X-ray speckle visibility spectroscopy (XSVS) is a method for studying dynamics in disordered systems. This method was originally developed in the visible-light region, in which an intense laser can be used. When applied in the X-ray region, where the number of photons is much smaller than in the visible-light region, it suffers from photon statistics. In this paper, we quantitatively discuss the effect of photon shot noise on XSVS analyses. The effect is experimentally confirmed using sequential speckle patterns from Brownian polystyrene nanospheres in glycerol.
© 2012 OSA
Coherent X-rays are attracting much attention in various scientific fields . When coherent X-rays impinge upon a disordered system, a grainy scattering pattern called speckle is observed . Such a speckle pattern reflects the exact configuration of the scatterers inside the system, whereas scattering patterns obtained by conventional scattering methods reflect the spatially averaged configuration of the scatterers. When the system evolves with time, the corresponding speckle pattern also changes. Temporal changes in the speckle patterns therefore provide information on the system dynamics. This technique is called X-ray photon correlation spectroscopy (XPCS) [3, 4]. In XPCS, time correlation of the X-ray intensities in speckle patterns is analyzed. XPCS has been applied to various disordered systems, such as colloidal suspensions [5–7], homopolymer blends , block copolymer micelles and vesicles [9, 10], nanoparticles in supercooled liquids , alloys [12, 13], and antiferromagnetic materials .
However, the application of XPCS is limited in two ways. First, its time resolution is limited by the frame rate of the detector. A two-dimensional (2D) detector is commonly used in XPCS, so the time resolution of XPCS experiments is at present limited to the order of milliseconds. Although the frame rates of zero- and one-dimensional detectors surpass those of 2D detectors, the use of a 2D detector is preferred because averaging over pixels significantly reduces the experimental time and is much more suitable for measuring transient phenomena . Secondly, the long exposure to X-rays often gives rise to radiation damage of the sample, which affects its dynamics. In XPCS, fixed positions on the sample are irradiated and the sample is easily damaged.
X-ray speckle visibility spectroscopy (XSVS) can overcome these problems. Speckle visibility spectroscopy (SVS) [16, 17] was originally developed in the visible-light region. In SVS or XSVS, the visibility of a speckle pattern is calculated and its dependence on exposure time is analyzed. When the system is static within the exposure time, the speckle patterns are stable. In contrast, when the scatterers move during exposure, the speckle pattern is blurred and the visibility decreases accordingly. Consequently, the dependence of the visibility on the exposure time is closely related to the dynamics of the system. The time resolution of XSVS is determined by the minimum exposure time and is not limited by the frame rate of the detector. XSVS can therefore achieve higher time resolution than XPCS can, even using a 2D detector. Moreover, it is possible to change the irradiation position on the sample for every exposure; this reduces the degree of radiation damage. In this regard, XSVS is expected to be a powerful tool in next-generation synchrotron X-ray facilities, in which much more intense X-rays can be used [18, 19]; this will open up a way of measuring microsecond/nanosecond dynamics, which is hard to achieve with current X-ray/neutron techniques. Compared to conventional SVS, XSVS enables us to study nanoscale structural fluctuation in virtue of a short wavelength of X-rays. Opaque specimens, which often show complex and interesting dynamics, can be investigated with XSVS due to the high penetration power of X-rays.
Previous studies have discussed the relationship between visibility and dynamics [16, 17]. This relationship, however, cannot be directly applied to XSVS analysis; XSVS is different from SVS in that the number of scattered photons is much smaller because of the small available flux of the incident X-ray beam. This will further decrease when XSVS is used for measuring fast dynamics. The effect of shot noise therefore needs to be properly included.
To the best of our knowledge, this is the first report of XSVS. In this paper, we quantitatively describe the effect of photon shot noise on the relationship between visibility and dynamics. Then, the importance of the effect is shown for the measurement of Brownian nanospheres.
2.1 X-ray photon correlation spectroscopy
In XPCS, a normalized intensity correlation function is measured; it is defined as
The function is related to the intermediate scattering function via the Siegert relation :
2.2 X-ray speckle visibility spectroscopy
In XSVS, the visibility of speckle patterns is calculated; this is defined asEq. (4) is substituted by the averages over pixels and time: assuming that the exposure time of the sequential images is , the visibility of the speckle pattern is calculated as
It has been shown  that is related to byEquation (6), however, only holds for the ideal condition, where the number of scattered photons is so large that the effect of shot noise is negligible. Let us consider the situation where the exposure time is much longer than the relaxation time of the system, for example. The speckle images from the system change with time and form various patterns on the detector during exposure to the X-ray beam. If the number of detected photons per pixel is large enough for the shot noise and the dark noise of the detector to be negligible, every is the same, and the visibility calculated by Eq. (4) becomes zero. Conversely, if the number of counted photons per pixel is small, has a nonzero variance because of the shot noise, thus the visibility is larger than zero. Since approaches zero as increases, the right-hand side of Eq. (6) becomes zero and therefore Eq. (6) does not hold. In SVS, the effect of the shot noise is negligible because of the high flux from laser sources, and Eq. (6) can simply be used. In XSVS, however, the effect cannot be ignored and a proper treatment of photon shot noise is required.
Now we show how to include the effect of shot noise in Eq. (6). First, let us assume that (i) the dark noise of the detector is much smaller than the scattering intensity and (ii) the size of an electron cloud on the detector produced by an incoming photon is small enough for the charge not to be shared between different pixels. Then, is proportional to the counted number of photons during a time , . Even if the configuration of the scatterers temporally does not change, is not always the same because of the shot noise. The configuration only determines the expected value of , . follows the Poisson distribution when the configuration of the scatterers temporally does not change. Ensemble average in Eq. (5) corresponds to average over all configurations of the scatterers and all cases of the photon statistics for each configuration. Let us define and as averages over configurations and photon statistics for each configuration, respectively. Then, the following relations hold when the relation of is taken into account:
Next, let us consider the case when the electron clouds are shared with different pixels. In this case, is proportional to effective counted number of photons during a time , , which is defined based on the ratio of the accumulated charge in a pixel to the total charge on the detector produced by a single X-ray photon. When the configuration of the scatterers temporally does not change, does not follow the Poisson distribution. For example, in the case of a spatially uniform exposure, the deviation of is smaller than by a factor of (0 ≤ ≤ 1), while ensemble-average of equals to . The factor is the constant determined by the spread of an electron cloud on the detector and equals to unity when the cloud is small enough for the charge not to be shared with different pixels . Considering the sharing effect of electron clouds, Eq. (7) and Eq. (8) are rewritten as
When the intensity of the incident X-ray beams is constant, is proportional to ; is also proportional to at a long exposure time , so is proportional to at large values of , regardless of the forms of , if the standard deviation of the dark noise is negligible. For a Brownian motion system with a diffusion coefficient , , with a relaxation time , and the visibility is given byFigure 1 shows the dependence of the visibility on the exposure time when = 1, calculated using Eq. (16), for a Brownian system: four cases, i.e., 0.1, 1, 10, and ∞, are compared. When is comparable to or shorter than , the dependence of on is greatly affected by the shot noise. The effect of the shot noise deserves careful attention, to obtain information on the dynamics of the system in XSVS experiments.
3. Materials and methods
Polystyrene latex nanospheres (Alfa Aesar, USA) in glycerol were used as the sample. The sphere diameter was 100 nm and the spheres were used without further purification. The latex was mixed with glycerol and kept in a vacuum desiccator for 205 h to prepare a colloidal suspension with a volume fraction of 5%. A sample cell of thickness 1 mm covered with Mylar film was filled with the suspension. XPCS and XSVS experiments were performed at BL40XU, SPring-8 (Hyogo, Japan). The experimental setup has been described elsewhere . An X-ray beam of 10.5 keV from a helical undulator  was used and a pinhole of diameter 5 μm and a guard pinhole of diameter 25 μm were installed upstream of the sample to make the X-ray beam coherent enough to produce speckle patterns. The temperature of the sample was controlled to be −20 °C using a temperature-controlled stage (THMS600, Linkam Scientific Instruments Ltd., UK) and the speckle patterns were recorded using an interline charge-coupled device detector (CCD) (C4880-80, Hamamatsu Photonics Ltd.) coupled with an X-ray image intensifier . Each pixel of the detector consists of a photodiode and a CCD. Photoelectrons produced by photons incoming to the pixel are accumulated in the photodiode and they are rapidly transferred to the CCD when each exposure is finished. Since the next exposure can be started as soon as the photoelectrons is transferred to CCD, the time gap between the end of a frame and the start of the next one is very small (microseconds) . In our experiment, the exposure time per frame was 36 ms (36 ms) and the time gap between frames was negligible compared with the exposure time. The distance between the sample and the detector, L, was 3 m and the beam size at the sample, d, was measured by a knife-edge method to be 5 μm. The speckle size s was estimated to be 71 µm using the relation , where λ is the wavelength of the X-ray. The pixels were binned so that the effective pixel size, 80 µm, corresponds to the speckle size. Two hundred sequential speckle patterns, after subtraction of an averaged dark image, were used as scattering images in the following analysis. In our experimental setup, the scattering intensity from the sample was high enough for the effect of the dark noise of the detector to be negligible in the XSVS analysis.
4. Results and discussion
In XPCS analysis, the normalized intensity correlation function is calculated using Eq. (2). The calculated is well fitted by
XSVS analysis is performed as follows. To calculate the visibility of the patterns for exposure time (), the scattering images are divided into units of successive images. The images in each unit are accumulated to produce a single scattering image of , as shown in Fig. 2 , and a series of scattering images is produced. Then, the visibility of the images for the exposure time is calculated according to Eq. (5). More than 3600 pixels are used for the calculation of each . Recently, it was pointed out that the number of pixels and the relation between the speckle size and the pixel size lead to differences between the variances of the expected values of the pixel-averaged and the ensemble-averaged intensities, whereas the expected value of the intensity averaged over pixels is the same as that of the ensemble-averaged intensity . The difference in the intensity variance is, however, negligible in our XSVS experiments because the relative difference scales as the inverse of the total pixels used for the calculation. The calculated visibility is therefore considered to be a good approximation to the real visibility defined by Eq. (4). The calculated visibility at each is fitted by a model function that includes the effect of the shot noise for a Brownian motion system:Figure 3(b) is a typical fitting result for = 0.0335 nm−1. The fitting result for the case when the effect of shot noise is not considered ( in Eq. (18) is infinity) is also shown for comparison. The fitted curve well describes the calculated visibility and this result is favorable compared with the data analysis, which does not include the effect of the shot noise. Relaxation time determined by the Eq. (18) is 198.7 s ± 19.5 s and well agrees that by XPCS (206.6 s ± 21.7 s). On the contrary, determined by XSVS without consideration of the effect of the shot noise is 38.2 s 1.5 s and this disagreement with the result of XPCS means that the effect of the shot noise deserves careful attention in XSVS analysis. The results at other values are also fitted by Eq. (18).
The relaxation times determined by XPCS and XSVS are shown in Fig. 3(b). The good agreements of the relaxation times determined by both methods indicate the validity of Eq. (14). Figure 3(c) shows the circularly averaged intensity of the speckle patterns and at each determined by XSVS analysis. The fact that is proportional to the circularly averaged intensity also supports the validity of Eq. (14). In order to obtain the value of , it is required to measure the spread of the electron cloud on the detector produced by a single counted photon. However, this is out of the scope of the present work.
As described above, the time resolution of XSVS is not limited by the frame rate of detector. This makes it possible to study microsecond/nanosecond dynamics which is hard to be studied using conventional XPCS, although the present study has demonstrated XSVS in millisecond dynamics using the technique summing successive images. Recently, a split-pulse technique using an X-ray free-electron laser was reported and is also expected to improve the time resolution of XPCS . The timescale over which the split-pulse technique can perform measurements is determined by the minimum and maximum delay times of the split pulses. Since the path difference of the pulses determines the delay time [28, 29], the split-pulse technique covers dynamics on a timescale shorter than sub-microseconds. XSVS is expected to measure nano- and micro-second dynamics by controlling the exposure time. Complementary use of XSVS and the split-pulse technique therefore gives us further opportunities for understanding hierarchical dynamics in various disordered systems. Moreover, XSVS can reduce the degree of radiation damage by changing the irradiation position on the sample for every exposure. The effect of the damage will become severe in next-generation synchrotron X-ray facilities and XSVS will be a powerful method in the future.
In this paper, we quantitatively describe the effect of shot noise on XSVS. It is shown that the shot noise significantly affects the visibility of speckle patterns. The effect of shot noise on visibility is properly included in the relationship between visibility and the corresponding intermediate scattering function. XSVS is successfully used to measure the relaxation times of polystyrene nanospheres in glycerol. The effect of shot noise is important in XSVS since the number of X-ray photons available is limited. The present study shows that the effect is properly accounted for in analyses, even when the number of photons is small. By combining XSVS and future X-ray sources, it will be possible to fill the gap between XPCS and quasi-elastic scattering techniques without radiation damage to samples.
The authors thank Drs. N. Yagi and N. Ohta (JASRI) and Mr. H. Kishimoto (Sumitomo Rubber Industry Ltd.) for their support in performing the experiments at SPring-8. The authors also thank the anonymous reviewers for their valuable comments. The XSVS experiments were performed under the approval of JASRI (Proposal Number: 2011A1112, 2011B1131). This study has been partly supported by JSPS KAKENHI (Grant Number 24710092).
References and links
1. K. A. Nugent, “Coherent methods in the X-ray sciences,” Adv. Phys. 59(1), 1–99 (2010). [CrossRef]
2. M. Sutton, S. G. J. Mochrie, T. Greytak, S. E. Nagler, L. E. Berman, G. A. Held, and G. B. Stephenson, “Observation of speckle by diffraction with coherent X-rays,” Nature 352(6336), 608–610 (1991). [CrossRef]
3. G. Grübel and F. Zontone, “Correlation spectroscopy with coherent X-rays,” J. Alloy. Comp. 362(1-2), 3–11 (2004). [CrossRef]
4. M. Sutton, “A review of X-ray intensity fluctuation spectroscopy,” C. R. Phys. 9(5-6), 657–667 (2008). [CrossRef]
5. S. B. Dierker, R. Pindak, R. M. Fleming, I. K. Robinson, and L. Berman, “X-Ray photon correlation spectroscopy study of Brownian motion of gold colloids in glycerol,” Phys. Rev. Lett. 75(3), 449–452 (1995). [CrossRef]
6. T. Thurn-Albrecht, W. Steffen, A. Patkowski, G. Meier, E. W. Fischer, G. Grübel, and D. L. Abernathy, “Photon correlation spectroscopy of colloidal palladium using a coherent x-ray beam,” Phys. Rev. Lett. 77(27), 5437–5440 (1996). [CrossRef]
7. L. B. Lurio, D. Lumma, A. R. Sandy, M. A. Borthwick, P. Falus, S. G. J. Mochrie, J. F. Pelletier, M. Sutton, L. Regan, A. Malik, and G. B. Stephenson, “Absence of scaling for the intermediate scattering function of a hard-sphere suspension: Static and dynamic x-ray scattering from concentrated polystyrene latex spheres,” Phys. Rev. Lett. 84(4), 785–788 (2000). [CrossRef]
8. D. Lumma, M. A. Borthwick, P. Falus, L. B. Lurio, and S. G. J. Mochrie, “Equilibrium dynamics in the nondiffusive regime of an entangled polymer blend,” Phys. Rev. Lett. 86(10), 2042–2045 (2001). [CrossRef]
9. S. G. J. Mochrie, A. M. Mayes, A. R. Sandy, M. Sutton, S. Brauer, G. B. Stephenson, D. L. Abernathy, and G. Grübel, “Dynamics of block copolymer micelles revealed by x-ray intensity fluctuation spectroscopy,” Phys. Rev. Lett. 78(7), 1275–1278 (1997). [CrossRef]
12. S. Brauer, G. B. Stephenson, M. Sutton, R. Brüning, E. Dufresne, S. G. J. Mochrie, G. Grübel, J. Als-Nielsen, and D. L. Abernathy, “X-ray intensity fluctuation spectroscopy observations of critical dynamics in Fe3Al,” Phys. Rev. Lett. 74(11), 2010–2013 (1995). [CrossRef]
14. O. G. Shpyrko, E. D. Isaacs, J. M. Logan, Y. Feng, G. Aeppli, R. Jaramillo, H. C. Kim, T. F. Rosenbaum, P. Zschack, M. Sprung, S. Narayanan, and A. R. Sandy, “Direct measurement of antiferromagnetic domain fluctuations,” Nature 447(7140), 68–71 (2007). [CrossRef]
17. R. Bandyopadhyay, A. S. Gittings, S. S. Suh, P. K. Dixon, and D. J. Durian, “Speckle-visibility spectroscopy: A tool to study time-varying dynamics,” Rev. Sci. Instrum. 76(9), 093110 (2005). [CrossRef]
18. H. Abe, S. Adachi, K. Amemiya, K. Asakura, L. Chavas, N. Funamori, T. Furuya, R. Hajima, K. Hirano, T. Honda, E. Kako, H. Kawata, S. Kishimoto, Y. Kitajima, Y. Kobayashi, S. Michizono, Y. Murakami, H. Nakai, N. Nakamura, M. Nomura, T. Obina, S. Sakanaka, N. Sei, and Y. Shinohara, Yoshio Takahashi, Yukio Takahashi, K. Tsuchiya, S. Wakatsuki, M. Yamamoto, and W. Yashiro, “Energy Recovery Linac preliminary design report” (2011), http://pfwww.kek.jp/ERLoffice/detabase/ERL%20Preliminary%20Design%20Report_web.pdf.
19. A. Fluerasu, L. Wiegart, and M. C. Dayton, “Preliminary design report for the coherent hard X-ray (CHX) beamline at NSLS-II” (2010), http://www.bnl.gov/nsls2/beamlines/files/pdf/CHX_pdr.pdf.
20. B. J. Berne and R. Pecora, Dynamic Light Scattering With Application to Chemistry, Biology, and Physics (Dover Publications, 2000).
21. E. Dufresne, R. Brüning, M. Sutton, B. Rodricks, and G. B. Stephenson, “A statistical technique for characterizing X-ray position sensitive detectors,” Nucl. Instrum. Methods Phys. Res. Sect. A 364(2), 380–393 (1995). [CrossRef]
22. Y. Shinohara, H. Kishimoto, T. Maejima, H. Nishikawa, N. Yagi, and Y. Amemiya, “X-ray photon correlation spectroscopy of filler in rubber,” Jpn. J. Appl. Phys. 46(4), L300–L302 (2007). [CrossRef]
23. T. Hara, T. Tanaka, T. Seike, T. Bizen, X. Maréchal, T. Kohda, K. Inoue, T. Oka, T. Suzuki, N. Yagi, and H. Kitamura, “In-vacuum X-ray helical undulator for high flux beamline at SPring-8,” Nucl. Instrum. Methods Phys. Res. Sect. A 467–468, 165–168 (2001). [CrossRef]
24. Y. Shinohara, R. Imai, H. Kishimoto, N. Yagi, and Y. Amemiya, “Indirectly illuminated X-ray area detector for X-ray photon correlation spectroscopy,” J. Synchrotron Radiat. 17(6), 737–742 (2010). [CrossRef]
25. S. M. Gruner, M. W. Tate, and E. F. Eikenberry, “Charge-coupled device area x-ray detectors,” Rev. Sci. Instrum. 73(8), 2815–2842 (2002). [CrossRef]
26. S. E. Skipetrov, J. Peuser, R. Cerbino, P. Zakharov, B. Weber, and F. Scheffold, “Noise in laser speckle correlation and imaging techniques,” Opt. Express 18(14), 14519–14534 (2010). [CrossRef]
27. C. Gutt, L.-M. Stadler, A. Duri, T. Autenrieth, O. Leupold, Y. Chushkin, and G. Grübel, “Measuring temporal speckle correlations at ultrafast x-ray sources,” Opt. Express 17(1), 55–61 (2009). [CrossRef]
28. W. Roseker, H. Franz, H. Schulte-Schrepping, A. Ehnes, O. Leupold, F. Zontone, A. Robert, and G. Grübel, “Performance of a picosecond x-ray delay line unit at 8.39 keV,” Opt. Lett. 34(12), 1768–1770 (2009). [CrossRef]
29. W. Roseker, H. Franz, H. Schulte-Schrepping, A. Ehnes, O. Leupold, F. Zontone, S. Lee, A. Robert, and G. Grübel, “Development of a hard X-ray delay line for X-ray photon correlation spectroscopy and jitter-free pump-probe experiments at X-ray free-electron laser sources,” J. Synchrotron Radiat. 18(3), 481–491 (2011). [CrossRef]