SDO (Split-and-Delay Optics)
The BL3 OH has split-and-delay optics (SDO) permanently installed. By using SDO, a double pulse XFEL with a time delay can be generated by dividing one XFEL pulse, which enables branches with different optical path lengths to be propagated, then recombined on the same optical axis. By varying the optical path length for one branch, the time difference between split pulses can be controlled from approximately 0 fs to 100 ps, and X-ray excitation dynamics on the femtosecond to picosecond scale, as well as spontaneous fluctuations in the material, can be measured. This page introduces the elemental configuration and performance of SDO.
If you are planning an experiment using SDO, please be sure to contact the XFEL Utilization Division for information on the available wavelength conditions before applying for an assignment.
※SDO was developed in collaboration with the group led by Professor Kazuto Yamauchi at Osaka University.
Configuration of the optical systems
Overall
Figure 1 displays the layout of SDO. The SDO at SACLA consists of a total of six Si (220) crystal optical devices. The XFEL pulse incident on the SDO is split into two replicated pulses by the first crystal beam splitter (BS). The first branch (blue) is reflected four times by two channel-cut crystals (CC1/ CC2) in the (+, –, –, +)arrangement (CC branch). The optical path length of the CC branch is fixed at a certain photon energy. In the other branch (red), in addition to the reflection from the BS, it returns to the same optical axis as the CC branch after a total of four reflections by two reflecting elements (BR1/ BR2) and the beam merger (BM). BR1 and BR2 are moveable in the direction perpendicular to the crystal surface, enabling control over the optical path length (delay branch). This optical system covers an energy range from 5 to 15 keV. The delay time range differs depending on the energy, please see figure 2 for information. The delay time can be set with a time step of 1 fs or less over the entire energy range. In addition, it is not effected by jitter, which are fluctuations in time synchronization which is a general problem for normal pump-probe measurements, and this enables delay time stability on the order of attoseconds.
This optical system is installed in a highly-hermetically sealed box, and the interior of the box can be replaced with He. When replacing with He, the box is feedback controlled to maintain the He concentration at 80%. In addition, no deterioration in stability has been observed with the introduction of He gas.

Fig. 1 Schematic optical layout of SDO system

Fig. 2 Range of delay time as a function of photon energy. At positive delays the CC-branch pulse comes earlier.
Beam splitter / merger (BS / BM)
The SDO at SACLA uses a wavefront splitting crystal (figure 3) with a crystal edge polished to high-precision as the BS/BM element [1]. The crystal edge is inserted on the optical axis of the XFEL, and only the part irradiated with the BS is reflected and divided into the delay branch, with the rest divided into the CC branch (Similarly in BM, only the delay branch is reflected with the BM). With this wavefront splitting crystal,
- split pulses with the same spectral distribution can be generated
- easy intensity matching control is possible by controlling the insertion position of the crystal edge
However, immediately after the BM, it is necessary to overlap the spatially separated split beams on the sample surface, to obtain a slight horizontal difference in the angle of incidence on the sample.

Fig. 3 Conceptual picture of wavefront division with an edge-polished crystal (left) and an example beam profile measured just downstream of the SDO (right).
Speckle free channel-cut crystals (CC1/ CC2)
The channel-cut crystal used in the CC branch was developed using a processing technology called Plasma Chemical Vaporization Machining (PCVM), which originated in the Yamauchi Laboratory of Osaka University [2]. With channel-cut crystals, X-rays are reflected twice within a single device, so they are very useful devices that can stabilize the parallelism between the incident optical axis and the reflected optical axis on the level of nanoradians. However, due to the structure, it is difficult to apply high-precision, distortion-free polishing techniques, such as CMP (Chemical Mechanical Polishing), so polishing with strong mechanical action or etching with a solution has been used for the final treatment of the surface. As a result, it is difficult to achieve a distortion-free, smooth surface, allowing for disruptions of the wave surface of the reflected beam, resulting in modulations in intensity (speckle). By applying PCVM to the final surface treatment, since it is a pure chemical processing method, it is possible to achieve distortion-free, smooth surface conditions, and maintain well-organized wave surfaces for XFEL (figure 3). PVCM is also applied to the Si(111) DCCM installed in BL3 EH1, and is used for improving pump-probe measurements [3].
SDO performance
Pulse energy
Each branch of SDO is equipped with a non-destructive intensity monitor that is commonly used at SACLA, and the relative pulse energy of each split pulse can be measured for each shot. In addition, by using the beam monitor equipped in the experimental hutch, which has the conversion coefficient calibrated to the absolute value, the absolute value of the pulse energy for each shot can be derived. Generally, the pulse energy of the split pulse is calculated using two stable CC branch intensity monitors and a calibrated beam monitor. In this case, the measurement error is approximately 5% rms (figure 4).

Fig. 4 Results of pulse energy diagnostics of SDO. (left) correlation between a calibrated beam monitor and a monitor at the CC branch measured by blocking the delay branch; (middle) Error distribution of the CC branch monitor; (right) sum of pulse energy vs normalized intensity difference between split pulses.
Since this optical system is comprised of Si(220) crystals, it also functions as an X-ray monochromator. While it has the advantage of being able to use XFEL with monochromaticity and a relative bandwidth ΔE/E = 5.6 × 10-5, it has the disadvantage of losing many photons. In the SASE mode, the total of both split pulses is approximately 0.5 μJ. With the recently developed self-seed mode [4], high-intensity XFEL with high monochromaticity can be generated, drastically improving the pulse energy when SDO is used, averaging between 3 - 8 μJ.
Accuracy of the delay time
The change in the delay time due to the amount of movement of the BR1/BR2 crystal elements is precisely calibrated, and the delay time can be controlled with an error of less than 1%. In addition, the zero time difference is determined for each experiment by the autocorrelation method [5]. Partial overlapping of unfocused split beams causes interference fringes in the overlapped region when the delay time is less than or equal to the coherence time. By evaluating the visibility of the interference fringes, it is possible to determine the zero time difference with an accuracy of 1 to 2 fs (figure 5).

Fig. 5 Example result of autocorrelation measurement.
Stability of the focusing position
Focused beams are generally used in experiments with SDO. In the case of a focused beam, the focused position has sensitive changes due to fluctuations in the beam angle variations of several hundred nanoradians. With the SDO at SACLA, the optical system is build on a robust stone surface plate, and the control mechanism of each crystal is comprised of a highly rigid stage, shown in figure 6, enabling the focused position to remain stable in the short and long term.

Fig. 6 (Left) long-term stability of relative focal pointing and (right) distribution of short-term pointing jitter. During this measurement, split beams are focused to 1 μm spot with a KB mirror system at EH4c.
Even if the delay time is changed, the focused position will change due to the change in the position of the stage. Since the change in the position of the stage is reproducible, the reproducibility is confirmed in the change of the focused position with respect to the delay time (figure 7). In principle, pointing fluctuations occurring when the delay time changes can be cancelled out, but the drift at the focused position increases due to the heat generated by driving the continuous stage. Therefore, continuous scanning of the delay time is currently difficult when pointing needs to be stabilized on the order of a few μm.

Fig. 7 Pointing stability during delay scans.
Energy scan
With this optical system, not only the delay time, but also the X-ray energy can be changed. This enables X-ray pump/X-ray probe spectroscopy measurements and measurements of the dynamics of the electronic state changes associated with X-ray excitation. It is possible to change the energy for any branch, but from the perspective of optical system stability, the energy of the CC branch is typically scanned. Therefore, the positive and negative delay time are switched. Also, if you change the energy of only one branch, the delay time will change at the same time, but this can be compensated for. Such measurements are only possible in the wide bandwidth SASE mode, and the measurable energy range depends on the bandwidth of the SASE, but is approximately 30eV.

Fig. 8 Delay time dependence of absorption spectra of Cu.
SDO control
SDO is a complex optical system, but it can be controlled with a user-friendly Python module (sdopy). With sdopy, the following operations can be performed semi-automatically:
- Changes of delay time and changes of the position of various equipment
- Semi-automatic realignments of each crystal element
- Automatic measurements of the conversion coefficient to absolute pulse energy of the intensity monitor
- Measurements of the intensity ratio distribution between the split pulses
- Changes of the energy of each branch and position changes of various equipment
source: /xdaq/work/share/SDO/sdopy.py
References:
[1] T. Hirano et al., J. Synchrotron Rad. 25, 20 (2018).
[2] T. Hirano et al., Rev. Sci. Instrum. 87, 063118 (2016).
[3] T. Katayama et al., J. Synchrotron Rad. 26, 333 (2019).
[4] I. Inoue et al., Nat. Photon. 13</strong style="color: #0000ff;">, 319 (2019).
[5] T. Osaka et al., IUCrJ 4, 728 (2017).