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

Подождите немного. Документ загружается.

298 Thin ﬁlm growth

as sample quality is concerned, the AlN layer formed at low temperature is

homogeneous and smooth, which acts as a high quality template and results

in a less defective and smooth buffer gan layer, leading to high quality gan

lm. On the other hand, the AlN layer formed at high temperature is rough

and inhomogeneous, which leads to a defective GaN buffer layer including

inversion domains, resulting in low quality GaN lm.

it should be noted that the aln formed by nitridation is mainly n-polarity.

On this N-polar AlN ultra-thin layer, GaN buffer layer grown at 600°C has

been proved to be mainly N-polarity (90%) by in-situ CaiCiss measurement

independent of the III/V ratios. This means that with Ga-rich growth or even

Ga bilayer, it is difcult to invert the polarity. However, the behavior of the

aln layer is different from that of gan. it has been found that the polarity

of AlN layer grown at low temperature depends critically on both the surface

stoichiometry in the initial growth stage and during epitaxy. The N-polar

AlN buffer or epitaxial layer has been grown under N-rich conditions while

the Al-polar buffer has been obtained under Al-rich growth conditions. The

N-polar AlN layer even changes to Al-polar after the growth conditions

change from N-rich to Al-rich as shown in Fig. 12.8(a). Even at the high

temperature (820°C) growth of AlN, the polarity of AlN is inverted from

N-polarity to Al-polarity by changing the growth conditions from N-rich to

al-rich. This polarity inversion is due to the formation of al bilayer in the

Al-rich growth condition as shown in Fig. 12.8(b). In the case of Ga-rich

growth for GaN epitaxy, Ga bilayer is also easily formed. However, it is

also easily broken by the strike of n atoms. The al bilayer is different. once

the Al bilayer is formed, it is more difcult to break than the Ga bilayer

and results in Al-polarity. This is also conrmed by the polarity inversion

of GaN through an insertion of Al bilayer. Figure 12.9 shows the polarity

dependence of GaN and AlN on growth conditions. The original polarity

of GaN is not inuenced by Ga/N ratios while that of AlN is maintained in

N-rich growth conditions. Al-polar AlN is usually obtained in Al-rich growth

conditions. so in the case of gan epitaxy, n-polarity is easily obtained by

using low temperature GaN buffer layer while Ga-polarity is achieved by

using AlN buffer layer grown under Al-rich conditions.

The growth behavior and properties of GaN in different polarities differ

largely. Growth is usually performed under Ga-rich conditions to improve

sample quality, which is independent of polarity. However, the Ga-bilayer

always remains at the growth front for Ga-polarity while a single Ga atom

layer exists on the top surface during epitaxy of n-polarity. The ga-polar

GaN usually shows better crystal quality and smoother surface than that of

N-polar GaN. The surface of Ga-polar GaN is essentially featureless between

steps while the N-polar surface exhibits an island structure. The Ga-polar

gan surface is more stable than that of the n-polar one in vacuum. The

growth rate of Ga-polar GaN is sometime higher than that of the N-polar one,

ThinFilm-Zexian-12.indd 298 7/1/11 9:44:16 AM



299Polarity controlled epitaxy of III-nitrides and ZnO

which is one of the reasons for the formation of pits with inversion domains.

The impurity incorporation and doping efciency are also inuenced by

the polarity. P-type doping using Mg is much easier on the ga-polar gan

than on that of the n-polar one. Li et al. (2000) reported that the p-type

conductivity with a free-hole concentration up to 5 ¥ 10

–3

was realized

on Ga-polar GaN while N-polar GaN was highly resistive or semi-insulating.

In the case of n-type doping, Ng and Cho (2002) found that there was no

signicant difference for the incorporation of Si into GaN of either polarity.

However, the background impurities such as C and O were found to be more

Signal intensity of AI (a.u.)

AIN epilayer grown in Al-rich condition

AIN epilayer grown in N-rich condition

0 20 40 60 80

Incident angle q (deg)

(a)

Ga-polarity

2ML-Al

N-polarity

(b)

12.8 (a) CAICISS spectrum of N-polar AlN epilayer grown in N-rich

conditions, and the spectrum of AlN epilayer after the growth

condition is switched to Al-rich. The thickness of an AlN layer grown

in Al-rich conditions is about 30 nm. (b) The schematic atomic

structure of polarity inversion by an Al bilayer.

ThinFilm-Zexian-12.indd 299 7/1/11 9:44:16 AM



300 Thin ﬁlm growth

easily incorporated into n-polar gan. sumiya et al. (2000) also reported

that the impurities related to C, O and Al were more readily incorporated

into N-polar GaN lm. Oxygen was incorporated at a rate ten times faster

into the n-polar gan than into the ga-polar gan, as reported by Park et al.

(2001). The effect of polarity on the AlN layer is very similar to that on the

gan.

12.5 Polarity controlled epitaxy of InN

InN epitaxy is the most difcult among the III-nitrides, and the effect of

polarity on it is more serious than on gan and aln. This is mainly because

the dissociation temperature of InN is very low, much lower than the re-

evaporation temperature of in metal. one of the most obvious effects of polarity

on InN epitaxy is the maximum epitaxy temperature (Xu and Yoshikawa,

2003). As shown in Fig. 12.10, the maximum epitaxy temperature is about

600°C for N-polarity, which is about 100°C higher than for In-polarity.

above these temperatures, inn epilayers start to decompose and in droplets

appear on the surface. once those in droplets appear on the surface, they

do not re-evaporate and the only way to remove them is nitrogen radical

irradiation. It means that the growth process is limited by the dissociation

of inn itself but not by the re-evaporation of excess in on the surface. This

is different from the other III-nitrides such as GaN. The growth regimes are

also different between InN and GaN. As shown in Fig. 12.11, there are two

regimes for inn, i.e. in-rich and n-rich regimes. Under in-rich regimes, there

is no way to avoid the formation of In droplets. Under N-rich regimes, the

surface is rough and the quality is not good. The two growth regimes of InN

epitaxy are different from those of GaN, where three growth regimes exist,

i.e. ga-rich, intermediate (ga-rich but free of ga droplets) and n-rich.

The growth mode and surface morphology are also greatly affected by

polarity. Step-ow growth mode is obtained in the In-polarity growth regime

N rich

Al rich

Ga rich

= 1

Original

polarity

Original

polarity

(10

–6

torr)

(10

–6

torr)

0.0 0.5 1.0 1.5 2.0 2.5

Ga ﬂux (Å/s)

0.00 0.25 0.50 0.75

Al ﬂux (Å/s)

Al-polarity

12.9 Schematic diagrams showing effect of surface stoichiometry on

polarity control processes of (a) GaN and (b) AlN in MbE growth.

(a) (b)

ThinFilm-Zexian-12.indd 300 7/1/11 9:44:17 AM



301Polarity controlled epitaxy of III-nitrides and ZnO

but not in the n-polarity regime (Wang et al., 2006). Figure 12.12 shows the

typical surface morphology of InN with In- and N-polarities, respectively.

An atomically at surface with a step and terrace feature formed in step-ow

growth mode can be observed for In-polar InN. The step height is 0.28–0.29

nm, which coincides with 1/2 of the c lattice constant of InN, showing that

the step height is a single monolayer. The surface roughness is 0.9 nm in

N polarity

In polarity

350 400 450 500 550 600 650

Substrate temperature (°C)

InN growth rate (Å)

2.5

2.0

1.5

1.0

0.5

0.0

12.10 Growth rate of InN ﬁlms with different polarities as a function

of growth temperature.

> or <e

l = 700nm

N-polarity

3 4 5 6 7 8 9 10 11 12

beam pressure of Indium (E-7 toor)

N-rich

In-rich

> of InN

Convergent point of <e

> of InN

Convergent point of <e

12.11 Growth regime diagram illustrated by <e

> and <e

> at InN

thickness of 100 nm and the convergent points. Solid and dotted

lines show guidelines for values of <e

> and <e

> at InN thickness of

100 nm and at convergent point respectively. As shown in the ﬁgure,

there are two growth regimes, In-rich and N-rich.

ThinFilm-Zexian-12.indd 301 7/1/11 9:44:17 AM



302 Thin ﬁlm growth

a 10 mm ¥ 10 mm area. This surface morphology is very similar to that of

homoepitaxial GaN grown by MBE on MOVPE-grown GaN.

on the other

hand, grain-like morphology with step-ow features within each grain is

observed in n-polar samples. The step height is usually 2 ML or multiple

MLs.

Thus, it is better to grow InN epilayers under In-polarity in order to achieve

an atomically at surface. In addition, it was observed that the following

factors are necessary: (1) the growth temperature should be as high as possible;

(2) growth should be kept under slightly In-rich conditions; and (3) keeping

the screw-component threading dislocation as low as possible. It should be

noted here that it is impossible to avoid the formation of in droplets to get

a at surface at such low temperatures as we mentioned above. However, a

regular growth interruption under N beam irradiation can be used to eliminate

the in droplets, though one should carefully control the irradiation time since

longer irradiation is found to result in a rough surface.

The control of polarity for inn is basically achieved by using gan buffer

layers with controlled polarity. The reason is that the lattice matched substrate

for inn is not available and sapphire is the most popular substrate in inn

epitaxy. Since the lattice mismatch between InN and sapphire is very big

(~25%), gan is usually used as a buffer layer or template for inn epitaxy,

which shows better crystalline quality than the InN layer grown on sapphire

by only using low temperature InN buffer layer (Chen et al., 2006). Then, the

In- and N-polarity InN layers are grown on Ga- and N-polarity GaN buffer

layer/template. For the Ga-polar layer/template, either MOVPE-grown GaN

epilayer or MBE-grown GaN/AlN double buffer layers can work well (Heying

et al., 1999; Ide et al., 2003). As for the N-polar GaN layer, it is grown on

GaN buffer layer using nitridation process by MBE as we discussed in the

last section.

In-polarity N-polarity

1 µm 1 µm

0.00 [nm] 4.03 0.00 [nm] 19.34

12.12 AfM images of surface morphology of InN at In- and

N-polarities, respectively.

ThinFilm-Zexian-12.indd 302 7/1/11 9:44:17 AM



303Polarity controlled epitaxy of III-nitrides and ZnO

A summary of MBE-grown high quality and high purity InN epilayers

under different polarities is shown in Table 12.2. The In-polar samples are

grown on 3.7 mm thick MOVPE-grown GaN templates and the N-polar

samples are grown by the whole MBE process on the 500 nm thick N-polar

GaN epilayers. The N-polar samples show better crystalline structural quality,

which is due to the higher maximum epitaxial temperature. The N-polarity

growth is preferable from the viewpoint of better matching with epitaxy

temperatures among other III-nitrides. From the viewpoints of surface atness

and sharp interface, however, In-polarity is denitely a better choice. As

shown in Table 12.2, the lowest n

is at 10

–3

levels for both polarities,

where the In-polar samples show a little lower residual electron density

and higher mobility. although the residual donor levels are still high, the

data imply that the extrinsic donor impurities in the inn layers are pretty

low considering the contribution from high density edge-type threading

dislocations.

it is important to investigate the origin of fairly high residual electron

concentration in InN. Since the Fermi stabilization energy is located in

the conduction band for inn, unintentionally doped impurities and native

defects tend to be ‘shallow’ donors. It is believed that the nitrogen vacancy

would be a major native defect due to the lowest formation energy among

native defects. in addition, oxygen and hydrogen are also expected to

be a signicant source of free electrons in InN. Thus, the growth should

be performed under growth chamber conditions as pure as possible and

under controlled surface stoichiometry as well. Furthermore, the edge-type

threading dislocations (ETDs) should be reduced to as low as possible. It

has been found that this type of threading dislocation is the dominant one

in the InN layer which basically originates from large lattice mist with the

substrate. Its density is about 1–2 orders higher than that of the screw-type

one. It is known that the edge-type threading dislocations in GaN introduce

deep levels in the forbidden band. However, it was theoretically conrmed

that they generate donor levels above the conduction band bottom in inn,

Table 12.2 Comparison of MbE grown In-polarity and N-polarity InN epilayers

In-polarity InN N-polarity InN

Growth temperature Less than ~500°C Less than ~600°C

Surface ﬂatness/

morphology

Atomically ﬂat, monolayer

steps

Columnar/grain-like rough,

bilayer steps

XRd-fWhM (002) 200–540 arcsec 200–250 arcsec

XRd-fWhM (102) 800–1100 arcsec 1000–650 arcsec

Residual electron density

at RT

2–30 ¥ 10

–3

4–50 ¥ 10

–3

Electron mobility at RT 1500–2200 cm

/Vs 1500–2150 cm

/Vs

ThinFilm-Zexian-12.indd 303 7/1/11 9:44:17 AM



304 Thin ﬁlm growth

though their exact levels are slightly different depending on their detailed

core structures (Takei and Nakayama, 2009). Then it is considered that

the edge-type threading dislocation is one of residual donor species in inn

when the residual impurities and point defects are reduced below 10

–3

levels. This is also conrmed by experimental observation as shown in Fig.

12.13, where the electron concentrations (n

) and mobilities (m

) of dozens

of In-polar InN lms with thicknesses of 0.5–5 mm are shown as a function

of full width at half maximum (FWHM) measurements of (102) InN w

scans. These n

’s are net for the bulk, i.e, the surface/interface charges of

3–5 ¥ 10

–2

has been removed.

As shown in Fig. 12.13, n

increases from 4.0 ¥ 10

to 2.4 ¥ 10

–3

with

increasing FWhM

(102)

from 950 to 2200 arcsec, indicating the increase of

with increasing ETD density since the FWHM

(102)

is correlated with the

ETDs. The dashed line indicates the contribution of dangling bonds to the

in the inn epilayers, assuming that each dangling bond at the ETDs acts

as a singly ionizable donor, i.e., two electrons are supplied every monolayer

along each ETD. It is shown that the dependence of the dashed line agrees

well with that of the n

on FWhM

(102)

and even the values are almost

exactly the same as the experimental data too. This conrms the theoretical

prediction that ETDs act as donors and are probably a dominant source of

in high purity InN epilayers. The discrepancy between the measured n

and the density of dangling bonds is probably caused by the contributions

from impurity- and point-defects-generated carrier concentrations not being

taken into account. in addition, m

decreases from 2150 to 1000 cm

/Vs with

increasing FWhM

(102)

from 950 to 2200 arcsec, showing that the m

is also

inuenced by the ETDs.

Electron concentration (cm

–3

)

hall mobility (cm

/Vs)

2.5 ¥ 1 0

2.0 ¥ 1 0

1.5 ¥ 1 0

1.0 ¥ 1 0

5.0 ¥ 1 0

0.0

1000 1200 1400 1600 1800 2000 2200

fWhM of (102) w scans (arcsec)

2200

2000

1800

1600

1400

1200

1000

800

12.13 Electron concentrations and hall mobilities of many InN ﬁlms

as a function of fWhMs of (102) w scans. The estimated density of

dangling bonds originating from edge-type threading dislocations is

shown by the dashed line.

ThinFilm-Zexian-12.indd 304 7/1/11 9:44:18 AM



305Polarity controlled epitaxy of III-nitrides and ZnO

P-type doping is one of the most important topics. although the residual

donor levels are still fairly high, it is considered that extrinsic donor impurity

levels are almost low enough to try p-type doping. Mg is a popular and

effective dopant in p-type doping of gan and aln, and it is also the most

studied one in InN. It was found that the effects of Mg doping on properties

of InN with different polarities are very similar (Wang et al., 2007a). The

sticking coefcient of Mg on In-polarity InN is almost unity, which is

probably attributed to the growth temperature being low and thus Mg is

effectively adsorbed and incorporated into inn. The Mg concentrations for

the N-polarity InN at the same Mg beam ux are almost the same as those

for the In-polarity case, indicating that the Mg sticking coefcient in InN is

independent of polarity.

An important factor in evaluating the doping difculty is the activation

energy of Mg acceptors. To evaluate it, photoluminescence (PL) properties

of Mg-doped InN are investigated. Figure 12.14 shows PL spectra and

corresponding PL intensity measured at 16 K as a function of [Mg]. It is

shown that PL intensity changes in more than ve orders of magnitude. At

[Mg] ~ 10

–3

, the intensity is already lower by about four orders in

magnitude than that of the undoped one, and no PL emission is detected at

[Mg] ~5.6–29 ¥ 10

–3

, where it is proven to be p-type region as shown

PL intensity (a.u.)

PL intensity

No detectable PL

Mg concentration (cm

–3

)

0.55 0.60 0.65 0.70 0.75 0.80 0.85

E (eV)

PL (a.u.)

[Mg]: cm

–3

3.9 ¥ 1 0

8.0 ¥ 1 0

1.8 ¥ 1 0

2.9 ¥ 1 0

5.6 ¥ 1 0

1.1 ¥ 1 0

1.3 ¥ 1 0

1.2 ¥ 1 0

Undoped

12.14 PL intensity and PL spectra (inset) of InN layers at 16 K as a

function of Mg concentrations.

ThinFilm-Zexian-12.indd 305 7/1/11 9:44:18 AM



306 Thin ﬁlm growth

later. Weak PL emission is observed again with increasing [Mg]. Further,

no PL emission can be detected at [Mg] ~3.9 ¥ 10

–3

again, which is

probably because of unknown luminescence-killer processes in these poor

quality samples. The exact reason for weak emission in InN:Mg at [Mg]

~5.6–29 ¥ 10

–3

as well as p-type InN reported later are not currently

clear, but one possible reason is the long diffusion length of minority carriers

(electrons) in p-InN resulting in their easy annihilation at killer defects and/

or the introduction of high-density complex defect levels by Mg-overdoping

resulting in about ve orders of magnitude higher carrier relaxation paths.

In the InN:Mg sample at [Mg] ~1.3 ¥ 10

–3

, two emission peaks are

clearly observed at around 0.67 (labelled as I

in Fig. 12.15) and 0.61

) eV, respectively. Excitation power dependent PL study of this sample

is shown in Fig. 12.15. It is clear that intensity of I

is linearly increased

with increasing excitation power while that of I

tends to be saturated. This

indicates that I

originates from the band-to-band transitions while I

comes

from the free-to-acceptor transitions, where the activation energy for Mg

acceptors is estimated to be about 61 meV.

The observation of such shallow Mg acceptor levels implies that the

PL intensity (a.u.)

Photon energy (eV)

0.55 0.60 0.65 0.70 0.75 0.80

Photon energy (eV)

(a)

0 10 20 30 40

Excitation power (mW)

(b)

40 mW

30 mW

20 mW

10 mW

5 mW

1 mW

0.5 mw

0.1 mW

0.680

0.675

0.670

0.620

0.615

0.610

12.15 (a) 16 K PL spectra of partly compensated InN:Mg under

different excitation powers. (b) Intensity and energy of I

(free to

acceptor recombination) and I

(band to band) as a function of

excitation power.

ThinFilm-Zexian-12.indd 306 7/1/11 9:44:18 AM



307Polarity controlled epitaxy of III-nitrides and ZnO

p-type doping may not be as difcult as we imagined previously. It seems

that the problem of how to detect the p-type conduction is more serious

at this moment. The p-type conduction has not been directly conrmed

by conventional hall effect measurements. This is due to the existence of

surface electrons as well as the interface accumulation layer with an electron

density of 3–5 ¥ 10

–2

, which is independent of bulk layer conductivity

(Mahboob et al., 2004). Therefore, the most commonly used method to

detect p-type conduction is currently electrolyte-based capacitance voltage

(ECV) and thermal power (Jones et al., 2006; Yim et al., 2007; Wang et

al., 2007a).

ECV measurements are usually performed by using an electrolyte, for

example KOH or NaOH solutions, to form a Helmholtz-layer contact on InN.

This Helmholtz layer behaves as a very thin insulator and the total structure

is like a ‘metal-insulator-semiconductor (MIS) structure’ in principle. To

successfully modulate the surface Fermi level, the electric eld intensity in the

Helmholtz layer must be strong enough to deplete the surface electrons. The

principle of how to detect the buried p-type region under high density surface

electron is shown in Fig. 12.16. As shown in Fig. 12.16(a), the surface Fermi

level in inn is located inside the conduction band due to high-density surface

electrons. since the inn is highly conductive in this case, the situation for

bias

= 0

n-type InN

p-type InN p-type InN

n-type InN

(a)

(b)

bias

mini

(=V

peak

)

12.16 Energy band diagram for ECV characterization on detecting

the conduction type of bulk InN region located below high-density

surface electrons. The vertical thick lines indicate the helmholtz layer

in the electrolyte. (a) and (c) are for the zero bias voltage for n-type

and p-type samples, respectively. Correspondingly, (b) and (d) are for

those under voltage biased for detecting the minimum capacitance.

ThinFilm-Zexian-12.indd 307 7/1/11 9:44:19 AM

