity, respectively. A small discrepancy between the two configurations is due to the difference between the bandwidths obtained with the CCs and DCM (Suzuki et al., 2014). The monochromatic beam generated with the CCs has a bandwidth about 0.8 times smaller than that obtained with the DCM due to the difference in their reflection geometries: (+, À, À, +) for the CCs and (+, À) for the DCM. Next, we estimated P C in scheme B by measuring the XMCD of the FePtPd film at the Pt L 3 edge. A magnetic asymmetry ratio, R, was determined by the equation where I þ (I À ) is the absorbed intensity in the FePtPd film detected with the MPCCD for a magnetic field applied in the direction antiparallel (parallel) to the X-ray incident direction. I þ 0 (I À 0 ) is the intensity of the incident XFEL beam, which is the sum of the output of the BMs. Fig. 3 shows the values of R as a function of the offset angle of the XPR crystal in schemes A and B, with the calculation curves obtained by the same method as that used in a previous study (Suzuki et al., 2014). This result indicates that P C in scheme B is almost the same as that of scheme A with a high value of 0.97, which was obtained in the previous study (Suzuki et al., 2014). The uncertainty in P C in scheme B was estimated to be AE0.06. The discrepancy in the offset angles between À20 and 20 arcsec is due to the influence of the bandwidth, which was seen in P L (Suzuki et al., 2014). For negative offset angles, R obtained in scheme B is lower than that in scheme A, which is probably due to a glitch as discussed in the previous study (Suzuki et al., 2014).
From the experiments, we confirmed that a high degree of P C was retained in scheme B at 11.567 keV. However, according to the dynamical theory of X-ray diffraction (Batterman & Cole, 1964), a CCs monochromator could introduce a phase difference () between theand -polarization components, which leads to degradation of the polarization states produced with the XPR crystal. To investigate the effect of CCs on P C over a wide photon energy range, we calculated the values of P C based on the dynamical theory of diffraction. Since the angular divergence of the XFEL beam ($2 mrad) is sufficiently smaller than the Darwin width of the Si(111) reflection ($20 mrad), we only considered the effect of the energy spread. Fig. 4(a) shows rocking curves for theand -polarization components and the resulting and P C after the four-fold reflections of the CCs at E 0 = 11.567 keV. For the calculation of P C , we assumed that the incident XFEL beam has circular polarization with P C = 1. Although P C decreased at the edges of the rocking curves, the   The magnetic asymmetry ratio obtained from the FePtPd film at the Pt L 3 edge as a function of the offset angle of the XPR crystal. The red solid circles and blue open circles represent the values in the mono-XPR (scheme A) and XPR-mono (scheme B) configurations, respectively. The green and black solid curves represent the calculated values for the scheme A and B configurations assuming Gaussian bandwidths of ÁE/E = 1.1 Â 10 À4 and 8.6 Â 10 À5 , respectively. weighted average value was maintained at 0.96 at 11.567 keV. This result is consistent with our experimental results, showing that P C in scheme B has almost the same value as that in scheme A. Note that the experimental value of P C ' 0.97 in scheme B included the effects of not only the degradation by the CCs but also the narrower bandwidth than that obtained by the DCM. Fig. 4(c) shows the weight-averaged values of P C after the four-fold reflections of the CCs as a function of photon energy. This result shows that a high P C can be maintained in a high photon energy region (larger than $10 keV), which includes XMCD for the L edges of 5d elements such as Pt, Ir and Os (Wienke et al., 1991;Schü tz et al., 1989). On the other hand, the value of P C decreases drastically in scheme B in a lower energy region of less than $10 keV, which includes the K edges of 3d transition metals and the L edges of rare earth elements (Schü tz et al., 1987(Schü tz et al., , 1989Giles et al., 1994). For example, Fig. 4(b) shows rocking curves with and the values of P C after the four-fold reflections of the CCs at E 0 = 7 keV. To improve P C in this energy region, a single CC monochromator can be used in scheme B. Alternatively, a high-intensity monochromatic beam generated with the self-seeded XFEL scheme (Amann et al., 2012;Inoue et al., 2019) is available for ultrafast magnetic measurements.