GaP collector development for SiGe heterojunction bipolar transistorperformance increase: A heterostructure growth studyO. Skibitzki, F. Hatami, Y. Yamamoto, P. Zaumseil, A. Trampert et al. Citation: J. Appl. Phys. 111, 073515 (2012); doi: 10.1063/1.3701583 View online: http://dx.doi.org/10.1063/1.3701583 View Table of Contents: http://jap.aip.org/resource/1/JAPIAU/v111/i7 Published by the American Institute of Physics. Related ArticlesChromium-oxide enhancement of photo-oxidation of CdSe/ZnS quantum dot solids J. Appl. Phys. 111, 084308 (2012) Three-step H2Se/Ar/H2S reaction of Cu-In-Ga precursors for controlled composition and adhesion ofCu(In,Ga)(Se,S)2 thin films J. Appl. Phys. 111, 083710 (2012) NO-assisted molecular-beam epitaxial growth of nitrogen substituted EuO Appl. Phys. Lett. 100, 162405 (2012) Ge atom distribution in buried dome islands Appl. Phys. Lett. 100, 164105 (2012) Fabrication and characterization of controllable grain boundary arrays in solution-processed small moleculeorganic semiconductor films J. Appl. Phys. 111, 073716 (2012) Additional information on J. Appl. Phys.Journal Homepage: http://jap.aip.org/ Journal Information: http://jap.aip.org/about/about_the_journal Top downloads: http://jap.aip.org/features/most_downloaded Information for Authors: http://jap.aip.org/authors

Downloaded 20 Apr 2012 to 62.141.165.1. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions

GaP collector development for SiGe heterojunction bipolar transistorperformance increase: A heterostructure growth study

O. Skibitzki,1,a) F. Hatami,2 Y. Yamamoto,1 P. Zaumseil,1 A. Trampert,3 M. A. Schubert,1

B. Tillack,1,4 W. T. Masselink,2 and T. Schroeder1,5

1IHP, Im Technologiepark 25, 15236 Frankfurt (Oder), Germany2Humboldt Universitat zu Berlin, MNF1, Newtonstrasse 15, 12489 Berlin, Germany3Paul Drude Institut fur Festkorperelektronik, Hausvogteiplatz 5-7, 10117 Berlin, Germany4Technische Universitat Berlin, HFT4, Einsteinufer 25, 10587 Berlin, Germany5Brandenburgische Technische Universitat, Konrad-Wachsmann-Allee 1, 03046 Cottbus, Germany

(Received 7 November 2011; accepted 20 February 2012; published online 9 April 2012)

To develop a III/V wide bandgap collector concept for future SiGe heterobipolar transistor

performance increase, a heterostructure growth study of GaP on pseudomorphic 4� off-oriented

Si0.8Ge0.2/Si(001) substrates was performed. For pseudomorphic GaP/Si0.8Ge0.2/Si(001)

heterostructure growth, critical thickness of GaP on Si and maximum thermal budget for GaP

deposition were evaluated. A detailed structure and defect characterization study by x-ray

diffraction, atomic force microscopy, and transmission electron microscopy is reported on single

crystalline 170 nm GaP/20 nm Si0.8Ge0.2/Si(001). Results show that 20 nm Si0.8Ge0.2/Si(001) can

be overgrown by 170 nm GaP without affecting the pseudomorphism of the Si0.8Ge0.2/Si(001)

layer. The GaP layer grows however partially relaxed, mainly due to defect nucleation at the GaP/

Si0.8Ge0.2 interface during initial island coalescence. The achievement of 2D GaP growth

conditions on Si0.8Ge0.2/Si(001) systems is thus a crucial step for achieving fully pseudomorphic

heterostructures. Anti-phase domain-free GaP growth is observed for film thicknesses beyond

70 nm. VC 2012 American Institute of Physics. [http://dx.doi.org/10.1063/1.3701583]

I. INTRODUCTION

State of the art silicon-germanium (SiGe) heterojunction

bipolar transistor (HBT) technologies are combined with

standard silicon (Si) complementary metal oxide semicon-

ductor devices (so called BiCMOS technologies) to realize

high levels of integration and functionality on single digital

systems on chip for wireless and broadband communica-

tion.1,2 High-speed SiGe:C HBT BiCMOS technology can

be fabricated up to cut-off frequency (fT)/maximum fre-

quency of oscillation/common-emitter breakdown voltage

(BVCEO) values of 300 GHz/500 GHz/1.6 eV.3 Future devel-

opments could reach even higher frequency values, entering

further into the terahertz regime. To reach maximum device

performance, HBTs run in reverse bias at the collector-base

(CB) junction with large current densities. In this condition,

the CB electric field increases and the carrier transit time

over the junction will be reduced. But in contrast to this posi-

tive effect, avalanche breakdown will be increased due to

increased impact ionization in the collector. So by increasing

fT, we observe a drop of BVCEO as described in the Johnson

limit.4 This trade-off behavior makes it difficult to combine

high-speed with high-power performance for SiGe:C HBTs.

To counteract this effect, one possible approach is to

change the Si-based collector material of the HBT device to-

ward new material compounds with suitable properties to

improve fT and BVCEO simultaneously. Here, indium gal-

lium phosphide (InGaP) is identified as potential collector

material to replace Si in future SiGe HBTs, creating so called

III-V/Si hybrid devices.5,6 Table I shows the main physical

parameters of the binary materials InP and GaP (at 300 K),

important for SiGe HBT speed and power performance

increase, in comparison to Si.7,8 On the one hand, InP has a

three times higher saturation velocity than Si, offering the

potential to increase the HBT speed performance by reduc-

ing the CB junction transit time. On the other hand, the wide

bandgap semiconductor GaP with a two times bigger

bandgap than Si decreases impact ionization rates in the col-

lector, offering that way promising parameters to increase

HBT power performance. In other words, ternary In1�xGaxP

(x¼ 0�1) systems as potential collector material provide the

vision to adjust speed and power of HBTs in a flexible way

as a function of chemical composition x.

InGaP growth on thin (�20 nm) pseudomorphic SiGe/

Si(001) for potential HBT application was not reported in the

literature so far. Interestingly, Fitzgerald et al. reported the

growth of InGaP heterostructures on Si substrates using a

relaxed thick (>1 lm) graded SiGe buffer to setup virtual

InGaP substrates for LED applications.9 In this paper, we

present a growth and material science characterization study

of GaP/SiGe/Si(001) heterostructures as a starting point for

evaluating InGaP as a potential SiGe HBT collector. GaP is

chosen as starting point due to its small lattice mismatch with

respect to Si (0.36% at 300 K) (Table I). As known from GaP

on Si growth studies for photonics and microelectronics,6,10–24

heteroepitaxy challenges for achieving high quality GaP/

SiGe/Si(001) heterostacks are given by crystallographic and

thermal lattice mismatch, polar/non-polar interfaces, 3D

a)Author to whom correspondence should be addressed. Electronic mail:

skibitzki@ihp-microelectronics.com. Tel.: þ49 335 5625 766. Fax:

þ49 335 5625 681.

0021-8979/2012/111(7)/073515/9/$30.00 VC 2012 American Institute of Physics111, 073515-1

JOURNAL OF APPLIED PHYSICS 111, 073515 (2012)

Downloaded 20 Apr 2012 to 62.141.165.1. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions

versus 2D growth modes, interdiffusion as well as defect for-

mation (stacking faults (SFs), microtwins (MTs), anti-phase

domains (APDs), etc.).

II. EXPERIMENTAL DETAILS

A. SiGe deposition

Si(001) wafers (200 mm size) with 4� off- (toward

[110]) orientation were used as substrates. Before epitaxial

SiGe growth, Si(001) substrates were wet chemically

cleaned by a Radio Corporation of America (RCA) solution.

After RCA cleaning, the Si(001) surface was covered by a

defined oxide layer, which was removed subsequently by a

prebake in the ASM Epsilon 2000 lamp-heated single wafer

reduced pressure chemical vapor deposition (RPCVD) cham-

ber at 1000 �C for 10 min in H2. This rather long annealing

time was necessary to additionally ensure the creation of pre-

ferred double atomic steps on the surface.11,17,25 Afterwards,

20 nm pseudomorphic Si0.8Ge0.2 was grown by RPCVD on

top of the Si(001) wafers at 600 �C growth temperature and

80 Torr chamber pressure, applying H2 as carrier gas for the

reactant gas sources SiH4 and GeH4.

B. GaP deposition

Before subsequent GaP deposition, the 4� off-oriented

Si0.8Ge0.2/Si(001) substrates (cut in 1� 1 cm2 pieces) were

cleaned again in a RCA solution combined with HF last

clean and then transferred into a Riber Compact 21 GSMBE

UHV system for 15 min bake-out at 800 �C. Using 4�

off-oriented substrates associated with a bake-out is a well-

known method to reduce APD formation during III/V-

deposition.11,14–20 After pretreatment, 170 nm GaP was

deposited on top of Si0.8Ge0.2/Si(001) substrates, using phos-

phine gas (PH3) thermally cracked at 920 �C and elemental

Ga as source materials. The temperature and dose profile of

the GaP growth procedure is illustrated in Fig. 1: First, a

two-monolayer (ML) Ga prelayer was created by opening

the shutter for 6 s at Ga temperature (TGa) of 830 �C and sub-

strate temperature (Tsub) of 400 �C. Afterwards, continuous

gas flow of 4 sccm PH3 was applied. Further GaP growth

was initiated by two successive growth steps: In the first

low-flux growth step, the adjusted TGa and Tsub parameters

were kept constant and a growth circle with a 3 s open Ga

shutter and a subsequent 1 s break for all 60 cycles was used

to grow a closed GaP seed layer. In the second high-flux

growth step for faster GaP growth, the temperature for Ga

crucible and substrate were increased to 890 �C and 450 �Ccorrespondingly, using this time a growth cycle with a rota-

tion of 1 s open Ga shutter and a subsequent 1 s break for all

600 remaining cycles.

C. Characterization

A SmartLab diffractometer from Rigaku equipped with

a 9 kW rotating anode Cu source (Cua¼ 0.1541 nm) was

used to apply X-ray diffraction (XRD) measurements on the

GaP/Si0.8Ge0.2/Si(001) heterostructure. High resolution XRD

curves were recorded using a Ge(400)� 2 collimator crystal

behind a x-ray mirror and a Ge(220)� 2 crystal analyzer.

Both specular h/2h and in-plane XRD measurements were

performed with a 0.114� soller slit on detector side. Pole fig-

ures were measured without crystal collimator and 0.5� sol-

ler slits on source and detector side. All measured samples

were adjusted in such a way that the normal of the (004) net

plane is parallel to the u-axis of the diffractometer. This

allows us to measure (004) diffraction curves under symmet-

rical as well as asymmetrical Bragg condition depending on

the chosen u-direction. To perform in situ temperature de-

pendent XRD studies under 1 bar N2 atmosphere, a DHS

1100 furnace (Anton Paar) was mounted on the SmartLab

diffractometer. Veeco Digital Instruments Dimension 5000

atomic force microscope (AFM) and a Philips CM 200 trans-

mission electron microscope (TEM) operating at 200 kV

were used to gain additional information about surface to-

pography, crystal quality, and defect formation. Dark field

high resolution TEM (HRTEM) imaging was performed in a

JEOL 3010 UHV TEM microscope operating at 300 kV. All

TEM lamellas were prepared by mechanical thinning fol-

lowed by an argon ion milling process.

III. RESULTS AND DISCUSSION

A. Theory

We start our evaluation study of GaP/Si0.8Ge0.2/Si(001)

heterostructures for potential HBT applications by theoretical

TABLE I. Important physical parameters of Si, InP, and GaP (at 300 K) for

SiGe HBT high frequency and power performance increase as well as heter-

ostructure growth.a

Si InP GaP

Band gap [eV] 1.12 1.33 2.26

Breakdown field [V cm�1] 3� 105 5� 105 1� 106

Electron mobility [cm2 V�1 s�1] 1400 5400 250

Saturation velocity [cm/s] 1� 107 3� 107 1.3� 107

Lattice constant [nm] 0.5431 0.5869 0.5451

aReference 7.

FIG. 1. Deposition process for 170 nm GaP growth on pseudomorphic 4�

off-oriented Si0.8Ge0.2/Si(001), displaying substrate temperature (Tsub), Ga

crucible temperature (TGa), Ga pulse program, and PH3 gas flow.

073515-2 Skibitzki et al. J. Appl. Phys. 111, 073515 (2012)

Downloaded 20 Apr 2012 to 62.141.165.1. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions

consideration of critical thickness for pseudomorphic GaP

growth on pseudomorphic Si0.8Ge0.2/Si(001) substrates.

1. Critical thickness of GaP on Si

Figure 2 shows the critical thickness (hcrit) values of

GaP on Si for different lattice mismatch calculated from the

equilibrium theory for strain relaxation in metastable hetero-

epitaxial semiconductor structures, following the approach

by Fischer et al.26 Accordingly, hcrit is given by

f ¼ b� cosk2hcrit

� 1þ 1� ð�=4Þ4p� cos2k� cos /� ð1þ �Þ

� ��

� lnhcrit

b

� ��;

where f is the lattice misfit, b is the magnitude of the Burgers

vector of the dislocations (12

ah110i in fcc lattices, inclined at

45� to (001), Ref. 27), � is the Poisson ratio, k the angle

between the Burgers vector and the h110i dislocation lines,

and / is the angle between the h111i slip plane normals and

the h110i azimuthal axis. Under the assumption that pseudo-

morphic Si0.8Ge0.2 on Si(001) counts as a layer with the in-

plane lattice constant of Si, and inserting appropriate material

parameters in this equation (f¼ 0.0036, b¼ 3.840 A, �¼ 0.31,

cos k¼ 0.5, and cos /¼ 0.816 for growth on Si(001) surfa-

ces), the value of hcrit for GaP is 64 nm (Fig. 2). In comparison

with published experimental data for GaP/Si(001) systems,

hcrit is reported to vary between 45 and 95 nm,28,29 consistent

with the magnitude of our theoretical result. Although the cal-

culated value here is a thermodynamical value and experimen-

tal parameters are usually higher due to kinetic hindrance for

defect injection in pseudomorphic layers, hcrit value of GaP

might be too low, as typical HBT designs require a collector

thickness in the range of several hundred nanometres.1–3 In

this respect, it is interesting to point out (Fig. 2) that nitrogen

incorporation in GaP can be used to reduce the misfit and thus

to substantially increase hcrit.30,31 Figure 2 shows the decrease

of lattice mismatch as a function of N-content in GaP1�xNy

systems. For example, hcrit moves to �300 nm for �1.57% N

incorporation where the lattice mismatch of GaP1�xNy with Si

decreases to 0.09%. Next, after these theoretical considera-

tions, we present the experimental results of the GaP/

Si0.8Ge0.2/Si(001) heterostructure study.

B. Experimental results

1. Investigation of thermal budget

An XRD thermal budget study was performed to ensure

that GaP growth does not induce relaxation processes in the

pseudomorphic 20 nm Si0.8Ge0.2/Si(001) substrate (Fig.

3(a)). For this reason, pseudomorphic 20 nm Si0.8Ge0.2/

Si(001) samples were annealed in an ex situ furnace for

30 min in N2 atmosphere at temperatures ranging from 500

to 1000 �C. Afterwards, specular h/2h XRD measurements in

/-position with symmetrical Bragg case were performed to

investigate changes in the Si0.8Ge0.2(004) diffraction curves

of the annealed samples with respect to as-grown sample to

detect possible relaxation processes.

The samples, annealed between 500 and 900 �C, indicate

no change in the (004) Bragg peak position of the pseudo-

morphic Si0.8Ge0.2. However, at 900 �C elevated diffraction

minima at the Si0.8Ge0.2 Bragg peak could be observed.

Reaching 1000 �C, a clear shift of the Si0.8Ge0.2(004) peak

toward a larger angle closer to the Si(004) peak position is

visible. Three insights can be gained from this analysis.

Firstly, applying a thermal budget of up to 800 �C does not

show any effect of relaxation on the samples and the 20 nm

Si0.8Ge0.2/Si(001) system remains pseudomorphic. Secondly,

at 900 �C the elevated diffraction minima of the Si0.8Ge0.2

Bragg peak indicate the onset of relaxation processes due to

misfit dislocation generation between Si and SiGe. Thirdly,

at 1000 �C the degree of relaxation of the Si0.8Ge0.2 layer

increases strongly and reaches a value of about 60%. This

observed behavior is a typical example for relaxation proc-

esses in pseudomorphic SiGe/Si layers during annealing.32

Consequently, it can be concluded that GaP growth tempera-

ture on 20 nm Si0.8Ge0.2/Si(001) must not exceed 800 �C to

maintain pseudomorphism of the Si0.8Ge0.2 layer on Si(001).

2. Epitaxy relationship and relaxation characterization

Next, we report the experimental results of 170 nm GaP

on 20 nm pseudomorphic Si0.8Ge0.2/Si(001) heterostructures.

To determine the pseudomorphic character of the GaP/

Si0.8Ge0.2/Si(001) heterostructure, we performed specular

h/2h XRD measurements in the /-position with symmetrical

Bragg case near the Si(004) Bragg peak position (Fig. 3(b)).

In comparison with Fig. 3(a)), a broad GaP(004) reflection

appears, which is neither situated at its bulk position nor at

the value estimated with the help of the Poisson ratio for

pseudomorphic GaP.7 This result indicates that the 170 nm

deposited GaP layer is crystalline and (001) oriented, but

contains structural defects and grows partially relaxed on the

Si0.8Ge0.2/Si(001) substrate. Interestingly, the Si0.8Ge0.2(004)

Bragg peak position shows a slight shift to larger angles after

GaP deposition. Considering the position and width of the

GaP(004) peak in close vicinity to the Si0.8Ge0.2(004) reflec-

tion, this slight shift can be explained by mutual interference

of the thickness fringes of both layers. To confirm that 20 nmFIG. 2. Critical thickness hc vs lattice misfit as a function of N-content in

GaP1�xNx (x¼ 0 – 0.02) systems.

073515-3 Skibitzki et al. J. Appl. Phys. 111, 073515 (2012)

Downloaded 20 Apr 2012 to 62.141.165.1. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions

Si0.8Ge0.2/Si(001) is still pseudomorphic despite the misfit

strain exerted by 170 nm GaP, reciprocal space mapping

(RSM) of the asymmetric (�2�24) reflections of Si, GaP, and

Si0.8Ge0.2 was performed (Fig. 3(c)). Figure 3(c) shows a

sharp Si(�2�24) signal from the high quality Si(001) substrate.

The Si0.8Ge0.2(�2�24) reflection shows otherwise a lower sig-

nal intensity and exhibits an ellipsoidal shape. The small full

width at half maximum (FWHM) in the Qx direction is com-

parable to Si(�2�24), indicating a high crystal quality of the

SiGe layer. The bigger FWHM in Qz direction is due to the

finite thickness of 20 nm. The Qx positions of the

Si0.8Ge0.2(�2�24) and the Si(�2�24) reflection are identical, dem-

onstrating that both layers have the same in-plane lattice

constant. A full relaxation of Si0.8Ge0.2 would otherwise be

expressed by a Qx peak shift to a position between the

Si(�2�24) peak position and the (0,0) origin direction of recip-

rocal space (indicated by arrow in Fig. 3(c)). As a result, no

relaxation processes has taken place so that the 20 nm

Si0.8Ge0.2 layer remains pseudomorphic on Si(001) after

170 nm GaP deposition. This result is positive for a potential

HBT collector application of GaP, which requires a stable

pseudomorphic base layer enduring a deposition of several

hundred nanometers of collector material on top.1–3 In con-

trast, the deposited GaP layer is characterized by a broad

GaP(�2�24) reflection with diffuse scattering, suggesting the

presence of structural defects.33 In addition, the GaP(�2�24)

Qx position is partly shifted (toward arrow in Fig. 3(c)), con-

firming a partial relaxation of the deposited GaP layer. To

determine the relaxation degree in detail, in-plane measure-

ments in /-position with symmetrical Bragg case around the

Si(220) peak position were also performed (Fig. 3(d)). After

all, from the GaP(004) out-of-plane and GaP(220) in-plane

peak positions, strain relaxation degree of about 40% was

determined for the 170 nm thick GaP film. It is noted, that

the experimentally derived Poisson ratio of 0.33, using the

in- and out-of-plane lattice constants, fits well to litera-

ture.7,34 In conclusion, the epitaxial relationship of the single

crystalline heterostructure is given by GaP[001];h110ikSi0.8Ge0.2[001];h110ikSi[001];h110i. While the pseudomor-

phism of the 20 nm Si0.8Ge0.2/Si(001) system is maintained

during heteroepitaxial overgrowth of 170 nm GaP, the GaP

structure itself grows partially relaxed.

3. Thermal expansion coefficient study

A possible origin of the partial GaP relaxation is the lat-

tice mismatch between GaP and the pseudomorphic

Si0.8Ge0.2/Si(001), which additionally increases at higher

FIG. 3. (a) Specular h/2h XRD scans of as-grown Si0.8Ge0.2/Si(001) samples and after annealing at 500 – 1000 �C in N2 atmosphere. Process pressure and

annealing time were 1 atm and 30 min, respectively. (b) Specular h/2h XRD scan after 170 nm GaP deposition on top. (c) RSM of asymmetric (�2�24) reflections

of Si, GaP, and Si0.8Ge0.2 measured on the same sample. Qz-axis is parallel to (004) net plane normal and Qx-axis is perpendicular to Qz in the diffraction

plane. (d) In-plane (220) XRD scan of the same sample.

073515-4 Skibitzki et al. J. Appl. Phys. 111, 073515 (2012)

Downloaded 20 Apr 2012 to 62.141.165.1. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions

temperature due to different coefficients of thermal expan-

sion (CTE). Therefore, we studied the impact of different

CTE on the relaxation behavior of a 170 nm GaP/20 nm

Si0.8Ge0.2/Si(001) heterostructure. For this purpose, the sam-

ple was placed in a furnace on the SmartLab diffractometer

under 1 bar N2 atmosphere. The sample was heated in 50 �Csteps up to 550 �C and in situ h/2h XRD measurements were

performed after 15 min annealing time at every step. Higher

postannealing temperatures are not suitable, because uncov-

ered GaP layers start to decompose.35,36 After cooling down

back to 50 �C, a final h/2h XRD measurement was carried

out. Figure 4(a) shows only the 50 �C, 250 �C, 550 �C and

the back to 50 �C results of this experiment for the sake of

clarity. Due to increasing annealing temperatures, the (004)

peak positions of Si, Si0.8Ge0.2, and GaP will shift according

to their CTE values to smaller angles. However, for a better

comparison and depiction of GaP(004) and Si0.8Ge0.2(004)

peak position changes relative to the Si(004) peak, the col-

lected XRD curves were aligned to the Si(004) peak position

at 50 �C in Fig. 4(a). Two main insights were drawn from

these data. First, the graphs before and after the applied

annealing procedure perfectly superimpose. Suppose that the

relaxation of GaP occurs at the growth temperature of

450 �C by generation of misfit dislocations, the 100 �C higher

temperature in this experiment and following higher misfit

should lead to a continuation of the relaxation process. How-

ever, since this is not the case here, a different relaxation

mechanism is needed to explain the partial relaxation of

GaP. Secondly, the GaP(004) peak position changes with

increasing temperature over a higher angular range than the

Si0.8Ge0.2(004) Bragg peak, pointing to a higher GaP CTE

value. To derive the CTE values, we plot in Fig. 5(b) the

out-of-plane lattice constant of Si, Si0.8Ge0.2, and GaP from

the out-of-plane high temperature in situ XRD study. As all

out-of-plane lattice constant values of all three layer materi-

als increase linearly with raising annealing temperature, the

linear CTE values were extracted by using the equation37

CTESi ¼DaSi

aSi� 1

DT;

where a is the lattice constant at 300 K and Da the change

with increasing temperature. A CTE value of

(3.8 6 0.7)� 10�6 K�1 was found for Si, which agrees

(within the error range) with literature for free standing bulk

Si (3.6� 10�6 K�1).38 To derive the CTE values of the thin

GaP and Si0.8Ge0.2 layers, we must correct for the influence

of the bulk Si substrate by the equation37

CTEm¼1

DT

� a1;m;T1þða0;m;T0þaSi;T0

�CTESi�DT�Km

am;T0�ð1þKmÞ

�1

� �

Km¼2�m

ð1��mÞ;

where a1,m,T1 is the out-of-plane lattice constant at T1, a0,m,T0

is the in-plane lattice constant at T0, and am,T0 is the value of

the bulk lattice constant at T0 of the top layers (m either

Si0.8Ge0.2 or GaP). aSi,T0 and CTESi,T0 are the lattice constant

and the linear CTE of the Si substrate at T0, respectively. The

Poisson ratio �m of the top layer material used in this calcula-

tion was taken from the literature (0.28 for Si0.8Ge0.2 and 0.33

for GaP).7 Using the in-plane lattice constant of Si(001) for

pseudomorphic Si0.8Ge0.2 on top and the in-plane lattice con-

stant for GaP measured at 300 K (Fig. 3(d)), it is possible to

derive the linear CTE related to free-standing bulk material

averaged over the range from RT to 550 �C as

CTESi0.8Ge0.2¼ (4.1 6 0.7)� 10�6 K�1 and CTEGaP¼ (5.9

6 0.7)� 10�6 K�1. In conclusion, although GaP exhibits a

higher CTE than Si0.8Ge0.2 and Si, no additional plastic relax-

ation occurred by the thermal budget applied during the high

temperature treatment (up to 550 �C; 15 min). Therefore the

observed partial relaxation of 170 nm GaP cannot be

explained by lattice mismatch effects only and has to follow a

different relaxation process mechanism.

4. XRD defect study

Relaxation processes in semiconductor films are caused

by different kinds of defect formation. Plastic relaxations

FIG. 4. (a) Specular h/2h XRD scan of 170 nm GaP/20 nm Si0.8Ge0.2/

Si(001) postannealed at 50 �C, 250 �C, 550 �C and cooled down back to

50 �C. Postannealing applied under N2 atmosphere at a pressure of 1 atm.

For better comparison and depiction, all XRD graphs were aligned to the

Si(004) reflection. (b) (001) lattice constants of GaP, Si0.8Ge0.2, and Si layers

vs temperature.

073515-5 Skibitzki et al. J. Appl. Phys. 111, 073515 (2012)

Downloaded 20 Apr 2012 to 62.141.165.1. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions

occur due to growth of closed films beyond hcrit, creating

misfit dislocations on the heterostructure interface. A differ-

ent possible relaxation mechanism in semiconductor films

are coalescence processes of initial 3D islands during

growth, which form growth defects like SFs and MTs.11,42

To learn more about the influence of these growth defects on

the partial relaxation of the GaP structure, XRD was applied

for defect characterization. To verify the presence of MTs in

the GaP(001) layer,39 a XRD GaP (1 1 1) pole figure (PF)

study was carried out on the same 170 nm GaP/20 nm

Si0.8Ge0.2/Si(001) sample, which was annealed in the ther-

mal expansion coefficient study. Figure 5(a) shows the result

of the PF measurement. Due to the fourfold symmetry, four

symmetric GaP(111) Bragg peaks at v � 55� were measured

(indicated by the red circles), corresponding to the sketched

angle between the GaP(001) surface orientation and {111}

GaP lattice planes in Fig. 5(b). Additionally, four symmetri-

cally orientated GaP{111} Bragg peaks with far lower inten-

sity were detected at v � 16� (indicated by black circles and

by close-up), matching the sketched angle between {111}MT

planes of microtwins and [001] surface normal in Fig. 5(b).

In summary, it can be stated that microtwin formation in the

grown GaP(001) layer appears, creating additional {111}MT

planes, which are tilted by 39� in the v-direction away from

the original {111} planes of the ideal GaP film structure.

(Fig. 5(b)).39,40

To study the microtwin formation in GaP layer in more

detail, a circular /-scan (0��360�) at fixed position v¼ 16�

and on the 2h value of GaP(111) was performed. Figure 5(c)

clearly depicts an anisotropic MT nucleation behavior: MT

density is generally higher along the [110] direction (bold

arrow in Figs. 5(a) and 5(d)), A and C) than along the [1�10]

direction (dashed arrow in Figs. 5(a) and 5(d), B and D).

Taking the use of a 4� off-orientated Si(001) substrate into

account (Fig. 5(d)), it is thus demonstrated that microtwin

formation along the [110] double step direction (solid arrow

directions) is higher in comparison to the [1�10] direction par-

allel to the step edge (dashed arrow directions). For example,

the highest amount of microtwins (A) emerge therefore

along the [110] direction oriented away from the double

steps. It is noted that this anisotropic behavior of the micro-

twin formation in GSMBE grown GaP on 4� off-oriented

Si0.8Ge0.2/Si(001) is different from the results observed for

GaP on Si grown by MOCVD.23 Finally, no difference in

MT concentration and anisotropic behavior were found com-

paring as-deposited and annealed samples (data not shown).

This result confirms that the applied postannealing process

does not reduce the number of existing microtwins. It is

noted that one possible and recently reported approach for

future attempts to remove microtwins could be laser

annealing.41

5. TEM defect study

For a further investigation of the defect structure in the

deposited 170 nm thick GaP layers on top of the 20 nm

Si0.8Ge0.2/Si(001) substrate, a cross section TEM study

was carried out. Figure 6(a) shows the TEM images of the

FIG. 5. (a) XRD pole figure measurement adjusted on the GaP(111) reflection performed after postannealing of a 170 nm GaP/20 nm Si0.8Ge0.2/Si(001) sample

at 550 �C for 15 min under 1 atm N2. (b) Sketch of microtwin formation in (001) oriented GaP layers. (c) XRD /-scan on GaP(1 1 1) Bragg reflection at

v¼ 16�. (d) Schematic sketch of microtwin orientation with respect to 4� off-oriented substrates.

073515-6 Skibitzki et al. J. Appl. Phys. 111, 073515 (2012)

Downloaded 20 Apr 2012 to 62.141.165.1. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions

GaP/Si0.8Ge0.2/Si(001) 4�-off oriented heterostructure pro-

jected along the h1�10i azimuth (parallel to the step edges).

The TEM image indicates a very high crystal quality of the

Si(001) substrate and Si0.8Ge0.2 layer combined with very

sharp interface between these two layers without any visible

defects (or residual oxide interfacial layers). Furthermore,

Fig. 6(a) depicts a crystalline and continuous GaP layer

grown on top of Si0.8Ge0.2. However, AFM images in

Fig. 6(b) show over a bigger scale (2� 2 lm2) an increase in

surface root mean squared (rms) roughness after GaP deposi-

tion from 0.2 nm to 19.6 nm. Most interestingly, the interface

between GaP and Si0.8Ge0.2 is more defective than the upper

GaP part. Most of these observed defects are annihilated af-

ter about 70 nm GaP thickness. Only a few defects, situated

on {111} glide planes, are located at larger thicknesses or

even reach the surface. This TEM result reports strong evi-

dence that these defects are mainly growth defects. Such

growth defects do not nucleate by plastic relaxation of

strained, closed 2D thin film structures, but mostly during

the coalescence process of a film structure formed by initial

3D island nucleation processes.23,24,42

To determine the interface quality between GaP and

Si0.8Ge0.2 in more detail, a HRTEM image is shown in

Fig. 6(c). Firstly, an enlarged section of the GaP/Si0.8Ge0.2

interface is displayed in Fig. 6(d) to demonstrate the high

quality of the GaP/Si0.8Ge0.2 heteroepitaxy. Due to the weak

contrast, no clear GaP/Si0.8Ge0.2 interface transition (indi-

cated by arrows) and no double step characteristics resulting

from the use of 4�-off oriented substrates can be identified.

Secondly, Fig. 6(c) demonstrates in addition that the GaP

interface layer contains stacking faults on GaP{111} planes

(also indicated by arrows). Some of these propagating stack-

ing faults are annihilated after their creation near the inter-

face by building a triangular structure that inhibits further

expansion of this defect. Figure 6(e) shows as an example an

intrinsic stacking fault on a (111) plane. Assuming the nor-

mal stacking order ABCABCABC along the [111] direction,

stacking disorder is depicted as ABC_BCABCA (A-plane

missing). It is noted that no clear indication of misfit disloca-

tions were found in our HRTEM images for the 170 nm

GaP/20 nm Si0.8Ge0.2/Si(001) heterostructure. This is

expected because, due to the small lattice mismatch, misfit

dislocations are separated by about 152 nm for fully relaxed

GaP on pseudomorphic Si0.8Ge0.2/Si(001).43

Finally, APD defect characterization of the deposited

170 nm GaP layers was performed using the {002} dark field

TEM imaging technique. It is known that {002} reflections

are especially sensitive to APDs, revealing this defect type in

form of reversed contrast changes.22–24 Figure 7 shows a

cross section dark field HRTEM image pair of the 170 nm

GaP/20 nm Si0.8Ge0.2/Si(001) heterostructure projected along

the 4� miscut direction (h110i azimuth) taken by slightly

FIG. 6. Cross section TEM image (a) and AFM surface images (before and

after GaP deposition) (b) of 170 nm GaP on pseudomorphic 4� off-oriented

Si0.8Ge0.2/Si(001). High resolution TEM image of the interface region

between Si0.8Ge0.2 and GaP layers (c), as well as close-up images of a well

grown interface area (d) and an intrinsic stacking fault (open circles labeled

with ABC show stacking order along the {1 1 1} direction) (e).

FIG. 7. Cross section dark field HRTEM image pair of APDs at the GaP/

Si0.8Ge0.2 interface of the 170 nm GaP/20 nm Si0.8Ge0.2/Si(001) heterostruc-

ture taken by slightly tilted (0 0 2) (a) and (0 0 �2) (b) reflection.

073515-7 Skibitzki et al. J. Appl. Phys. 111, 073515 (2012)

Downloaded 20 Apr 2012 to 62.141.165.1. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions

tilted (002) (Fig. 7(a)) and (00�2) (Fig. 7(b)) reflections. This

dark field image pair confirms the presence of APDs in the

GaP layer in form of triangular shaped structures located

near the defective GaP/Si0.8Ge0.2 interface by the character-

istic contrast change. The APDs are limited by anti-phase

boundaries, forming mainly {111} and {211} facets.

According to theoretical calculations, {211} facets are ener-

getically favored over {111} facets.23 However, observed

APDs disappear by self-annihilation of crossed anti-phase

boundaries after about 70 nm GaP thickness. It is mentioned

that, besides APD detection, SFs are found inside and out-

side of APDs (indicate by arrows in Fig. 7(a)). It is noted

that APD defect free growth of GaP after about 70 nm in our

study on 20 nm Si0.8Ge0.2/Si(001) heterostructure corre-

sponds well to similar results for GaP on Si.22–24 Differences

in APD defect nucleation might however exist and require

further investigation.

IV. CONCLUSION

A heterostructure growth study of GaP on pseudomor-

phic 4� off-oriented Si0.8Ge0.2/Si(001) substrates, using a

combination of RPCVD for SiGe and GSMBE for GaP depo-

sition, was performed in order to develop a wide bandgap

GaP collector concept for future SiGe HBTs. The following

main results were reported:

1. Theoretical model calculation was applied to evaluate the

feasibility of the approach to prepare truly pseudomorphic

GaP/Si0.8Ge0.2/Si(001) heterostructures suitable for HBT

applications. It is found that the calculated critical thick-

ness of about 64 nm for GaP on pseudomorphic

Si0.8Ge0.2/Si(001) might be too low for a HBT wide

bandgap collector application. Consequently, nitrogen

incorporation in GaP can be a viable way for increasing

the critical GaP thickness.

2. To determine the maximal thermal budget for GaP over-

growth on pseudomorphic 20 nm Si0.8Ge0.2/Si(001), a

detailed XRD analysis was performed. A maximal GaP

growth temperature of 800 �C was identified, because

plastic relaxation of pseudomorphic 20 nm Si0.8Ge0.2/

Si(001) starts beyond this process temperature.

3. XRD was used to characterize the epitaxial relationship and

structure quality of the 170 nm GaP/20 nm Si0.8Ge0.2/

Si(001) heterostructure system. The epitaxial relationship of

the single crystalline heterostructure is given by

GaP[001];h110ikSi0.8Ge0.2[001];h110ikSi[001];h110i.However, we did not succeed to establish growth conditions

for fully pseudomorphic growth of the heterostructure:

Although the 20 nm Si0.8Ge0.2 base stays pseudomorphic

underneath 170 nm GaP, the GaP layer grows partially

relaxed (� 40%).

4. XRD and TEM revealed that partial relaxation is due to

the presence of mainly stacking faults and microtwins,

and are primarily located at the GaP/Si0.8Ge0.2 interface

region. This result in combination with high temperature

XRD studies, which revealed no plastic relaxation within

the applied thermal budget, point to the formation of

so-called growth defects during the initial 3D island

nucleation of the GaP film as the main origin of the partial

relaxation process in the GaP thin film.

5. APD-free GaP growth is observed for layer thicknesses

beyond 70 nm, in line with the literature for GaP on Si.

Future work will focus on improved 2D GaP layer

growth conditions in order to prepare truly pseudomorphic

GaP/Si0.8Ge0.2/Si(001) heterostructures with low defect den-

sities. For this purpose, (selective) GaP heteroepitaxy studies

in local HBT Si0.8Ge0.2/Si(001) mesa structures are currently

under way by GSMBE and MOCVD.

ACKNOWLEDGMENTS

The authors would like to acknowledge the support of

A. Fischer in the frame of theoretical studies for this work.

1J. D. Cressler, The Silicon Heterostructure Handbook (CRC Press, New

York City, New York, 2005).2G. S. May and S. M. Sze, Fundamentals of Semiconductor Fabrication(Wiley, Hoboken, New Jersey, 2004).

3B. Heinemann, R. Barth, D. Bolze, J. Drews, G. G. Fischer, A. Fox, O.

Fursenko, T. Grabolla, U. Haak, D. Knoll, R. Kurps, M. Lisker, S. Marsch-

meyer, H. Rucker, D. Schmidt, J. Schmidt, M. A. Schubert, C. Wipf, D.

Wolansky, Y. Yamamoto, and B. Tillack, Tech. Dig. – Int. Electron Devi-

ces Meet. p. 688 (2010).4E. O. Johnson, “Physical limitations on frequency and power parameters

of transistors,” RCA Rev. 26, 163 (1965).5T. Kraemer, M. Rudolph, F. J. Schmueckle, J. Wuerfl, and G. Traenkle,

IEEE Trans. Electron Devices 56(9), 1897 (2009).6S. F. Fang, K. Adomi, S. Iyer, H. Mordoc, C. Choi, H. Zabel, and N.

Otsuka, J. Appl. Phys. 68(7), R31 (1990).7M. Levinshtein, S. Rumyantsev, and M. Shur, Handbook Series on Semi-conductor Parameters (World Scientific, Singapore/New Jersey/London/

Hong Kong, 1996).8M. V. Fischetti, IEEE Trans. Electron Devices 38(3), 634 (1991).9E. A. Fitzgerald, Y.-H. Xie, D. Monroe, P. J. Silverman, J. M. Kuo, A. R.

Kortan, F. A. Thiel, and B. E. Weir, J. Vac. Sci. Technol. B 10(4), 1807

(1992).10H. Doscher, B. Kunert, A. Beyer, O. Supplie, K. Volz, W. Stolz, and T.

Hannappel, J. Vac. Sci. Technol. B 28(4), C5H1 (2010).11T. J. Grassman, M. R. Brenner, S. Rajagopalan, R. Unocic, R. Rehoff, M.

Mills, H. Fraser, and S. A. Ringel, Appl. Phys. Lett. 94, 232106 (2009).12H. Doscher and T. Hannappel, J. Appl. Phys. 107(12), 123523 (2010).13W. T. Masselink, T. Henderson, J. Klem, R. Fischer, P. Pearah, M. Hafich,

P. D. Wang, G. Y. Robinson, and H. Morkoc, Appl. Phys. Lett. 45(12),

1309 (1984).14T. Soga, T. Jimbo, and M. Umeno, J. Cryst. Growth 163, 165 (1996).15Y. Takagi, H. Yonezu, K. Samonji, T. Tsuji, and N. Ohshima, J. Cryst.

Growth 187, 42 (1998).16M. Sadeghi and S. Wang, J. Cryst. Growth 227–228, 279 (2001)17S. L. Wright, H. Kroemer, and M. Inada, J. Appl. Phys. 55(5), 2916

(1983).18A. Letoublon, W. Guo, C. Cornet, A. Boulle, M. Veron, A. Bondi, O.

Durand, T. Rohel, O. Dehaese, N. Chevalier, N. Bertru, and A. LeCorre,

J. Cryst. Growth 323, 409 (2011).19K. Yamane, T. Kobayashi, Y. Furukawa, H. Okada, H. Yonezu, and A.

Wakahara, J. Cryst. Growth 311, 794 (2009).20X. Yu, P. S. Kuo, K. Mai, O. Levi, M. M. Fejer, and J. S. Harris Jr.,

J. Vac. Sci. Technol. B 22(3), 1450 (2010).21W. G. Bi, X. B. Mei, and C. W. Tu, J. Cryst. Growth 164, 256 (1996).22V. Narayanan, S. Mahajan, K. J. Bachmann, V. Woods, and N. Dietz, Acta

Mater. 50, 1275 (2002).23A. Beyer, I. Nemeth, S. Liebich, J. Ohlmann, W. Stolz, and K. Volz,

J. Appl. Phys. 109, 083529 (2011).24K. Volz, A. Beyer, W. Witte, J. Ohlmann, I. Nemeth, B. Kunert, and W.

Stolz, J. Cryst. Growth 315, 37 (2011).25R. Kaplan, Surf. Sci. 93, 145–158 (1980).26A. Fischer, H. Kuhne, M. Eichler, F. Hollander, and H. Richter, Phys. Rev.

B 54(12), 8761 (1996).

073515-8 Skibitzki et al. J. Appl. Phys. 111, 073515 (2012)

Downloaded 20 Apr 2012 to 62.141.165.1. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions

27J. W. Matthews and A. E. Blakeslee, J. Cryst. Growth 27, 118 (1974).28T. Soga, T. Jimbo, and M. Umeno, Appl. Phys. Lett. 63(18), 2543

(1993).29Y. Takagi, Y. Furukawa, A. Wakahara, and H. Kan, J. Appl. Phys. 107,

063506 (2010).30W. G. Bi and C. W. Tu, Appl. Phys. Lett. 49(24), 3710 (1996).31K. Momose, H. Yonezu, Y. Fujimoto, K. Ojima, Y. Furukawa, A. Utsumi,

and K. Aiki, Jpn. J. Appl. Phys. 41(12), 1 (2002).32G. Bhagavannarayana and P. Zaumseil, J. Appl. Phys. 82, 1172 (1997).33V. M. Kaganer, R. Koehler, M. Schmidbauer, and R. Opitz, Phys. Rev. B

55(3), 1793 (1996).34S. N. G. Chu, A. T. Macrander, K. E. Strege, and W. D. Johnston, Jr.,

J. Appl. Phys. 57(2), 249 (1985).

35C. R. Bayliss and D. L. Kirk, J. Phys. D: Appl. Phys. 9, 233 (1975).36T. Ogawa, H. Morota, and S. Adachi, J. Phys. D: Appl. Phys. 40, 4603

(2007).37P. Zaumseil and T. Schroeder, J. Phys. D: Appl. Phys. 44, 055403 (2011).38F. Ishizuka and T. Itoh, J. Appl. Phys. 66(7), 3007 (1989).39T. Schroeder, A. Giussani, J. Dabrowski, P. Zaumseil, H.-J. Mussig, O.

Seifath, and P. Storck, Phys. Status Solidi C 6(3), 653 (2009).40P. Zaumseil and T. Schroeder, J. Appl. Phys. 104, 023532 (2008).41G. Wang, E. Rosseel, R. Loo, P. Favia, H. Bender, M. Caymax, M. M.

Heyns, and W. Vandervorst, Appl. Phys. Lett. 96(11), 111903 (2010).42F. Ernst and P. Pirouz, J. Appl. Phys. 64, 4526 (1988).43I. V. Markov, Crystal Growth for Beginners (World Scientific Publishing

Co., Singapore, 2003).

073515-9 Skibitzki et al. J. Appl. Phys. 111, 073515 (2012)

Downloaded 20 Apr 2012 to 62.141.165.1. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions