SURFACE AND INTERFACE ANALYSIS
Surf. Interface Anal. 2005; 37: 673–682
Published online 17 June 2005 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/sia.2062
Characterization of wet-etched GaAs (100) surfaces
M. Rei Vilar,1∗ J. El Beghdadi,1 F. Debontridder,1 R. Artzi,2 R. Naaman,2 A. M. Ferraria3
and A. M. Botelho do Rego3
1
2
3
ITODYS, CNRS-Université Denis Diderot, Paris, France
Chemical Physics Department, Weizmann Institute of Science, Rehovot, Israel
CQFM, Instituto Superior Técnico, Lisboa, Portugal
Received 22 November 2004; Revised 24 April 2005; Accepted 28 April 2005
To enable the use of GaAs-based devices as chemical sensors, their surfaces must be chemically modified.
Reproducible adsorption of molecules in the liquid phase on the GaAs surfaces requires controlled etching
procedures. Several analytical methods were applied, including Fourier transform infrared spectroscopy
(FTIRS) in attenuated total reflection and multiple internal reflection mode (ATR/MIR), high-resolution
electron energy loss spectroscopy (HREELS), X-ray photoelectron spectroscopy (XPS) and atomic force
microscopy (AFM) for the analysis of GaAs (100) samples treated with different wet-etching procedures.
The assignment of the different features due to surface oxides present in the vibrational and XPS spectra
was made by comparison with those of powdered oxides (Ga2 O3 , As2 O3 and As2 O5 ). The etching procedures
here described, namely, those using low concentration HF solutions, substantially decrease the amount
of arsenic oxides and aliphatic contaminants present in the GaAs (100) surfaces and completely remove
gallium oxides. The mean thickness of the surface oxide layer drops from 1.6 nm in the raw sample to
0.1 nm after etching. However, in presence of light, water dissolution of arsenic oxides is enhanced, and
oxidized species of gallium cover the surface. Copyright  2005 John Wiley & Sons, Ltd.
KEYWORDS: GaAs (100); oxide; wet-etching; ATR/MIR; XPS; HREELS; AFM
INTRODUCTION
GaAs (100) surfaces have a high surface energy.1 As a
consequence, they are very reactive and chemically unstable.
Chemical characterization of the different oxides present
on the GaAs (100) surface is problematic since only a few
attempts for identification of the oxides were performed by
infrared spectroscopy2 and high-resolution electron energy
loss spectroscopy (HREELS),3 and a great dispersion in the
X-ray photoelectron spectroscopy (XPS) values exists.4
In the present study, we provide detailed identification
and assignment of XPS and infrared spectral features of
the surface oxides by both using the data found in the
literature and comparing the spectra obtained with GaAs
(100) surfaces with those recorded with different powdered
oxides of gallium and arsenic.
The elimination of contaminants and oxide present on
the GaAs surface can be accomplished by the usual procedures of degreasing and etching. Degreasing is achieved
with acetone and ethanol.5 In the case of etching, several dry and wet methods have been used and are
described in the literature. In addition to those procedures
Ł Correspondence to: M. Rei Vilar, ITODYS, 1 rue Guy de la Brosse,
F-75005 Paris, France. E-mail: [email protected]
Contract/grant sponsor: EU Growth Programme.
Contract/grant sponsor: Financiamento Plurianual, FCT, Portugal.
Contract/grant sponsor: GRICES/Frech Embassy, Portugal;
Contract/grant number: GRD2-2000-30012 SENTIMATS.
performed under vacuum conditions, like the atomic hydrogen cleaning (AHC),6,7 other procedures were proposed
for wet etching. Cho8 pioneered the wet-etching procedures on GaAs surfaces aiming to introduce the treated
samples into ultrahigh vacuum (UHV) for further epitaxial growth. Currently, wet-etching methods are based on
various solutions of Br2 /methanol,9 H2 O2 /NH4 OH/H2 O,10
H2 SO4 /H2 O2 /H2 O,11 – 13 H2 SO4 /H2 O2 /HCl,14 or HCl/
H2 O.15 Another commonly used procedure to remove surface
oxides is based on aqueous solutions of hydrofluoric acid.5
This method has been successfully employed for removal of
oxides from GaAs surfaces.3,15
The results of all the above-mentioned methods, in terms
of surface composition and roughness, have never been
compared. In fact, these methods were mainly employed
for molecular beam epitaxy (MBE), where a second step of
oxide elimination should be performed by annealing the
GaAs samples under vacuum above the thermal cleaning
temperature of 580 ° C.8 In contrast, in the present work,
since we are interested in adsorption from liquid solutions
at ambient temperature, the second step is not performed.
In order to choose the most appropriate wet etching for
our purposes, we have studied wet-etched GaAs (100)
surfaces using some of the procedures described above.
Furthermore, the subsequent adsorption of molecules from
liquid solutions, the action of solvents, namely, of water, on
substrate composition and morphology had to be studied.
To obtain a complete characterization of the surface in
its different states, a set of techniques were applied. The
Copyright  2005 John Wiley & Sons, Ltd.
674
M. Rei Vilar et al.
methods include Fourier transform infrared spectroscopy
(FTIRS) in ATR/MIR, HREELS, and XPS. These spectroscopic
methods are complementary since they have different
sensitivities to the superficial region. Surface roughness was
estimated through atomic force microscopy (AFM).
EXPERIMENTAL
Chemicals
Undoped semi-insulating single crystal GaAs wafers with
orientation (100) were used. Diarsenic trioxide, As2 O3 ,
99% pure, diarsenic pentoxide, As2 O5 , 97% and digallium
trioxide, Ga2 O3 , 99.999% were obtained from Sigma-Aldrich.
Acetone and ethanol purchased from SDS were anhydric.
Hydrofluoric acid (40%) was acquired from Sigma-Aldrich.
Hydrochloric acid (37%), hydrogen peroxide (30%), sulphuric acid (98%) and ammonia (28%) were obtained from
Prolabo. All these chemicals were analytical grade and used
without further purification.
Deionized water (DIW) with 18.2 M cm resistivity was
produced by a Millipore deionising system previously fed
with distilled water.
Sample preparation
The GaAs samples were cleaned by sonication for 5 minutes
first in acetone, then in ethanol. The degreased samples
were subsequently etched using different procedures by
immersing them in aqueous solutions (composition is given
in volumic proportions):
(a) H2 SO4 : H2 O2 : H2 O (4 : 1 : 1 at 70 ° C) for 10 min followed
by HCl for 5 min then rinsed for 5 s;
(b) Modified RCA treatment: NH4 OH : H2 O2 : H2 O (1 : 1 :
100) for 5 min followed by rinsing for 5 s and by another
treatment of NH4 OH : H2 O2 : H2 O (1 : 1 : 30) for 5 min
then rinsed for 5 s;
(c) HCl : H2 O (1 : 1) for 1 min and rinsed for 1 min;
(d) HF : H2 O (different concentrations) for 5 s and rinsed for
3 s.
DIW was always used for the rinsing steps.
Surface oxidation was performed by immersing the
samples in H2 O2 for 30 s, and by rinsing them in DIW
for 2 min. Samples were then dried with a nitrogen flux and,
those analysed by HREELS and/or XPS transferred into the
UHV chambers under an argon atmosphere and introduced
under a nitrogen flux.
Special care was taken regarding the cleaning of glassware to avoid contamination and to prevent the presence of
ions from the detergent, such as NaC and phosphates. After
washing carefully, all glassware was rinsed for 1.5 hours in
several baths of boiling deionized water.
Apparatus
Major faces of the ATR/MIR samples 800 š 40 µm thick
were cut from the wafers in 40 mm ð 15 mm rectangles
and optically polished to 45° bevelled edges in an isosceles
trapezoidal configuration. This enables the transmitted
infrared beam to be subjected to about 25 total internal
reflections on each side of the sample. Infrared spectra were
Copyright  2005 John Wiley & Sons, Ltd.
recorded using a Magna-IR Nicolet 850 equipped with a
MCT detector. Spectral resolution was 4 cm1 .
Two independent UHV set-ups were used. The first one,
described elsewhere,16 is composed of the HREELS chamber
(¾108 Pa) and a preparation chamber equipped with an
ion gun and an Auger spectrometer. The second UHV setup (¾107 Pa) contains the XPS spectrometer, a XSAM800
(KRATOS).
HREELS spectra were recorded using a double pass
spectrometer, Leybold-Heraeus ELS-22 described before.17
Spectral resolution, as defined by the FWHM of the elastic
peak, is 3 meV (¾24 cm1 in a straight beam and around
10 meV after the electron interaction with the sample surface.
Spectra were recorded using 5 eV incident electrons. The
electron energy losses ranged from 1000 to 5000 cm1 .
The sweeping voltage step was 0.34 meV (2.7 cm1 . Current
intensity across the sample was in the order of 1011 A.
Samples were analysed by XPS, using 90° and 30°
take-off angles (TOA) relative to the surface. Unmonochromatized Mg K˛ and Al K˛ X-radiations (h D 1253.6 eV and
1486.6 eV, respectively) with a power of 10 mA ð 13 kV D
130 W were used. The electron energy analyser was operated
in the fixed analyser transmission (FAT) mode using a pass
energy of 10 eV. Spectra were recorded by a Sun SPARC
Station 4 with Vision software (KRATOS) using sweep steps
of 0.1 eV. No flood gun was used for charge neutralization.
The curve fitting for component peaks was carried out with a
non-linear least-squares algorithm using Voigt profiles and a
Shirley background subtraction. For quantification purposes,
sensitivity factors, provided by KRATOS, were 0.66 for O 1s,
0.25 for C 1s, 0.53 for As 3d, 6.3 for As 2p3/2 , 0.31 for Ga 3d
and 4.74 for Ga 2p3/2 .
In the first UHV set-up, one of the samples was sputtered
under a 500-eV argon ion beam for 5 min in ultrahigh vacuum
(UHV), using a RIBER CI10 ion gun. The current through the
sample was around 50 µA. Auger electron spectroscopy was
used for the analysis of elements present on the substrate,
particularly those existing in carbonaceous contamination
(KLL peaks located at 278 eV) and in the oxide (KLL peaks
located at 508 eV).
AFM measurements (NT-MDT P47) were carried out
with NSC-12 tips. Semi-contact mode was used to evaluate
the morphology of the surface.
All the analyses were performed at room temperature.
RESULTS AND DISCUSSION
In order to identify the oxides present on the GaAs surface,
some commercially available, powdered oxides of gallium
and arsenic were investigated using FTIRS and XPS. The
comparison between spectra of the powders and those of
GaAs surface helps to identify the stoichiometry of the
surface oxides.
The spectra of KBr pellets containing Ga2 O3 , As2 O3 and
As2 O5 are shown in Fig. 1. Digallium trioxide spectrum
shows a broad band centred at 670 cm1 containing a
shoulder at around 760 cm1 . The positions of both bands
are in agreement with data obtained for polycrystalline ˇGa2 O3. 18 Both spectra of As2 O3 and As2 O5 are dominated
Surf. Interface Anal. 2005; 37: 673–682
Characterization of etched GaAs surfaces
802
1.2
Absorbance
670
760
840
0.8
913
Ga2O3
0.4
0.0
1050
As2O3
As2O5
500
750
1000
1250
Wavenumber, cm−1
Figure 1. FTIRS spectra of powdered digallium trioxide (bold
line), diarsenic trioxide (line) and diarsenic pentoxide (dotted
line).
for As2 O3 and for Ga2 O3 at 531.8 eV, which was used as
a reference for As2 O5 . These results agree with previous
studies indicating that the surface oxide layer is not a
simple mixture of gallium and arsenic oxides, but should
also contain a significant amount of mixed oxides of gallium
and arsenic, including GaO4 and AsO3 units linked by oxygen
atoms.4,11
The removal of most of the organic contaminants
covering the surface of GaAs (100) samples can be achieved
by sonication of the semiconducting wafers in different
organic solvents. After degreasing, the ATR/MIR spectrum
of the GaAs element shown in Fig. 2 exhibits negative
absorption peaks corresponding to the symmetric and
asymmetric CH stretching modes of methylene groups
of aliphatic contaminants located at 2853 and 2925 cm1 ,
respectively, and bending modes at 1462 cm1 . These
767
5.0 × 10
Absorbance
by a peak at 802 cm1 with a shoulder at around 760 cm1 .
As2 O3 spectrum contains a weak shoulder at 840 cm1 and
an isolated band of medium intensity at 1050 cm1 .
In the same way, Ga2 O3 , As2 O3 and As2 O5 powders were
analysed by XPS, prior to the characterization of raw (as
received) GaAs (100) surfaces. The different assignments
of the binding energies obtained for these powders are
displayed in Table 1 and compared with those obtained for
the native GaAs oxides. In GaAs samples, binding energies
were corrected by using that of Ga 3d5/2 (D19.1 eV), an
average value obtained from the literature.3,19 For As2 O3
and Ga2 O3 , the C 1s peak of carbonaceous contamination
(binding energy D 285 eV) was used. However, in the case
of As2 O5 the contamination level was too low to be detected.
But, as for the oxide mixture, only an O 1s peak was obtained
3735
−3
0.0
1734
1462
−5.0 ×10−3
−1.0 × 10−2
2853
1061
−1.5 × 10−2
2925
1000
2000
3000
4000
Wavenumber, cm−1
Figure 2. ATR/MIR spectrum of a GaAs (100) surface after
degreasing. Negative and positive peaks are associated with
the removal and with the appearance of different species,
respectively.
Table 1. XPS assignments for arsenic and gallium in GaAs (100) surfaces covered with the native oxide and in powders, compared
with those found in the literature. Values in parentheses correspond to oxides in 1% HF etched surfaces (when present) and different
from those found in non-etched surfaces
Binding energy (eV)
Compound
Photoelectron
Ga2 O3
Ga 2p3/2
Ga 3d5/2
O 1s
As 2p3/2
As 3d5/2
O 1s
As 2p3/2
As 3d5/2
O 1s
Ga 2p3/2
Ga 3d5/2
As 2p3/2
As 3d5/2
O 1s
As2 O3
As2 O5
GaAs native oxide
4,12,19,20
Literat.
1116.9–1118.9
19.7–20.9
530.7–532
1326.6–1327.3
44.9 (43.9–46.3)a
531.6–532.3
1328
45.9–46.5a
531.6–532.3
1117.5–1119
19.6–21.6
1325.9–1327.0
42.0–45.8
529.8–532.3
This work
1118.7 š 0.2
0.7 š 0.2
531.8 š 0.2
1326.9 š 0.2
44.9 š 0.2
531.8 š 0.2
1326.1 š 0.2 and 1327.9 š 0.2b
45.5 š 0.2 and 47.4 š 0.2
531.8 š 0.2
1118.9 š 0.2
20.5 š 0.2
1327.0 š 0.2 (1326.1 š 0.2)
44.2 š 0.2 (43.1 š 0.2)
531.2 š 0.2
a
Values for As 3d without resolving into 3d5/2 and 3d3/2 .
As2 O5 is very hygroscopic; the high O/As ratio (D3.38) indicates that As2 O5 mixes with water forming species richer in oxygen, like
hydroxides.
b
Copyright  2005 John Wiley & Sons, Ltd.
Surf. Interface Anal. 2005; 37: 673–682
675
M. Rei Vilar et al.
negative peaks confirm the removal of surface aliphatic
contamination. Another negative peak at 1734 cm1 is most
likely related to the disappearance of carboxylic species. It is
also interesting that some oxide is already etched during the
degreasing process, as shown by the negative peak located at
1061 cm1 . At 767 cm1 , a narrow positive peak, also found
in the KBr pellets spectra of all powdered oxides, appears.
It can be assigned to both Ga–O and As–O vibrations.
At higher wavenumbers, there is a positive narrow peak
located at 3735 cm1 and a broad band with a maximum
centred around 3660 cm1 , associated with free and vicinal
hydroxyl groups, respectively. This is a consequence from
the emergence of As–OH and/or Ga–OH groups at the
surface, now liberated from the bound water, as shown by
the corresponding negative broad band around 3300 cm1 .
These results agree with those obtained by XPS.
Table 2 shows the quantitative analysis of the GaAs (100)
surface after degreasing, as obtained by XPS. The relative
atomic ratios, Y, were obtained from the XPS peak-areas
divided by the sensitivity factors, and oxide-film average
thicknesses were estimated considering a flat oxide layer
adsorbed on a flat GaAs surface and fitting the ratio
computed by Eqn. (1) for both X-radiation sources to the
experimental XPS ratios.
nYox 1 exploxide /Y2p or 3d sin Yox
2p or 3d D
Ynox
nYnox exploxide /Y2p or 3d sin 1
where Y is Ga or As, loxide is the oxide layer thickness,
Y2p or 3d is the inelastic mean free path for a given
photoelectron, excited by a given X-ray anode that was
calculated from the expression TPP-2M (the modified Bethe
equation) proposed by Tanuma et al.,21 and is the TOA. The
n
density ratio n Yox is hard to estimate since nYox depends on
Ynox
the composition of the oxide. Moreover, it is not certain that
the density of the oxide on the GaAs surface is the same as
the density of free oxide. It is, then, a fitted parameter and
its value varied from 0.6 to 1.0 (nGa (in free Ga2 O3 /nGa (in
GaAs) D1.79 and nAs (in free As2 O3 /nAs (in GaAs) D 1.03)).
These results show that cleaning in acetone or in ethanol
effectively removes one-third of the amount of carbonaceous
contamination on the surface. Comparatively, two thirds of
the contamination is removed if both solvents are employed.
However, the role of degreasing goes beyond the removal of
the carbon contaminants, as indicated by a strong decrease
of arsenic and gallium contributions. The oxide-film average
thickness decreases from 1.6 š 0.3 to 1.2 š 0.2. A decrease
in the arsenic contribution was also detected in ATR/MIR
spectra of the degreased element (refer Fig. 2). XPS also
reveals that degreased samples are richer in arsenic at
the extreme surface compared to the unprocessed ones,
indicating that gallium was removed or redistributed.
For the assignment of the oxides of GaAs (100) surfaces,
samples were oxidized purposely using hydrogen peroxide
and then etched with HF 40%. Since the penetration depth of
the evanescent wave in the ATR/MIR experiment is of the
order of 1000 nm, the analyses are sensitive both to surface
and bulk oxides. The corresponding spectra are presented in
Fig. 3. Spectrum a is that of the oxidized GaAs (100). Peaks
at 846 and 1058 cm1 , also identified in powdered As2 O3 ,
are present, indicating that arsenic oxide is preferentially
formed at the surface. Two intense positive bands of water,
a broad band centred at 3250 cm1 and a narrower one
located at 1620 cm1 , are due to hydrogen-bonding among
adsorbed water molecules on the surface. Spectrum b shows
that oxides are efficiently removed by etching the substrate
with HF 40%. The peaks at 846 and 1058 cm1 corresponding
to the arsenic oxide are now negative with about the same
intensity as in a. The disappearance of the surface oxide layer
also implies that the bound water disappears, leaving the
hydroxyl groups liberated. As a result, at 3733 cm1 , one
can distinguish the characteristic peak of the OH stretching
corresponding to free hydroxyl groups. The disappearance
of water from the surface is corroborated by the negative
bands, with maximum intensities situated around 1620 and
3250 cm1 . Spectrum c, which is the sum of spectra a and
b, shows that the oxides resulting from the oxidation are
completely removed. As a matter of fact, the peak at 846 cm1
completely disappears and the negative peak at 1058 cm1
confirms the removal of a part of the native oxide.
oxides
690 767
846
Table 2. XPS quantitative analysis of the degreasing effect of
acetone and/or ethanol on raw GaAs (100) surfaces. XPS ratios
presented were obtained with Al K˛ radiation at TOA D 90°
except the last row which contains values computed from data
obtained with both Al K˛ and Mg K˛ radiations
Ethanol
Acetone
C ethanol
0.3
0.2
0.2
0.7
0.6
0.5
1.0
1.6
1.5
3.7
2.1
2.0
1.2
0.7
0.9
1.6 š 0.3 1.2 š 0.2 1.2 š 0.2
0.1
0.6
1.3
2.2
0.9
0.9 š 0.3
Raw
C/(Ga3d C As3d)
O/(Ga3d C As3d)
As/Ga (2p)
Asox /Asnox (2p)
Gaox /Ganox (2p)
Oxide film average
thickness (nm)
Acetone
Copyright  2005 John Wiley & Sons, Ltd.
bound water
1058
free hydroxyl
groups
3250
1620
1457
0.04
Absorbance
676
a
3733
c
0.00
b
−0.04
−0.08
1000
2000
Wavenumber,
3000
4000
cm−1
Figure 3. ATR/MIR spectra of a GaAs (100) surface: a – after
oxidation with H2 O2 (black); (b – after HF etching (black bold);
c – sum of both spectra (grey).
Surf. Interface Anal. 2005; 37: 673–682
Characterization of etched GaAs surfaces
Different wet-etching processes were tested and analysed using different analysis techniques. The classical
Br2 –methanol etching was not included in this study, as
it ostensibly increases the surface roughness to a degree visible to the naked eye. Figure 4 compares the HREELS spectra
of raw GaAs (100) surfaces, etched by the wet-etching methods described in the experimental section, and by argon ion
sputtering under UHV. All spectra exhibit the characteristic peak located on both sides of the elastic peak around
290 cm1 . These peaks, corresponding to the electron energy
loss and the energy gain due to the surface phonons constitute an unambiguous signature of the GaAs (100) surface.
Energy losses associated with this optical surface phonon in
sputter-annealed GaAs (100) surfaces have been described in
the literature.22,23
In the case of the sample treated by method a, the energy
loss peak due to the phonon almost vanishes, revealing the
destruction of the crystallinity of the GaAs surface, where
undulations are visible to the naked eye. In contrast, the
other wet-etching methods do not significantly alter the
smoothness of the optical polished GaAs (100) surfaces. In
all HREELS studies performed on etched wafers, electrons
are backscattered by the surface in broad lobes of FWHM
values between 20° and 40° , indicating that electrons are
backscattered quasi isotropically.
A band corresponding to the electron energy loss due
to a double scattering with the phonon is detected at
580 cm1 . The background level in the spectra, usually
assigned to multiple scattering, is different in each of the
recorded spectra. It is the strongest for the sample treated
with procedure a, second in intensity is the spectrum
taken with the raw GaAs sample and the weakest for
the sample sputtered under vacuum. All spectra reveal
oxides and contamination present on the sample surface.
By far, the argon sputtered sample is the least oxidized
or contaminated one. Characteristic peaks present in most
spectra corresponding to energy losses of 800 and 1060 cm1 ,
are assigned to the oxides. A small shoulder around 750 cm1
broadens the peak at 800 cm1 . Skeletal C–C stretching
Raman active modes from aliphatic contaminants can also
be found between 1037 and 1168 cm1 .24 These peaks are
overlapping with those related to the arsenic oxide and
therefore prevent the use of this region of the spectrum to
quantify the content of arsenic oxide. The presence of arsenic
oxide can be deduced from the band at 800 cm1 . Aliphatic
contaminants can be clearly identified through the typical
bending and stretching CH modes appearing in the regions
around 1425 and 2900 cm1 , respectively.
In the case of the sputtered sample, we monitored the
surface contamination process under UHV conditions. The
two spectra, displayed in Fig. 5, illustrate the evolution after
a 4-day and an 8-day period. Both spectra have very low
intensities and the inelastic losses were multiplied by a
factor of 1600. After a 4-day storage in the UHV, aliphatic
contaminants appear on the ion-etched surface, but only
a small oxide signal is detected. After an 8-day storage,
the initiation of surface oxidation by the residual oxygen is
clearly evident through the appearance of features located at
800 and 1060 cm1 . Aliphatic contamination also increases.
This spectrum reveals that oxidation occurs even under UHV
conditions because of the high reactivity of the clean GaAs
(100) surfaces.
Table 3 presents the peak positions in vibrational spectra
of GaAs surfaces obtained using ATR/MIR, and HREELS.
These peaks also appear in spectra of As2 O3 and As2 O5 .
However, the peak at 913 cm1 appears in the As2 O5
spectrum but not in ATR/MIR or HREELS spectra. This
leads us to exclude the existence of As2 O5 units on the
surface.
By comparing the different spectra displayed in Fig. 4, it
is evident that all features diminish after wet etching. It is
also clear that spectra of etched samples (excluding the one
treated with method a appear to have less intense features
associated with the oxides and contamination.
The HREELS spectra confirm the efficiency of the
wet-etching methods for removing oxides and organic
−290 290
1385
800
580 1060
0.20
1060
2900
800
2900
0.15
580
a
0.10
raw
0.05
Intensity, a.u.
Normalized Intensity, a.u.
1425
b
d
c
Ar +
0.00
0
1000
2000
Energy loss,
3000
4000
cm−1
Figure 4. HREELS spectra of raw GaAs (100) compared with
wet-etched samples using methods a, b, c, d (see
experimental section) and after argon ion sputtering. Spectra
are normalized to the elastic peak intensity taken as unity.
Copyright  2005 John Wiley & Sons, Ltd.
0
1000
2000
3000
4000
Energy loss, cm−1
Figure 5. HREELS spectra of a clean GaAs (100) surface
stored in the UHV chamber for a 4-day period (bold black) and
an 8-day period (grey). Losses were magnified by a factor of
1600 relative to the elastic peak.
Surf. Interface Anal. 2005; 37: 673–682
677
678
M. Rei Vilar et al.
Table 3. Peak positions of oxides found in powders and on the GaAs (100) surface
detected by infrared and HREELS spectroscopy compared with those described in the
literature
Wavenumber (cm1 This work
Literature
2,25,26
Powders
690
760
826
843
896
1058
1390–1370
1465–1440
2870–2840
2940–2915
ATR/MIR
HREELS
768
760
800
670
760
802
840
913
1050
846
1061
1462
2853
2925
Table 4. Comparison of the efficiency of different etching
procedures based on HREELS and AFM data. In parentheses,
absolute values of the different contributions are shown for the
raw sample. All the other contributions presented are
normalized to the raw sample
HREELS
Wavenumber/cm1
Samples
NH4 OH
HCl
HF
ArC
sputtered
AFM
Area
690
800
1060
2900
RMS (nm)
1.00
0.59
0.60
0.54
0.28
(1.76)
1.00
0.59
0.17
1.46
0.79
(6.62)
1.00
1.22
0.33
0.71
0.13
(8.17)
1.00
0.68
0.68
0.64
0.10
1.00
0.43
0.59
0.49
0.05
0.2 š 0.02
0.3 š 0.03
0.3 š 0.03
0.3 š 0.03
–
Raw
1060
1385
1425
2900
contamination. An attempt of a quantitative analysis of these
spectra, excluding that of the sample treated by the method a,
is presented in Table 4. The electron backscattering in impact
geometry is mostly caused here by the sample roughness
and/or by the presence of oxides and contaminants on
the sample. Consequently, the quality of the GaAs surface
can be evaluated from the backscattered electron intensity
calculated by the integral of the electron intensity between 0
and 4000 cm1 . As expected, the argon-sputtered sample has
the lowest value. The HF (1%) etched samples present the
best quality among the wet-etched samples, followed by the
Assignments2,25,26
Ga–O
Ga–O and As–O
As–O
As–O
As–O (As2 O5 )
As–O (As2 O3 )
Sym. def. CH2
Asym. def. CH2
Sym. stretching CH2
Asym. stretching CH2
HCl- (18.5%) and NH4 OH- treated samples. From the AFM
images obtained using the different methods, also excluding
method a, we conclude that roughness (rms values) is very
low for all samples and is of the order of 0.3 nm. This value
is comparable with that obtained for the raw sample (see
Table 4).
The relative concentration of the different species on each
sample was obtained through the analysis of the spectra
presented in Fig. 4, using a fitting based on a Gaussian sum.
The FWHM of the contributions was fixed to that of the
elastic peak and their positions are presented in Table 3.
Those related to the oxides appear at 690 cm1 , 800 cm1 and
1060 cm1 . The background multiple scattering level can be
well fitted by a decreasing exponential. In Table 4, these
values are compared relatively to the raw sample, for which
the contribution of the backscattered electron intensity in
each region was normalized to unity. Absolute values of
the different contributions in the raw sample are presented
in parentheses in Table 4. One can see that the absolute
contribution of gallium oxide (peak at 690 cm1 is very
low. Arsenic oxide is tenaciously present in all samples but
especially weak in the HCl-treated ones. Concerning surface
contamination, values shown in Table 4 were determined
through the energy loss peak related to the CH aliphatic
stretching mode corresponding to the band centred at
2900 cm1 . Here, the efficiency of the NH4 OH followed by HF
treatment is evident. In a global view, one can consider that
the HF wet-etching procedure is clearly the preferred one.
XPS results were obtained on samples treated with
identical processes and are presented in Table 5. Average
Table 5. Comparison of the efficiency of different etching procedures based on XPS data. Ratios presented were obtained with Al
K˛ radiation at TOA D 30° to enhance surface sensitivity
Samples and etchings
Raw
NH4 OH
HCl
HF
Gaox /Ganox (2p)
Asox /Asnox (2p)
O/(As C Ga) (2p)
C/(Ga C As) (2p)
Oxide average thickness (nm)
3.04
0
0
0
4.7
0.35
0.09
0.3
0.88
0.21
0.21
0.31
0.8
0.43
0.55
0.62
1.60 š 0.30
0.11 š 0.04
0.07 š 0.05
0.11 š 0.02
Copyright  2005 John Wiley & Sons, Ltd.
Surf. Interface Anal. 2005; 37: 673–682
Characterization of etched GaAs surfaces
thicknesses were estimated from Eqn. (1) applied to arsenic.
After etching, no gallium oxide is detected and the arsenic
oxide concentration also drops to very low values in all
samples. Carbon contamination significantly diminishes as
well. The average thickness of the oxide layer, of the order of
1.6 nm for the raw sample, substantially decreases to about
0.1 nm after etching.
In Fig. 6, the 2p and 3d XPS spectral regions of arsenic
and gallium are displayed for the unprocessed wafer and
that etched with 1% HF. It is clear that the etched sample
is less oxidized than the unprocessed one. The HF etching
treatment leaves the surface free of oxidized gallium species
and strongly diminishes the amount of oxidized arsenic
species: the ratio between oxidized and non-oxidized species
of arsenic, Asox /Asnox , in the region 2p3/2 , drops from 4.7
for the unprocessed sample to 0.3 after etching. In the
case of the unprocessed sample, the peak in the gallium
region is fitted with 2 components: the non-oxidized gallium
(i.e. Ga (-As)) at 1117.8 š 0.2 eV and that at 1118.9 š 0.2 eV
assignable to Ga2 O3 units (Table 1). After etching, the Ga
2p3/2 region is fitted with just one peak corresponding to a
non-oxidized gallium. In the arsenic spectral region, the nonoxidized arsenic (i.e. As (-Ga)) appears at 1323.8 š 0.2 eV for
both samples. However, the binding energy of the oxidized
arsenic changes with the etching treatment: for the raw
sample the oxidized peak is centred at 1327.0 š 0.2 eV, but for
the etched sample it is lower by 0.9 eV (1326.1 š 0.2 eV). This
means that before etching the surface is mainly composed of
As2 O3 (and Ga2 O3 ). Etching completely removes the Ga2 O3
and most of the As2 O3 , but the remaining oxidized arsenic
seems to be a mixture of arsenic oxide in different oxidation
states. Peaks FWHM and Lorentzian percentage were: As
2p: 2.05 š 0.15 eV, 72 š 2%; Ga 2p: 1.60 š 0.1 eV, 86 š 0.1%.
The same conclusions are arrived at from the analysis of
3d regions. 3d regions are fitted with doublets. The spin
orbit splitting of As and Ga 3d regions is 0.7 and 0.44 eV,
respectively.19 The peaks FWHM and Lorentzian percentage
were: As 3d: 1.15 š 0.05 eV, 63 š 15%; Ga 3d: 1.10 š 0.05 eV,
85 š 1%. Peak positions of Ga 3d5/2 and As3d5/2 in GaAs
were for the raw surface 19.1 š 0.2 eV and 41.0 š 0.2 eV,
respectively; and for the corresponding oxide 20.5 š 0.2 eV
and 44.2 š 0.2 eV, respectively. For the etched sample, the
GaAs components are in the same position as in the raw
surface. No oxidized Ga 3d component is present and the As
3d5/2 is lower by 1.1 eV.
The effect of the HF concentration in the etching solution
was also monitored by XPS. In Table 6, the composition
of GaAs surfaces is presented for substrates subjected to
HF etching treatments using different dilutions: 1%, 5%
and 10%. The results are based on the 3d and 2p spectral
regions. In the 2p region, the photoelectrons have lower
kinetic energies and therefore, the spectrum is more surfacesensitive. In contrast, in the 3d region, the photoelectrons
have higher kinetic energy and consequently can escape
from deeper layers. As expected, the Asox /Asnox (3d) values
are always lower than those calculated from the 2p region
and those of Asox /Asnox at a TOA of 30° are higher than
those obtained at a TOA of 90° . With the increasing of the
HF concentration, at a TOA of 90° , one can observe a slight
decrease of the ratio Asox /Asnox (2p) indicating a higher
efficiency of etching for more concentrated solutions. On
the other hand, for TOA D 30° , Asox /Asnox (2p) does not
1333
Intensity, a.u.
Ga 2p3/2
Intensity, a.u.
As 2p3/2
1328
1323
Binding Energy, eV
As 3d5/2
Ga 3d3/2
1112
Ga 3d5/2
Intensity, a.u.
Intensity, a.u.
As 3d3/2
1121
1118
1115
Binding Energy, eV
1318
46
43
40
Binding Energy, eV
37
24
21
18
Binding Energy, eV
15
Figure 6. Arsenic and gallium XPS regions (2p3/2 and 3d) in GaAs samples: raw (full symbols) and treated with a 1% HF (empty
symbols). 3d regions are fitted with doublets. Only the doublets of the GaAs components are labelled for the sake of clarity. Spectra
were recorded using AlK˛ radiation and a TOA of 90° .
Copyright  2005 John Wiley & Sons, Ltd.
Surf. Interface Anal. 2005; 37: 673–682
679
M. Rei Vilar et al.
Table 6. Composition of GaAs surfaces submitted to treatments with etching aqueous solutions of hydrofluoric acid of different
concentrations. Values were computed from the XPS 3d and 2p regions using Al K˛ radiation
90°
TOA
HF%
Asox /Asnox (2p)
Asox /Asnox (3d)
As/Ga (2p)
As/Ga (3d)
O/(Ga 3d+As 3d)
O/Asox (3d)
5%
10%
1%
5%
10%
0.54 š 0.06
0.10 š 0.01
1.7 š 0.2
0.89 š 0.09
0.13 š 0.02
3.0 š 0.3
0.42 š 0.05
0.12 š 0.02
1.7 š 0.2
0.91 š 0.09
0.09 š 0.01
1.9 š 0.2
0.36 š 0.04
0.10 š 0.01
1.6 š 0.2
0.89 š 0.09
0.15 š 0.02
3.6 š 0.4
0.92 š 0.1
0.19 š 0.02
1.90 š 0.2
0.90 š 0.09
0.24 š 0.03
3.2 š 0.3
0.73 š 0.08
0.20 š 0.02
1.9 š 0.2
0.95 š 0.1
0.17 š 0.02
2.0 š 0.2
0.83 š 0.09
0.17 š 0.02
1.9 š 0.2
1.02 š 0.1
0.16 š 0.02
2.1 š 0.2
exhibit the same behaviour with the HF concentration. This
could be explained by shadow effects due to the increase
of the roughness at higher HF content. This is in agreement
with other studies on the etching of GaAs surfaces using HF
solutions.27 Table 6 also shows that the ratio As/Ga (2p) is
always greater than 1, and that it increases for lower TOA (of
30° ), indicating that arsenic is more superficial than gallium.
One should expect the ratio As/Ga (3d) to be lower but
never to be less than 1, the stoichiometric factor of GaAs.
This indicates that the surface enrichment in arsenic results
from its depletion underneath the top oxide layers, and
that in addition, some arsenic disappeared from the surface
during etching.
An additional observation is that the ratio O/Asox does
not follow the decrease of Asox /Asnox (2p) for samples treated
with different concentrations of HF. This can be rationalized
by analysing the oxygen peak present in Fig. 7. In fact, for the
10% HF etching, one can observe, besides the oxygen feature
centred at 531 š 0.2 eV, another component at 532.8 š 0.2 eV.
This component, which is much less important in the spectra
of samples treated with lower HF concentrations, is assigned
to hydroxyl groups28 bound to arsenic pendant groups. These
arseniols formed on the GaAs surfaces should be AsOH or
537
534
30°
1%
Intensity, a.u.
680
531
528
525
Binding Energy, eV
Figure 7. XPS O 1s region for GaAs substrate etched in
different diluted solutions: 1% HF (empty squares), 10% HF
(full triangles). Peaks FWHM and Lorentzian percentage were:
1.90 š 0.2 eV, (45 š 13) %. Spectra were acquired with MgK˛
radiation and a TOA of 90° .
Copyright  2005 John Wiley & Sons, Ltd.
even As(OH)2 . The existence of these species was already
observed in the ATR/MIR spectra of etched samples (Figs. 2
and 3). The assignment of this XPS feature to hydroxyl
groups could explain the fact that the ratio O/Asox is larger
in the 10% HF treated sample than those obtained for lower
HF concentrations, since the O/As stoichiometric ratio in the
arseniol groups is higher than that in the oxide. The decrease
of the hydroxyl contribution from TOA 90° to 30° indicates
that OH groups are not located at the extreme surface.
This suggests a rough surface where hydroxyl groups are
preferentially buried in deep valleys. This is consistent with
the fact that HREELS spectra recorded in impact regime,
where the depth of analysis is the lowest, do not exhibit any
peak corresponding to stretching modes of hydroxyl groups.
For various applications GaAs surfaces are modified, by
adsorbing different molecules from solutions.27 Typically,
these solutions are very dilute and therefore the samples
are mainly in contact with the solvents. We investigated
the effect of solvents, and particularly that of water, on the
surface morphology of the sample. Etched samples were
immersed either in water or in a solvent composed of 10%
water and 90% acetonitrile (ACN90%). Figure 8 presents the
surface morphology of samples etched in a 1% HF solution
(a), then being immersed for 4 h in water (b) or in ACN90%
(c). The value of the roughness (rms) for the 1% HF etched
surface is 0.3 š 0.03 nm (Table 4). This value is in agreement
with previous results measured by Adachi and Kikuchi.20
These findings indicate that the use of low concentration HF
solutions results in relatively smooth surfaces. The roughness
of samples immersed in pure water is 1.9 š 0.2 nm (Fig. 8b),
comparatively higher than 1.3 š 0.1 nm obtained for samples
immersed in ACN 90% solution (Fig. 8(c)). The roughness
of these samples increases relative to that of samples that
were only etched. Such an increase can be attributed to
the re-oxidation of the etched sample and a subsequent
selective dissolution of the formed oxides when the samples
are immersed in solutions containing water. This increase in
the roughness should be considered as a possible source
for inhomogeneity in adsorbed organic layers on these
substrates.
Figure 9 displays two series of ATR/MIR spectra
recorded on samples previously oxidized, and then left in
water for a certain amount of time. In presence of light,
the dissolution of arsenic oxide is enhanced, as the intensity
of the peaks located at 848 and 1060 cm1 diminishes. At
the same time, there is an increase in the intensity of the
Surf. Interface Anal. 2005; 37: 673–682
Characterization of etched GaAs surfaces
(a)
(a)
848
Absorbance
0.06
(b)
1058
1018 1089
1161
760
0.04
3′
8′
20′
30′
45′
75′
0.02
0.00
800
1000
1200
Wave number, cm−1
1060
(b)
850
0.08
Absorbance
(c)
Figure 8. AFM images of GaAs (100) surfaces etched with:
1%HF (a); 1%HF followed by a 4 hour immersion in water
(b) 1%HF followed by a 4 hour immersion in a solution
containing 90% acetonitrile and 10% water (c).
Normalized Intensity, a.u.
1h25
45′
30′
20′
12′
3′
800
1000
1200
Wave number, cm−1
Ga 2p3/2
1122
0.04
0.00
band located at 760 cm1 , attesting to a further oxidation of
gallium atoms. In contrast, in the dark, spectra do not practically evolve even after 1 h and 25 min. It is interesting to note
that the broad band located around 1060 cm1 assigned to
arsenic oxide becomes structured as the sample is left in the
water in presence of light. This band can be fitted by a Gaussian sum containing five contributions centred at 990, 1016,
1058, 1108 and 1170 cm1 . All these contributions correspond
to different oxidation states of the surface. The contribution
centred at 1058 cm1 , corresponding to As2 O3 in the powder
spectrum (Fig. 1) is the most affected and strongly decreases
after immersion in water during 1 h and 15 min in the presence of light. In contrast, the other contributions change less.
This strongly suggests a preferential dissolution of the less
1125
760
Figure 9. ATR/MIR spectrum of a GaAs (100) sample oxidized
and immersed in deionized water for different periods under
white light (a) and in darkness (b). Spectra are off-set for clarity
of presentation.
oxidized arsenic species, in accordance with the following
XPS results.
XPS spectra of As and Ga 2p3/2 regions for samples that
were in contact with water (after being degreased and etched)
are displayed in Fig. 10. One of the samples was left in water
for 1 h in the dark, the second left for 1 h in water under
white light. Finally, a third one, previously oxidized, was left
in water 6 h under white light. One can see in spectral regions
corresponding both to gallium and arsenic that the etched
sample that was kept in the dark, is less oxidized than the one
kept under illumination. However, the relative amount of
As 2p3/2
1119
1116
Binding Energy, eV
1113
1331
1328
1325
1322
1319
Binding Energy, eV
Figure 10. XPS As 2p3/2 and GaAs 2p3/2 regions for samples etched with an aqueous solution of 40% HF and immersed in
deionized water. From top to bottom: for 1 h in darkness, for 1 h under white light and for 6 h under white light. Peaks FWHM and
Lorentzian percentage for all the samples: Ga 2p: 1.94 š 0.05 eV, 76 š 4%; As 2p: 2.20 š 0.15, 35 š 2%. AlK˛ radiation and TOA of
90° were used. Spectra were off-set for clarity of presentation.
Copyright  2005 John Wiley & Sons, Ltd.
Surf. Interface Anal. 2005; 37: 673–682
681
682
M. Rei Vilar et al.
oxidized species is always much larger for gallium than for
arsenic. The gallium binding energies do not change with the
time of interaction or with the illumination conditions: The
non-oxidized gallium appears at 1117.8 eV and the oxidized
one at 1119.4 eV, which is 0.5 eV higher than the value found
for the native oxide. This may be due to the incorporation
of water molecules in the oxide. However, for the arsenic,
the binding energies of the oxidized species show that their
quality varies during the oxidation. In spectra of samples
which were in contact with water in the dark, after 1 hour
interaction, for sample A, the As 2p3/2 region is fitted with a
single peak centred at 1326.8 eV, which is 2.7 eV higher than
the non-oxidized form, and assignable to As2 O3 (Table 1).
However, for the sample, which was in contact with water
under light for the same time interval, at least two peaks
appear at 1326.5 and 1328.4 eV exhibiting the same distance
as the two peaks found for As2 O5 (Table 1) but centred
at binding energies 0.4–0.5 eV higher. The reason for this
deviation may be the same as in the case of the gallium
oxide. For the sample which was previously oxidized and
left in water for 6 hours under light, the arsenic peak is
very small and could be fitted by a single peak centred at
1327.7 eV, assignable to As2 O5 . As the non-oxidized species
are practically not detected in the 2p and in the 3d regions,
one should conclude that the thickness of the oxide layer
after immersion in water and under white light must be
larger than 9 nm (around three free mean paths of the 3d
electrons).
CONCLUSIONS
The goal of the present study was to establish a preparation
method required to ensure good adsorption of organic
layers on GaAs (100) surfaces for sensor technology. We
focussed this work on the comparison of different etching
procedures and establishing the surface quality both in
terms of its chemical composition and morphology. The
GaAs samples were studied by different methods, and the
ATR/MIR and XPS assignment of the surface oxides was
obtained by comparison with those obtained in the analysis
of powdered oxides. To improve the analysis, GaAs wafers
were also oxidized using hydrogen peroxide and then etched
so the different oxides could be more easily identified. It was
also found that degreasing performed with organic solvents
removes not only the aliphatic contamination but also some
of the native oxide.
The action of the different wet-etching methods was
compared, and the extreme surface was analysed by HREELS
and XPS. Both methods show that after etching the presence
of gallium oxide is negligible and that arsenic oxide is still
remaining on the surface. XPS also shows that etching
reduces the oxide thickness from 1.6 to 0.1 nm. The effect
of HF concentration on the final quality of the surface,
namely, its chemical composition and its roughness, was
also determined. AFM images reveal that etched surfaces,
using different wet-etching methods (HF, HCl or NH3 ,
have a very low roughness. The presence of intense surface
phonons of GaAs (100) in the HREELS spectra observed in
samples etched with the three procedures also attests to very
regular surfaces.
Copyright  2005 John Wiley & Sons, Ltd.
Finally, it was confirmed that light is an important
parameter on the modification of GaAs (100) surfaces. When
immersed in water, light accelerates the rate of dissolution of
arsenic oxide, giving rise to an increase of gallium oxide
on the surface. This effect must be taken into account
when water-containing solutions are used for the adsorption
process.
Acknowledgements
This work is part of the Growth project n° GRD2-2000-30012
SENTIMATS financed by the EU. We would like to acknowledge
the ‘Financiamento Plurianual’ from FCT, Portugal and the project
GRICES/French Embassy in Portugal. We also thank Y. Feldman for
the AFM measurements.
REFERENCES
1. Freiberger General Specifications, issue 2000 (www.fcmsemicon.com/pdf/gen spec.pdf).
2. Lenczycki CT, Burrows A. Thin Solid Films 1990; 193–194: 610.
3. Gräf D, Grundner M, Lüdecke D, Schulz R. J. Vac. Sci. Technol. A
1990; 8: 1955.
4. Surdu-Bob CC, Saied SO, Sullivan JL. Appl. Surf. Sci. 2001; 183:
126. This reference is an example of a search in the ISI Web of
Knowledge with the terms XPS, GaAs and oxide(s) finding 177
papers.
5. Köhler M. Etching in Microsystem Technology (1st edn). WileyVCH: Weinheim, 1999; 32.
6. Razek N, Otte K, Chassé T, Hirsch D, Schindler A, Frost F,
Rauschenbach B. J. Vac. Sci. Technol. 2002; 20: 1492.
7. Bell GR, Kaijaks NS, Dixon RJ, McConville CF. Surf. Sci. 1998;
401: 125.
8. Cho AY. Thin Solid Films 1983; 100: 291.
9. Bertrand PA. J. Vac. Sci. Technol. 1981; 18: 28.
10. Song JS, Choi YC, Seo SH, Oh DC, Cho MW, Yao T, Oh MH. J.
Cryst. Growth 2004; 264: 98, references therein.
11. Liu HC, Tsai SH, HSU JW, Shih HC. Mat. Chem. Phys. 1999; 61:
117.
12. Massies J, Contour JP. J. Appl. Physiol. 1985; 58: 806.
13. Muñoz-Yagüe A, Piqueras J, Fabre N. J. Electr. Soc. 1981; 128:
149.
14. Lu ZH, Lagarde C, Sacher E, Currie JF, Yelon A. J. Vac. Sci.
Technol. A 1989; 7: 646.
15. Matsushita K, Miyazaki S, Okuyama S, Kumagai Y. Jpn. J. Appl.
Phys. 1994; 33: 4576.
16. Rei Vilar M, Uluçaili K, Dubot P. J. Adhesion 1999; 71: 279.
17. Dannetun P, Schott M, Rei Vilar M. Thin Solid Films 1996; 286:
321.
18. Palik ED, Ginsburg N, Holm RT, Gibson JW. J. Vac. Sci. Technol.
1978; 15: 1488.
19. Wagner CD, Naumkin AV, Kraut-Vass A, Allison JW, Powell CJ, Rumble JR. NIST X-ray Photoelectron Spectroscopy Database,
NIST Standard Reference Database 20, Version 3.3 (Web Version),
http://srdata.nist.gov/xps/ [2003].
20. Adachi S, Kikuchi D. J. Electrochem. Soc. 2000; 147: 4618.
21. Tanuma S, Powell CJ, Penn DR. Surf. Interface Anal. 1994; 21: 165.
22. Dubois LH, Schwartz GP. Phys. Rev. B 1982; 26: 794.
23. Matz R, Lüth H. Phys. Rev. Lett. 1981; 46: 500.
24. Snyder RG, Schachtschneider JH. Spectrochim. Acta 1963; 19: 85.
25. Devitt NT, Baum WL. Spectrochim. Acta 1964; 20: 799.
26. Socrates G. Infrared Characteristic Group Frequencies (2nd edn).
Wiley International Publication, John Wiley & Sons: Chichester,
1994.
27. Artzi R, Daube SS, Cohen H, Naaman R. Langmuir 2003; 19: 7392,
and references therein.
28. Beamson G, Briggs D. High Resolution XPS of Organic Polymers.
The Scienta ESCA300 Database. John Wiley & Sons: New York,
1992.
Surf. Interface Anal. 2005; 37: 673–682