Etching AlAs with HF for epitaxial lift-off applications

Article abstract of DOI:10.1149/1.1690293

The epitaxial lift-off process allows the separation of a thin layer of III/V material from the substrate by selective etching of an intermediate AlAs layer with HF In a theory proposed for this process, it was assumed that for every mole of AIAs dissolved three moles of H₂ gas are formed. In order to verify this assumption the reaction mechanism and stoichiometry were investigated in the present work. The solid, solution and gaseous reaction products of the etch process have been examined by a number of techniques, It was found that aluminum fluoride is formed, both in the solid form as well as in solution. Furthermore, instead of H₂ arsine (AsH₃) is formed in the etch process. Some oxygen-related arsenic compounds like AsO, AsOH, and AsO₂ have also been detected with gas chromatography/mass spectroscopy. The presence of oxygen in the etching environment accelerates the etching process, while a total absence of oxygen resulted in the process coming to a premature halt. It is argued that, in the absence of oxygen, the etching surface is stabilized, possibly by the sparingly soluble A1F₃ or by solid arsenic.

Full text of DOI:10.1149/1.1690293

Journal of The Electrochemical Society, 151 ͑5͒ G347-G352 ͑2004͒
G347
0
013-4651/2004/151͑5͒/G347/6/$7.00 © The Electrochemical Society, Inc.
Etching AlAs with HF for Epitaxial Lift-Off Applications
a,z
a
a
a
M. M. A. J. Voncken, J. J. Schermer, A. T. J. van Niftrik, G. J. Bauhuis,
a
a
b
b
c
P. Mulder, P. K. Larsen, T. P. J. Peters, B. de Bruin, A. Klaassen,
d,
and J. J. Kelly *
a
b
c
Department of Experimental Solid State Physics III, Department of Inorganic Chemistry and Department
of Solid State Nuclear Magnetic Resonance, University of Nijmegen, 6525 ED Nijmegen, The Netherlands
d
Debye Institute, Utrecht University, 3584 CC Utrecht, The Netherlands
The epitaxial lift-off process allows the separation of a thin layer of III/V material from the substrate by selective etching of an
intermediate AlAs layer with HF. In a theory proposed for this process, it was assumed that for every mole of AlAs dissolved three
moles of H gas are formed. In order to verify this assumption the reaction mechanism and stoichiometry were investigated in the
2
present work. The solid, solution and gaseous reaction products of the etch process have been examined by a number of
techniques. It was found that aluminum fluoride is formed, both in the solid form as well as in solution. Furthermore, instead of
H arsine (AsH ) is formed in the etch process. Some oxygen-related arsenic compounds like AsO, AsOH, and AsO have also
2
3
2
been detected with gas chromatography/mass spectroscopy. The presence of oxygen in the etching environment accelerates the
etching process, while a total absence of oxygen resulted in the process coming to a premature halt. It is argued that, in the absence
of oxygen, the etching surface is stabilized, possibly by the sparingly soluble AlF or by solid arsenic.
3
©
2004 The Electrochemical Society. ͓DOI: 10.1149/1.1690293͔ All rights reserved.
Manuscript received June 9, 2003. Available electronically March 26, 2004.
The epitaxial lift-off ͑ELO͒ process allows the production of
single-crystalline thin films of III/V materials. The technique is in-
teresting for the optoelectronics industry, because the use of thin
film devices results in a more efficient transfer of generated heat
from device to carrier or heat sink and significantly reduces the
amount of material needed by reuse of the substrates. Furthermore,
ELO allows the integration of III/V-based components with, e.g.,
silicon-based devices.
with D0 the diffusion coefficient at infinite temperature, Ea the
activation energy, and k the Boltzmann constant. Using E
a
5
Ϫ5
2
6
ϭ 0.22 eV and D(298 K) ϭ 4.8 • 10
cm /s, D is found to
0
2
be 0.25 cm /s. The combination of Eq. 1 and 2 and the
substitution of constants, with N ϭ 0.037 mol/cm37 and
n
Ϫ6
38
ϭ 0.63 • 10
mol/cm yield Ve,max as a function of the process
parameters R, h, and T
1
In 1978, Konagai et al. first reported on peeled-film technology
0
.23
Ve,max
ϭ
eϪ2551/T
͓3͔
͑
PFT͒; they separated a Ϯ5 ␮m thick GaAs epilayer from the GaAs
substrate by etching a thin intermediate AlGaAs release layer with
aqueous HF solution. It was found that this process stopped at cer-
tain depths, because etchant and reaction products could not be ex-
ͱ
Rh
9
with Ve,max in mm/h, R and h in mm, and T in K. By application of
2
changed sufficiently fast through the narrow etch slit. In 1987,
5,9
a weight-induced epitaxial lift-off process ͑WI-ELO͒, the relation
3
Yablonovitch et al. reported that for thinner epilayers with a thick-
between V and the relevant process parameters, according to Eq. 3,
e
ness in the order of 1 ␮m this problem could be overcome by plac-
ing a droplet of black wax on top of the GaAs layer. The GaAs
epilayers experience some stress due to the wax and curl up, thereby
forcing open the small crevice between substrate and epilayer. As a
result, the etch process, now referred to as ELO, no longer stopped
at a certain depth. In a model to describe this process, Yablonovitch
has been investigated. As far as the dependence of V on h and T is
e
concerned, good qualitative agreement between theory and experi-
ment was found, but a qualitative discrepancy was encountered for
the relation between V and R. Furthermore, the lateral etch rate, as
e
predicted by the model, showed a very large quantitative discrep-
ancy with experiments. For a radius of curvature of 50 mm and a
release layer thickness of 5 nm the model predicts a maximum at-
tainable etch rate of approximately 3 ␮m/h at room temperature,
3
et al. assumed that in etching AlAs release layers with HF solution
in water each mole of AlAs forms three moles of H gas and that the
2
out-diffusion of this H gas through the etch crevice is the limiting
9
2
while the experiments show values around 3 mm/h. From this it
factor for the lateral etch rate. By assuming the rate of diffusion of
5,9
was concluded that the diffusion of hydrogen gas out of the etch
H out of the etch slit to be equal to the rate of production at the etch
2
crevice is not the limiting factor in the process, or that hydrogen
may not even be a reaction product. This shows that for a correct
description of the ELO process information about the actual etching
reaction of AlAs with HF is essential.
The present study aims to clarify this reaction. For this purpose,
layers of AlAs were etched with aqueous HF solution. The solid,
aqueous, and gaseous reaction products were examined with a num-
ber of techniques. In previous WI-ELO experiments it was found
that oxygen has an influence on the etch rate. For this reason the
effect of ambient oxygen on the etch process has also been investi-
gated.
front, the maximum attainable etch rate was found to be
1
Dn
Ve,max
ϭ
͓1͔
ͱ
Rh/2
3
N
␲
where N and n are the molar concentrations of AlAs and dissolved
H , respectively, D the diffusion constant of H in the solution, R
2
2
the radius of curvature of the film, and h the thickness of the AlAs
release layer, which generally varies between 3 and 50 nm. Maeda
4
et al. noted that, because of the diffusion coefficient in Eq. 1, V
e,max
Experimental
is expected to increase exponentially with temperature. This diffu-
sion coefficient is given by
All samples used were grown on 2 in. undoped GaAs wafers
with crystal orientation ͑100͒, 2 degrees off towards ͗110͘ using
low-pressure metallorganic vapor-phase epitaxy ͑MOVPE͒ in an
Aixtron 200 reactor. Source materials were trimethyl-gallium and
trimethyl-aluminum for the group III and arsine for the group V
elements. Disilane was used as the precursor of the n-type dopant
silicon. The sample for the solid and aqueous reaction products
ϪE /kT
a
D ϭ D0e
͓2͔
*
Electrochemical Society Active Member.
E-mail: maartenv@sci.kun.nl
z
Downloaded on 2014-11-21 to IP 128.111.121.42 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

G348
Journal of The Electrochemical Society, 151 ͑5͒ G347-G352 ͑2004͒
analysis consisted of a layer of n-type AlAs of 50 ␮m thickness,
capped with a 0.3 ␮m undoped GaAs layer. This cap layer was
applied to prevent premature oxidation of the AlAs by oxygen in the
air. The samples used for studying the gaseous reaction products and
the influence of oxygen in the environment on the etch rate had a
similar structure but an AlAs layer of only 5 ␮m thickness.
Table I. Conditions applied during preparation and etching of
the samples used in the gas composition experiments.
Processing conditions
Sample Grinding atmosphere Etching atmosphere
HF solution
The sample for the solid and aqueous reaction products analysis
was broken into smaller pieces, which were placed in an airtight
stainless steel container, together with two stainless steel balls. This
container was mounted in a shaker machine ͑Retsch MM200͒ and
after 3 min of milling at a frequency of 30 Hz, the wafer was ground
up finely enough for further processing. The AlAs/GaAs powder
mixture was transferred to a glass vessel, and a stirring magnet was
added. The vessel was closed with a rubber septum to prevent es-
cape of any poisonous gas, which could be produced during the etch
process. 2 mL of an aqueous 12% HF solution was added to the
glass vessel with a syringe through the septum. This amount of HF
ensures that four moles of HF are present for each mole of AlAs in
the powder. Therefore, all the AlAs could react with the HF while
leaving only a small amount of HF, so that the glass of the vessel
would hardly be etched. Note that, compared to AlAs, the reaction
1
2
3
N2
Air
N2
N2
O2
N2
Kept under air
Kept under air
N bubbling
2
the septum, thereby preserving the atmosphere in the vessel. The
third sample was etched with an HF solution that was placed in a
nitrogen environment and through which nitrogen gas was bubbled
for an hour. In this procedure, almost all the oxygen was removed
from the solution and hardly any residual oxygen was expected to be
present during the etching process. A titration measurement after-
wards confirmed that the nitrogen bubbling procedure had not
changed the concentration of the HF solution. For an overview of
the samples and the processing conditions, see Table I.
For each of the three samples, 10 ␮L gas samples were taken
from the vessel in which the AlAs and HF were reacting, every 10 to
15 min during the first hour, and subsequently every 30 to 60 min,
and fed into both the GC/MS and the GC/FID.
6
10
rate of GaAs with HF is 10 times lower. After the reaction was
complete ͑1.5 h, as found in explorative experiments͒, the slurry,
containing the fluid together with part of the fine solid remains of
the wafer and some solid reaction products, was extracted and trans-
ferred to another glass vessel. This vessel was placed in a centrifuge
for 5 min at 7000 rpm to separate the different fractions in the slurry.
After the separation process, solid debris from the wafer, a white
powder, a clear solution, and, floating on this solution, a thin film
with a metal-like appearance were found form the bottom to the top
of the vessel. The solution was examined using both aluminum
nuclear magnetic resonance ͑NMR, Bruker Avance DRX500͒ and
fluorine NMR ͑Bruker Avance DMX300͒ analyses. The NMR
Results and Discussion
Solid reaction products; SEM-EDS and powder diffrac-
tion.—The SEM-EDS measurements revealed the presence of alu-
minum and fluorine in the white powder. Oxygen could not be de-
tected with this particular SEM-EDS setup; it is therefore not pos-
sible to decide whether the powder consists of aluminum fluoride,
aluminum oxide, or an oxyfluoride. The thin film contained gallium
and arsenic. This is probably some very fine GaAs powder, which,
due to surface tension, floated on the solution and thereby formed a
film with a metal-like appearance.
3
ϩ
equipment was calibrated with an ͓Al(H O) ͔ solution and with
2
6
trifluoro-acetic acid, respectively. The white powder and the metal-
like film were taken from the vessel to be examined with scanning
electron microscopy ͑SEM͒, combined with energy dispersive spec-
troscopy ͑SEM-EDS͒. Additionally, an X-ray powder diffraction
spectrum was obtained from the white powder, using a Philips
PW1820 automatic powder diffractometer.
X-ray powder diffraction analysis of the white powder revealed
two different crystal forms of the compound AlF • 3H O ͑rosen-
3
2
bergite and aluminum fluoride hydrate͒. It can therefore be con-
cluded that aluminum fluoride is the only solid reaction product.
Aqueous reaction products; aluminum and fluorine NMR.—The
result of the Al-NMR measurement is given in Fig. 1. A sharp peak
around 0 ppm, identical to the reference signal from the reference
measurement is attributed to ͓Al(H O) ͔ , the underlying wide
band with a fwhm of ϳ650 Hz to either an ͓AlF (H O)6Ϫn͔
For the gas-chromatography measurements, half a wafer with a 5
␮
m thick AlAs layer was milled for 3 min at 30 Hz and etched with
2.5 mL of an aqueous 0.5% HF solution in a closed vessel. This
3
ϩ
2
6
ensures that a sufficient amount of HF is present to etch all the AlAs.
After the reaction was complete, a 10 ␮L sample from the gas phase
above the solution was extracted through the septum using a syringe
and fed into both a gas chromatograph/mass spectrometer ͑GC/MS,
Interscience Polaris Q with a 30 m Rtx-1 column͒ and a gas chro-
matograph with a flame ionization detector ͑GC/FID, Hewlett Pack-
ard HP6890 Plus with a 30 m HP-1 column͒. The latter gives flame
conductivity as a function of the retention time. To test whether or
(
3Ϫn) ϩ
n
2
4
ϩ
with n ϭ 0 . . . 3 or an ͓Al (OH) (H O) ͔
compound. Comparing the tabulated data for the latter two com-
with n ϭ 0 . . . 4
2
2
2
n
11
not AsH is formed in the reaction, a sample of arsine gas, taken
3
directly from the gas system of the MOVPE reactor, was also fed
into both the GCs for comparison of the signals. To test whether or
not hydrogen is formed during the etch process, a sample from the
reaction gas and a hydrogen calibration sample were fed into a gas
chromatograph with a thermal conductivity detector ͑GC/TCD,
Hewlett Packard HP 5890A with a Porapack Q ͑80-100 mesh, Su-
pelco͒ column͒, which is particularly suited for the detection of
hydrogen gas.
In a final series of experiments, the gas composition in the glass
vessel was varied to find out whether or not the oxygen of the
ambient atmosphere plays an active role in the etch process of AlAs
with HF. For this purpose, three samples, each consisting of half a
wafer with 5 ␮m of AlAs, were processed. The first and third
samples were ground and etched under pure nitrogen to prevent
oxidation of the AlAs, while the second sample was processed in air
and etched in a pure oxygen atmosphere. A syringe was used to
apply 2.5 mL aqueous 0.5% HF solution to samples 1 and 2 through
Figure 1. Aluminum NMR on the reaction products in solution. The mea-
surement was performed at a frequency of 130 MHz. The sharp peak is
3
ϩ
attributed to ͓Al(H O) ͔ , the wide peak with a fwhm of ϳ650 Hz is
2
6
(
3Ϫn)ϩ
attributed to an ͓AlF (H O)6Ϫn͔
with n ϭ 1 . . . 3 compound.
n
2
Downloaded on 2014-11-21 to IP 128.111.121.42 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

Journal of The Electrochemical Society, 151 ͑5͒ G347-G352 ͑2004͒
G349
Figure 2. Fluorine NMR on the reaction products in solution. The measure-
ment was performed at a frequency of 282 MHz. The peak at Ϫ79.6 ppm is
2
ϩ
attributed to an AlF compound. Independent measurements on an aqueous
HF solution and on NaF in water confirmed that the strong peak at Ϫ53.0
Ϫ
ppm stems from the F ion, while the peak at Ϫ80.3 ppm most likely
originates from the undissociated HF. The origin of the other peaks is un-
known.
pounds, we conclude that the chemical shift ͑ϩ2 to ϩ15 ppm for the
fluoride and ϩ3 ppm for the hydroxide͒ cannot be used to distin-
guish between the two. The tabulated line widths ͑550 Hz for the
fluoride and 450 Hz for the hydroxide͒ indicate a better match of the
fluoride with the measurements. Additionally, since the formation of
a hydroxide compound in a strongly acidic environment (pH Ͻ 2)
is rather unlikely,12 it can be deduced that the fluoride compound is
the reaction product measured. Further evidence for this was found
in the fluorine-NMR measurement ͑see Fig. 2͒. Several peaks are
Figure 4. Mass spectrum for the gaseous reaction products ͑a͒ and for arsine
͑b͒. Values below mass 35 have been omitted. For mass values of 101 to 200,
the intensity has been multiplied by 20, for mass values of 201 and higher,
the intensity has been multiplied by 200.
2
ϩ
detected, and the peak at Ϫ79.6 ppm is attributed to an AlF
compound.13 From these results it is concluded that during HF etch-
ing of AlAs, solid aluminum fluoride hydrate (AlF •3H O) and
3
2
(
3Ϫn)ϩ
dissolved aluminum fluoride compounds (͓AlF (H O)
͔
n
2
6Ϫn
with n ϭ 0 . . . 3) are produced.
Gaseous reaction products; gas chromatography.—The GC/MS
measurements yield a three-dimensional diagram showing retention
time vs. mass number vs. intensity ͑counts͒. In a two-dimensional
projection, as given in Fig. 3, the intensities for all mass numbers
are added. The reaction gas signal appears to lie between 1 min, 38
s and 4 min. For further analysis, the average signal before injection
of the reaction gas and after the reaction gas has passed the chro-
matography column is subtracted from the reaction gas signal in
order to eliminate all possible background signals. Furthermore,
only the signal collected between 2 and 2.5 min retention time is
used, in order to obtain measurements that are as reproducible as
possible. The mass spectrum for this time frame is given in Fig. 4a.
Mass numbers below 35 are omitted, because some of the main
peaks in this region ͑nitrogen, oxygen, and water͒ are so strong that
they render the peaks of several of the reaction products virtually
indistinguishable on the same scale. Figure 4b shows a spectrum for
arsine gas, diluted in nitrogen, taken from the gas system of the
MOVPE reactor. Arsine has a mass number of 78; the peaks at 79,
7
7, 76, and 75 are attributed to several ionized forms of AsH
3
ϩ
ϩ
ϩ
ϩ
(
AsH , AsH , AsH , and As , respectively͒, which are easily
4 2
formed in the mass spectrometer. The peaks around 150 and around
25 are attributed to arsenic dimers and trimers. A comparison of
Fig. 4a and b shows that the arsine signature from Fig. 4b is pre-
dominantly present in Fig. 4a, indicating that arsine gas is a major
reaction product in the reaction of AlAs with an aqueous HF solu-
2
Figure 3. Retention time vs. total intensity for a GC/MS measurement. It is
shown that the typical retention time for the gaseous reaction products is
between 1 min, 38 s and 4 min. Signals from 2 to 2.5 min were considered
for further analysis.
Downloaded on 2014-11-21 to IP 128.111.121.42 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

G350
Journal of The Electrochemical Society, 151 ͑5͒ G347-G352 ͑2004͒
arsenic compounds reaches the lowest level if the sample is etched
with an oxygen-free HF solution ͑sample 3͒. The fact that an
oxygen-containing species is still formed, although no oxygen what-
soever should be present, can be explained by possible small leaks
in the rubber septum, resulting in very small amounts of oxygen
diffusing into the vessel, and by the fact that it is not possible to
remove all oxygen from the HF solution, simply by bubbling nitro-
gen gas through it. The lower intensity of the oxygen-related arsenic
compounds in the case of sample 3 compared to sample 1 may
partly be due to the fact that less oxygen is present, and therefore
less of these compounds can be formed. Another effect is that, in
case of sample 3, probably not all AlAs was etched in the reaction,
as was observed in the intensity plots for arsine. As a result, less
arsenic is freed and less of the compounds can be formed.
For all experiments, except that of sample 3, GC/FID measure-
ments show two peaks at retention times of 2 min, 6 s and at 2 min,
Figure 5. Time dependence of the GC/MS sum intensity for arsine ͑mass
numbers 75 to 79, 150 to 153, and 225 to 229͒ and the oxygen-related
arsenic compounds ͑mass numbers 91, 92, and 107͒. Samples 1 and 3 were
milled and etched in a nitrogen environment ͑the etchant for sample 3 was
deaerated͒, sample 2 was milled in air and etched in an oxygen environment,
respectively. Sample 3 was etched in deaerated HF solution.
9
3
.6 s, respectively. For the gas retrieved from the etching of sample
, only one peak at a retention time of 2 min, 6 s is observed.
Injection of pure arsine from the gas system of the MOVPE reactor
also resulted in one peak at 2 min, 6 s. From these results it can be
concluded that arsine is formed in all cases. For samples 1 and 2,
processed with a significant amount of oxygen, the formation of one
other component is detected. This is most likely the set of oxygen-
related arsenic components with mass numbers 91, 92, and 107. For
sample 3 the concentration of oxygen-related arsenic species, which
according to GC/MS measurements is half of that obtained by
sample 1, is obviously below the detection limit of the GC/FID
equipment.
tion. Other reaction products with mass numbers of 91, 92, and 107
ϩ
can be coupled to the oxygen-related arsenic species AsO ,
ϩ
ϩ
2
AsOH , and AsO , respectively. These species are most likely
formed by ionization in the GC/MS of a single compound, like
As O . This assumption is supported by the fact that the GC/FID
2
3
only shows one peak for the whole range of oxygen-related arsenic
compounds. For arsine, this is also the case, since the GC/MS shows
peaks at 75 to 79, 150 to 153, and at 225 to 229, whereas the
GC/FID only shows one peak.
Reaction mechanism and stoichiometry.—From the investigation
of solid reaction products and the reaction products in solution it can
3
ϩ
be concluded that ͓Al(H2O)6͔ , and aluminum fluoride com-
(
3Ϫn)ϩ
Injection of the reaction gas and a calibration sample into the
pounds ͓AlF (H O)6Ϫn͔
with n ϭ 1 . . . 3 are formed. Judg-
n
2
GC/TCD, suited for hydrogen detection, showed an H concentra-
tion of 0.053% in a 55 mL vessel, resulting in a total amount of 1.19
ing from the formation of AlF •3H O powder, present in the vessel,
2
3
2
we must conclude that the fluorine compound does not have a high
solubility in water or in an HF solution.
␮
mol. Because 170 ␮mol AlAs was dissolved in the vessel, this
would mean that each mole of AlAs yielded only 0.007 mole of
hydrogen gas. From this it can be concluded that, contrary to previ-
ous assumptions, H2 is not a major reaction product in the etch
The analysis of the gaseous reaction products clearly shows the
formation of arsine (AsH ) gas. A number of oxygen-related arsenic
3
3
compounds, such as AsO, AsOH, and AsO , are also detected at low
2
process of AlAs with HF. Diffusion of hydrogen out of the etch
crevice therefore cannot be the limiting factor for lateral etching in
the ELO process. The hydrogen detected is either the background
concentration or it is formed as a secondary reaction product by
concentrations with the GC/MS, if oxygen is not excluded during
the preparation of the samples. From the analysis of gaseous reac-
tion products, formed in the etching process under different atmo-
spheres, it can be concluded that ͑i͒ no hydrogen is formed in etch-
ing AlAs with HF and ͑ii͒ the presence of oxygen is necessary for
completing and maintaining the reaction.
partial decomposition of AsH to solid arsenic and hydrogen gas.
3
Influence of oxygen on the etch process.—The influence of oxy-
gen on the reaction was found by examining the total intensity of
arsine ͑75 to 79, 150 to 153, and 225 to 229 added͒ and of the
oxygen-related arsenic compounds ͑91, 92, and 107 added͒ over
time. In Fig. 5, the summation of these mass-spectrometer intensities
vs. time are given for samples 1 to 3. Note that the vertical axis in
Fig. 5 has a logarithmic scale. Comparison of the increase in AsH3
intensity over time for the three samples shows that sample 1 and
sample 2 reach the same final intensity. In the etching of sample 1,
however, more time is needed to reach this final intensity, indicating
that the presence of oxygen enhances the reaction rate. From the fact
that the arsine intensity of sample 1 reaches the same final level as
that of sample 2, it can be concluded that oxygen in the HF solution
Etching of AlAs in HF solution with the formation of AsH3
seems similar to the etching of InP in concentrated HCl solution.
In the latter case chemical attack by undissociated HCl on InP sur-
14
face bonds gives rise to the formation of PH and InCl3
3
InP ϩ 3HCl → InCl3 ϩ PH3
͓4͔
The InP is subsequently hydrated. In principle, etching of AlAs
in the present work could be due to either the undissociated acid ͑HF
Ϫ
2
ϩ
Ϫ
or HF ) or the dissociated species (H and F ions͒. In order to
͑
about 1 mole of oxygen for every 50 moles of AlAs͒ acts catalyti-
check these possibilities, AlAs etch experiments were performed in
cally, or that it reacts in very small quantities. Sample 3, which was
etched with HF solution purged of dissolved oxygen, shows a very
fast increase up to its maximum arsine intensity, but this maximum
intensity is significantly lower than that of the other two samples.
This indicates that the etching process was blocked before all the
AlAs was etched.
both concentrated HCl solution ͑with a high content of undissoci-
1
4
ϩ
ated acid͒ and in 10% HCl solution ͑which only contains H and
Ϫ
Cl ions͒. In both cases fast etching of AlAs was observed with the
formation of considerable amounts of arsine. This result suggests
that dissociated HF should be capable of dissolving AlAs.
Comparison of the intensity increase with time for the oxygen-
related arsenic compounds shows a clear difference between sample
Because of the difference in electronegativity of Al and As, the
Al-As surface bond is expected to be polarized. Proton attack on the
negatively charged As leads to the rupture of the Al-As bond and the
formation of a new As-H bond ͑step 1 of Fig. 6a͒. At the same time,
2
and the other two samples. If oxygen is present in excess, these
compounds are formed in much larger quantities than when very
little or no oxygen is present. The formation of the oxygen-related
Ϫ
the positively charged Al can be complexed by either F ͑step 2͒ or
Downloaded on 2014-11-21 to IP 128.111.121.42 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

Journal of The Electrochemical Society, 151 ͑5͒ G347-G352 ͑2004͒
G351
present in the solution, might oxidize the arsenic to form As O as
2
3
detected in the GC/MS and thus prevent passivation and maintain
the etch process.
Conclusions
In previous reports on the epitaxial lift-off process it was as-
sumed that in etching AlAs with HF solution three moles of hydro-
gen are formed for each mole of AlAs. In the present study, this has
been investigated by a structural analysis of all reaction products,
solid, in solution, and gaseous. SEM-EDS and X-ray diffraction
measurements show that a white powder, formed during the etching,
is an aluminum fluoride compound. Aluminum and fluorine NMR
experiments on the solution showed the presence of the aluminum
͓AlF (H O)6Ϫn͔ 3Ϫn)ϩ,
(
with
n ϭ 1 . . . 3
and
species
n
2
3
ϩ
͓
Al(H O) ͔ . Gas chromatography measurements with GC/MS
2 6
and with GC/FID show the formation of arsine and oxygen-related
arsenic compounds like AsO, AsOH, and AsO . The latter com-
2
pounds are probably formed when As O decomposes in the GC/
2
3
MS, an assumption that is supported by the fact that the GC/FID
shows only one peak for these compounds. The GC/TCD experi-
ments refute the assumption that hydrogen gas is formed in the etch
process. Consequently, the out-diffusion of hydrogen through the
narrow etch slit cannot be the limiting step in the lateral etch rate of
the epitaxial lift-off process.
Oxygen is shown to play a significant role in the ELO etch pro-
cess. During an in situ measurement in an aerated solution it is
found that the amount of arsine gas increases faster to its maximum,
if a sample is etched in an oxygen atmosphere. If the process is
performed in a nitrogen atmosphere, the same maximum value is
reached, but the time needed to reach this value is considerably
longer. If, in addition, the oxygen dissolved in the HF solution is
first removed, the arsine concentration reaches its maximum consid-
erably faster, but this maximum value is much lower than that found
in the presence of oxygen.
Figure 6. Reaction mechanism for the etching of ͑a͒ AlAs with dissociated
HF and ͑b͒ for undissociated HF.
From experiments in which AlAs was etched with both concen-
trated and diluted ͑10%͒ HCl, it was found that the dissociated acid
is capable of dissolving AlAs. The mechanism describing the etch-
ing of AlAs with aqueous HF solution is therefore similar. In a first
step, the proton attack on the As leads to a rupture of the As-Al bond
H O ͑step 3͒. If the remaining back bonds to As and Al react in the
2
and an As-H bond is formed. The positively charged Al is at the
Ϫ
same way as in steps 1 to 3, then AsH and either AlF3 or
3
same time complexed by either F or H O. The remaining Al and
2
3
ϩ
Al(H O) will be formed. Alternating attack on a surface Al atom
by F and H O will give rise to a mixed fluoro-aqua complex
As back bonds may react in a similar way to form AlF3 or
2
n
Ϫ
3ϩ
2
Al(H O) compounds. Alternating attacks on a surface Al atom by
2
n
(
3Ϫn)ϩ
Ϫ
͓
AlF (H O)(6Ϫn)͔
, with n ϭ 1 . . . 3.
F
and H O will give rise to a mixed fluoro aqua complex
2
n
2
(
3Ϫn)ϩ
In principle, attack on a surface Al-As bond by the undissociated
͓AlF (H O)(6Ϫn)͔
, with n ϭ 1 . . . 3. Etching of AlAs by un-
n
2
acid, as occurs for InP in HCl, is also possible ͑see Fig. 6b͒. The
dissociated HF is in principle also possible. The reaction mechanism
is then similar to the etching of InP in concentrated HCl and the
process yields the same reaction products as for the etching of AlAs
in a dissociated HF solution.
result will be the same as for step 2 of Fig. 6a. Subsequent reactions
Ϫ
with F or HF and/or H O will lead to the same products.
2
The results of the present work clearly show that the AlF prod-
3
The AlF compound that is formed in the reaction is shown to be
uct is only sparingly soluble. The formation of an insoluble trifluo-
ride ͑which may be in the hydrated form, see the NMR results͒ can
explain the passivation, which is observed in this system. The fact
that passivation does not occur if AlAs is etched with HCl resulting
in AsH3 and AlCl3 shows that arsenic hydride is unlikely to be
3
sparingly soluble in water. The formation of this compound at the
surface can therefore explain the passivation, as observed in this
process. Another possible passivating agent is solid arsenic at the
surface. The oxygen in the solution might oxidize the arsenic to
form As O as is detected in the GC/MS and thus prevent passiva-
2
3
responsible for the termination of etching. Also the AlCl formed in
3
tion.
this case will rapidly hydrolyze and dissolve ͑similar to HCl etching
of InP͒. From the results obtained in aerated and deaerated solutions
it is clear that oxygen plays a role in the surface chemistry. Direct
oxidation of A1As by oxygen can occur to a limited extent ͑this
would explain the As O detected in the GC/MS experiments͒. The
The model, as described above, is confirmed by examination of
the ELO process in its normal implementation. In that case, the
release layer is extremely thin ͑ϳ5 nm͒, so that blocking of the AlAs
surface plays an extremely important role. In experiments, it was
found that this process comes to a premature halt, when absolutely
no oxygen is present during etching. With oxygen present, the pro-
cess continues and the III/V films are freed from their substrate.
It should be clear that an understanding of the epitaxial lift-off
process requires more than knowledge of the reaction mechanism
and stoichiometry. Interesting points for further investigation in-
2
3
presence of adsorbed oxygen is expected to influence the etching
steps, as shown in Fig. 6. The results suggest that passive film
(
A1F ) formation is hindered; why this is so is not clear at present.
3
Another possible explanation for the observed passivation is the
formation of solid arsenic at the surface. The oxygen, which is
Downloaded on 2014-11-21 to IP 128.111.121.42 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

G352
Journal of The Electrochemical Society, 151 ͑5͒ G347-G352 ͑2004͒
clude the influence of stress and strain on the lateral etch rate and
the escape of the gaseous reaction products.
References
1
2
3
. M. Konagai, M. Sugimoto, and K. Takahashi, J. Cryst. Growth, 45, 277 ͑1978͒.
. J. C. C. Fan, J. de Physique, 43, C1-327 ͑1982͒.
. E. Yablonovitch, T. Gmitter, J. P. Harbison, and R. Bhat, Appl. Phys. Lett., 51, 2222
͑
1987͒.
Acknowledgments
4
5
6
. J. Maeda, Y. Sasaki, N. Dietz, K. Shibahara, S. Yokoyama, S. Miyazaki, and M.
Hirose, Jpn. J. Appl. Phys., Part 1, 36, 1554 ͑1997͒.
. J. J. Schermer, G. J. Bauhuis, P. Mulder, W. J. Meulemeesters, E. Haverkamp, M.
M. A. J. Voncken, and P. K. Larsen, Appl. Phys. Lett., 76, 2131 ͑2000͒.
. Kirk-Othmer Encyclopedia of Chemical Technology, Vol. 13, 4th ed., Wiley, New
York ͑1995͒.
7. Properties of Aluminium Gallium Arsenide, S. Adachi, Editor, EMIS Datareviews
Series No. 7, INSPEC, London, ͑1993͒.
. Treatise on Analytical Chemistry, I. M. Kolthoff, P. J. Elving, and E. B. Sandell,
Editors, Part II, Vol. 1, Interscience, New York ͑1961͒.
. M. M. A. J. Voncken, J. J. Schermer, G. Maduro, G. J. Bauhuis, P. Mulder, and P.
K. Larsen, Mater. Sci. Eng., B, 95, 242 ͑2002͒.
0. M. M. A. J. Voncken, J. J. Schermer, G. J. Bauhuis, P. Mulder, and P. K. Larsen,
Appl. Phys. A: Mater. Sci. Process, In press.
1. NMR and the Periodic Table, R. K. Harris and B. E. Mann, Editors, Academic
The authors thank Dr. H. op den Camp ͑Department of Microbi-
ology, University of Nijmegen͒ for the GC/TCD measurement on
hydrogen and J. Joordens and A. Swolfs ͑Department of Biophysical
Chemistry, University of Nijmegen͒ for their help on the aluminum
and fluorine NMR measurements, respectively. Dr. J. Fransen
͑
Nijmegen Center for Molecular Life Sciences, University of
8
Nijmegen͒ is acknowledged for the SEM-EDS measurements. J. M.
M. Smits and Dr. R. de Gelder ͑Crystallography Group, Department
of Inorganic Chemistry, University of Nijmegen͒ are acknowledged
for the powder diffraction measurement. The authors thank Dr. P. H.
M. Budzelaar ͑Department of Inorganic Chemistry, University of
Nijmegen͒ for helpful discussions on the reaction mechanism and
the critical reading of the manuscript.
9
1
1
Press, London ͑1978͒.
12. Advanced Inorganic Chemistry, 6th ed., F. A. Cotton, G. Wilkinson, C. A. Marillo,
and M. Bochmann, Editors, p. 183, John Wiley & Sons, New York ͑1999͒.
13. N. A. Matwiyoff and W. E. Wageman, Inorg. Chem., 9, 1031 ͑1970͒.
14. P. H. L. Notten, J. Electrochem. Soc., 131, 2641 ͑1984͒.
University of Nijmegen assisted in meeting the publication costs of this
article.
Downloaded on 2014-11-21 to IP 128.111.121.42 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).