Cao Z. (Ed.) Thin Film Growth: Physics, materials science and applications

Подождите немного. Документ загружается.

308 Thin ﬁlm growth

measuring the capacitance value is almost the same as that for the electrical

double layer capacitor. The measured capacitance value corresponds only

to the Helmholtz layer itself and is the largest. Figure 12.16(b) shows the

situation where the surface Fermi level is effectively modied into the inside

forbidden gap by applying bias voltage. The measured capacitance value in this

case is the synthetic capacitance value for serial connection of the Helmholtz

layer and the depletion layer in inn. By further increasing the applied bias

voltage (V

bias

), the measured capacitance value nally approaches that for

the much smaller side capacitance, i.e. the depletion layer in inn. The net

donor (or acceptor in the p-type sample) concentration can be estimated by

the minimum capacitance value or the slope for the relationship between the

capacitance and applied voltage. Figures 12.16(c) and (d) show the similar

surface Fermi level modulation for p-type InN samples where the surface

is also covered by high-density electrons, i.e., Figs 12.16(c) and (d) are for

under zero bias voltage and for the surface Fermi level modied into the

forbidden gap, respectively. as stated above, the minimum capacitance value

reects the net donor/acceptor concentration inside the InN bulk layer.

The important point for detecting the conduction type of the semiconductor

sample by ECV is as follows. The applied bias voltage necessary for modulating

the surface Fermi level to detect the minimum capacitance is different between

n- and p-type samples as shown in Figs 12.16(a) and (c). The reason is that

the Fermi level position inside the layer is different for n- and p-type samples.

The Mott–Schottky plot is usually prepared by 1/C

vs V

bias

and the bulk

doping level and type can be estimated from the peak height of 1/C

and the

position of V

bias

. During the ECV measurements, it was found that interface

states existed on the surface of inn and contributed to the measured C

min

and thus the directly estimated N

or N

are overestimated. The effect of

the interface states on ECV results is estimated by the difference between

theoretically predicted and measured capacitance, where the average DC of

634 and 1320 nF/cm

were estimated for In- and N-polarity, respectively.

The DC can be regarded as a contribution from the interface states, and

their densities corresponding to DC are roughly estimated to be 4.0 ¥ 10

and 8.4 ¥ 10

–2

for In- and N-polarity. The difference between the

densities is probably due to different surface properties of InN with opposite

polarities.

Figure 12.17 summarizes the p-type doping regions of Mg-doped InN by

ECV measurements. in the in-polarity case, N

increases from 7.0 ¥ 10

to 1.3 ¥ 10

–3

with increasing [Mg] from 1.2 ¥ 10

to 2.9 ¥ 10

–3

For the over-doped InN at [Mg] ≥ 1.8 ¥ 10

–3

, they change to n-type

and N

increases from 2.2 ¥ 10

to 2.4 ¥ 10

–3

with increasing [Mg]

from 1.8 ¥ 10

to 3.9 ¥ 10

–3

. For InN:Mg at [Mg] < 1.2 ¥ 10

–3

compensation of residual donors has been observed and N

decreases with

increasing [Mg]. Similar tendency of N

’s and N

’s as a function of [Mg]

ThinFilm-Zexian-12.indd 308 7/1/11 9:44:19 AM



309Polarity controlled epitaxy of III-nitrides and ZnO

has also been observed in the n-polarity case. here, it should be noticed

that there is a relatively large error for the estimated N

’s and N

’s due to

the complexity of the ECV measurement. Anyway, we can conclude that the

p-type conduction is achieved at [Mg] ~ 10

–3 ¥ 10

–3

while the InN:Mg

at other regions are n-type conduction. in addition, the p-type conduction is

also conrmed by thermal electric measurement for In-polar samples grown

at [Mg] ~6 ¥ 10

–3

12.6 Polarity controlled epitaxy of ZnO

as has already been discussed in section 12.2, surface treatment of sapphire

substrate is important for polarity controlled epitaxy of high quality Zno

epilayers. During the ZnO epitaxy, the initial stage growth of the ZnO buffer

layer determines the polarity of the ZnO lm while the growth condition of

the ZnO epilayer has no inuence. Table 12.3 shows the polarities of single

Net acceptor or donor concentration (cm

–3

)

n-type p-type n-type

N-polarity

In-polarity

Calibrated N

Mg concentration (cm

–3

)

12.17 Conduction regions and calibrated net acceptor/donor

concentrations of InN:Mg layers for In- and N-polarity as a function

of Mg concentration.

ThinFilm-Zexian-12.indd 309 7/1/11 9:44:19 AM



310 Thin ﬁlm growth

domain ZnO epilayers. As shown in the table, ZnO samples using Ga pre-

exposure are all o-polar. in this case, the surface is terminated at ga atoms

no matter what kind of pre-treatment methods are used. When the growth

of the ZnO buffer layer begins, the Ga–O bond is formed rst followed by

the formation of the Zn–o bond. This ga–o bond belongs to the structure

of ga

and results in o-polarity, acting as a template for the subsequent

ZnO growth and resulting in the O-polar ZnO epilayer. In addition, the Ga

pre-exposure greatly enhances the sample quality, where the FWHM of

w-rocking curve (002) symmetric plane is as narrow as 67 arcsec in X-ray

diffraction (XRD) measurement.

sapphire nitridation also greatly improves the crystalline quality of Zno

since the formed thin AlN layer reduces the lattice mismatch between ZnO

and sapphire. The FWHMs of (002) and (102) w-scans decrease from 912

arcsec to 95 arcsec and 2870 arcsec to 445 arcsec, respectively. In the case

of the polarity, the ultra-thin aln layer formed after nitridation is n-polar.

However, the following ZnO buffer layer is found in Zn polarity, as shown

in Table 12.3. It means that the polarity is inverted from N-polarity to Zn-

polarity at the initial stage of ZnO buffer layer growth. At the initial growth

stage, it seems that the Zn atom should bond with the N atom rst because

the cation–anion bond would be easily formed. In this case, if the Zn–N bond

has a hexagonal structure, the polarity should not be inverted. However, if the

Zn–n bond has the same structure as that of Zn

, it is possible to invert the

polarity because the Zn

may play the same role as the Mg

which has

been found to invert the polarity during gan epitaxy. Unfortunately, there is

no experimental evidence of the formation of Zn

. in our in-situ RhEED

monitoring of the ZnO buffer layer growth, a halo pattern was observed

shortly before the appearance of a Zno RhEED pattern. This indicates that

an amorphous/chaotic layer including Al, N, O and Zn is formed at the

interface. This is reasonable because the o atom may replace the n atom

partly at the interface due to the stronger bond of al–o in comparison to

that of al–n. on this layer, the Zn-polar Zno is more thermodynamically

stable due to the low growth temperature, resulting in Zn-polar lm (Ohkubo

et al., 1999). Here, the formation of the amorphous/chaotic layer is due to

the aln layer under it being terminated by the n atom plane. Therefore, if

Table 12.3 Growth conditions and polarities of single-domain ZnO epilayers

Samples TC h* O* Ga N* Polarity

30 mins

10 mins

30 mins 30 mins

2 ML

O-polarity

I 30 mins 10 mins 60 mins Zn-polarity

ML = monolayer.

ThinFilm-Zexian-12.indd 310 7/1/11 9:44:19 AM



311Polarity controlled epitaxy of III-nitrides and ZnO

we deposit only one ML Ga, the top layer is not N but Ga. In this case, the

defective layer should not be formed because the Ga–O bond will form at

the interface. During the growth, the halo pattern does not appear from the

RhEED in-situ investigation. Hence, the polarity follows the N-polarity of

ultra-thin AlN and O-polar ZnO layer is grown. These polarities are conrmed

by both CaiCiss and chemical etching measurements.

Zn-polar ZnO layers usually show a higher growth rate, about 1.4 times

to that of O-polar layers. The reason for the different growth rates is the

different dangling bond congurations of the growing surfaces. Since each

o atom on an o-polar Zno surface has only one dangling bond along the

c-axis while each O atom has three dangling bonds on the Zn-polar one, the

Zn sticking coefcient on the O atom plane of the Zn-polar ZnO is higher

than that of the O-polar one. Therefore, the growth rate of the Zn-polar ZnO

is higher than that of the O-polar one. The Zn-polar samples also show a

slightly better crystalline quality. The FWHM values of (002) and (102)

w scans for Zn-polar ZnO were 119 arcsec and 486 arcsec while those for

O-polar one were 73 arcsec and 673 arcsec, respectively. Of course, the

quality is also related to the buffer layer or interface modication.

The crystalline quality and surface morphology can be further improved

by using a thin GaN interlayer. A quite thin (3 nm) GaN layer grown under

heavily n-rich conditions has been used to greatly improve the surface

atness and also the crystalline quality. Figure 12.18 describes the detailed

growth processes of the ZnO epilayers grown on nitridated sapphires

with and without the 3 nm thick GaN interlayer based on in-situ RhEED

investigation. It is clear that the RHEED pattern of [101

0]Al

changed

to [112

0]AlN at the same e-beam azimuth after sapphire nitridation. Then,

during the growth of the GaN layer, the streaky RHEED pattern of AlN

changed into a spotty pattern, indicating the three-dimensional growth

of GaN. This spotty pattern was maintained even after high-temperature

thermal annealing as shown in Fig. 12.18(c). During the subsequent growth

of the Zno buffer layer, the RhEED spots became gradually elongated

at rst and remained so until the end of buffer layer growth as shown in

Fig. 12.18(d). This RHEED pattern changed to a streaky pattern when the

temperature was increased to 600°C and became sharper as shown in Fig.

12.18(e) after the temperature was increased to 680°C, i.e., the epilayer

growth temperature. Thus, higher temperature annealing of the ZnO buffer

layer is not necessary in this case. This streaky pattern became more intense

after several minutes of epilayer growth and was maintained until the end

of growth as shown in Fig. 12.18(f), indicating a at surface. In the case

of the sample grown without the LT–GaN layer, however, a spotty RHEED

pattern was observed at the end of the growth of the ZnO buffer layer as

shown in Fig. 12.18(g). From Fig. 12.18(h), we found that these RHEED

spotty patterns did not change into streaky patterns when the temperature

ThinFilm-Zexian-12.indd 311 7/1/11 9:44:19 AM



312 Thin ﬁlm growth

was increased to 680°C. They became more spotty after several minutes

growth and were maintained until the end of growth, indicating a rough

surface. The surface morphology is shown in Fig. 12.19, where several

large triangular terraces can be clearly seen. The step height is conrmed

to be 0.26 nm by the line scan of the surface. This value corresponds to the

distance between the neighboring Zn layers (single layer or half a unit cell)

in the ZnO crystal structure. These step edges are along the (101

0), (0110)

and (1

100) planes and the steps in the following layer are rotated by 60° as

shown in the gure. This type of triangular terrace with single atomic steps

has also been observed on the surface of Zn-polar Zno bulk by scanning

tunneling microscopy (STM) and was proved to be favorable to the surface

stability. small hexagonal pits can also be found in the surface as denoted

in the AFM image. The surface is very at with an rms roughness of 0.13

nm in a scanned area of 3 ¥ 3 mm. This surface morphology was different

from that of the O-polar ZnO epilayer, where a hexagonal morphology with

double-atomic-layer step (0.52 nm) is usually observed.

(a) before nitridation

[101

0]Al

[1120]ZnO

[112

0]AlN

[112

0]ZnO

[1120]ZnO

[112

0]GaN

[112

0]ZnO

[1120]ZnO

(d) ZnO buffer with GaN

(g) ZnO buffer without

Gan

(b) After nitridation

(e) before Epi (with GaN)

(h) before Epi (without

Gan)

(f) After Epi (with GaN)

(i) After Epi (without

Gan)

12.18 RhEEd patterns along [1010]Al

e-beam azimuth recorded

during growth. (a) the (0001) Al

surface before nitridation; (b)

the surface after nitridation, it is clear that a smooth AlN layer was

formed with an epitaxial relationship of [112

0]AlN//[1010]Al

; (c)–

(f), growth with the LT-GaN layer; (g)–(i), growth without the LT–GaN

layer. (c) the thin GaN layer after thermal annealing; (d) the ﬁnished

ZnO buffer layer (at 400°C); (e) the ZnO buffer layer before epitaxy

(at 680°C); (f) the ﬁnished ZnO epilayer grown with the GaN layer; (g)

the ﬁnished ZnO buffer layer grown without the GaN layer (at 400°C);

(h) the ZnO buffer layer before epitaxy (at 680°C); (i) the ﬁnished ZnO

epilayer grown without the GaN layer.

ThinFilm-Zexian-12.indd 312 7/1/11 9:44:19 AM



313Polarity controlled epitaxy of III-nitrides and ZnO

The crystalline quality was also improved in comparison with that of

the ZnO epilayer grown without the LT–GaN layer. The FWHM values of

(002) and (102) symmetric and asymmetric w-scans were 41 and 378 arcsec,

respectively. These narrow FWHM values indicate that the densities of both

the edge-type and screw-type threading dislocations are low, which was also

reected by the high mobility (about 148 cm

–1

at room temperature).

12.7 Conclusions

Polarity does have great inuence on the epitaxy and properties of III-nitrides

and Zno. several methods such as CBED, chemical etching, CaiCiss,

and CPGE can effectively detect the polarity, with each having their own

particular advantages in different specic areas. C-plane sapphire is the

most commonly used substrate. on this non-polar substrate, the skillful

modication of growing surface is very important to improve quality and

control the polarity. sapphire nitridation is a very popular and successful

method for growing high quality, single polarity III-nitrides layers by MBE.

The ultra-thin AlN layer with the thickness of several monolayers is formed

by nitridation and it exhibits n-polarity. on this n-polar aln ultra-thin

layer, high temperature-grown AlN layers lead to the Al-polar layers and

thus result in the ga-polar gan if the aln layer is used as the buffer layer

for the GaN. On the other hand, GaN epilayer with a low temperature GaN

buffer layer on nitrided sapphire leads to a n-polar gan layer. Basically,

Ga-polar samples show better morphology and most nitride-based devices

[1010]

[0110]

100]

Pit

12.19 AfM surface image of the ZnO epilayer grown with the LT–GaN

layer in a scanned area of 3 ¥ 3 mm. Triangular terraces can be

observed.

ThinFilm-Zexian-12.indd 313 7/1/11 9:44:20 AM



314 Thin ﬁlm growth

are under Ga-polarity. However, N-polarity shows its advantage on electronic

devices such as HEMTs due to the different polarizations from the Ga-polar

one.

InN is still the most mysterious among the III-nitrides. The growth of InN

is also the most difcult. The polarity of InN is controlled by the downside

GaN or AlN buffer layer. The effect of polarity on InN growth is more

serious, in particular, the maximum growth temperature of the N-polarity

sample is 100°C higher than that of In-polarity. This leads to a better match

for epitaxy of alloys and quantum structures with GaN. On the other hand,

In-polar samples show better surface morphology and relatively lower

residual electron concentration. P-type doping has been performed on inn

layers, where the doping efciency is almost independent of polarity due to

the low growth temperature. The activation energy of Mg in InN is about 61

meV. The p-type conduction has been conrmed by ECV and thermal power

measurement. Further investigation on demonstration of inn pn junction

should be carried on.

on non-polar c-plane sapphire substrate, polarity controlled epitaxy of Zno

is performed. ga pretreatment and sapphire nitridation are very effective in

improving sample quality. on nitrided sapphire, Zn-polar Zno is obtained

because Zn-polarity is more thermal-dynamically stable at relatively low

temperatures. on the other hand, 1 ML ga pre-exposure is enough to prevent

the formation of defective layers and results in the O-polar ZnO. A 3 nm

thick chaotic gan interlayer is found to greatly improve the quality and

surface atness of ZnO.

12.8 References

Barth C and Reichling M (2001), ‘Imaging the atomic arrangements on the high-temperature

reconstructed a-al

(0001) surface’, Nature 414, 54.

Chen T C P, Thomidis C, Abell J, Li W and Moustakas T D (2006), ‘Growth of InN

lms by RF plasma-assisted MBE and cluster beam epitaxy’ J. Crystal Growth 288,

254.

Green D S, Haus E, Wu F, Chen L, Mishra U K and Speck J S (2003), ‘Polarity control

during molecular beam epitaxy growth of Mg-doped GaN’, J. Vac. Sci. Technol. B,

21, 1804–1811.

Heying B, Tarsa E J, Elsass C R, Fini P, DenBaars S P and Speck J S (1999), ‘Dislocation

mediated surface morphology of GaN’, J. Appl. Phys., 85, 6470.

Ide T, Shimizu M, Kuo J, Jeganathan K, Shen X Q and Okumura H (2003), ‘Surface

morphology of GaN layer grown by plasma-assisted molecular beam epitaxy on

MOCVD grown GaN template’, Phys. Status Solidi. C, 0, 2549.

Jones R E, Yu K M, Li S X, Walukiewicz W, Ager J W, Haller E E, Lu H and Schaff

W J (2006), ‘Evidence for p-type doping of InN’, Phys. Rev. Lett. 96, 125505.

Lebedev V, Schroter B, Kipshidze G and Richter W (1999), ‘The polarity of AlN lms

grown on Si(111)’, J. Crystal Growth 207, 266.

ThinFilm-Zexian-12.indd 314 7/1/11 9:44:20 AM



315Polarity controlled epitaxy of III-nitrides and ZnO

Li L K, Jurkovic M J, Wang W I, Van Hove J M and Chow P P (2000), ‘Surface polarity

dependence of Mg doping in GaN grown by molecular beam epitaxy’, Appl. Phys.

Lett. 76, 1740.

Mahboob I, Veal T D, McConville C F, Lu H and Schaff W J (2004), ‘Intrinsic electron

accumulation at clean InN surfaces’, Phys. Rev. Lett. 92, 036804.

Namkoong G, Doolittle W A, Brown A S, Losurdo M, Capezzuto P and Bruno G (2002),

‘Role of sapphire nitridation temperature on GaN growth by plasma assisted molecular

beam epitaxy: Part I. Impact of the nitridation chemistry on material characteristics’,

J. Appl. Phys. 91, 2499.

Ng H M and Cho A Y (2002), ‘Investigation of Si doping and impurity incorporation

dependence on the polarity of GaN by molecular beam epitaxy’, J. Vac. Sci. Technol.

B, 20, 1217.

Ohkubo I, Ohtomo A, Ohnishi T, Mastumoto Y, Koinuma H and Kawasaki M (1999),

‘In-plane and polar orientations of ZnO thin lms grown on atomically at sapphire’,

Surf. Sci. 443, L1043.

Park A J, Holbert L J, Ting L, Swartz C H, Moldovan M, Giles N C, Myers T H, Van

Lierde P, Tian C, hockett R a, Mitha s, Wickenden a E, koleske D D and henry R

L (2001), ‘Controlled oxygen doping of GaN using plasma assisted molecular beam

epitaxy’, Appl. Phys. Lett. 79, 2740.

Ponce F A, Bour D P, Young W T, Saunders M and Steeds J W (1996), ‘Determination

of lattice polarity for growth of GaN bulk single crystals and epitaxial layers’, Appl.

Phys. Lett. 69, 337.

Smith A R, Feenstra R M, Greve D W, Shin M S, Skowronski M, Neugebauer J and

Northrup J E (1998), ‘Determination of wurtzite GaN lattice polarity based on surface

reconstruction’ , Appl. Phys. Lett. 72, 2114.

Sumiya M, Yoshimura K, Ohtsuka K and Fuke S (2000), ‘Dependence of impurity

incorporation on the polar direction of GaN lm growth’, Appl. Phys. Lett. 76,

2098.

Takei Y and Nakayama T (2009), ‘Electron carrier generation by edge dislocations in

InN lms: rst-principles study’, J. Crystal Growth 311, 2767.

Wang X and Yoshikawa A (2004), ‘Molecular beam epitaxy growth of GaN, AlN and

InN’, Prog. Cryst. Growth Charact. Mater., 48/49, 42.

Wang X, Che S B, Ishitani Y and Yoshikawa A (2006), ‘Step-ow growth of In-polar

InN by molecular beam epitaxy’, Jpn. J. Appl. Phys. Part 2, Letters, 45, L730.

Wang X, Che S B, Ishitani Y and Yoshikawa A (2007a), ‘Systematic study of p-type

doping control of InN with different Mg concentrations with both In and N polarities’,

Appl. Phys. Lett. 91, 242111.

Wang X, Che S B, Ishitani Y, Yoshikawa A, Sasaki H, Shinagawa T and Yoshida S

(2007b), ‘Polarity inversion in high Mg-doped In-polar InN epitaxial layers’, Appl.

Phys. Lett. 91, 81912.

Xu K and Yoshikawa A (2003), ‘Effect of lm polarities on InN grown by molecular

beam epitaxy’, Appl. Phys. Lett. 83, 251–253.

Xu K, Yano N, Jia A W, Yoshikawa A and Takahashi K (2002), ‘Polarity control of GaN

grown on sapphire substrate by RF-MBE’, J. Crystal Growth, 237, 1003.

Yim J W L, Jones R E, Yu K M, Ager J W, Walukiezicz W, Schaff W J and Wu J

(2007), ‘Effects of surface states on electrical characteristics of InN and InGaN’,

Phys. Rev. B. 76, 041303(R).

Yoshikawa A, Wang X Q, Tomita Y, Roh O, Iwaki H and Ishitani Y (2004), ‘Rotation-

ThinFilm-Zexian-12.indd 315 7/1/11 9:44:20 AM



316 Thin ﬁlm growth

domains suppression and polarity control of ZnO epilayers grown on skillfully treated

C-al

surface’, Phys. Stat. Solidi B 241, 620.

Zhang Q, Wang X Q, he X W, Yin C M, Xu F J, shen B, Chen Y h, Wang Z g,

Ishitani Y and Yoshikawa A (2009), ‘Lattice polarity detection of InN by circular

photogalvanic effect’, Appl. Phys. Lett. 95, 031902.

ThinFilm-Zexian-12.indd 316 7/1/11 9:44:20 AM



317

Understanding substrate plasticity and

buckling of thin films

F. Foucher, centre de Biophysique Moléculaire – cNrS,

France and c. coupeau, J. coliN, a. ciMetière,

and J. Grilhé, université de poitiers, France

Abstract: the aim of this chapter is to study the effect of crystalline

substrate plasticity on thin lm buckling. A general description of buckling

is given rst within the framework of the Föppl–von Kármán theory of thin

plates. The model is then used to describe the buckling induced by plasticity

of the crystalline substrate. The localization of the buckling structures

on the areas of dislocation emergence is explained. Finally, the model is

generalized in order to obtain all the equilibrium states of a deformed thin

lm on a plastically strained crystal.

Key words: thin lm buckling, crystalline plasticity, atomic force

microscopy, Föppl–von Kármán theory of thin plates.

13.1 Introduction

Thin lms are used in a wide range of domains in order to confer new

properties to materials. To be efcient, the lm must be fully adhesive to the

substrate surface. The thin lms may, however, peel off from their substrate

after deposition, forming buckling structures. This phenomenon is due to the

relaxation of the internal stresses in the lm that can reach several GPa in

compression. Several studies have been carried out in order to understand

the underlying mechanisms of buckling since this phenomenon is associated

with a loss of properties initially conferred to the system.

The rst observations of buckling structures in thin lms were carried out

at the beginning of the 1980s by Matuda et al. (1981). The authors realized

that the mechanical properties of thin lms, such as Young’s modulus,

internal stress or adhesion energy, i.e. properties that are relatively difcult

to determine experimentally, could be estimated through the phenomenon of

buckling. In 1992, Hutchinson and Suo described the mechanical processes

involved during buckling and delamination. Their model was based on the

theory of thin plates developed at the beginning of the twentieth century by

Föppl (1907) and von Kármán (1910).

In most of the present applications, the lm thicknesses rarely exceed

one micrometer, leading to the formation of micrometric buckles. Their

ThinFilm-Zexian-13.indd 317 7/1/11 9:44:39 AM

