`M. C. Shih, M. B. Freiler, R. Scarmozzino, and R. M. Osgood, Jr.
`Microelectronics Sciences Laboratories, Columbia University, New York, New York 10027
`共Received 25 May 1994; accepted 19 October 1994兲
`
`We present a high-resolution, damage-free etching technique for GaAs and related compound
`semiconductors which utilizes surface-specific photochemistry at 193 nm to excite a physisorbed
`layer of Cl2 on a cryogenically cooled 共⬃140 K兲 sample. Etch rates as high as 0.25 Å/pulse
`共corresponding to 0.09 m/min兲 have been achieved. Etching is anisotropic, and etched features of
`0.2–0.3 m linewidth have been routinely obtained. The etch rate has been characterized as a
`function of several ‘‘system’’ parameters including Cl2-partial pressure, substrate-temperature, laser
`repetition rate and fluence, and the addition of rare gases. A phenomenological model of this
`cryoetching has been developed which agrees well with the experimental data. The etch damage and
`contamination have been studied with Auger electron spectroscopy, photoluminescence, and
`Schottky-barrier measurements. All results indicate that there is minimal if any damage induced by
`the cryoetching process. © 1995 American Vacuum Society.
`
`I. INTRODUCTION
`High-resolution, damage-free etching of GaAs and related
`compound semiconductor materials is an increasingly impor-
`tant requirement in the fabrication of advanced photonic and
`electronic device structures. For example, both characteris-
`tics are crucial for the fabrication of very high speed
`recessed-gate metal–semiconductor field-effect
`transistor
`共MESFET兲 and heterostructure field-effect transistor 共HFET兲
`devices,1,2 and especially for the etching of one-dimensional
`共‘‘wires’’兲 and zero-dimensional 共‘‘dots’’兲 quantum-confined
`structures.3,4 The current methods of anisotropic etching rely
`heavily on reactive ion etching 共RIE兲 or chemically assisted
`ion-beam etching 共CAIBE兲, in which charged-particle bom-
`bardment is utilized to achieve desorption of the reacted
`products and to provide anisotropic etching along the beam
`direction. However, etching studies of III–V 共and II–VI兲 ma-
`terials have demonstrated that ion bombardment can result in
`damage to the etched surface, thus causing degradation of
`the electrical properties of patterned material.5–9 For ex-
`ample, in the case of etch-defined quantum-confined struc-
`tures, which is an issue of particular emphasis for the work
`described in this article, damage may actually render the de-
`vices inoperable for feature sizes less than 100 nm and can
`severely affect performance at even greater sizes.5,10,11 In
`these small structures, the depletion or recombination of car-
`riers due to the surface damage incurred during processing is
`particularly important and typically causes such an etched
`microstructure to be electrically nonconductive. In fact, dam-
`age may even limit the performance of large-scale devices.
`For example, recent studies by Thirstrup et al.12 demon-
`strated that etched waveguide structures
`in integrated
`GaInAs/InP electro-optic devices may have their perfor-
`mance degraded by etch-induced damage on the sidewall of
`the waveguide. Although several methods have been pro-
`posed to passivate or remove such damage by application of
`postprocessing steps such as regrowth or wet-chemical
`etching,13,14 their application adds complexity to the fabrica-
`tion process.
`Photo-induced surface chemistry provides an alternate
`
`route to achieving anisotropic etching, and the application of
`this technique to both liquid and gas-phase etching has been
`demonstrated.15–19 Typically, the etching chemistry is initi-
`ated as a result of direct absorption of a photon by the ab-
`sorbate molecule in the vicinity of a semiconductor substrate,
`or in the case of wet etching by the generation of charge
`carriers from the near-surface region of the semiconductor
`bulk. Thus, the process generally confines the reaction to the
`laser illuminated zone. This confinement provides process
`resolution as well as etching anisotropy. An important advan-
`tage of this form of etching, which is purely chemical in
`origin, is that it does not rely on the bombardment of mas-
`sive particles to achieve etching and thus damage is not in-
`curred. Moreover, the high flux of atomic species or carriers
`that is generated by laser irradiation makes it possible to
`obtain high surface-reaction rates. Etching resolution in such
`a neutral-atom, laser-assisted method is determined both by
`the quality of the imaging optics and by the confinement of
`the reactive species to the illuminated surface area. In the
`case of laser-assisted gas-phase or wet etching, diffusive mo-
`tion of the reactive photoactivated species or semiconductor
`mobile carriers, respectively, can reduce etching resolution.
`Recently it has been reported that cryogenic cooling of a
`substrate
`during
`silicon RIE can
`enhance
`etching
`anisotropy20,21 by suppressing any thermally initiated etch
`reactions, particularly on the etched sidewalls. We have de-
`scribed the use of cryogenic-etching technology to enhance
`process resolution for the case of photon-assisted etching.22
`In this etching, a cryogenic 共⬃140 K兲 physisorbed layer of
`molecular chlorine is formed on the surface of a GaAs wafer.
`The use of a continuously formed, then laser-dissociated,
`physisorbed molecular source contrasts with the use of laser
`desorption of a chemisorbed chloride layer reported in Ref.
`16. Illumination of this physisorbed adlayer by a deep-UV
`laser, e.g., at 193 nm, generates a surface-localized source of
`atomic chlorine that, along with the suppression of the etch
`rate in the nonirradiated area at low temperature, gives rise to
`an anisotropic etching process with high resolution. Since the
`etch reaction is purely photochemically initiated and the
`
`43
`
`J. Vac. Sci. Technol. B 13(1), Jan/Feb 1995
`
`0734-211X/95/13(1)/43/12/$1.00
`
`©1995 American Vacuum Society
`
`43
`
`Page 1 of 12
`
`Samsung Exhibit 1012
`Samsung Electronics Co., Ltd. v. Daniel L. Flamm
` IPR2016-01512
`
`
`
`44
`
`Shih etal.: Patterned, photon-driven cryoetching of GaAs and AlGaAs
`
`44
`
`etched surface is not subjected to excessive laser heating,
`etch-induced damage should be minimized. In this article,
`we present a detailed discussion of this etching, focusing on
`its etch chemistry and pattern transfer capabilities, its para-
`metric behavior and the modeling of this behavior, the use of
`this etching in the fabrication of a semiconductor laser, and
`the results of various measurements, such as Auger electron
`spectroscopy 共AES兲, photoluminescence, and Schottky-
`barrier characteristics on etched surfaces.
`
`II. DESCRIPTION OF CRYOETCHING
`
`A. Summary of the relevant etch chemistry
`
`The reaction of molecular and atomic chlorine with vari-
`ous surfaces of GaAs has recently been the subject of several
`studies. A detailed understanding of the process is made dif-
`ficult by the fact that multiple sets of etch reactions dominate
`for different conditions of chlorine adsorbate coverage, ex-
`tent of reaction, and substrate temperature. While a complete
`understanding is not yet at hand, a reasonable understanding
`can be drawn from the common findings of these experi-
`ments.
`Chlorine in its atomic and molecular forms exhibits very
`different behavior in so far as its reaction with and sticking
`to the surface of GaAs is concerned. For example, earlier
`research has shown clearly that molecular chlorine exhibits a
`much smaller
`reactive sticking coefficient
`than atomic
`chlorine.23,24 Recent molecular-beam experiments have
`shown strong evidence that when Cl2 impinges on a GaAs
`surface it sticks and reacts via two separate mechanisms.25 At
`low translational energies, i.e., E i⭐0.05 eV, such as those
`predominating at temperatures below room temperature, ad-
`sorption is via a physisorbed precursor-mediated mechanism.
`At higher kinetic energies, i.e., E i⭓0.05 eV, adsorption is
`via direct chemisorption. Since the reaction barrier to chemi-
`sorption from the precursor-mediated state is larger than 0.05
`eV, condensed chlorine layers at cryogenic temperatures are
`expected to remain predominantly in their physisorbed state.
`The surface-bound reaction products of chlorine on GaAs
`depend on the degree of exposure to the reactant flux as well
`as the form, i.e., atomic or molecular, of the reactant. For
`example, thermal desorption studies by French, Balch, and
`Ford26 showed that for GaAs at room temperature, exposure
`to low dosages of Cl2 resulted in the formation of only GaCl
`and free arsenic, while at higher exposures the more volatile
`triatomic species, AsCl3 and GaCl3, dominated. On the other
`hand, x-ray photoelectron spectroscopy 共XPS兲 studies by
`Freedman and Stinespring23,24 have shown that surface reac-
`tion with atomic chlorine leads to the formation of both gal-
`lium and arsenic chlorides—even at the low substrate tem-
`peratures of 130 K in that study. Again, higher coverages of
`chlorine lead to the formation of the di- or trichloride spe-
`cies.
`Desorption of product species also plays a key role in
`determining the etching characteristics. For example, in the
`molecular-beam studies alluded to above,24 it was found that
`on a carefully prepared, fully crystalline surface, low volatil-
`ity reaction products can effectively passivate a surface by
`blocking surface-reaction sites. Progressive chlorination of
`
`J. Vac. Sci. Technol. B, Vol. 13, No. 1, Jan/Feb 1995
`
`the surface can alleviate this problem by forming the more
`volatile AsCl3 or GaCl3. However, at lower temperatures,
`even these species will remain surface bound. In fact, a sur-
`face temperature of 175 K is needed to desorb AsCl3 and 210
`K to desorb GaCl3.27 It has been suggested separately28,29
`that the difference in the volatility of these two compounds
`can lead to surface roughening during etching.
`Finally, in many of these prior studies, the specific surface
`crystallographic orientation of the GaAs substrate has been
`found to be important to the etching mechanism. For ex-
`ample, while the initial sticking coefficient of chlorine mol-
`ecules on a bare, well prepared GaAs surface is relatively
`insensitive to surface orientation, the passivation phenomena
`mentioned in the previous paragraph is surface-orientation
`dependent. Note, however, that such passivation appears to
`be only important in inhibiting reaction for the case of the
`molecular chlorine.
`In addition to the general studies of the surface chemistry
`of chlorine, there has been prior investigation of the photo-
`chemistry of physisorbed chlorine on GaAs; this work, in
`fact, helped motivate the etching experiments discussed here.
`In this study, Liberman, Haase, and Osgood27 reported that in
`the absence of UV irradiation, multilayers of molecular chlo-
`rine could be readily physisorbed on a GaAs共110兲 surface at
`85 K. In addition to this physisorbed layer, ⬃1 ML 共mono-
`layer兲 of chlorine was dissociatively chemisorbed. The com-
`position of this chemisorbed species was found to be in
`agreement with the molecular-beam results discussed above.
`In particular, temperature programmed desorption 共TPD兲 of
`this reacted layer suggested that GaCl, As2, and As4 were the
`predominant desorption products. Irradiation with 193, 248,
`or 350 nm light was found to induce a much more extensive
`chlorination of the GaAs共110兲 surface and, after extended
`exposure to the UV radiation, AsCl3 appeared in the thermal
`desorption spectrum at 175 K. The enhanced chlorination
`was attributed to the reaction of chlorine atoms which were
`generated by both direct UV photolysis, and for 248 and 193
`nm irradiation, by a UV-induced intermolecular-electron-
`transfer process in the condensed chlorine film.30 This
`intermolecular-electron-transfer process is possible only in
`the condensed film of chlorine and was found to be the domi-
`nant dissociation channel only at 193 nm. Further, the time-
`of-flight 共TOF兲 signatures of the desorbed products in our
`previous study27 indicated that in the presence of UV illumi-
`nation, enhanced dissociation of the etch products occurred,
`thus providing a route for photoactivated desorption. This
`enhanced dissociation is most probably due to direct or
`substrate-electron-initiated bond cleavage of the reaction
`products 共see below兲.
`
`B. Photon-assisted cryoetching
`
`The above results suggest that an etching reaction can be
`achieved by forming monolayers of physisorbed Cl2 on a
`GaAs surface, followed by subsequent irradiation of this sur-
`face with a laser wavelength selected so as to excite domi-
`nantly the physisorbed layer of chlorine, thus creating a re-
`active and localized atomic chlorine source on the GaAs
`surface. This atomic chlorine source allows the multiply
`chlorinated compounds, AsCl3 and GaCl3, to form on the
`
`Page 2 of 12
`
`
`
`45
`
`Shih etal.: Patterned, photon-driven cryoetching of GaAs and AlGaAs
`
`45
`
`FIG. 2. A sketch of the cryoetching system with sample loadlock. The
`sample is preheated to 210 °C in the preparation chamber to desorb residual
`moisture.
`
`fects. Finally, after each pulse a new physisorbed layer of Cl2
`forms on the GaAs surface.
`
`III. ETCHING CHAMBER AND EXPERIMENTAL
`PROCEDURE
`Figure 2 presents a sketch of the etching chamber used in
`our experiment. The chamber is made of stainless steel and
`can be pumped to a base pressure of ⬃10⫺7 Torr with a
`turbomolecular pump 共pumping speed ⬃600 l /s兲. The
`sample is mounted on a molybdenum plate by indium bond-
`ing to achieve good thermal contact. Before etching, the
`sample is heated to 210 °C in a preparation chamber to melt
`the indium foil and to desorb residual moisture on the
`sample. The molybdenum sample plate is then transferred
`into the chamber through a loadlock mechanism and fixed on
`a molybdenum etching platform, which is in contact with a
`copper heat sink cooled by a liquid-nitrogen dewar. The
`sample temperature is measured by a K-type thermocouple,
`spot welded close to the sample, and a digital temperature
`controller maintains the sample temperature to within 1 K of
`its set value. During etching, the etching chamber is continu-
`ously pumped while chlorine gas 共99.999% pure兲 is intro-
`duced through a small tube directed toward the quartz win-
`dow ⬃2.5 cm above the sample; this gas-flow direction
`serves to homogenize the condensation of chlorine molecules
`on the sample. The sample is illuminated with the output
`from a 15 ns pulse-length excimer laser operating at 193 nm.
`The light from the laser is collimated and homogenized using
`a pair of cylindrical lenses, then passed through a square
`aperture which is projected onto the sample surface such that
`a uniformly illuminated region is achieved on the sample
`surface. All samples were chemically cleaned by sequential
`immersion in trichloroethylene, acetone, methanol, and
`deionized water. After this cleaning, they were rinsed in a
`NH4OH:H2O2共1:1兲 solution for 10 min to remove any native
`oxide. Most etching experiments utilized samples of
`GaAs共100兲 共n type, 4⫻1017 cm⫺3兲; however, a limited num-
`ber of samples with different crystal orientation, particularly
`n type, 4⫻1017 cm⫺3 GaAs共110兲, and different doping type
`and doping level were used to determine if there were any
`strong dopant-dependent effects. In order to reveal the selec-
`
`FIG. 1. The laser-assisted cryoetching process. 共a兲 Before laser irradiation,
`molecular chlorine is physisorbed on the cryogenically cooled substrate. 共b兲
`Laser irradiation dissociates the physisorbed chlorine molecules which then
`react with the surface, and desorbs the chloride reaction products. 共c兲 After
`laser irradiation, the bare, etched GaAs surface again physisorbs Cl2 mol-
`ecules.
`
`semiconductor surface. These may be then desorbed as AsCl3
`and GaClx (x⫽2,3) via surface heating or photolysis, thus
`achieving continuous etching. In such a process, the reactive
`area is confined to the laser-beam-irradiated zone and the
`thermal background reaction in the nonirradiated area is sup-
`pressed by the low ambient temperature. Further, since the
`laser fluence is relatively low, the etched surface structure
`should not be damaged. In effect, the physisorbed molecules
`act as an atomically thin, ⬃1–3 ML, dry ‘‘photoresist’’
`which is continuously replenished between pulses by the
`low-pressure gas-phase ambient. Finally, note that flood illu-
`mination can be used if patterning is accomplished with sur-
`face masking.
`Figure 1 illustrates the general features of the UV-assisted
`cryogenic etching process. At the beginning of each etching
`cycle, a physisorbed layer of molecular chlorine is formed on
`the cooled GaAs surface by impingement of a flux of mo-
`lecular chlorine from a gas-phase ambient. Next, a 193 nm
`laser pulse irradiates the physisorbed layer of chlorine pro-
`ducing a surface-localized source of atomic chlorine; a per-
`centage of these reactive chlorine atoms react with the non-
`masked GaAs surface to form AsCl3 and GaClx (x⫽2,3),
`which are nonvolatile at the cryogenic temperatures used
`here. Also during laser illumination, desorption of the chlo-
`rine products occurs at the laser fluence, 15–20 mJ/cm2,
`which is typical in our experiments. This process appears to
`occur primarily through low-temperature photochemical ef-
`
`JVST B - Microelectronics and Nanometer Structures
`
`Page 3 of 12
`
`
`
`46
`
`Shih etal.: Patterned, photon-driven cryoetching of GaAs and AlGaAs
`
`46
`
`FIG. 3. SEM photos of submicrometer cryoetching of GaAs共100兲 using an
`Au surface mask. The etching was done at a fluence of 21 mJ/cm2, the
`exposure time was 30 min at 60-Hz repetition rate, the sample temperature
`was 140 K, and the chlorine-ambient pressure was 5 mTorr.
`
`tivity of etching, samples of molecular-beam epitaxy grown
`共MBE兲 AlxGa1⫺xAs on GaAs, with differing Al content were
`also examined.
`
`FIG. 4. Cross sectional SEM photograph showing crystallographic etching.
`Etched mesa profiles along the 共11¯0兲 directions on the GaAS 共110兲 surface
`are shown. The etching was done at 共a兲 T s⫽140 K and 共b兲 T s⫽300 K, the
`laser fluence 共193 nm兲 was 21 mJ/cm2 and the chlorine pressure was 5
`mTorr.
`
`IV. PATTERN TRANSFER
`An important requirement of a viable dry etching process
`is its ability to transfer images of high spatial resolution.
`Pattern transfer must rely on noncrystallographic etching for
`the process to be generally applicable as a fabrication tech-
`nique. In addition, the etching chemistry must be sufficiently
`controlled that a practical surface-mask material is available
`and that undercutting of the surface mask does not occur.
`These properties were investigated using GaAs samples that
`were patterned with either 1 m scale optical lithography or,
`for features ⬍1 m, via electron-beam patterned surface
`masks formed at Cornell University’s National Nanofabrica-
`tion Facility.
`For example, in order to study the resolution limits of this
`etching process, we investigated the use of various forms of
`surface masks to pattern features as small as 0.2 m. Several
`different materials were used, including Au/Cr and Si3N4.
`Both of these mask material systems were suitable for use as
`surface masks since both were inert to the etching chemistry
`used here. However, a close examination of the etched fea-
`tures formed at 0.6 m resolution indicated that the Si3N4
`mask showed a significant near-surface undercut. We at-
`tribute this undercutting to the enhanced reactivity of the
`material just under the nitride mask which we believe to be
`strained, as has been observed in Ref. 31. When the Au/Cr
`mask was used, undercutting was significantly reduced and
`features as small as 0.2 m could be made reproducibly.
`Figure 3 is an electron micrograph of a typical Au/Cr
`
`J. Vac. Sci. Technol. B, Vol. 13, No. 1, Jan/Feb 1995
`
`patterned and etched GaAs共100兲 sample with 0.2–0.3 m
`pattern features etched to a depth of 0.4 m. The etching was
`done with 20 mJ/cm2 laser fluence 共60 Hz兲 at 140 K substrate
`temperature and 5 mTorr chlorine ambient pressure for 30
`min. Note that the etched morphology is smooth and exhibits
`little undercutting.
`For some Cl2-based plasma etching and for thermal etch-
`ing of GaAs, the etching rate exhibits crystallographic an-
`in this etching, k 具111典B⬎k 具110典⫽k 具100典
`isotropy. Typically,
`⬎k 具111典A, where k具x典 is the reaction rate for surface x.32–34
`However, in our case, no such crystallographic anisotropy of
`etching was observed. Figure 4 shows two electron micro-
`graphs of mesa cross sections which are formed by pattern-
`ing along the 具11¯0典 direction on a GaAs共110兲 surface by
`UV-assisted chlorine etching. In both cases, the same laser
`fluence and chlorine partial pressure were used. In one case
`the substrate is at T⫽140 K and in the other 300 K; thus
`allowing us to compare photochannel etching from a cryo-
`genic film and from a gas ambient, respectively. We note
`here that the etching reaction at the higher temperatures is
`primarily thermally driven.18 The results show that in con-
`trast to gas-phase Cl etching, the cryoetching is not crystal-
`lographic. The lack of crystallographic etching is in agree-
`ment with the results of Freedman and Stinespring23,24 who
`showed that there is no preference in Cl-atom chemisorption
`for particular crystalline surfaces. In contrast to the results
`seen for thermally activated etching of GaAs by Cl2, the
`
`Page 4 of 12
`
`
`
`47
`
`Shih etal.: Patterned, photon-driven cryoetching of GaAs and AlGaAs
`
`47
`
`slope of the etched sidewall in our cryoetching is the same
`for all crystal orientations. There are a number of possible
`explanations for the origin of the sidewall slope, the most
`obvious stemming from the fact that the shape of the trench
`should reflect the local intensity distribution of the etching
`radiation. Thus as is seen in some photoresist exposure, dif-
`fractive effects at the edge of small features would reduce the
`light intensity at the wall and thus reduce the local etch rate.
`The fact that in some cases the contour of the etch feature
`showed features suggesting Fresnel diffraction supports this
`explanation. Note that once a sloped sidewall is formed, the
`well known reduction of the flux with incident angle would
`ensure a continued slower etch rate.
`Surface roughness and etch-depth uniformity provide an
`important measurement of the quality of any device-etching
`technique, since both influence the fidelity of pattern transfer
`as well the amount of surface carrier recombination. At low
`substrate temperatures, the condensation of gas-phase impu-
`rities on the surface leads to the blockage of surface sites or
`possibly even microscopic surface regions if impurity islands
`are formed. Thus, in our experiments, it was found that the
`lower the etching temperature, the more the morphology de-
`graded. After installing the loadlock assembly to eliminate
`any introduction of water vapor, etc., into the etching cham-
`ber while loading and unloading the sample, a significant
`improvement in morphology was achieved. However, an in-
`trinsic roughness of ⭐50 Å was still detected on etched sur-
`faces upon scanning electron microscopy 共SEM兲 and profilo-
`meter examination. This ⭐50 Å roughness appears to be
`intrinsic to the process chemistry.
`
`V. PARAMETRIC BEHAVIOR AND MODEL
`
`A. Etch rate measurements
`
`In order to understand the properties of the etching pro-
`cess, sets of experiments were done to characterize its para-
`metric behavior. In these experiments,
`the excimer-laser
`beam was patterned by a shadow mask into a narrow rect-
`angle of illumination. After irradiating the sample for a set
`time, the mask and the laser beam were translated together
`by a fraction of the length of the rectangle, and a second
`exposure was made. The etch depths of the various regions
`were then measured by stylus profilometry, and the etch rate
`was determined by the slope of a graph of the etch depth
`versus exposure time. Figure 5 shows the plot of etch depth
`versus etch time that was done under typical etching condi-
`tions. The linear increase of the etched depth with exposure
`time is consistent with the absence of a buildup of etch-
`resistant products such as often occurs in plasma-etch
`techniques.35 Generally, there is a short induction period seen
`at the beginning of exposure time. Since this induction time
`was affected by the details of the surface cleaning procedure,
`for example the time of rinsing in NH4OH:H2O⫽1:1 solu-
`tion, we believe that the induction time results from the pres-
`ence of a native oxide layer which is more resistant to the
`etching chemistry than is the pure semiconductor. If care was
`taken to remove most of the surface native oxide 共ex situ兲,
`the induction period was reduced.
`
`JVST B - Microelectronics and Nanometer Structures
`
`FIG. 5. Etched depth of GaAs共100兲 vs laser-exposure time. The etching was
`done at 120 K, 15 mJ/cm2 共60 Hz兲, and 2 mTorr of Cl2 .
`
`In most dry etching systems, the pressure of the reactive
`gases is an important factor in determining the etch rate. As
`mentioned above, the chlorine partial pressure determines
`the coverage of the physisorbed layer of chlorine on the
`GaAs surface; this layer is then the source for chlorine at-
`oms. Thus for our low-temperature etching, the functional
`dependence of the etch rate on Cl2 partial pressure will stem
`from the variation of the coverage of the physisorption mol-
`ecules with the partial pressure of the ambient gas. Figure 6
`shows the variation of the etch rate with chlorine partial pres-
`sure, with the substrate temperature maintained at 150 K and
`laser fluence at 31.5 mJ/cm2. The figure also contains the
`results of a curve calculated by a model based on multilayer
`physisorption to be described later. At low pressures, the etch
`rate increased rapidly with pressure until it reached a plateau
`at about 10 mTorr. This pressure corresponds to a steady-
`state coverage of 1 ML as calculated from our model. After
`this plateau, the etch rate increases more rapidly with pres-
`sure, as the physisorbed chlorine begins to exhibit multilayer
`condensation. A typical etch rate is 0.08 Å/pulse at 1 ML
`
`FIG. 6. Etch rate vs chlorine pressure for etching of GaAs共110兲 at 150 K and
`31.5 mJ/cm2 共60 Hz兲. The plateau at 10–25 mTorr is indicative of 1 ML of
`Cl2 coverage.
`
`Page 5 of 12
`
`
`
`48
`
`Shih etal.: Patterned, photon-driven cryoetching of GaAs and AlGaAs
`
`48
`
`FIG. 7. The dependence of etch rate on substrate temperature for the etching
`of GaAs共110兲. The etching was done at 10 mTorr Cl2 and 21 mJ/cm2 共60
`Hz兲.
`
`FIG. 8. Dependence of the etch rate of GaAS共110兲 on laser fluence at 60-Hz
`repetition rate, 140 K, and 10 mTorr Cl2 pressure.
`
`coverage and at 31.5 mJ/cm2 193 nm irradiation with a 20
`Hz repetition rate; this would correspond to 96 Å/min. This
`rate is somewhat slower than for thermal or RIE etching;
`however, the amount of etching is precisely controlled by the
`number of laser pulses. The fact that the etch rate is strongly
`dependent on chlorine pressures for p⬍10 mTorr suggests
`that etching is source-limited in this pressure regime and that
`desorption process is nearly complete after each pulse. The
`rate-limiting factor is thus the formation rate of products,
`which is proportional to the number of chlorine atoms cre-
`ated on the surface, and directly related to physisorbed-
`chlorine coverage.
`Intuitively, based on the usual activation energy for a re-
`action sequence, we expect that a dry etching process will
`exhibit a dependence on the substrate temperature. Typically,
`the higher the substrate temperature, the higher the etch rate,
`a behavior generally seen in the thermal and RIE etching.
`However, in our low-temperature etching process we found
`the opposite behavior, namely the etch rate decreased as sub-
`strate temperature increased. Figure 7 shows the dependence
`of the etch rate on substrate temperature while keeping chlo-
`rine partial pressure at 10 mTorr and laser fluence at 21
`mJ/cm2 and 60 Hz. The etch rate decreases nearly exponen-
`tially with increasing temperature. This increase in etch rate
`with increasing pressure and decreasing temperature is con-
`sistent with an etching process that is initiated by Cl atoms
`photogenerated from the physisorbed chlorine layer. As we
`will show below, the pressure and temperature dependences
`also indicate that the etch rate is surface-reaction limited.
`Again, this etching behavior clearly shows that there is an
`important difference in the etching mechanism from other
`gas-phase laser-assisted etching or RIE in which the etch rate
`generally increases exponentially with temperature.
`An important feature of the etching process is that the 193
`nm laser irradiation directly initiates a photoreaction in the
`physisorbed chlorine layer, as well as achieving desorption
`of the chloride compounds 关AsCl3 and GaClx (x⫽2,3)兴.
`Figure 8 shows the etch depth per pulse as a function of laser
`fluence. For fluences below 30 mJ/cm2, the etch rate in-
`
`J. Vac. Sci. Technol. B, Vol. 13, No. 1, Jan/Feb 1995
`
`creases linearly with fluence, as would be consistent with a
`single-photon photodissociation process. Above 30 mJ/cm2,
`the etch rate increases superlinearly with fluence, which may
`indicate an enhancement of the etch rate due to the onset of
`thermally initiated reactions or a contribution from multipho-
`ton photodissociation processes at higher laser fluence 共a
`phenomenon that also has been found in other much higher-
`laser-fluence processing兲. Initially, we found that no etching
`occurred for fluences below 7 mJ/cm2. However, this thresh-
`old was reduced to 3.5 mJ/cm2 by first etching the sample at
`high fluence, then reducing the laser fluence. We believe that
`the threshold at 7 mJ/cm2 is determined by the need to etch
`through the native oxide layer. The threshold at 3.5 mJ/cm2
`may be due to the inefficient desorption of the reaction prod-
`ucts at low laser fluences, or the presence of an etch resistant
`layer formed from redeposition of the etched products during
`the etching process. Regarding the desorption of etch prod-
`ucts, while it is clear that pulsed irradiation can result in
`limited surface heating, in our case this heating appears to be
`of inadequate magnitude, in itself,26 to cause removal of the
`etch products. In particular, a calculation of the thermal de-
`sorption due to the short-pulse laser heating,36 and using
`published desorption data and GaAs thermal properties.37
`suggests that the desorption of etch products due to laser
`heating at 20 mJ/cm2 is insufficient in itself to desorb the
`etch products. Electron-hole pairs are well known to be im-
`portant in many areas of surface photochemistry and may
`thus be important in this case.
`In order to probe the dynamics of the etching process, the
`etch rate was characterized as a function of the reciprocal of
`the repetition rate; this measurement thus samples the pro-
`cess rate as the time between each consecutive pulse is de-
`creased. As shown in Fig. 9, at low repetition rates 共below 10
`Hz兲, the etch depth per pulse is insensitive to the time be-
`tween two successive laser pulses. At high repetition rates
`共above 50 Hz兲, the etch depth per pulse was observed to fall
`off quickly with decreasing interpulse time. Assuming that
`each laser pulse desorbs all physisorbed species and reacted
`chloride products, the etch rate should be proportional to
`
`Page 6 of 12
`
`
`
`49
`
`Shih etal.: Patterned, photon-driven cryoetching of GaAs and AlGaAs
`
`49
`
`FIG. 9. Etch rate versus laser repetition rate for the etching of GaAs共110兲 at
`140 K and 21 mJ/cm2 共60 Hz兲 using two different chlorine pressures, 10 and
`25 mTorr.
`
`FIG. 10. Etch depth versus exposure time for AlxGa1⫺xAs alloys with dif-
`ferent Al content, showing an increase in etch rate with increased Al con-
`centration. The etching was done at 140 K, 5 mTorr of chlorine, and a laser
`fluence of 21 mJ/cm2 共60 Hz兲.
`
`physisorbed chlorine coverage, , for Ⰶ1. The low etch rate
`at the high repetition rates suggests that full physisorbed
`chlorine coverage is not obtained with short interpulse spac-
`ings.
`in ion etching of GaAs heterostructures,
`Typically,
`AlGaAs is observed to etch at a rate less than or equal to that
`of GaAs.
`For
`example,
`in
`fluorine-based
`etching
`chemistries38,39 共such as CHClF2, CHCl2F, or SiF4/SiCl4兲,
`very high selectivity of GaAs over AlGaAs has been found
`and attributed in part to the formation of a passivating layer
`of AlF3. For Chlorine-based etching chemistries, the etch-
`rate difference has been attributed entirely to the formation
`of a passivating Al2O3 layer, originating either from residual
`water vapor40 or the deliberate addition of O2 to the etching
`mixture41 rather than from differential volatility of the etch
`products. However, in high vacuum studies of ion etching
`with Cl2 or BCl3, in which presumably less impurity levels
`of H2O, etc., are present, researchers have found near equal
`etching rates for GaA