DOI: 10.1002/adma.200703097
Simultaneous Use of Small- and Wide-Angle X-ray
Techniques to Analyze Nanometerscale Phase Separation
in Polymer Heterojunction Solar Cells**
By Mao-Yuan Chiu, U-Ser Jeng, Chiu-Hun Su, Keng S. Liang, and Kung-Hwa Wei*
The development of conjugated polymers for use in organic optoelectronic devices has been an area of extensive inves-tigation. Heterojunction polymer solar cells have been reported utilizing poly( p-phenylenevinylene) derivatives,[1] poly (3-alkylthiophene),[2] or low-bandgap polymers[3] as donors and fullerene derivatives as acceptors. Notably, poly(3-hexylthiophene)/[6,6]-phenyl-C61-butyric acid methyl
ester (P3HT/PCBM) bulk heterojunctions display power conversion efficiencies of up ca. 5%.[4]The improved power
conversion efficiencies of these devices results from the use of thermal annealing[4a] and solvent annealing[4b] processes, which enhance the film morphology relative to that of the as-cast film. The resulting improvement in performance of these systems have been attributed to (i) self-organization of P3HT into a crystalline structure exhibiting enhanced absorp-tion and high hole mobility/transport, and (ii) diffusion of PCBM molecules into PCBM-rich clusters to form large PCBM clusters. The nanostructured phase separation of P3HT/PCBM creates percolated pathways for transporting the holes and electrons to their respective electrodes. There-fore, the performance of these heterojunction solar cells depends critically on the morphology of their active layers.
Grazing-incidence X-ray diffraction (GIXRD) has been used previously to establish the dimensions of P3HT crystal-lites in P3HT/PCBM bulk heterojunction polymer solar cells;[5] morphological studies have also been undertaken using transmission electron microscopy (TEM) and electron diffraction.[6]The morphology of P3HT/PCBM bulk hetero-junction solar cells after solvent annealing can also be analyzed using atomic force microscopy (AFM),[7] and scanning
transmission X-ray microscopy (STXM)[8] has been applied to study the phase separation of PCBM on the order of several hundred nanometers, with a resolution limit of several tens of nanometers. These analytical tools, however, provide only a
local view of the morphology; they cannot represent the full morphology of the active layer in the devices.
Small-angle X-ray scattering (SAXS), which utilizes the elastic scattering of X-rays to probe nanostructures having sizes ranging from 1 to 100 nm, provides statistically averaged morphologies of analyzed samples.[9] For polymer/fullerene
bulk heterojunction solar cells, SAXS can provide global information regarding the internal structure of the dispersion of fullerene units within the polymer matrix much more effectively than can TEM or the other analysis techniques. In this present study, we simultaneously applied grazing-incidence small-angle X-ray scattering (GISAXS) and wide-angle X-ray diffraction (GIWAXD) to study the morphology of P3HT/PCBM bulk heterojunction solar cells after their thermal annealing. Utilizing this approach, we could therefore elucidate the relationship between the relative length scales of the PCBM clusters and P3HT crystallites and the devices’ performance.
Figure 1a displays the GIWAXD profiles of several P3HT/ PCBM films. After annealing the P3HT/PCBM film at 150 8C for 15 min, a strong peak appeared representing the (100) reflections of the P3HT lamella layer structure along with two small peaks contributed by the (300) and (400) or (010) plane reflections of the P3HT lamella structure.[10]
The sizes of the P3HT crystallites at (100) reflection (D100)
can be obtained using Scherrer’s relation:[5,11]
Dhkl¼
0:9l
bhklcos u (1)
where Dhkl is the apparent crystallite size along the [hkl]
direction and bhklis the full width at half maximum of an (hkl)
diffraction (in radians). The P3HT crystallites dramatically increased in size – from 9.6 nm in the as-cast film to 18.0 nm – after annealing at 100 8C for 15 min. This phenomenon can be explained by considering that the glass transition temperature in regioregular poly(3-alkylthiophene) is ca. 100 8C.[12]Table 1 lists the values of D100for the P3HT crystallites obtained after
annealing at various temperatures.
In Figure 1a, in the low-Q range (ca. 0.006 A˚1) of the GISAXS curves for the as-cast film, we observe a small scattering peak that is possibly attributable to the presence of PCBM in P3HT. For confirming this scattering peak, we have carried out control experiments by measuring GISAXS for both the pure P3HT and the P3HT/PCBM thin films
TION
[*] Prof. K.-H. Wei, M.-Y. Chiu
Department of Materials Science and Engineering National Chiao Tung University
1001 Ta Hsueh Road, Hsinchu 30050, Taiwan (ROC) E-mail: khwei@mail.nctu.edu.tw
Dr. U.-S. Jeng, Dr. C.-H. Su, Dr. K. S. Liang National Synchrotron Radiation Research Center 101 Hsin-Ann Road, Science-Based Industrial Park, Hsinchu 30077, Taiwan (ROC)
[**] We thank the National Science Council of Taiwan for funding (NSC 96-2120-M-009-005). Supporting Information is available online from Wiley InterScience or from the authors.
COMMUNICA
(ca. 100 nm thickness; on Si substrate). Figure 1b shows that the annealing at 150 8C for 15 min does not affect the low-Q intensity of the GISAXS profile of the P3HT film, indicating that even if there is local phase separation inside the P3HT
the as-cast or annealed pure P3HT thin film, probably due to the large electron density difference between PCBM and P3HT (0.7 versus 0.4 eA˚3). Therefore, the GISAXS features in the low-Q region of P3HT/PCBM thin film are clearly dominated by PCBM, and we attribute the large GISAXS intensity change in the P3HT/PCBM thin film after annealing at 150 8C for 15 min to a significant growth of the PCBM aggregates. The PCBM scattering intensity increased drama-tically as a result of the PCBM molecules diffusing out of the P3HT matrix and forming larger PCBM clusters.[14]
Figure 2 displays the morphologies of P3HT/PCBM thin films as determined through phase-contrast AFM measure-ments. Phase separation was more apparent – with domain sizes ranging from between 30 to 60 nm – for the sample that had experienced annealing at 150 8C for 15 min relative to that of the as-cast film. These images are consistent with the results obtained from the simultaneous GISAXS/GIWAXD analyses. The radius of gyration (Rg) of a PCBM cluster can be
determined from the scattering peak intensity using the Guinier approximation (Equation 2):
IðQÞ ¼ Ið0Þ exp Q
2R g2
3
(2)
where I(Q) is the scattering intensity, I(0) is the zero-angle scattering intensity, Q is defined as 4psin(u/2)/l in terms of the scattering angle u and the wavelength l, and Rgis the radius of
gyration of the PCBM clusters. Figure 3 presents plots of lnI(Q) versus Q2that were fit using Equation 2 (solid lines) in the low-Q range (0.006–0.012 A˚1).[15]The values of Rgcan be extracted
from the slopes (Rg2/3) of the fitted lines. We fitted each set of
data with two slopes for two Rgvalues that were then used to
give an averaged Rgand the spread of Rgfor the aggregation size
of PCBM. Table 1 lists these values for the PCBM clusters obtained under various annealing conditions. After annealing at 100 and 120 8C, the values of Rgof the PCBM clusters increased
marginally to (16 2) nm from (15 2) nm in the as-cast film. After annealing at 150 8C for 15 min, Rgincreased dramatically
to (23 3) nm. Thus, it appears that large-scale diffusion of PCBM molecules into the P3HT matrix to form aggregates occurred only at temperatures at or above 150 8C. The mean diameter of the PCBM clusters obtained after annealing at 150 8C for 15 min was (60 20) nm, as deduced from the value of Rg, assuming that the PCBM clusters were spherical.
[16]
This result is consistent with those obtained from TEM studies.[6]
Table 2 summarizes the current density–voltage character-istics of the P3HT/PCBM devices under illumination with respect to their annealing temperatures. Under 1 equivalent sun intensity (100 mW cm2), the power conversion efficiency (h) from our as-cast device was 0.53%, with values of Jsc, Voc,
and FF of 4.0 mA cm2, 0.38 V, and 35%, respectively. After annealing at 100 8C for 15 min, the value of h became 1.6%. We
Figure 1. a) GISAXS/GIWAXD curves of as-cast and annealed (150 8C, 15 min) P3HT/PCBM films (not to scale). b) GISAXS curves of as-cast and annealed (150 8C, 15 min) P3HT and P3HT/PCBM thin films.
Table 1. Values ofRgandD100for PCBM cluster aggregates and P3HT
crystallites obtained after annealing for 15 min at various temperatures.
Annealing temperature Rgof PCBM[a] [nm] Number of PCBM molecules[b] D100of P3HT [c] [nm] As-cast 15 2 10640 9.6 100 8C 16 2 12320 18.0 120 8C 16 2 14168 17.7 150 8C 23 3 39608 16.0
[a] Determined using Guinier approximations to obtain the average value ofRgin the
low-Q range. [b] Calculation byRgof fullerenes for 0.38 nm, assuming a cubic lattice
TION
attribute this increase in h to the growth of the P3HT crystallites, which increased to 18.0 nm from 9.6 nm after annealing, since the hole mobility increased upon increasing the size of the P3HT crystallites,[17]as evidenced by the current density increasing from 4.0 to 7.7 mA cm2. Furthermore, the
change in the morphology of P3HT/ PCBM film results in increased carrier delocalization.[18]
Annealing the P3HT/PCBM blend at 120 8C for 15 min, increased the value of h to 3.0%; we believe this behavior resulted from an increase in the number of P3HT crystallites, even though the average crystallite size remained about the same as that in the sample annealed at 100 8C (cf. stronger (100) peak in Fig. S1). The sizes of the PCBM clusters remained virtually unchanged when the annealing of P3HT/PCBM was performed at either 100 or 120 8C for 15 min. Annealing for 15 min at 150 8C, however, led to a dramatic increase in the value of Rg of the
PCBM cluster aggregates to (23 3) nm; accordingly, the h increased to 4.0%, with values of Jsc, Voc, and FF of
9.9 mA cm2, 0.63 V, and 63%, respec-tively. In this case, the P3HT crystal-lite size reduced slightly relative to that for the sample annealed at 120 8C, whereas the absolute intensity of the (100) reflection remained roughly the same. Therefore, we interpret the increase in Jscas resulting
from larger PCBM clusters forming better pathways for electron transport. Hence, different mechanisms dominate the development of the P3HT/PCBM blend morphology at different temperatures. At the glass transition temperature (100 8C), increased planar stacking of P3HT chains occurred, thereby increasing the P3HT/PCBM crystallite size but not affecting the cluster size of the PCBM molecules. At 150 8C, the PCBM molecules diffused to form large clusters; therefore, phase separation of P3HT and PCBM occurred to provide improved continuous pathways for the holes and electrons. Figure 4 illustrates the morphologies of the as-cast and optimized devices.
Figure 5 displays the relative length scales of the P3HT crystallites and the radii of gyration of the PCBM clusters. The
Figure 2. Phase-contrast AFM images of a) as-cast and b) annealed (150 8C, 15 min) P3HT/PCBM films.
Figure 3. Plots of lnI(Q) veruss Q2for P3HT/PCBM films annealed at
various temperatures, fitted using the Guinier approximation (solid lines).
Table 2. Comparison of the P3HT crystal size, PCBM cluster size, and device electronic parameters at different annealed temperature for 15 min.
D100of P3HT [nm] Rgof PCBM [nm] Jsc[a] [mA cm2] Voc[b] [V] FF[c] [%] h[d] [%] As-cast 9.6 15 2 4.0 0.38 35 0.53 100 8C 18.0 16 2 7.7 0.41 52 1.6 120 8C 17.7 16 2 8.7 0.59 58 3.0 150 8C 16.0 23 3 9.9 0.63 63 4.0
[a] Short-circuit current density. [b] Open-circuit voltage. [c] Filling factor. [d] Power-conversion efficiency.
COMMUNICA
value of the corresponding lattice constant, d100, can be
calculated using Bragg’s law (nl¼ 2dsinu at 2u ¼ 5.3 8)[11]
to be 1.67 nm. From Scherrer’s relation, the value of D100is equal to
md100, where m is the number of layers in the P3HT lamellar
structure. For the device exhibiting the optimal performance (annealing temperature, 150 8C; P3HT:PCBM blend weight ratio, 1:0.8), each P3HT crystallite contained ca. 10 layers of P3HT main chains p-stacked parallel to the substrate with their alkyl chains positioned normal to substrate at the (100) reflection, and each cluster contained ca. 39600 PCBM molecules.
To discern the surface confinement effect on thin film by electrodes,[19] we have carried out SAXS measurements on
bulk P3HT/PCBM composite that have the same composition and experienced the same annealing process as that of 100 nm thin film with one-side confined on a Si substrate. In Figure 6, in
with Rg of (22 3) nm and (23 3) nm
for PCBM clusters for the bulk and the thin film case, respectively. This reveals that more than 95% accuracy for our results. We expected that the size differ-ence in PCBM cluster between double-side confinement and one-double-side confine-ment on a 100 nm thin film should not differ more than that we have observed between the bulk and the one-side confined thin film of P3HT/PCBM. Furthermore, one report indicated the strength of substrate interactions on a polymer thin films was minimum when the film thickness is more than 17 nm, judging from the glass transition temperature.[20]Therefore,
our basic conclusions on the correlation between the PCBM aggregate size and the device efficiency should be correct.
In summary, we applied GISAXS and GIWAXD simulta-neously to study the morphologies of P3HT/PCBM active layers in bulk heterojunction solar cells annealed at various temperatures. This approach allowed us to investigate the effects of the sizes of the PCBM clusters and P3HT crystallites on the power conversion efficiencies of bulk heterojunction solar cells. It appears that improved power conversion efficiency requires the value of Rg of the PCBM clusters to
be greater than 20 nm and the value of D100 of the P3HT
crystallites to be greater than 16 nm for an active layer thickness of ca. 100nm.
Figure 4. Schematic representation of P3HT/PCBM blends as active layers (ca. 100 nm) in solar cell devices. a) Non-annealed devices: holes and electrons recombine as a result of percolation. b) After annealing the devices at 150 8C for 15 min, the holes and electrons experience a good percolated network for their passage to the ITO and Al electrodes, respectively.
Figure 5. Relative length scales of the P3HT crystallites, the radii of gyration of the PCBM clusters, and the power conversion efficiencies of the various P3HT/PCBM film devices.
Figure 6. Comparison of the SAXS and GISAXS data for the bulk P3HT/ PCBM composite and the thin film (100 nm) on a Si substrate at annealing 150 8C for 15 min.
TION
Experimental
Device Fabrication: Regioregular poly(3-hexylthiophene) (Mw
50000 g mol1, used as received) and [6,6]-phenyl-C61-butyric acid
methyl ester were obtained from Rieke Metals and Nano-C Inc.,
respectively. P3HT was dissolved in chlorobenzene (15 mg mL1) and
blended with a solution of PCBM in chlorobenzene (12 mg mL1). The
P3HT/PCBM composite weight ratio was 1:0.8. The indium tin oxide (ITO)-coated glass substrate (UID Co., Ltd.; surface resistance: 15 V/ &) was cleaned through ultrasonic treatment in detergent, methanol, acetone, and isopropyl alcohol and then it was dried in an oven overnight. PEDOT:PSS (Baytron P VP A1 4083) was spin-coated on the ultraviolet ozone-treated ITO. After annealing the PEDOT:PSS film (ca. 40 nm) at 140 8C for 30 min in air, the sample was transferred to a glove box. The P3HT/PCBM solution blend was spin-coated on top of the PEDOT:PSS layer. The thickness of film was ca. 100 nm (AFM). Finally, the Al electrode (100 nm) was deposited on the P3HT/PCBM
layer under high vacuum (ca. 107Torr (1 Torr¼ 1.333 102Pa)) using
a thermal evaporator. For thermal annealing, the completed devices were placed directly on a digitally controlled hotplate and heated at various temperatures for 15 min. After annealing, the devices were encapsulated in UV-curing epoxy and analyzed in an air atmosphere.
Calibration and Measurement: Current density–voltage
charac-teristics of the P3HT/PCBM devices were measured under simulated
AM1.5G irradiation (100 mW cm2) using a Xe lamp-based Newport
66902 150-W solar simulator. A Xe lamp equipped with an AM1.5G filter was used as the white light source. The spectrum of the solar simulator was calibrated by a PV-measurement (PVM-154) mono-Si solar cell (NREL calibrated), and Si photo diode (Hamamatsu S1133)
was used to check the irradiation of the exposed area (100 mW cm2).
The mismatch factor (M) was 1.34, which was obtained by taking the PVM-154 as the reference cell and our devices as the test cells and calculating the spectrum from 300 nm to 900 nm with intervals of 10 nm. The PVM-154 combined with KG-5 filter (350–700 nm passed, Newport) was to simulate a reference solar cell with spectral responsivity from 350 nm to 700 nm. The calibration was based on the IEC-69094-1 spectrum [21]. The J–V characteristics were measured using a Keithley 236 source measurement unit.
Characterization: Grazing-incidence small-angle X-ray
scatter-ing (GISAXS) and wide-angle X-ray diffraction (GIWAXD) mea-surements were performed at the SWAXS endstation of the BL17B3 beamline of the Taiwan light source at the National Synchrotron Radiation Research Center (NSRRC). The BL17B3 beamline [22], equipped with a SAXS area detector and a WAXS linear detector (in the meridian direction) connected to two data acquisition systems operated in master-slave mode, allows one to perform simultaneously SAXS/WAXS measurements for correlated changes of the crystalline structures and nanostructure of the interesting sample thin films. The wave vector transfer Q was equal to 4p sin(u/2)/l, which was defined in terms of the scattering angle u and wavelength l of the X-rays. The X-rays passed through a slit having a width of 0.5 mm; the photon
energy was 10.5 keV (l¼ 1.1809 A˚ ); the sample-to-detector distances
were 3.2 and 0.75 m for the GISAXS and GIWAXS systems, respectively; the angle of incidence was 0.3 8 with an accuracy of 0.005 8.
The samples were prepared on 3 3.5 cm2silicon substrates through
the spin-coating of a P3HT/PCBM solution; the film thickness was the same for each sample (ca. 100 nm). X-ray diffraction measurements (Q
ranging from 0.25 to 1.1 A˚1) were performed using a Bede D1
high-resolution X-ray diffractometer and fixed angle of incidence (0.3 8;
l¼ 1.54 A˚ ). The devices’ morphologies were recorded under ambient
conditions using an atomic force microscope operated in tapping-mode and a Digital Nanoscope IIIa instrument.
Received: December 13, 2007 Revised: January 11, 2008 Published online: June 2, 2008
[1] a) M. M. Wienk, J. M. Kroon, W. J. H. Verhees, J. Knol, J. C. Hummelen, P. A. Van Hal, R. A. J. Janssen, Angew. Chem. Int. Ed. 2003, 42, 3371. b) H. Hoppe, M. Niggemann, C. Winder, J. Kraut, R. Hiesgen, A. Hinsch, D. Meissner, N. S. Sariciftci, Adv. Funct. Mater. 2004, 14, 1005.
[2] a) F. Padinger, R. S. Rittberger, N. S. Sariciftci, Adv. Funct. Mater. 2003, 13, 85. b) M. Reyes-Reyes, K. Kim, D. L. Carroll, Appl. Phys. Lett. 2005, 87, 083506. c) C. N. Hoth, S. A. Choulis, P. Schilinsky, C. J. Brabec, Adv. Mater. 2007, 19, 3973.
[3] a) Y. T. Chang, S. L. Hsu, M. H. Su, K. H. Wei, Adv. Funct. Mater. 2007, 17, 3326. b) J. Peet, J. Y. Kim, N. E. Coates, W. L. Ma, D. Moses, A. J. Heeger, G. C. Bazan, Nat. Mater. 2007, 6, 497. c) F. Zhang, W. Mammo, L. M. Andersson, S. Admassie, M. R. Andersson, O. Ingana¨s, Adv. Mater. 2006, 18, 2169. d) D. Mu¨hlbacher, M. Scharber, M. Morana, Z. Zhu, D. Waller, R. Gaudiana, C. Brabec, Adv. Mater. 2006, 18, 2884. e) A. M. Ballantyne, L. Chen, J. Nelson, D. D. C. Bradley, Y. Astuti, A. Maurano, C. G. Shuttle, J. R. Durrant, M. Heeney, W. Duffy, I. McCulloch, Adv. Mater. 2007, 19, 4544. f) N. Blouin, A. Michaud, M. Leclerc, Adv. Mater. 2007, 19, 2295. [4] a) W. Ma, C. Yang, X. Gong, K. Lee, A. J. Heeger, Adv. Funct. Mater.
2005, 15, 1617. b) G. Li, V. Shrotriya, J. Huang, Y. Yao, T. Moriarty, K. Emery, Y. Yang, Nat. Mater. 2005, 4, 864. c) J. Y. Kim, S. H. Kim, H. H. Lee, K. Lee, W. Ma, X. Gong, A. J. Heeger, Adv. Mater. 2006, 18, 572. d) M. C. Scharber, D. Mu¨hlbacher, M. Koppe, P. Denk, C. Waldauf, A. J. Heeger, C. J. Brabec, Adv. Mater. 2006, 18, 789. [5] T. Erb, U. Zhokhavets, G. Gobsch, S. Raleva, B. Stuhn, P. Schilinsky,
C. Waldauf, C. J. Brabec, Adv. Funct. Mater. 2005, 15, 1193. [6] a) X. Yang, J. Loos, S. C. Veenstra, W. J. H. Verhees, M. M. Wienk,
J. M. Kroon, M. A. J. Michels, R. A. J. Janssen, Nano Lett. 2005, 5, 579. b) W. Ma, C. Yang, A. J. Heeger, Adv. Mater. 2007, 19, 1387. c) W. Ma, A. Gopinathan, A. J. Heeger, Adv. Mater. 2007, 19, 3656. d) X. Yang, G. Lu, L. Li, E. Zhou, Small 2007, 3, 611.
[7] V. Shrotriya, Y. Yao, G. Li, Y. Yang, Appl. Phys. Lett. 2006, 89, 063505.
[8] C. R. McNeill, B. Watts, L. Thomsen, W. J. Belcher, A. L. D. Kilcoyne, N. C. Greenham, P. C. Dastoor, Small 2006, 2, 1432.
[9] a) C. P. Li, C. H. Wu, K. H. Wei, J. T. Sheu, J. Y. Huang, U. S. Jeng, K. S. Liang, Adv. Funct. Mater. 2007, 17, 2283. b) S. W. Yeh, K. H. Wei, Y. S. Sun, U. S. Jeng, K. S. Liang, Macromolecules 2003, 36, 7903. [10] a) H. Sirringhaus, P. J. Brown, R. H. Friend, M. M. Nielsen, K.
Bechgaard, B. M. W. Langeveld-Voss, A. J. H. Spiering, R. A. J. Janssen, E. W. Meijer, P. Herwig, D. M. De Leeuw, Nature 1999, 401, 685. b) T. J. Prosa, M. J. Winokur, J. Moulton, P. Smith, A. J. Heeger, Macromolecules 1992, 25, 4364. c) R. J. Kline, D. M. DeLongchamp, D. A. Fischer, E. K. Lin, L. J. Richter, M. L. Chabinyc, M. F. Toney, M. Heeney, I. McCulloch, Macromolecules 2007, 40, 7960. [11] B. D. Cullity, Elements of X-Ray Diffraction, Addison-Wesley,
Read-ing, MA 1956.
[12] a) K. Yazawa, Y. Inoue, T. Yamamoto, N. Asakawa, Phys. Rev. B 2006, 74, 094204. b) V. Causin, C. Marega, A. Marigo, L. Valentini, J. M. Kenny, Macromolecules 2005, 38, 409.
[13] K. A. Affholter, S. J. Henderson, G. D. Wignall, G. J. Bunick, R. E. Haufler, R. N. Compton, J. Chem. Phys. 1993, 99, 9224.
[14] D. Chirvase, J. Parisi, J. C. Hummelen, V. Dyakonov, Nanotechnology 2004, 15, 1317.
[15] For general aggregate estimation using the Guinier approximation,
where QRgis less than 1, the value of Q (0.006 A˚1) provided by
GISAXS can be used to determine that the maximum size of Rgwill be
16 nm; the sample was consistent with that obtained using bulk sample
SAXS data (Qmin, 0.0038 A˚1; maximum size of Rg, 26 nm).
[16] For a sphere, Rg2¼ (3/5)R2, where R is the radius.
[17] V. D. Mihailetchi, H. Xie, B. de Boer, L. J. A. Koster, P. W. M. Blom, Adv. Funct. Mater. 2006, 16, 699.
COMMUNICA
[20] S. Ge, Y. Pu, W. Zhang, M. Rafailovich, J. Sokolov, C. Buenviaje, R. Buckmaster, R. M. Overney, Phys. Rev. Lett. 2000, 85, 2340.
W. T. Chuang, Y. S. Huang, C. H. Hsu, M. T. Lee, H. Y. Lee, K. S. Liang, A. Gabriel, M. H. J. Koch, J. Appl. Crystallogr. 2006, 39, 871.
