QUANTUM SIZE EFFECTS IN CDS THIN-FILMS

PHYSICAL REVIEW

B

VOLUME 45, NUMBER 20

Quantum

size

effects

in

CdS

thin films

15MAY 1992-II

Der-San Chuu

Department

of

Electrophysics, National Chiao Tung University, Hsinchu, Taiwan 30050, Republic ofChina

Chang-Ming Dai

Electronics Research & Service Organization, Industrial Technology Research Institute, Chutung, Hsinchu, Taiwan 31015,Republic ofChina

(Received 2October 1991;revised manuscript received 17 December 1991)

Resonant Raman spectroscopy is used to study quantum size effects in CdSfilms. The lattice

soften-ing ofthe CdS LO-phonon mode in a CdS film with a thickness less than 800 A is observed. As the

0 ~ 0

thickness is less than 410A, the TO-phonon mode isobserved at 4880A excitation wavelength, which is

above the band gap ofbulk CdS (2.42 eV)at room temperature. These phenomena are attributed to the size quantization effects ofthe grain size and the low-dimensional thin-film structure. The quantum size

effects cause a blueshift ofthe band gap in the as-deposited CdS thin film. The peak ofthe TO-phonon-mode Raman line ofthe CdSfilm is observed around 220 cm ',which has ashift of8cm 'from the

Ra-man line ofthe most active TO-phonon mode ofbulk CdS. The magnitude ofsoftening energy ofthe TO-phonon mode isobserved tobeindependent ofthe film thickness.

I.

INTRODUCTION

Quantum size effects in semiconductors have attracted

much attention in recent years. ' ' In these investiga-tions, vibrational spectroscopy has played an important

role. The quantum size effect is usually defined as the dependence

of

certain physical properties

of

a solid on its

characteristic geometric dimensions when the latter be-comes comparable to the de Broglie wavelength

of

the charge carriers. As the film thickness (measured along the zdirection) ismuch less than the two in-plane dimen-sions (measured in the x-y plane) and is comparable to

the de Broglie wavelength

of

the carriers, quantization

of

the carrier motion in the perpendicular direction to the film plane occurs, although the carrier motion is particle-like in the other two dimensions. Quantum size effects are usually attributed to the small size

of

microcrystal-lites in the films, ' ' the lower-dimensional thin-film structure,

"

' and the effect

of

mechanical stress. '

In the last few years, the studies

of

quantum size effects were mainly focused on the observation

of

the shift

of

Raman peaks. This is because Raman spectroscopy can probe the local vibrational environment and thus can

determine the presence

of

semiconductor microcrystal-lites inthe films.

To

date, several works have discussed the size effects

of

small CdS crystallites with a diameter ranging from 30to 1000

A.

Meanwhile, some authors studied the relation-ship between the Raman shift and the film

thick-ness.

"

' Ekimov, Efros, and Onushchenko developed a growth technique

of

the semiconductor microcrystals in aglassy dielectric matrix which permitted one to vary the size

of

the grown microcrystals in a controlled manner from some tens to thousands

of

angstroms and thus to

study the size dependence

of

absorption spectra

of

a

num-ber

of

semiconductors. Their results showed a consider-able short-wavelength shift

of

the exciton lines was ob-served as the microcrystal size decreased. Ramsteiner

et al.

"'

studied the hard hydrogenated amorphous

car-bon films on Si by using Raman spectroscopy. They ob-served that the frequency

of

the main Raman peak de-creased for layers thinner than 100 A and applied the term "mode softening" to describe the result. Tu, Wong and Ketterson' studied the thickness dependence

of

the band gap by measuring the resistivity in the intrinsic re-gion. Briiesch et

al.

' studied the vibrational properties

of

thin A1203 films sputtered onto Au, Al, and Si sub-strates by using infrared reflection and absorption

spec-troscopy. They found a softening

of

the

longitudinal-optical

Al-0

stretching mode for a layer thickness less than

-500

A.

In this paper we report that the thickness dependence

of

the LO vibrational mode softening

of

the CdS thin films can be observed asfor film thicknesses less than 800

A.

Furthermore, the TO-phonon mode

of

CdS film can be observed at 4880 A excitation wavelength, which is above the band gap

of

bulk CdS (2.42 eV) at room tem-perature asforfilm thicknesses less than

410

A.

II.

EXPERIMENT

The pulsed laser evaporation (PLE)technique was used

to

produce the CdS thin film on a p-type (100)silicon wafer. In general, the Raman signal is quite sensitive

to

crystal orientation due to selection rules. As shown in a previous paper, the

PLE

technique could easily produce the high-quality thin films that yielded sharp Raman peaks. The CdS thin films were grown by a

PLE

system which contains a homemade Xe-ion laser'7 (A,

=495

nm) operating at about 1 pulse per second with maximum pulsed energy

of

10mJ. The laser beam was focused onto a target at about 20' from the surface normal

of

the tar-get. The laser fluence was about 1 J/cm and the target

was moved slightly after hundreds

of

shots

to

prevent causing a sunken spot. A clean vacuum chamber, which was pumped by a 25 liter/sec ion pump, having a base

11806 DER-SAN CHUU AND CHANG-MING DAI

pressure

of

3X10

Torr was used. CdS thin films were deposited on the p-type (100)silicon wafers at

150'C

sub-strate temperature. The temperature was controlled to fix at

150'C

with the aid

of

a temperature controller. The deposition rate was about

0.

07A/pulse. CdS powder

of

99.

999%

purity was pressed into a pellet which was deliberately left rough so that the coupling

of

the laser beam into the target was quite efficient.

The crystallinity and surface morphology were identified by x-ray diffraction

(XRD)

and scanning elec-tron microscopy (SEM). These high-orientation films were prepared with a thickness ranging from

-80

to

-800

A.

The thickness

of

film was measured by using an ellipsometer which has a resolution

of

+10

A.

The

Ra-man spectra were recorded by using the unpolarized line

of

4880 A

of

an

Ar+

laser. Spectra were recorded at

room temperature in 45' reflection geometry with sam-ples being placed under the micrometer stage

of

a triple-grating spectrograph (Spex 1877C) equipped with a liquid-nitrogen-cooled charge coupled devices detector

array (Photometrics CC200). The incident power on the sample was about 100 mW. The emission spectra

of

Ne and

Xe

lamps were used for frequency calibration. The

slit width (100pm) used in the experiment led toan

accu-racy

of

3-cm ' resolution.

film has wurtzite structure and high orientation in (002) direction even when the substrate isat room temperature.

This result manifests itself in that the clusters in the films have acolumnar structure in the normal direction to the film surface. Figure 2 shows the SEMphotographs

of

the films at substrate temperatures

30'C

and

150'C.

The re-sults reveal that the surfaces

of

the as-deposited CdS films are smooth and uniform. The crystallite sizes

of

the CdS films can be either estimated from the SEM photograph

or calculated _by the Scherrer equation:

d=kk, lD

cos8,

where kistaken to be 1forhexagonal CdS, A, isthe x-ray

wavelength,

D

isthe angular linewidth forhalf-maximum intensity (in radians), and

8

is the Bragg angle. These re-sults obtained are

-290

A for

30'C

and

-380

A for

150'C.

These are very close tothe results estimated from

the transmission electron microscopy photographs.

Figure 3 shows the Raman spectra

of

the as-deposited

CdS thin films with film thicknesses ranging from 85 to 800A. These Raman spectra were excited by the 4880-A line

of

an Ar+ laser. The spectra show a main CdS

Ra-man peak (1LO)around 300cm ' as well as the intrinsic Raman scattering (Or)

of

p-type (100)silicon substrate around 520 cm

'.

The 1LOphonon frequency for a sin-gle crystal

of

CdS was given' as 305 cm

',

while the values measured by us were around 300cm

'.

This low-frequency shift

of

the 1LORaman peak in CdS thin film

III.

RESULTSAND DISCUSSIONS

The structure and crystallinity were analyzed by

XRD

with a Cu

Ea

source (A,

=1.

54 A). Figure 1 shows the

XRD

spectrum

of

as-deposited CdS film on ap-type (100)

silicon wafer at room temperature. One can see that the

(a)

(b)

cV C) CO CD cu

0+

hl l 10 I 20 I 30 40 I 50 I 60 70 2&9 (deg)

FIG.

1. (a) The XRDspectrum ofthe as-deposited CdSfilms

on silicon substrate at room temperature. (b) The pattern of CdS powders.

FIG.

2. The SEMphotographs for the films grown on silicon substrates at temperature (a)30 Cand (b) 100 C.

45 QUANTUM SIZE EFFECTSIN CdSTHIN FILMS I I I I 1000

6000-

Si OI _800— C

&4000-CO ~~ M C

I

c

2000 CdS 1LO 800A 500A 600 g

400-8

200-12 15 i'I I ~ ~ 300 150I g5A 400 500 Raman shift (cm )

FIG.

3. Thickness dependence ofthe 1LO phonon energies. These RRS spectra ofCdS films deposited on a p-type silicon

0

wafer were excited at 4880A and were detected at room tem-perature. CdS film thickness is indicated in the figure. The dashed line marks the frequency ofthe main Raman peak (300 cm ') for a thick (800 A)film.

has been studied before and is mainly ascribed to the grain-size effect. In

Fig.

3 the spectra reveal as decreas-ing the thickness

of

the CdS films, the silicon Raman peak shifts to the low-frequency region although the amount

of

shift is not very evident (around 6 cin ' for 85-A-thick film). The low-frequency shift

of

the silicon Raman peak as the film thickness is less than 800 A

might be ascribed to the stress effect caused by the cou-pling

of

the CdS film and the silicon substrate at the

in-terface. As the film thickness

of

CdS increases, the cou-pling effect between the CdS film and the silicon substrate

at the _{interface becomes less prominent.} Hence, for CdS films thicker than 800 A, the optical zone center

of

the Raman line

of

silicon resumes eventually to the value

of

520cm '

of

a p-type (100)silicon substrate.

From

Fig.

3 it is also found that the main CdS 1LO

Raman peak shows an increasing shift to the lower-frequency region with decreasing layer thickness. The

shifts

of

the Raman peak

of

the 1LOphonon mode as a function

of

the film thicknesses at three excitation ener-gies, 4579, 4880, and 5145 A, are shown in

Fig. 4.

One

can note that these shifts are independent

of

the photon

excitation energies and are consistent with the previous works.

"'

From

Fig.

4 it can be seen that the Raman shifts

of

the 1LOphonon peak

of

CdS films are saturated

to the value

of

5 cm ' when the films are thicker than

800 A and the 1LORaman peak has a shift

—

15 cm

for a thin film with 85 A thickness. The saturation

of

1LORarnan shift as film thickness exceeds 800 A is due

to the grain-size effect, as mentioned above. The low-frequency shift

of

the 1LORaman peak

of

CdS films as the film isthinner than 800 A might be attributed tothe combination

of

(1) the size effect,

'

i.

e., the effect

of

size on the vibrational properties in small crystallites and the low-dimensional thin-film structure, and (2)the stress effect induced by the film-substrate interface on the

vibra-Shift ofPeak Position (cm')

FIG.

4. Frequency downshifts ofthe 1LO Raman peaks rela-tive toits position forbulk CdS as a function offilm thicknesses. The full curve is drawn to guide the eye. The Raman spectra

were excited at4579,4880,and 5145A.

tional spectrum.

It

is also known that the vibrational mode frequencies vary with increasing temperature. Thus, the Raman shift might be caused by the higher temperature due tothe higher absorption

of

the probe en-ergy in the thicker films. However, this possibility can be ruled out because as we will show later (see

Fig.

6) a

significant difference in the absorption cannot happen for

the films having a thickness difference within only an

or-der

of

several hundred angstroms. Thus, the thickness variations

of

the films cannot make a meaningful difference in temperature.

The vibrational properties

of

the A1203 thin films on Si and Au have been studied by Briiesch et

al.

' They found

a softening

of

the longitudinal-optical

Al-0

stretching mode for a layer thickness less than

—

100

A. For

a 10-A-thick A1203 film a relative mode softening

of

-3%

with respect to the mode

of

the bulk crystal was reported.

Ramsteiner et

al.

"'

studied

a-C:H

on Si using Raman

spectroscopy. They reported that the relative softening

of

a 10-A

a-C:H

film on Siis

-4.

5%.

They also found a

clear substrate dependence

of

the softening. Our present results yield a relative softening

of

-4.

4%

(with respect

to the mode

of

bulk CdS) for a 100-A CdS film on Si wafer. The discrepancy between our results and previous work might be caused by the following. (1) The lattice

mismatch in CdS/Si system (23.

8%,

lattice constant

of

CdS, a axis

=4.

136 A; for Si, 5.43A) islarger than those in a-C:H/Si or A1203/Si systems. The energy

of

the

LO-phonon mode contains a higher contribution from the

lattice mismatch. (2) The grain-size effect was combined with the thickness effect in our work which was not

con-sidered by the previous works.

"

Figure 5shows the Raman spectra

of

as-deposited CdS

thin films with film thicknesses ranging from 700 to 100

A.

These Raman spectra were also excited by the 4880-A

line

of

an

Ar+

laser. The spectra reveal that there is a

TO

mode accompanying the main CdS Raman peak

(1LO) around 300 cm ' when the film thickness is less than

410 A.

This TO Raman line

of

the CdS thin film shifts from 228 cm

',

the most active

TO

mode

of

bulk

thick-ll

808 DER-SAN CHUU AND CHANG-MING DAI 45

ness. Since the deformation potential is the main mecha-nism

of

the TO Raman scattering, the TO phonon line can be degraded only by the grain-size and stress effects. Therefore, the softening

of

the TO mode (shift from 228 cm '

of

the most active TO mode

of

bulk CdS to 220

cm ' for a film thinner than 800 A) shown in

Fig.

5 might be ascribed to the combined effect

of

the stress and the grain size.

It

is well known that the resonant Raman scattering (RRS) cross section

of

the TO mode has the highest efficiency when the resonant condition is satisfied. ' ' As the photon energy approaches the band gap

of

cadmium sulfide, the pronounced decrease in the Raman scattering efficiency

of

the TO mode will be ob-served prior to the onset

of

resonant enhancement. '

The observed decrease in the TOmode as the photon en-ergy approaches the band gap

of

CdS can be interpreted by extending the Loudon theory to include in the Raman-scattering amplitude the destructive interference term between resonant and nonresonant contributions.

To

date, TO mode scattering was observed' for pho-ton energy

of

a laser beam larger than the band-gap ener-gy. Although the phenomenon

of

the disappearance

of

the scattering intensity

of

the TO mode for photon ener-gy larger than the band-gap energy is well known,

howev-er, the real mechanism still remains unknown. Since the band-gap energy

of

bulk CdS at room temperature (i.e.,

25'C)

is 2.42 eV, the TO mode

of

the bulk CdS single crystal cannot be observed by using 4880-A (2.54-eV)

ex-citation.₀ The abnormal occurrence

of

the TO mode at 4880-A excitation in this work might be ascribed to the blueshift

of

the band gap

of

a CdS film when the film thickness is less than 700

A.

The effect

of

reducing film thickness will normally result in the observable high-energy shifts

of

the absorption and transmittance. As shown in Fig. 6,the transmittance properties

of

CdS thin films deposited on the glasses with various thicknesses

100 60

g

40 8 cg 20 C p&p' ~ T . g 400 t 500 1 600 700 800 900 0

ranging from 1000 to 4000 A were investigated. We can note from the figure that the absorption edge

of

thinner film is shifted to the smaller-wavelength region (blue-shift). One can also note that the transmittance curves below absorption edges are highly thickness dependent, and the features

of

these curves agree with the work

of

El-Nahass et al.

Figure 7 shows the curves

of

the square

of

the absorp-tion coefficient

a

versus the photon energy hv by using the results in Fig.

6.

As shown in Fig. 7,

a

varies linear-ly with hv,and thus the interband transition can be inter-preted by the relation

a

=

A(hv

—

E

).

To

obtain the value

of

the optical band gap we extrapolated the straight lines in

Fig.

7 to zero absorption. Results showed that the band-gap energy

E

decreases as the thickness

of

the

CdS film increases. The blueshifts

of

E

of

CdS films were estimated at about

0.

06and

0.

16 eVfor thicknesses

of

3700and

—

1000 A, respectively (comparing with the band gap

of

2.42eV

of

bulk CdS). Recently, the blueshift

AE caused by the quantum size efFects

of

CdS crystal-lites was measured by Tanahashi et

al.

They reported

that the blueshifts AE are

0.

13 eV for 76-A and

-0.

05

0

eV for 313-Amicrocrystallites embedded in the Si02

ma-trix. Since the grain size

of

our CdS films for different

Wavelength (nm)

FIG.

6. Spectral dependence ofthe transmittance for the as-deposited films deposited on glass substrate at different thick-ness. CU ~~ V) CD 50 230A 100A O ~& CL O tA

30-20

10-200

300

400

500

Raman shift (cm )

FIG.

5. Thickness dependence ofthe 1TO phonon intensities

and energies. The RRS spectra ofCdS films deposited on the p-type (100) silicon wafer were detected at room temperature.

0

The spectra were excited at4880 A.

Q

2.46 2.5P 2.58

Photon Energy (ev)

P..64 P..70

FIG.7. Square ofthe absorption coefficient ofCdSthin films

45 QUANTUM SIZEEFFECTSIN CdSTHIN FILMS 11809 2.0 0 I 1.5 0.5 1.0 8 4P

masked by the envelope

of

the LO signal; and (3) the scattering efficiency

of

the LOmode is larger than the TO

mode in the resonant Raman-scattering condition.

It

is worth noting that the scattering efficiency

of

the TO

mode overshadows that

of

the LO mode when thickness is less than 150

A.

It

is evidence that the Frohlich

in-teraction is weak for very thin films, so that LO-mode scattering is mainly caused by the deformation potential.

IV. CONCLUSION 0.0 100 200 300 400 Film Thickness (g) 500 4E WF 600

FIG.

8. Thickness dependence ofthe ratio ofthe 1TO inten-sity to the 1LO intensity. The solid line is only a guide to the eyes.

0

thicknesses are in the range

of

-300

A, therefore, we conclude that the blueshift

of

0.

06eV for film thickness

3700 A might be caused by the grain-size effect

of

the polycrystalline CdS film. Another blueshift

of

-0.

1 eV between the band gaps

of

films with thicknesses 3700and

—

_{1000 A can be attributed to the thickness effect}

_of

_the

CdS films. Therefore, as the thickness is less than 700 A,

the quantum size effect due to the low-dimensional thin-film structure becomes dominant and the band gap

of

the film will blueshift

to

the range

of

~2.

58eV. Under this

circumstance, the photon energy

of

2.

54eV (4880A) ly-ing below the band edge

of

the film becomes sufficient for

satisfying the necessary resonance condition

of

the TO

mode

of

CdS films.

In order to see the thickness effect more clearly, the relative intensity ratio

ITO/I,

„o

as a function

of

film thickness is plotted in

Fig.

8. We can see that the ratio

decreases when the film thickness increases. One can also note that the Raman signal

of

the

TO

mode is insufficient

for observing asfilm thickness is larger than

410 A.

This might be due to (1) the intrinsic thermal broadening

of

the LO Raman peak at room temperature; (2)the intensi-ty

of

the LO peak becomes stronger as the film thickness increases and finally the Raman signal

of

the TOmode is

In conclusion, the shift

of

the 1LOmode Raman peak

to the lower-frequency region as the thicknesses

of

as-deposited CdS films vary from 800 to 85A isascribed to

the size effect (including the lower-dimensional thin-film structure and the grain-size effect

of

the thin films) and the stress effect.

For

films thicker than 800 A, the peak

of

the 1LO phonon line is at 300 cm ' instead

of

305

cm ' for the bulk CdS. This saturation

of

the shift

of

1LORaman peak for films thicker than 800 A is attribut-ed to the effect

of

the grain size.

For

thicknesses less than

410

A, the TO-phonon mode around 220 cm ' in the CdS thin film can be observed by using 4880-A pho-ton excitation energy which is above the band-gap energy

of

bulk CdS. The occurrence

of

the TO-phonon mode is attributed to the quantum size effect because the size quantization

of

free carriers in the lower-dimensional thin-film structure will cause the blueshift

of

the band gap in thin films. The size quantization

of

free carriers is caused by either small-volume microcrystallities or low-dimensional thin-film structure as the film thickness

de-creases. The Raman shift

of

the TOmode

of

CdS films is about 8 cm ' from the Raman peak

of

the most active

TO mode

of

bulk CdS. The magnitude

of

this softened energy isfound tobe independent

of

the film thickness.

ACKNOWLEDGMENTS

One

of

the authors

(D.S.C.

) is grateful to Professor

Y.

C.

Leeat SUNY at Buffalo forhis helpful comments. We also thank Professor

M.

C.

Lee and Professor W.

F.

Hsieh for their helpful discussions. This work has been partially supported by the National Council

of

Science, ROCunder Grant No. NSC 80-0208-M-009-25.

J. F.

Scottand

T.

C.Damem, Opt.Commun. 5, 410 (1972).

D. S.Chuu, C. M. Dai, W.

F.

Hsieh, and C.

T.

Tsai,

J.

Appl. Phys. 69,8402(1991).

R.

Rossetti, S.Nakahara, and L.

E.

Brus,

J.

Chem. Phys. 79, 1086 (1983)~

4B.

_F.

Variano, N.

E.

Schlotter, D.M. Hwang, and C.

J.

San-droff,

J.

Chem. Phys. 88,2848 (1988).

5A. V.Baranov, Ya.S.Bobovich, N.

I.

Grebenshchikova, V.

I.

Petrov, and M. Ya.Tsenter, Opt. Spektrosk. 60, 1108 (1986) [Opt. Spectrosc. (USSR)60,685(1986)].

H. Jerominek, M. Pigeon, S.Patela, Z.Jakubczk, C. Delisle, and

R.

Tremblay,

J.

Appl. Phys. 63, 957(1988).

7E.

F.

Hilinski and P.A.Lucas,

J.

Chem. Phys. 89,3435(1988). 8A.

I.

Ekimov, Al. L.Efros, and A. A. Onushchenko, Solid

State Commun. 56,921(1985).

D.V.Murphy and S.

R.

J.

Brueck, Opt. Lett. 8, 494 (1983).

G.Kanellis,

J. F.

Morhange, and M.Balkanski, Phys. Rev. B

21,1543(1980).

M. Ramsteiner,

J.

Wagner, Ch. Wild, and P.Koidl,

J.

Appl. Phys. 62, 729(1987).

M. Ramsteiner,

J.

Wagner, Ch. Wild, and P. Koidl, Solid State Commun. 67, 15(1988).

Y.

H. Lee,

K.

J.

Bachmann,

J.

T.

Glass,

Y.

M. LeGrice, and

R.

J.

Nemanich, Appl. Phys. Lett. 57, 1916 (1990).

P.Bruesch,

R.

Kotz,H.Neff,and L.Pietronero, Phys. Rev. B

29, 4691(1984).

t5V.B.Sandormirskii, Zh. Eksp. Teor. Fiz.52, 158 (1967) [Sov. Phys. JETP25, 101(1967)].

ll

810 DER-SAN CHUU AND CHANG-MING DAI 45

L.W. Tu, G.

K.

Wong, and

J.

B.

Ketterson, Appl. Phys. Lett. 55,1327(1989)~

C. M. Qai,

K.

S.Wu, W.

F.

Hsieh, and D. S.Chuu, Rev. Sci. Instrum. 61, 3713 (1990).

R.

J.

Briggs and A.

K.

Ramdas, Phys. Rev.B13, 5518 (1976).

R.

C. C. Leite and S.P. S. Porto, Phys. Rev. Lett. 17, 10

(1966).

-"R:

H. Cailehaer', S.S.&ussma, ivi. SeitIer's, anti P.K..ChdIlg;— Phys. Rev.B 7,3788(1973).

'J.

M. Ralston,

R.

L.Wadsack, and

R. K.

Chang, Phys. Rev. Lett. 25,814 (1970).

R.

Loudon, Proc.

R.

Soc.London Ser.A 275,218(1963).

J.

F.

Scott,

R.

C.C.Leite, and

T.

C.Damen, Phys. Rev. 188,

1285(1969).

T. C. Damen and

J.

F.

Scott, Solid State Commun. 9,383 (1971).

T. C.Damen and

J.

Shah, Phys. Rev.Lett.27, 1506(1971).

R.

M.Martin, Phys. Rev. B10,2620(1974).

M. M. El-Nahass, O. Jamjoum, S. M. Al-Howaity, and

K.

Abdel-Hady,

J.

Mater. Sci. Lett. 9,79(1990).

--r".SaÃMni

mcus.

I .Fdngvicimimi ms6't.'i7vfllcSQQi eh phdBpil=

calTransitions in Solids, edited by

R.

A.Ballinger (Pergamon, Oxford, 1975).

I.

Tanahashi, A.Tsujimura,

T.

Mitsuyu, and A.Nishino. Jpn.

J.

Appl. Phys. 29, 2111 (1990).