Reprinted from
REGULAR PAPER
Heterojunction Diodes Comprising p-Type Ultrananocrystalline Diamond Films Prepared by Coaxial Arc Plasma Deposition and n-Type Silicon Substrates Yu¯ki Katamune, Shinya Ohmagari, Sausan Al-Riyami, Seishi Takagi, Mahmoud Shaban, and Tsuyoshi Yoshitake
Jpn. J. Appl. Phys. 52 (2013) 065801
# 2013 The Japan Society of Applied Physics
Person-to-person distribution (up to 10 persons) by the author only. Not permitted for publication for institutional repositories or on personal Web sites.
REGULAR PAPER
Japanese Journal of Applied Physics 52 (2013) 065801 http://dx.doi.org/10.7567/JJAP.52.065801
Heterojunction Diodes Comprising p-Type Ultrananocrystalline Diamond Films Prepared by Coaxial Arc Plasma Deposition and n-Type Silicon Substrates Yu¯ki Katamune1 , Shinya Ohmagari1 , Sausan Al-Riyami1 , Seishi Takagi2 , Mahmoud Shaban3 , and Tsuyoshi Yoshitake1 1
Department of Applied Science for Electronics and Materials, Kyushu University, Kasuga, Fukuoka 816-8580, Japan Faculty of Engineering, Kyushu Institute of Technology, Kitakyushu 804-8550, Japan 3 Department of Electrical Engineering, Aswan Faculty of Engineering, Aswan University, Aswan 81542, Egypt E-mail: tsuyoshi
[email protected] 2
Received December 5, 2012; accepted March 22, 2013; published online May 22, 2013 Heterojunction diodes, which comprise boron-doped p-type ultrananocrystalline diamond/hydrogenated amorphous carbon composite (UNCD/ a-C:H) films prepared by coaxial arc plasma deposition and n-type Si substrates, were electrically studied. The current–voltage characteristics showed a typical rectification action. An ideality factor of 3.7 in the forward-current implies that carrier transport is accompanied by some processes such as tunneling in addition to the generation–recombination process. From the capacitance–voltage measurements, the built-in potential was estimated to be approximately 0.6 eV, which is in agreement with that in a band diagram prepared on the assumption that carriers are transported in an a-C:H matrix in UNCD/a-C:H. Photodetection for 254 nm monochromatic light, which is predominantly attributable to photocurrents generated in UNCD grains, was evidently confirmed in heterojunctions. Since dangling bonds are detectable by electron spin resonance spectroscopy, their control might be an important key for improving the rectifying action and photodetection performance. # 2013 The Japan Society of Applied Physics
1. Introduction
Ultrananocrystalline diamond/hydrogenated amorphous carbon composite (UNCD/a-C:H) films, wherein nanosized diamond grains are embedded in an a-C:H matrix, contain a very large number of grain boundaries (GBs) in the films, which is a distinctive structure specific to UNCD/ a-C:H films. Here, GBs denote the interfaces between UNCD grains and those between UNCD grains and the a-C:H matrix. The electrical properties of UNCD/a-C:H films, mainly prepared by chemical vapor deposition, have received considerable attention. It has been reported that nitrogen-doped UNCD/a-C:H films exhibit n-type conduction with an increased electrical conductivity,1,2) whereas nitrogen does not act as a donor in a diamond crystal because it forms a deep donor level (1.7 eV below the conduction band minimum). Thus far, it has been explained that the increase in electrical conductivity is attributable to the increased number of sp2 bonds at GBs.3) Theoretical calculations have predicted the formation of n-type conduction as follows. Nitrogen atoms are preferentially incorporated into GBs4) and increase the number of threefoldcoordinated carbon atoms (sp2 bonding) at GBs. As a result, a shift in the Fermi energy level toward the conduction band by approximately 0.4 eV takes place at a high nitrogen concentration.5) We have reported that UNCD/a-C:H films prepared by pulsed laser deposition (PLD)6,7) possess large optical absorption coefficients of more than 105 cm1 in the photon energy range between 3 and 6 eV,8,9) which differs from those of UNCD/a-C:H films prepared by CVD. The physical properties of UNCD/a-C:H films seem to depend on the film preparation method. Since the growth mechanisms are clearly different from each other,10) there should be a difference in the chemical bonding structure. In addition, it has been demonstrated that their p- and n-type conduction accompanied by enhanced electrical conductivity are possible by boron and nitrogen doping, respectively. Their superior optical and electrical properties are advantageous for their application as photodiodes.
Recently, we have succeeded in the growth of UNCD/ a-C:H films by coaxial arc plasma deposition (CAPD).10,11) The formation of diamond grains by CAPD has the following major differences from that by CVD: (i) lower substrate temperatures, (ii) much higher deposition rates, and (iii) no requirement of substrate pretreatment using diamond powder.10) From these differences, the formation mechanism by CAPD should be markedly different from that by CVD. As compared with PLD, wherein the deposition is physically carried out in a pulsed process similarly to CAPD, both methods enable the growth at low substrate temperatures and require no pretreatment of substrates. On the other hand, there are two distinctive differences. One is that the deposition rate per pulsed process (1.33 nm/pulse) is two orders of magnitude larger than that in PLD. The other is that, although a hydrogen atmosphere during the deposition is necessary for diamond crystallite formation by PLD, it is not necessary for that by CAPD. The growth mechanism in CAPD should be distinct from not only that in CVD but also that in PLD. Heterostructures comprising semiconducting carbon films and Si substrates have been studied for the purpose of applying them in UV detectors,12,13) field emission devices,14,15) UV electroluminescence systems,16–19) and solar cells.20) Aside from their attractive applications, singlecrystalline Si substrates, whose physical parameters are well known, are suitable for investigating the electrical properties of pair materials in heterojunctions. In this study, boron doping for inducting p-type conduction was examined for UNCD/a-C:H films prepared by CAPD. From this study, the production of p-type conduction was confirmed, and the electrical properties of boron-doped UNCD/a-C:H films were studied through the electrical evaluation of heterojunctions comprising p-type UNCD/ a-C:H films and n-type Si substrates whose electrical properties are already known. 2. Experimental Procedure
Heterojunction diodes comprising boron-doped UNCD/ a-C:H films and n-type Si substrates, wherein boron-doped
065801-1
# 2013 The Japan Society of Applied Physics
Person-to-person distribution (up to 10 persons) by the author only. Not permitted for publication for institutional repositories or on personal Web sites.
Jpn. J. Appl. Phys. 52 (2013) 065801
Y. Katamune et al. –2
300 µm Au/Pd
10
–4
p-UNCD films
900 nm
n-Si
260 µm
10
0
–2
Ti
10
–6
–4
10
10
–8
–6
10 10
–10
–5 –4 –3 –2 –1 0 1 2 Voltage (V)
3
4
5
−4
10
10
−1
10
−2
−6
10
−3
10
−4
−7
10
−5
10
−6
10
−7
10
−8
−5
10
Current (A)
10 10
n = 3.7
−8
10
−9
10
n=1 −10
raw data processed data fitting
10
Current density (A/cm2)
Fig. 1. (Color online) I–V characteristics of heterojunction diode comprising boron-doped UNCD/a-C:H film and n-type Si substrate. The inset shows the schematic of the diode.
3. Results and Discussion
The boron content of the films deposited with 5 at. % boronblended targets was estimated to be 4 at. % from the X-ray photoemission spectra, similar to that in our previous study.21) The electrical conductivity at room temperature was 6 102 Scm1 , whereas that of undoped films was difficult to measure owing to the films being insulating. The p-type conduction of the boron-doped films was confirmed thermally, namely, from the Seebeck effects. Figure 1 shows the I–V characteristic curve for the heterojunction diode comprising the p-type UNCD/a-C:H film and n-type Si substrate. The inset illustrates the schematic of the diode. The I–V curve shows a typical rectification action of pn junction diodes. The boron-doped UNCD/a-C:H films prepared by CAPD clearly act as a p-type semiconductor. Figure 2 shows the magnification of the I–V curve in the low-forward-voltage range. Since the leakage current observed in the reverse-voltage range should symmetrically exist at the applied voltage, the background current, which should be the same as the reverse current, was subtracted from the forward current for the analysis of ideality factors. The forward current can be expressed by
10
Current density (A/cm2)
10
Current (A)
UNCD/a-C:H films were deposited on Si substrates by CAPD with a coaxial arc plasma gun (Ulvac ARL-300) equipped with 5 at. % boron-incorporated graphite targets, were electrically studied. The resistivity and thickness of the Si substrates were 1 cm and 260 m, respectively. The film preparation apparatus was evacuated down to a base pressure of less than 104 Pa using a turbomolecular pump, and then hydrogen gas was fed into the apparatus. The deposition of a 900-nm-thick film was conducted at a substrate temperature of 500 C and a hydrogen pressure of 53 Pa. A voltage of 100 V was applied to the arc plasma gun equipped with a capacitor of 1800 F. The repetition rate of pulsed discharge was 5 Hz. As ohmic back-surface electrodes, Ti films were deposited on Si substrates by radiofrequency magnetron sputtering. For the formation of front ohmic contacts of 300 m diameter on the films, Pd and Au films were successively deposited with a mask having 300-m-diameter holes by magnetron sputtering. After that, the samples were transferred into an etching apparatus and the film side was exposed to oxygen plasma to form cylindrical structures. The boron content of the films was estimated by X-ray photoemission spectroscopy using a Mg K line (h ¼ 1253:6 eV), and argon ion etching was employed to measure the boron content distribution in the depth direction. Current–voltage (I–V ) and capacitance– voltage (C–V ) measurements were performed at room temperature in the dark. C–V characteristic curves were measured at an applied modulation voltage of 50 mV and a signal frequency of 5 kHz. The positive voltage applied to the p-type UNCD/a-C:H side is generally defined as the forward bias. The ionization potential was measured by photoelectron spectroscopy (Riken Keiki AC-2). Electron spin resonance (ESR) signals were measured in the X-band frequency range at room temperature. The microwave power and magnetic field modulation were 1 mW and 0.1 mT, respectively. For calibration, signals from Mn2þ in MgO set together with the samples were used.
−11
10
0
0.5 Voltage (V)
1
Fig. 2. (Color online) Magnification of I–V characteristic curve in low forward-voltage range.
qV J ¼ Js exp ; nkT
ð1Þ
where J, Js , q, k, n, and T are the current density, saturation current density at zero bias, elementary electron charge, Boltzmann’s constant, ideality factor, and temperature, respectively. The ideality factor is determined by the dominant carrier conduction mechanism; it theoretically becomes 1 and 2 for diffusion and generation–recombination (G–R) processes, respectively.22) In the forward-voltage range from 0 to 0.1 V, the ideality factor becomes 1, which indicates that diffusion is predominant. In the range between 0.1 and 0.6 V, the ideality factor was estimated to be 3.7. This value is evidently larger than 2, and carrier transport might be accompanied by some processes such as tunneling
065801-2
# 2013 The Japan Society of Applied Physics
Person-to-person distribution (up to 10 persons) by the author only. Not permitted for publication for institutional repositories or on personal Web sites.
Jpn. J. Appl. Phys. 52 (2013) 065801
Y. Katamune et al. B doped UNCD film
100
2
2
1/C ×10
16
(F/cm )
2 –2
Atomic concentration (%)
2
Capasitance density (nF/cm )
R.T.
20
1 Vbi = 0.58 eV data fitting
0
–5
–4
–3
–2
–1
0
1
Reverse voltage (V)
10
80 60
O1s C1s B1s Si2p
40 20 0
–5
Si sub.
frequency 5 kHz
3
–4
–3
–2
–1
0
50 100 150 Sputter time (min)
200
Fig. 4. (Color online) Depth dependence of chemical composition of UNCD/a-C:H film, measured by X-ray photoemission spectroscopy combined with argon ion etching.
0
Reverse voltage (V)
(arb. unit)
Fig. 3. (Color online) C–V characteristics in reverse-voltage range of heterojunction diode comprising boron-doped UNCD/a-C:H films and n-type Si substrate. The inset shows the plot of 1=C2 –V .
(Emission Yield)
1 2ð"Si NSi þ "UNCD NUNCD ÞðVbi V Þ ¼ ; C2 q"Si NSi "UNCD NUNCD
graphite
1/2
in addition to the G–R process. UNCD/a-C:H contains a very large number of GBs, and they might serve as centers for tunneling. Figure 3 shows the C–V characteristic curve in the reverse-voltage range. The capacitance density decreases with increasing reverse voltage, which indicates the expansion of the depletion region with increasing reverse-voltage. The inset shows the 1=C2 –V plot. The experimental data was fitted with
UNCD/a−C:H
diamond
ð2Þ 0
where "Si and "UNCD are the dielectric constants, and NSi and NUNCD are the carrier concentrations of the Si and UNCD films, respectively. Vbi is the built-in potential of the heterojunction, estimated to be approximately 0.58 eV from the plot. In order to investigate the depth dependence of the boron content, X-ray photoemission spectroscopic measurement combined with argon ion etching was employed for a 50-nm film prepared under the same conditions as the heterojunction film. The spectroscopic measurement and etching were repeatedly carried out until the etching caused the penetraton of the film in the depth direction. The depth dependence of the chemical composition is shown in Fig. 4. On the film surface at zero sputtering time, a peak due to oxygen adsorption was detected. The decrease in the carbon content simultaneous with an increase in the silicon content indicates that the etching penetrates the film. The boron content, shown by a blue diamond, does not significantly change in the depth direction. Films prepared by CAPD are depositions resulting from accumulated depositions in pulsed processes that are independent of each other. No posttreatment such as high-temperature annealing was applied. Therefore, little depth dependence of the boron content is reasonable from the viewpoint of deposition. The dielectric constants of diamond and a-C:H are 5.68 and 3.5, respectively, from previous reports.23,24) The dielec-
4
5
6
Photon Energy (eV) Fig. 5. (Color online) Photoelectron spectrum of UNCD/a-C:H film prepared by CAPD.
tric constant of boron-doped UNCD/a-C:H was assumed to be between those of diamond and a-C:H because of UNCD/ a-C:H comprising diamond grains and an a-C:H matrix. By using the carrier concentration of the n-type Si substrate, 5 1015 cm3 , the active carrier concentration of the UNCD/ a-C:H film was calculated to be ð0:9{1:5Þ 1016 cm3 . The widths of expansion of a depletion region into the UNCD/ a-C:H and Si sides at a zero bias were estimated to be approximately 100 and 300 nm, respectively. Figure 5 shows the photoelectron spectrum of the UNCD/ a-C:H films. The threshold of the incident photon energy, corresponding to the ionization potential, was measured to be 5.22 eV. From this value and the known parameters of Si and UNCD/a-C:H, the energy band diagram of the p-type UNCD/a-C:H/n-type Si heterostructure was derived, as shown in Fig. 6. Here, the band gap of the UNCD/a-C:H film is 1.7 eV, which was estimated from the transmittance and reflectance spectra of UNCD/a-C:H films prepared by CAPD in our previous work.25) This band gap should be attributed to the a-C:H matrix in the UNCD/a-C:H films.25)
065801-3
# 2013 The Japan Society of Applied Physics
Person-to-person distribution (up to 10 persons) by the author only. Not permitted for publication for institutional repositories or on personal Web sites.
Jpn. J. Appl. Phys. 52 (2013) 065801
Y. Katamune et al.
0.88 -4
3.52
-6
10
–3
10
–4
10
–5
10
4.05
-7 1.70 -8 1.12
-9
–2
Current (A)
Energy (eV)
-5
10
under illumination (254 nm)
10
–7
10
–8
10
–9
10
–10
10
–11
1 Distance (µm)
1.5
2
Fig. 6. (Color online) Band diagram of heterojunction diode comprising boron-doped UNCD/a-C:H film and n-type Si substrate.
We have previously reported that a large number of GBs, whose structure is specific to UNCD/a-C:H films, act as centers for tunneling, and carriers might be transported between GBs in tunneling.13) Thus, this band diagram supposes that carriers are transported in the a-C:H matrix. Since the built-in potential is in good agreement with that estimated from the C–V measurement, this assumption is probably reasonable. The junction has no spike in the conduction band, which is beneficial for the delivery of photocarriers generated in the UNCD/a-C:H films to the Si-side electrode. The valence band discontinuity is also negligible. Figure 7 shows the I–V characteristics under light illumination with a 254 nm, 2 mW lamp. The effective light irradiation area in the heterojunction is extremely small, because the top surface of the cylindrical structural UNCD/ a-C:H film is entirely covered with the Au/Pd electrode and the effective irradiation area is limited to the side face of the cylindrical structure. However, photogenerated carriers are evidently observed. The 254 nm light detection including the signal/noise ratio was clearly better than that of commercial GaN photodiodes. It is considered that photocarriers might generate at a large number of UNCD grains in the UNCD/ a-C:H film and that they are transported through the a-C:H matrix. Since the photon energy of 254 nm light is too large for effective photocarrier generation in Si and a-C:H, the contribution of UNCD grains might be predominant. Since the I–V measurement predicted the existence of recombination centers, the presence of dangling bonds in the films was investigated by ESR. Figure 8 shows the ESR spectrum of the film. The ESR spectrum exhibits a clear signal from the film, which implies that a detectable number of dangling bonds exist in the film. The g-factor was estimated to be approximately 2:0026 0:0002 at room temperature, and this value is in agreement with that of carbon dangling bonds related to defects in polycrystalline diamond including nanocrystalline diamond26,27) and diamond-like carbon, namely, sp3 -rich a-C:H.28) From the above-mentioned results, possible origins of the signal are dangling bonds at GBs, defects in UNCD grains, and defects in an
in the dark
–5 –4 –3 –2 –1 0 1 2 Voltage (V)
10
0
10
–1
10
–2
10
–3
10
–4
10
–5
10
–6
10
–7
3
4
5
Fig. 7. (Color online) I–V characteristics of heterojunction diode comprising boron-doped UNCD/a-C:H film and n-type Si substrate under illumination of 254 nm monochromatic light and in the dark.
Intensity (arb. unit)
0.5
1
–6
-10 0
10
Current density (A/cm2)
-3
Mn
328
2+
Mn
330
332
334
336
2+
338
340
Magnetic field (mT) Fig. 8. ESR signal of boron-doped UNCD/a-C:H film, measured at room temperature.
a-C:H matrix. From our previous study, it was found that a large number of dangling bonds at GBs are terminated by hydrogen. However, residual dangling bonds that are not terminated may also be detected. Moreover, since the XRD measurement in our previous study indicated that diamond lattices should have defects in them,21) this possibility is also reasonable. The details are under investigation. 4. Conclusions
It was experimentally demonstrated that boron doping is effective for producing p-type conduction in UNCD/a-C:H films prepared by CAPD. From the C–V measurement of heterojunction diodes comprising p-type UNCD/a-C:H films and n-type Si substrates, it was confirmed that a depletion region clearly expands into the UNCD/a-C:H film and the built-in potential was estimated to be approximately 0.6 eV. This value is in good agreement with that in the band diagram prepared on the assumption that carriers are transported through an a-C:H matrix with an indirect band gap of 1.7 eV in UNCD/a-C:H films. The heterojunction diode evidently detects 254 nm monochromatic light, and photocarriers generated in UNCD grains with a band gap of 5.4 eV are expected to be transported through the a-C:H
065801-4
# 2013 The Japan Society of Applied Physics
Person-to-person distribution (up to 10 persons) by the author only. Not permitted for publication for institutional repositories or on personal Web sites.
Jpn. J. Appl. Phys. 52 (2013) 065801
Y. Katamune et al.
matrix. Since the UNCD/a-C:H films contain dangling bonds detectable by ESR, the appropriate termination of dangling bonds should be key for the application of borondoped UNCD/a-C:H films in optoelectronic devices.
7) T. Yoshitake, A. Nagano, S. Ohmagari, M. Itakura, N. Kuwano, R. Ohtani,
8) 9) 10)
Acknowledgements
This work was partially supported by ‘‘Nanotechnology Network Japan’’ (Kyushu Synchrotron Light Research Center, Proposal Nos. 081152N and 090423N) and ‘‘Open Advanced Research Facilities Initiative’’ (Kyushu Synchrotron Light Research Center, Proposal Nos. 100320AS and 1104035AS) of the Ministry of Education, Culture, Sports, Science and Technology, Japan (MEXT); and the Advanced Low Carbon Technology Research and Development Program (ALCA), Japan Science and Technology Agency (JST), a research grant from the Mazda Foundation, and a Grant-in-Aid for Scientific Research (24656389) from Japan Society for the Promotion of Science. Y.K. was supported by a research fellowship from the Japan Society for the Promotion of Science (JSPS) for Young Scientists.
11) 12) 13) 14) 15) 16) 17) 18) 19) 20) 21) 22)
1) S. Bhattacharyya, O. Auciello, J. Birrell, J. A. Carlisle, L. A. Curtiss, A. N.
2) 3)
4) 5) 6)
Goyette, D. M. Gruen, A. R. Krauss, J. Schlueter, A. Sumant, and P. Zapol: Appl. Phys. Lett. 79 (2001) 1441. S. Al-Riyami, S. Ohmagari, and T. Yoshitake: Appl. Phys. Express 3 (2010) 115102. O. A. Williams, M. Nesladek, M. Daenen, S. Michaelson, A. Hoffman, E. Osawa, K. Haenen, and R. B. Jackman: Diamond Relat. Mater. 17 (2008) 1080. P. Zapol, M. Sternberg, L. A. Curtiss, T. Frauenhein, and D. M. Gruen: Phys. Rev. B 65 (2001) 045403. J. Birrell, J. E. Gerbi, O. Auciello, J. M. Gibson, D. M. Gruen, and J. A. Carlisle: J. Appl. Phys. 93 (2003) 5606. K. Hanada, T. Nishiyama, T. Yoshitake, and K. Nagayama: J. Nanomater. 2009 (2009) 901241.
23) 24) 25)
26) 27) 28)
065801-5
H. Setoyama, E. Kobayashi, and K. Nagayama: Jpn. J. Appl. Phys. 48 (2009) 020222. T. Yoshitake, A. Nagano, M. Itakura, N. Kuwano, T. Hara, and K. Nagayama: Jpn. J. Appl. Phys. 46 (2007) L936. S. Ohmagari and T. Yoshitake: Appl. Phys. Express 5 (2012) 065202. K. Hanada, T. Yoshitake, T. Nishiyama, and K. Nagayama: Jpn. J. Appl. Phys. 49 (2010) 08JF09. K. Hanada, T. Yoshida, Y. Nakagawa, and T. Yoshitake: Jpn. J. Appl. Phys. 49 (2010) 125503. E. Pace, R. Di Benedetto, and S. Scuderi: Diamond Relat. Mater. 9 (2000) 987. S. Ohmagari and T. Yoshitake: Jpn. J. Appl. Phys. 51 (2012) 090123. Y. T. Jan, H. C. Hsieh, and C. F. Chen: Diamond Relat. Mater. 8 (1999) 772. A. M. Rodrigues, H. L. Gomes, P. Stallinga, L. Pereira, and E. Pereira: Diamond Relat. Mater. 10 (2001) 858. Z. C. Dong, A. S. Trifonov, N. V. Suetin, and P. V. Minakov: Surf. Sci. 549 (2004) 203. X. Liang, L. Wang, X. Ma, P. Chen, and D. Yang: Diamond Relat. Mater. 16 (2007) 306. Y. G. Chen, M. Ogura, M. Kondo, and H. Okushi: Appl. Phys. Lett. 85 (2004) 2110. X. Liang, L. Wang, X. Ma, D. Li, P. Cheng, and D. Yang: Appl. Phys. Lett. 90 (2007) 161123. H. Zhu, J. Wei, K. Wang, and D. Wu: Sol. Energy Mater. Sol. Cells 93 (2009) 1461. Y. Katamune, S. Ohmagari, and T. Yoshitake: Jpn. J. Appl. Phys. 51 (2012) 078003. K. Mayes, A. Yasan, R. McClintock, D. Shiell, S. R. Darvish, P. Kung, and M. Razeghi: Appl. Phys. Lett. 84 (2004) 1046. A. Grill: Thin Solid Films 355–356 (1999) 189. M. Guerino, M. Massi, and R. D. Mansano: Microelectron. J. 38 (2007) 915. T. Yoshitake, Y. Nakagawa, A. Nagano, R. Ohtani, H. Setoyama, E. Kobayashi, K. Sumitani, Y. Agawa, and K. Nagayama: Jpn. J. Appl. Phys. 49 (2010) 015503. G. K. Walters and T. L. Estle: J. Appl. Phys. 32 (1961) 1854. N. Mizuochi, H. Watanabe, J. Isoya, H. Okuchi, and S. Yamasaki: Diamond Relat. Mater. 13 (2004) 765. B. Druz, I. Zaritskiy, Y. Yevtukhov, A. Konchits, M. Valakh, B. Shanina, S. Kolesnik, I. Yanchuk, and Yu. Gromovoy: Diamond Relat. Mater. 13 (2004) 1592.
# 2013 The Japan Society of Applied Physics
