You are on page 1of 16

AMAN SPECTROSCOPY OF DOPANT IMPURITIES AND DEFECTS IN GaAs

YERS
Joachim Wagner
FraunhoferInstitut fUr Angewandle Festkiirperphysik
Tullastrasse 72, D7800 Freiburg, Federal Republic of Germany
yrRODUCTION
Raman spectroscopy has found widespread use for the investigation of intrinsic
'bonon modes and electronic excitations of free carriers in bulk semiconductors as well
is in semiconductor heterostructures 1-3, Raman scattering by impurities and defects 4, 5,
J contrast, has been used so far only in a limited number of examples. This also
otrasts to other techniques, such as photoluminescence and absorption spectroscopy,
'ch are used routinely for the investigation of defects and impurities.
This is, at least partially, due to the weakness of Raman signals from defects and
'mpurities because of their small concentrations as compared to the density of host
lltice atoms. In the last few years, the advent of optical multichannel detectors allowed
Dincrease the sensitivity of the Raman technique considerably 6, which opens new
JOssibilities for the study of defects and impurities by Raman scattering. This is, in
true for the recent work on Raman spectroscopy of dopant impurities in
llanar (5.) doped GaAs 7 which would have been hardly possible using conventional
;figle channel detection.
In Raman spectroscopy, impunues and defects can be observed either via
ttering by internal electronic excitations 4, 5 or via scattering by localized vibrational
nodes 2, 8- n. Both types of scattering probe directly individual impurities and defects.
e intensity of the scattered light per unit solid angle as/an can be written as 12
as/an - a n V (I)
.vbere a denotes the scattering cross section per impurity, n is the impurity concentra-
ion, and V is the scattering volume. If the sample is opaque for the light used to excite
Raman spectrum, V is proportional to the probing depth 1/(2") where" denotes the
!bsorption coefficient. Using light for which the material is fully transparent, V is for
lackscattering geometry proportional to the thickness of the sample. Eq. 1 shows that,
a given scattering cross section per impurity and a given scattering volume, the
lleasured Raman intensity is directly proportional to the impurity concentration. There-
t.ig
m
Scattering in Semiconductor Structures and Super/altius
tdittd by D.J. Lockwood and I.F. Y01mg, Plenum Press, New York, 1991
275
fore, it is straight forward to calibrate the Raman technique and to make it a verS.lll_
tool for quantitative materials characterization. This is in contrast to photoluminescenc;
spectroscopy, where the measured signal intensity is not only determined by the impUrit
concentration but also by the carrier life time. 1
From Eq. I it is also evident that there are two ways to enhance the Raman signal
from scattering by impurities and defects. One possibility is to use incident light for
which the material is transparent. This leads to a fairly large scattering volume limited
only by the thickness of the sample, which is typically in the range 0.5 - 5 mm. l'bi!
approach has been used successfully to study impurities in bulk semiconductors such as
Si 4, 5 and GaAs 4, 5, 13, 14 In the case of GaAs, a detection limit for Raman scattering by
electronic excitations of residual shallow acceptors of 5 x 10
14
cm,3 has been reo
ported t4 which demonstrates the sensitivity of this technique.
Alternatively, the incident photon energy can be chosen to match an electronic
interband transition which is well above the fundamental band gap of the material. For
this kind of optical excitation, the probing depth 1/(2,,) is, for e.g., GaAs in the range of
10 - 100 am 15, 16 The scattering cross section, on the other hand, can be enhanced signi
ficantly by this approach as it has been demonstrated for scallering by electronic excita
tions of acceptors incorporated in quantum well heterostructures 3,17-19 and for scatter
ing by impurity induced local vibrational modes (LVM) in GaAs Il In the latter case
optical excitation in resonance with the E
1
band gap has been used, leading to a probing
depth of only 10 nm IS, 16 This makes the above approach particularly well-suited 10
study, e.g., dopant impurities in thin highly doped epitaxial layers,ll, 20, 21 including 6
doped structures 7
Defects introduced in large concentrations by ion implantation or reactive ion
beam etching can also affect Raman scattering by intrinsic phonon modes. The firsl-
order phonon Raman spectrum gets modified by the relaxation of the
imposed by momentum conservation as studied, e.g., in detail for ion implanted
GaAs 22, 23 Resonant dipole forbidden first-order and allowed second-order scattering
by longitudinal optical (LO) phonons is sensitive to ion beam induced defects via the
broadening of the corresponding resonances in the Raman efficiency 23-25
In the following we concentrate on GaAs and discuss Raman spectroscopy of
dopant impurities and defects via vibronic excitations. Firstly, we focus on the effect of
ion beam damage on resonant Raman scallering by intrinsic LO phonons both in ion
implanted and in reactive ion etched GaAs. Secondly, recent work on the Raman
spectroscopy of impurity induced LVM in highly-doped GaAs is reviewed. It is further
demonstrated how this spectroscopic technique can be applied to the analysis of the
incorporation and spread of dopant atoms in 6-doped GaAs structures.
EFFECf OF ION BEAM INDUCED DEFECfS ON RAMAN SCATTERING BY
INTRINSIC PHONON MODES
Fig. 1 displays the effect of ion implantation on the first-order deformation poten-
tial scattering 22, 23 A series of Raman spectra is shown for material implanted with 29Si
at an ion energy of 100 keV with doses ranging from I x 1013 to I x 1O
t6
cm-
l.
With in'
creasing implantation dose, the LO(r) phonon line shifts towards lower energies and
broadens. These effects have been explained by Tiong et al. 26 by a "spatial correlation
276
600
,
GoAS:29Si+
LOcr)
100 keY
hVL ::: 3.00 eV
TOCr) I
ll' fl2
y:
2lO(r)
1x10
13
em-
2
~
, I
I
I
,
I
,
,
12 -, j
,
I
I
hl10 em J:\
;"
I'
I I
"
, ,
t
,
as graw,n
1,
100
CIl
cr
:'.
>-
~
Ul
Z
W
~
Z
z
..
:>:
..
cr
200 300 '00 500
RAMAN SHIFT{cm-')
Fig. 2 Room-temperature Raman
spectra of as-grown and 29Si +-implanted
GaAs. Implantation doses are given in
the figure. The spectra were excited at
3.00 eV and recorded in the x(y,y)i1 scat-
tering geometry with a spectral resolu-
tion of 5 cm-
I
.
GoAs:29Si+
100keV
hill::: 2.57 eV
h'O'6
em
-2
, 00 200 300 "10 500
RAMAN SHIFT lem-')
Z
::>
z I-c---I--
..
:E
..
cr
Fig. 1 Room-temperature Raman
spectra of 29Si +-implanted GaAs. Im-
plantation doses are indicated in the
figure. The spectra were excited at 2.57
eV with the incident light polarized along
a(110) crystallographic direction and the
scattered light not analyzed for its polari-
zation. Spectral resolution was 6 cm-
l
model". This model is based on a phonon confinement concept 27 with the phonon local-
I ization length being as small as - 50 Afor the highest damage level for which a LO
phonon peak is resolved. N. can be seen from Fig. 1 for doses exceeding 1 x 10
'5
cm
2
,
the LO phonon line disappears and three broad bands are observed which arise from
disorder activated first-order scattering in amorphized GaAJ.. These bands develop
gradually for doses in the range 10
13
- 10
15
cm-
2
, indicating that in this range the
material consists of damaged crystalline, as well as amorphized, regions.
The effect of ion implantation induced damage on dipole-forbidden but defect
induced lLO and resonant 2LO phonon scattering is shown in Fig. 2, which displays a
sequence of Raman spectra excited at 3.00 eV in resonance with the E
,
band gap en-
ergy 28, In as-grown GaN., the defect induced ILO phonon line intensity is weak and the
Raman spectrum is dominated by 2LO phonon scattering. Upon ion implantation the
defect induced 1LO phonon line increases and the 2LO phonon peak decreases in inten-
~ t y . Already for a moderate implantation dose of 1 x 1013 cm-
2
, where first-order defor-
malion potential scattering is only modified slightly (Fig. 1), resonant defect induced
ILo and 2LO phonon scattering shows a drastic change with the relative intensities of
both phonon peaks inverted.
277
- 0.59
o
o
1
,
o l(lOOP4 Imp)/I{250em'ljl
6I{210l/I(S'Ocm-') 1
o II210)/1(lO Imp) I
I
o 0
o
'0
,
0 .... -Slopt -0.73
,
o
,
GaAs:
29
Si+
'00 k.V
, ,
,
_00,
o
.... 0.
0 ,
10=-6 6, ........
0
'0 '
'. ,
........ ....::-- Stopt
'4 0 ....

Slopt
10
11
10'2 10
13
10" 10
'5
10
'6
IMPLANTATION DOSE I,m-')
Fig. 4 Relative intensities of allowed
first-order scattering I(LO
DP
+Imp)/I(250
cm"), allowed secondorder scanering
I(2LO)/1(540 cm-
I
), and allowed second
order normalized to symmetry-forbidden
but defect induced first-order scattering
I(2LO)/I(L0
lm
) versus implantation
dose. The ami-vs at the vertical scale
indicate. the corresponding relalive
intensities in as-grown GaAs.
100-
OJ)
Z
UJ

z 1.0-
UJ
>

J
O.H-
-.ll0
- .4 ... 210
"
, ,
, ,
, .
: \ __ 0.25

os_grown
2 CdTe:
'll
In+
in
1- 101 ......
Of----1'-------=="'----I
.,;'
'"

c:

'"

Of---'--------j
1
tel /...- .... ! ...
o .
2.5 2.55 2.6
INCIDENT PHOTON ENERGV (eV)
Fig. 3 Measured scanering efficiencies
as/an for firstorder (lLO) and second
order (2LO) Raman scanering in (a) as
grown and (b) and (c) 113In+.implanted
CdTe. The full (lLO) and dashed (2LO)
lines are drawn to guide the eye. In (a)
the 2LO resonance curve displayed has
been reduced by a factor of 4.
(.)
This slrong effect of implantation damage on the lLO and 2LO phonon scattering
intensities for excitation in resonance with, e.g., the E
t
band gap of GaAs, is caused by a
damage induced broadening of the corresponding resonance in the scattering effi-
ciency 24, 25 This has been demonstrated explicitly for 113In implanted CdTe 25, as
shown in Fig. 3. Here the Eo +"0 gap resonance in the 2LO phonon scanering efficienCY
broadens and decreases in height continuously with increasing implantation dose (Fig.
3a.c). The resonance for lLO phonon scattering also broadens wilh increasing dose, bu
l
its height first increases for low doses because this scanering process is defect induced 1J;
(Fig. 3a and b). For higher implantation doses the increase in defect concentration
overcompensated by the broadening of the resonance leading eventually also to a
decrease in the lLO phonon scattering intensity (Fig. 3c).
The damage induced broadening of the above resonances for lLO and 2LO
phonon scattering is understood as follows 23 It has been shown for GaAs that
structures in the spectrum of the dielectric function E, which are related to interbao
d
276
;taDsitions such as across the E
t
band gap, get broadened and smeared out upon ion
, plantation 29 The resonances in the efficiency for defect induced 1LO and 2LO
,banon scattering can be approximated by la2x/aw
2
1
2
, where X denotes the electric
'-'1 and wthe incident photon energy 28, Thus anychange in E, which is related
Xby E= 1+41TX, strongly affects the resonances for scattenng by LO phonons.
The effect of ion implantation damage on various scattering processes by intrinsic
"anon modes in GaAs is summarized in Fig. 4 24. It shows the intensity of resonant
\La phonon scattering normalized to that of dipole forbidden but defect induced I LO
,.,altering (I(2LO)/I(L0
1m
)) plotted versus implantation dose. This ratio decreases
;lfangly with increasing dost which has been exploited to study the lateral homogeneity
:[ian implantation in GaAs wafers 24. For comparison also, the 2LO scattering intensity
to the less resonant second-order scattering by two transverse optical (TO)
'banons (I(2LO)/I(540 cm-
1
)), as well as the scattering intensity of first-order scattering
LO phonons normalized to the TO phonon band in amorphized GaAs
I(LO
DP
+Imp)/I(250 em-I)), are plotted in Fig. 4 24.
The sensitivity of resonant LO phonon scattering to ion beam induced damage in
.lar semiconductors in general, and in GaAs in particular, can he used to assess the
led :amage induced by reactive ion etching (RIE) processes. This is demonstrated in Fig. 5
!50 ere the intensity of resonanl 2LO phonon scattering, normalized to that of dipole for-
ing 'dden but defect induced lLO scattering (I(2LO)/I(LO
lmp
))' is plotted versus the bias
,d- oltage applied during the RIE process. The material processed was n-type GaAs with a
len 'ee carrier concentration of 2.7 x 10
17
cm-
3
which was exposed to a CHF
3
plasma. For
ing 'as voltages exceeding 300 V, the ratio I(2LO)/I(L0
1mp
) decreases, indicating a degra-
lan ition of the material within the probing depth of 10 nm underneath the surface. The
ale
ive
O. 6
GoAs:Si
CHF,-RIE


o
H
-,0,5

E
H
o
-'
H
::: 0.4
o
-'
N
ing
,ya
:f/i.
as
ncy
Fig,
but
j28
n is
, a
0.3
L-__.....l---'
o 300 350 400
BIAS VOLTAGE Iv)
LO iig. 5 Relative intensity of allowed 2LO scattering normalized to symmetry-forbidden
III defect induced lLO scattering in n-type GaAs plotted versus bias voltage applied
reactive ion etching.
279
degradation of the material detected by Raman spectroscopy correlates with a degrada.
tion of the electrical properties of Schottky diodes fabricated on the same RIE
processed substrates 30. The above results clearly show that resonant Raman scattering
by intrinsic LO phonon modes provides a useful tool to analyze residual defects of RIE
processing steps. This is of particular interest for the lateral processing of semiconductor
heterostructures to fabricate, e.g., quantum wires and quantum dots 31
RAMAN SCATTERING FROM LOCALIZED VIBRATIONAL MODES
Dopant impurities which are significantly lighter than the host lattice atoms, give
rise to localized vibrational modes (LYM). These modes, which are observed in OaAs
doped, e.g., with Si, Be, or C, are higher in frequency than the intrinsic host lattice
phonon modes 32. For a given dopant atom, the frequency and fine structure of the LYM
indicate the lattice site occupied by the atom 32 Infrared absorption spectroscopy is the
commonly used technique to study LYM in bulk material and in relatively thick
(> 1 I'm) epitaxial layers 32 Recently Raman spectroscopy has been used to investigate
impurity induced LYM in heavily doped OaAs layers 7, 10, 11,20,21,3336.
In the following subsection, Raman scattering by LYM is discussed for homoge
neously doped OaAs layers, whereas in the subsection following it emphasis is laid on
tbe application of this Raman technique to Ihe study of dopant incorporation in <I-doped
OaAs structures.
Homogeneously Doped Layers
Fig. 6 displays the low-temperature Raman spectrum of a heavily doped OaAs
MBE layer. The upper spectrum, which was excited at 3.00 eV in resonance with the E
J
band gap, shows the LYM of 2llSi on a Ga site (SiGal at 384 cm-
I
32 superimposed on the
intrinsic second-order phonon spectrum 28, For comparison, the second-order pbonon
spectrum recorded under identical conditions from an undoped reference sample is
displayed in the lower half of Fig. 6.
The analysis of the incorporation of Si into OaAs is of particular interest because
Si can be incorporated on a Ga site (Si
G
.), where it acts as a donor, as well as on an A5
site (Si
As
) acting as an acceptor 32 In addition, electrically inactive Si
G
- SiAs next
neighbouring pairs can be formed, as well as complexes with, presumably, intrinsic de-
fects labelled Si-X and Si-Y 32 Si-X is believed to act as an acceptor and Si-Y might also
be an acceptor or electrically inaclive 32,33,37,38 Consequently, the electrical properties
of heavily Si doped GaAs depend strongly on the balance between these different
impurities and impurity complexes as shown in Fig. 7. There, a sequence of low tempe-
rature Raman spectra is displayed, which were recorded from MBE grown GaAs layerS
doped with Si at different concentrations. The total Si concentration lSi] measured by
secondary ion mass spectroscopy (SIMS) increases monotonically from 4.7 x 10
18
cm-) to
4 x 10
19
cm-) (Fig. 7a-c). For a total Si concentration of 4.7 x 10
18
and 1.2 x 10
19
crn-
l
(Fig. 7a and b), the free carrier concentration measured by Raman spectroscopy of the
coupled plasmon-pbonon modes 2 increases from 3.8 x 10
18
to 8.6 x 10
18
cm-
3
When ibe
total Si concentration is further increased to 4 x 10
19
cm-
3
(Fig. 7c), the free carrier con-
centration drops to S.4 x 10
18
cm-
3
These differences in the electrical aClivalion of the
280
GoAs:Si
'hwl'=3.00 eV
77 K
ISiJ=l.hl0'9 em-)
[5il" 4.10
19
em-)
(1:5. I; )1110 '8
ern
-)
[SiJ=4.7 ,,1018
cm
-3
",,3.8.,0\8 em-
3
,
'i
Ga
I "5'
'A.
I I
I
I
10)
350 400 450
RAMAN SHIFT {em-'}
Fig. 7 Low-temperature Raman spectra
of MBE grown GaAs:Si with different
dopant concentrations lSi] given in the
figure. The free carrier concentration n is
also given. For all spectra, which were
excited at 3.00 eV and recorded with a
spectral resolution of 5 cm-
I
, the second-
order phonon spectrum has been subtrac-
ted.
GoAs: 51
LSD 550
RAMAN SHIFT(cm"l
350
>-
>-
~
'::
iii
Vl
'"
'"
"'
W
~
~
Ib)
'"
=
'"
'"
..
..
1:
1:
..
..
a:
a:
Fig. 6 Low-temperature Raman spectra
of Si doped GaAs grown by MBE (lOp)
and of an undoped reference sample
(bottom). The spectra were exciled at
3.00 eV and recorded with a spectral re-
solution of 5 cm
l
dopant is explained by the Si site occupancy in the different samples. For the two lower
5i concentrations, only the LVM of 28SiGa is observed acting as a donor (Fig. 7a and b).
For the highest Si concentration, the 28SiGa donor gets compensated by the acceptors
28SiAs and Si-X, which produce LVM at 399 and 370 cm-
I
, respectively 32, 33
The incorporation of Si dopant atoms into different lattice sites depends very much on
the delails of the growth conditions used for the epilaxy of the GaAs layer 33, 37, 39 This
applies similarly to the incorporation of Si into ion implanted and thermally annealed
material 40. It is illustrated in Fig. 8, which shows a sequence of LVM Raman spectra
taken from a series of samples all implanted with 100 keV 29Si at a dose of 1 x 10
16
cm-
I.
Then the samples were subjected to different annealing processes. Wafers A and D were
rapid thermal annealed with the sample surface protected by a Si0
2
(wafer A) or a
5i
3
N
4
(wafer D) capping layer. Wafers Band C were furnace annealed under arsine
overpressure (wafer B) or with the surface protected by the proximity technique (wafer
C). It is evident that the Si site distribution within the probing depth of 10 nm
underneath the surface depends strongly on the details of the annealing process. The
relative concentrations of Si
Ga
and Si
As
vary considerably and the formation of the Si-X
complex is observed exclusively for annealing under a Si0
2
protective cap 40.
281
GcAs: 29
Si
+
1x1O'6
cm
-2
hWl:JOOeV
77 K
350 400 450 500
RAMAN SHIFT lem-')
Fig. 8 Low-temperature Raman spectra of GaAs implanted with 1 x IO
t6
29Sijcm
2
and
annealed under various conditions. Wafer A was capped with Si0
2
followed by rapid
thermal annealing. Wafers Band C were furnace annealed with the surface protected by
arsine overpressure (B) or the proximity technique (C). Wafer D capped with Si
3
N
4
was
processed by rapid thermal annealing. The Raman spectra excited at 3.00 eV were
recorded with a spectral resolution of 5 cm-].
Another interesting point is the dependence of the Si LVM Raman spectrum on
the incident photon energy. Fig. 9a shows a spectrum excited at 3.00 eV in resonance
with the E
I
band gap, whereas the spectrum displayed in Fig. 9b was excited below thai
resonance at 2.71 eV. The LVM of Si
As
and Si-X are observed for both incident photon
energies. The LVM of Si
Ga
, in contrast, is the dominant peak for excitation at 3.00 eV
but absent in the spectrum excited at 2.71 eV. The difference in behaviour for the Si
G

LVM on one hand, and the Si


As
and Si-X LVM on the other, migbt be either due to the
different lattice sites occupied or to the different electrical activity as a donor and an
acceptor, respectively 21
It is interesting to compare the above findings with the resonance behaviour of
Raman scattering by the 9BeGa LVM. Be is an acceptor occupying a Ga site. Tbe cor
responding LVM Raman spectra are displayed in Fig. 10. Both for excitation in reso
nance at 3.00 eV (Fig. lOa) and below resonance at 2.71 eV (Fig. lOb), the LVM of
9Be
Ga
is observed at 482 em-I, whicb indicates a resonance behaviour for Be
Ga
different
to the one for Si
Ga
but similar to the Si
As
and Si-X acceptors 21 This may lead to (he
conclusion that the difference in resonance bebaviour for Si
Ga
and Si
AS
is due to the
different electrical activity. However, more work is necessary in order to understand the
underlying scattering mechanisms.
282
GoAs:Se
3.5_10
19
em-)
17K
'"
z
UJ hUlL'::3.00eV
~
z
z
..
:>:
..
a:
,00 500 600
RAMAN SHIFT (em-I)
2B
S
'
GoAs: 51
'Go
Si-X
5 . ~ _10
18
em-)
I
17K
2B
SiAs
lo}
I
I
,.
I
!:
I
'"
I
%
uJ I
~
I
%
I
%
I
<
r
<
"
IbJ
350 400 450 500
RAMAN SHI F, lem-')
Fig.9 Low-temperature Raman spectra
of MBE grown GaAs:Si excited at diffe-
fOnt photon energies indicated in the
figure. Free carrier concentration is also
given. Spectral resolution was set to
Scm-I
Fig. 10 Low-temperature Raman
spectra of MBE grown GaAs:Be excited
at different photon energies indicated in
the figure. The nominal dopant concen-
tration is also given. Spectral resolution
was 5 em-I.
Recently, doping of GaAs with carbon bas received considerable interest for the
fabrication of thin heavily doped layers in, e.g., heterostructure bipolar transistors 41-43
Carbon in GaAs gives rise to a LVM with a frequency of 582 cm-] for 12CAS 32 The
Raman spectrum of a heavily carbon doped layer grown by metalorganic molecular
beam epitaxy (MOMBE) is displayed in Fig. 11a 44 Fig. Jib shows the corresponding
;pectrum of an undoped reference sample and the difference spectrum [(a) - (b)] is
otled in Fig. llc. The LVM of t2CAs is clearly resolved in the Raman spectrum. All
ilie spectra were excited at 2.71 eV, because for excitation at 3.00 eV in resonance with
ilie E
t
gap, the resonantly enhanced scattering by 2LO phonons strongly overlaps with
~ e 12CAs LVM.
Raman spectroscopy by LVM has also been made a quantitative technique. The
uoss section per impurity as/an has been determined for scattering by the 28Si
Ga
and
~ e 9Be
Ga
LVM to as/an = (3.2 0.6) x 10-
24
Sr']cm
2
and as/an = (l.! 0.6) x 10-
24
51-1cm2, respectively 20, 45 These cross sections are considerably larger than the cross
'<ction per host lattice atom of as/an = 9 x 10-
26
Sr-
I
cm
2
for dipole allowed intrinsic
lst,order scattering by LO phonons 46 The above values refer to low temperatures
<n K) and excitation at 3.00 eV.
Because it is difficult to measure accurately absolute scattering intensities, also
:aIibration factors have been derived using the strength of intrinsic second-order phonon
;callering as a "built-in" reference 33,45 Based on this internal reference, a linear rela-
:onship between the 9Be
Ga
LVM intensity and the BeGa concentration has been found
283
3S0 '00 450 500
RAMAN SHIFT (cm-
1
1
Fig. 12 Low-temperature Ramao
spectra of (a) a Si and Be .-doped GaAs
sawtooth superlattice and of (b) an
undoped reference sample. The spectrum
of the superlattice with the intrinsic
second-order phonon spectrum subtrac
ted is displayed in (c),
GoAs'Si,8e
5-doped
'"
z
w

z
(')
GoAs:C
p.L.L
l'l.La2.11 ,V
11K
500 600 700
RAMAN SHIFT (em -')
Fig. 11 Low-temperature Raman
spectrum of (a) carbon doped GaAs
grown by MOMBE and of (b) an
undoped reference sample. The diffe-
rence spectrum [(a) - (b)] is displayed in
(c). The spectra were excited at 2.71 eV
and recorded with a spectral resolution
of 5 em-I
for concentrations ranging from 3 x 10
18
up to 1 x 10
20
cm-
3
45 This proves the validity
of the above concept and the derived calibration factors, The detection limit for both
Si
Ga
and Be
Ga
is 2 - 3 x 10
18
cm-]
Dopant Incorporation in Doped Structures
The applicability of Raman scattering by LVM as a quantitative technique, corn'
bined wilh its sensitivity to even very thin doped layers due to the small probing depth of
10 nm for optical excitation in resonance with the E
t
band gap,15 allows us to study the
incorporation of dopants in .-doped GaAs 7 This is demonstrated in Fig, 12 for a Si and
Be .-doped GaAs sawtooth superlattice 47. This superlattice consists of a first layer
5.3 nm underneath the sample surface .-doped wilh 5 x 10
12
cm-
2
Be followed by an
alternating sequence of Si and Be .-doped layers with a spacing of 10.5 nm in-between
and a doping concentration of 1 x 1013 cm-
2
each. Due to the small probing depth of the
Raman experiment for excitation at 3,00 eV in resonance with the E
I
band gap,
284
essentially only the topmost two layers doped with Be and Si, respectively, are probed.
The spectrum (a) shows the 28Si
Ga
LYM at 384 cm-
l
as well as the 9Be
Ga
LYM at
482 em-I superimposed on a background of intrinsic second-order phonon scattering 28.
Fig. 12b displays the second-order phonon spectrum of an undoped reference sample
and Fig_ 12c shows the difference spectrum [(a) - (b)] after subtraction of the second-
order phonon spectrum. Both the 28SiGa and the 9Be
Ga
LYM are well resolved in Fig.
12c which demonstrates the sensitivity of the Raman technique to even a single .-doped
layer-
The .-doping technique has recently found considerable interest both for funda-
roental studies and for device applications 47_ An important question related to this
doping technique is the incorporation and the spread of the dopant atoms along the
growth direction_ A number of experimental techniques has been applied to characterize
.-doping, such as capacitance-voltage (C-Y) profiling 48, magnetotransport measure-
roents 49-51, secondary-ion mass spectroscopy (SIMS) 52-54, and Raman scattering by
LVM 7 In the following, a discussion of how Raman spectroscopy of LYM provides
information on the depth distribution of those dopant atoms incorporated on lattice sites
will be given.
The basic idea is to grow a series of GaAs layers containing a single .-doping spike
of, e.g., Si at a nominal depth Zo underneath the surface with all the samples identical
except for the different depths zo- The measured normalized Si
Ga
LYM intensity
I(Si
Ga
)/I(540 cm-
t
), where the intrinsic second-order phonon scattering strength at
540 em-I is used as a reference, can be written for each sample with a given Si depth
profile [SiGa(z-zo) as
I(Si
Ga
)/I(540 em-I) = k J[SiGa(z-zo)] e-
2az
dz_
o
(2)
Here k is a calibration factor and e-
2az
describes the decay of the Raman sensitivity
versus z, which is the coordinate normal to the surface. Q is the absorption coefficient
with 1/(2"') = 10 nm for excitation at 3.00 eY and 77 K 15 With the above mentioned
series of samples at hand, I(Si
Ga
)/I(540 em-I) can be measured for a variety of depths Zo
and the actual dopant depth profile is, in principle, obtained by SOlving Eq. 2 for
[SiGa(z)]-
In Fig. 13, the normalized Si
Ga
LYM intensity I(Si
Ga
)/1(540 em-I) is plotted versus the
nominal depth Zo for a series of samples grown by MBE at a substrate temperature of
580"C with a nominal dopant density of 8 x 10
12
cm-l. The LYM intensity is found to
increase up to a depth of Zo = 20 nm followed by a rapid decrease for larger values of zo0
These experimental data can neither be explained by a very narrow Gaussian depth dis-
tribution with a full width at half maximum (FWHM) of 2 nm (dotted curve in Fig. 13),
nor by a strongly broadened symmetric distribution with a FWHM of 23 nm (dashed
CUrve in Fig_ 13). The doping profiles, which have been assumed for the calculation [Eq.
2! 01 the I(Si
Ga
)/I(540 em-I) versus Zo curves, plotted in the main part of Fig. 13, are
displayed in the inset.
285
286
To fit the experimental data, one has to postulate that, after the in
growth and deposition of the dopant atoms, these atoms are incorporated in
grown material at a given 3dimensional dopant density 7, 55 Dopant atoms.
not incorporated in the material when finishing the growth of the layer, de
the growth interruption just sit on the surface and are not detected by tIIc
experiment 7 This leads to a strongly asymmetric depth profile with a
concentration over a certain distance in growth direction and an abrupl
towards the substrate (solid curve in the inset of Fig. 13). Assuming a
dopant atoms along the growth direction of z 20 nm, a good quantitative W.UlI-
the experimental data is obtained 7, as shown by the solid curve in the maID
13.
Fig. 13 Normalized 28Siaa LVM scattering intensity I(Si
aa
)/I(540 em
1
) ve
depth Zo of the doping spike. The samples were grown by MBE at
temperature of 580C with an intended doping level of 8 x 10
12
Si/cm
l
.
displays different depth profiles of the Si concentration [Siaal used to calculale
the corresponding I(Si
aa
)/I(S40 cm
l
) versus Zo curves which are displayed in
part of the figure.
0.025
~ 0.100
5'
'"
-;:: 0.075
"-
'J 0.050
-;::
0.125
10 20 30 40
1
0
(nm)
0.150'r:--,-----------
The actual spread of the dopant atoms in adoped structures depend>.
on the growth conditions, such as substrate temperature, and on the 2
dopant density 4954 A reduction in the 2.dimensional dopant density and ~
the substrate temperature are both expected to reduce the acmal wi.dth/
spike. This is illustrated by the LVM data of Fig. 14 which were obtalll
e
.
of Si a-doped samples grown at a nominal substrate temperature of sOO"C "'r
level of 2.8 x 10
12
cm
l
. The resulting doping profile is much nar
ro
"
measured width partially determined by the experimental depth resolunOn,
0.100
E
u 0.075
~
"'
""
....... 0.050
'8
iii
"" 0.025

~ ~
~ o o
I (nm)

I
10 20 30 40
I
o
(nm)
Fig. 14 Same as Fig. 13 for a sequence of samples grown at a nominal substrate tempe-
rature of 500C with an intended doping level of 2.8 x 10
12
cm-
2
the order of 1/(2Cl). Therefore the actual width of the doping profile can only be estima-
ted to S5 nm.
In n-type 6-doped structures space charge effects induce a potential well in wbich
quantized electron subbands are formed 47 The actual shape of this potential well de-
pends on the spread of the dopant atoms. Ideal o-doping with the dopant atoms confined
to one atomic plane leads to a V-shaped potential well whereas a large spread of the
dopant atoms results in a parabolic well 49, 50 The shape of lhis well, in turn, determines
lbe energy spacings of the eleclron subbands. Therefore Raman spectroscopy of inter-
subband transitions provides complementary information on the actual width of doping
spikes in nominally o-doped GaAs. For the present samples grown at a substrate tempe-
ralure of 580C with a doping density of 8 x 10
12
cm-
2
spin-density excitation energies of
24.8 and 36.7 meV were found for a nominal depth of Zo = 30 nm 57 Self-consistent
electronic subband calculations, using the actual spread of the dopant atoms measured
by the above described LVM Raman experiments (see Fig. 13) as an input parameter,
yield spin-density excitation energies of 25.1 and 36.5 meV for transitions between the
second and third and the third and fourth subband, respectively. This is in good agree-
ment with the experimental values 57 It demonstrates that a consistent picture is
obtained for the spread of dopant atoms in Si 6-doped GaAs from the combination of
LVM and subband Raman spectroscopy.
CONCLUSIONS
The effect of impurities and defects on the Raman spectroscopy of vibronic excita-
tions has been discussed. Resonant Raman scattering by intrinsic LO phonon modes has
287
been shown to be sensitive to ion beam induced damage which allows to assess, e.g.,
residual defects in reactive ion etched semiconductor structures. Dopant atoms signifi.
cantly lighter than the host lattice atoms are accessible by Raman spectroscopy via light
scattering by localized vibrational modes. This technique has a detection limit of s 2 - 3,
lO
t8
cm-
3
corresponding to an area density of s 2 - 3 x 10
12
cm-
2
for Si and Be in GaAs.
This opens the possibility to study dopant impurities in quasi-two-dimensional systems
such as, e.g., planar (5-) doped GaAs structures. For such structures the incorporation
and spread of the dopant atoms has been analyzed and correlated with the energy
spacings of the electron subbands formed in the space charge induced potential well.
ACKNOWLEDGMENTS
The author wants to thank M. Ramsteiner (IAF Freiburg), R. Murray, and R.c.
Newman (both with the Imperial College London) for their cooperation in the course of
this work as well as P. Koidl and H.S. Rupprecht for many helpful discussions and
continuous support of the work at the IAF. Thanks are further due to F. Eisen (IAF
Freiburg), K. Kohler (IAF Freihurg), W. PIetschen (IAF Freiburg), K. Ploog (MPI/FKF
Stuttgart), and M. Weyers (RWTH Aachen) for providing samples, as well as to A
Maier for help with the preparation of this manuscript.
REFERENCES
I. M. Cardona, in: "Light Scattering in Solids II", eds. M. Cardona and
G. Giintherodt, Springer, Berlin (1982) p. 49 and references therein.
2. G. Abstreiter, A. Pinczuk, and M. Cardona, in: "Light Scattering in Solids IV", eds.
M. Cardona and G. Giintherodt, Springer, Berlin (1984) p. 5 and references
therein.
3. A. Pinczuk and G. Abstreiter, in: "Light Scattering in Solids V", eds. M. Cardona
and G. Giintherodt, Springer, Berlin (1989) p. 153 and references therein.
4. G_B. Wright and A. Mooradian, in: "Proceedings of the 9'h Int. Conf. on the
Physics of Semiconductors", Nauka, Leningrad (1969) p. 1067; A. Mooradian.
in: "Laser Handbook", Vol. 2, eds. FT Arecchi and E.O. Schulz-Dubois, North
Holland, Amsterdam (1972) p. 1409_
5. M_V. Klein, in: "Light Scattering in Solids I", ed_ M. Cardona, Springer, Berlin
(1975) p. 148.
6. J. Tsang, in: "Light Scattering in Solids V", eds. M. Cardona and G. Giinterhord,
Springer, Berlin (1989) p. 233 and references therein_
7. J. Wagner, M. Ramsteiner, W. Stolz, M. Hauser, and K. Ploog, Appl. Phys. Lett.
55,978 (1989).
8. G. Contreras, M. Cardona, and A. Axmann, Solid State Commun. 52, 861 (1985).
9. G. Contreras, M_ Cardona, and A. Compaan, Solid State Comrnun. 53, 857 (1985).
10. T. Nakamura and T. Katoda, J. App!. Phys. 5], 1084 (1985). .
II. M. Ramsteiner, J. Wagner, H. Ennen, and M. Maier, Phys. Rev. B 38, 10669 (198
8
)
and references therein.
12. K. Jain, S. Lai, and M.V. Klein, Phys. Rev. B 13,5448 (1976).
288
13. K Wan and R. Bray, Phys. Rev. B:12., 5265 (1985).
14. J. Wagner, M. Ramsteiner, H. Seelewind, and J. Clark, J. Appl. Phys. 64, 802
(1988); J. Wagner, Physica Scripta 1'29, 167 (1989).
15. M. Cardona and G. Harbeke, J. Appl. Phys. 34, 813 (1962).
16. D.E. Aspnes and AA Studna, Phys. Rev. B 27, 985 (1983).
17. D. Gammon, R. Merlin, W.T. Masselink, and H. Morkoc, Phys. Rev. B 33, 2919
(1986).
18. KT. Tsen, J. KJem, and H. Morkoc, Solid State Commun. 59, 537 (1986).
19. P.O. Holtz, M. Sundaram, R. Simes, J.L. Merz, A.e. Gossard, and J.R. English,
Phys. Rev. B39,13293 (1989).
20. J. Wagner, M. Ramsteiner, R. Murray, and R.C. Newman, in: "Materials Science
Forum", Vols. 38-41, Trans Tech, Switzerland (1989), p. 815.
21. J. Wagner and M. Ramsteiner, IEEE J. Quantum 993 (1989).
22. T. Nakamura and T. Katoda, J. App!. 5870 (1982) and references therein.
23. J. Wagner and e.R. Fritzsche, J. Appl. Phys. 64, 808 (1988) and references therein.
24. J. Wagner, Appl. Phys. Lett. 1158 (1988).
25. A Lusson, J. Wagner, and M. Ramsteiner, Appl. Phys. Lett. 54, 1787 (1989).
26. KK Tiong, P.M. Amirtharaj. F.H. Pollak, and D.E. Aspnes, App!. Phys. Lett. 44,
122 (1984).
27. H. Richler, Z.P. Wang, and L. Ley, Solid State Commun. 22, 625 (1981).
28. R. Trommer and M. Cardona, Phys. Rev. BIZ, 1865 (1978).
29. D.E. Aspnes, S.M. Kelso, e.G. Olson, and D.W. Lynch, Phys. Rev. Lett. 48, 1863
(1982).
30. J. Wagner, W. Pletschen, and G. Kaufel, to be published.
31. For a recenl review, see: A. Forche!, H. Leier, B.E. Maile, and R. Germann, in:
"Advances in Solid State Physics", Vol. 28, Vieweg, Braunschweig (1988) p.99.
32. R.e. Newman, Mat. Res. Soc. Symp. Proc. Vol. 46, 459 (1985) and references
therein.
31 R. Murray, R.e. Newman, M.J.L. Sangster, R.B. Beall, J.J. Harris, PJ. Wright, J.
Wagner, and M. Ramsleiner, J. Appl. Phys. 66, 2589 (1989).
34. M. Holtz, R. Zallen, A.E. Geissberger, and R.A Sadler, J. Appl. Phys. ,l2, 1946
(1986).
35. T. Kamijoh, A Hashimoto, H. Takano, and M. Sakuta, J. App!. Phys. 59, 2382
(1986).
36. R. Ashokan, KP. Jain, H.S. Mavi, and M. Balkanski, J. Appl. Phys. 60, 1985 (1986).
37. J. Maguire, R. Murray, R.e. Newman, R.B. Beall, and JJ. Harris, Appl. Phys. Lett.
516 (1987).
38. H. Ono and R.e. Newman, J. Appl. Phys. 66, 141 (1989).
39. J.H. Neave, P.J. Dobson, U. Harris, P. Dawson, and B.A Joyce, Appl. Phys. A 11
195 (1983).
40. J. Wagner, H. Seelewind, and W. Jantz, J. Appl. Phys. fil., 1779 (1990).
41. T.F. Kuech, M.A Tischler, P.-J. Wang, G. Scilla, R. Potemski, and F. Cardone,
Appl. Phys. Lett. 53,1317 (1988).
42. K Saito, E. Tokumitsu, T. Akatsuka, M. Miyauchi, T. Yamada, M. Konagai, and K.
Takahashi, J. Appl. Phys. 64, 3975 (1988).
43. e.R. Abernathy, SJ. Peanon, R. Caruso, F. Ren, and J. Kovalchik, Appl. Phys.
Lett. 25., 1750 (1989).
44. J. Wagner and M. Weyers, to be published.
289
45. J. Wagner, M. Maier, R. Murray, R.C. Newman, R.B. Beall, and J.J. Harris, J.
Appl. Phys., to appear.
46. M.H. Grimsditch, D. Olego, and M. Cardona, Phys. Rev. B 20, 1758 (1979).
47. For a recent review, see: K. Ploog, M. Hauser, and A. Fischer, Appl. Phys. A~
233 (1988).
48. E.F. Schubert, J.B. Stark, B. Ullrich, and J.E. Cunningham, Appl. Phys. Leu. ~
1508 (1988).
49. F. Koch and A. Zrenner, Mater. Sci. Eng. B1, 221 (1989) and references therein.
50. A. Zrenner, F. Koch, R.J. Williams, R.A. Stradling, K. Ploog, and G. Weimann,
Semicond. Sci. Technol. J" 1203 (1988).
51. M. Santos, T. Sajoto, A. Zrenner, and M. Shayegan, Appl. Phys. Lett. ~ , 2504
(1988).
52. R.B. Beall, J.B. Clegg, and J.J. Harris, Semicond. Sci. Technol. J" 612 (1988).
53. J.B. Clegg and R.B. Beall, Surf. Interface Anal. 14,307 (1989).
54. A.-M. Lanzillotto, M. Santos, and M. Shayegan, Appl. Phys. Lett. 55, 1445 (1989).
55. A. Zrenner, F. Koch, and K. Ploog, Inst. Phys. Conf. Ser. 2.1, 171 (1988).
56. G. Ahstreiter, R. Merlin, and A. Pinczuk, IEEE J. Quantum Electron. no 1771
(1986).
57. J. Wagner, M. Ramsteiner, D. Richards, G. Fasol, and K. Ploog, Appl. Phys. Lett,
January 14 (1991).
290

You might also like