Search

Recent experimental studies of fluorocarbon (FC) plasmas etching various substrates suggest that ions will transport initially bound fluorine (F) through overlying FC films, thereby defluorinating these films and inducing fluorination reaction with the underlying substrate material. Simulations of thermal CF2 on Si with simultaneous bombardment by energetic Ar+ demonstrate this defluorination phenomenon, showing that F is separated from adsorbed CF2 and mixed into the underlying Si, initiating etching. Additionally, this creates dangling bonds on the surface where CF2 may adsorb. Thus, our simulations show that F and C uptake is enhanced by energetic rare gas ion impact, the number of Si–F bonds is greatly increased, and the resultant Si etch rate is higher than expected from physical sputtering alone. The results are compared to experimental measurements made under similar conditions, and the mechanisms of ion-induced F transport are identified.

When fabricating microscopic features in SiO2 layers using low pressure, high-density fluorocarbon plasmas, microtrenching has commonly been observed. Microtrenching has been explained either as due to ion scattering from sloped sidewalls or negative charging of the sidewalls by electrons, and the influence of the associated electric field on ion trajectories. In this work, we show that a weak magnetic field produces a significant asymmetry in microtrenching. Our results demonstrate unambiguously that electron-based sidewall charging is to a significant extent responsible for microtrenching, and, more generally, that differential charging is an important effect in microstructure fabrication using high-density plasmas.

The slowdown of the oxide etch rate with width of submicrometer structures is known as reactive ion etching (RIE) lag and has been explained by ion shadowing and differential charging of the sidewalls, among other effects [R. A. Gottscho and co‐workers, J. Vac. Sci. Technol. B 10, 2133 (1992)]. Here we show for an inductively coupled high density plasma reactor working in the pressure regime from 6 to 20 mTorr that inverse RIE lag is primarily observed, i.e., the etch rates increase as the width of the microstructures decrease. Inverse RIE lag, which was first discussed by Vitkavage et al. [Tegal Plasma Proceedings Symposium, San Francisco, 1991 (unpublished)], may be explained by considering the neutral flux distribution at the structure bottom. The neutral flux has a stronger dependence on the aspect ratio than the ion flux due to its isotropic velocity distribution. The neutral flux distribution has been modeled and is consistent with etching profiles observed at high pressure.

By comparing the etching characteristics of silicon and silicon nitride in CF4/O2/N2 microwave downstream plasmas we demonstrate clearly how low concentrations of energetic species can play a dominant role in remote plasma processing: Injection of 5% N2 to a CF4/O2 plasma increases the silicon nitride etch rate by a factor of 7, while not significantly affecting the bulk composition of the discharge. Downstream injection of N2 is ineffective. Using surface spectroscopies we directly show a dramatic enhancement of the reactivity of fluorine and oxygen atoms with silicon and silicon–nitride surfaces upon N2 injection to the discharge. Our results can be explained by the production of energetic metastable species in the discharge region which transport energy to the gas‐surface interface.

We characterize and distinguish two regimes of atmospheric pressure plasma (APP) polymer interactions depending on whether the electrical interaction of the plasma plume with the surface is significant (coupled) or not (remote). When the plasma is coupled to the surface, localized energy deposition by charged species in filaments dominates the interactions with the surface and produces contained damaged areas with high etch rates that decrease rapidly with plasma source-to-sample distance. For remote APP surface treatments, when only reactive neutral species interact with the surface, we established specific surface-chemical changes and very slow etching of polymer films. Remote treatments appear uniform with etch rates that are highly sensitive to feed gas chemistry and APP source temperature.

X‐ray photoelectron spectroscopy has been used to investigate the effect of varying substrate bias when silicon is etched in a CF4 electron cyclotron resonance plasma. After etching, the silicon surface is found to be covered by a two layer structure consisting of damaged, fluorinated silicon and a fluorocarbon overlayer. The thickness of the fluorinated silicon layer depends on the self‐bias voltage. Reduction of the ion energy leads to a thinner fluorinated layer as well as a lower etch rate, suggesting that damage and fluorination of the crystal lattice are important in the ion enhanced etching of silicon in fluorine containing plasmas.

In situ ellipsometry has been used to measure in real time the surface damage introduced during electron cyclotron resonance (ECR) plasma etching of silicon as a function of rf bias to the substrate. CF4 and SF6 plasmas were employed. For all ECR plasma operating conditions, the amount of Si surface damage increases with the rf bias voltage, without an apparent damage threshold. It is shown that the surface damage depends on the ion current to the substrate and the gas, with SF6 plasmas resulting in the least surface damage.

A picosecond ultrasonics technique has been used to detect interfacial fluorocarbon (CFx) layers as thin as 0.5 nm between aluminum and silicon. The presence of the CFx material reduces acoustic damping and heat loss from the Al film into the Si substrate. This provides a means for noninvasive identification of organic/polymeric contaminants at the buried interface and potentially for characterizing interfacial mechanical properties.

The complex refractive index at a wavelength of 632.8 nm of strained epitaxial SiGe layers on silicon substrates has been determined as a function of the germanium content using in situ ellipsometry during reactive ion etching. The germanium concentration was obtained from Rutherford backscattering. These index values are used to invert the ellipsometry equations. Using this principle, the Ge concentration depth profile of an unknown SiGe structure can be determined from an in situ ellipsometry measurement sequence that is taken while the unknown sample is being etched

Gallium‐implanted p+ layers in Si were exposed to atomic hydrogen from a plasma. It was found that very large hydrogen concentrations, up to 7.5 times larger than the peak Ga concentration of 7×1019/cm3, segregated into the p+ layer during treatment at 200 °C. The shape of the hydrogen concentration profile was similar to that of the Ga profile. Ion channeling showed that the H atoms did not occupy simple high‐symmetry sites in the lattice, and electron microscopy revealed the presence of extended {111} stacking fault defects associated with the layer of high hydrogen concentration. A mechanism to account for these findings is suggested.

Light emitted from a plasma during reactive ion etching and reflected by the wafer surface at a grazing angle is utilized to determine the remaining film thickness with an accuracy of ±30 A. This promises a more flexible etching approach, e.g., tailoring the final stage of etching to minimize lattice damage.

Reactive ion etching of SiGe alloys is desirable for pattern transfer in the fabrication of heterojunction bipolar transistors using a SiGe base region. Such devices are of rapidly increasing technological importance, as it has been demonstrated that they may be used at speeds previously unattainable by Si-based devices. * Of particular technolog- ical importance are strained epitaxial SiGe alloys with a proportion of Ge of 20% or less.' Here, the reactive ion etching of SiGe, Si, and Ge in HBr plasmas has been stud- ied. HBr etching plasmas have previously been examined for Si etching,3-5 and offer anisotropic etching which may be very selective with respect to an oxide mask. Another attractive feature of HBr etching processes on Si is the very thin reaction layer which is formed-about one monolayer thick.5*6 ing reactor by the atmosphere. Patterned SiGe, Si, and Ge samples employing photoresist or oxide masks were used to examine the profile obtained using HBr RIE. Figure 1 shows reactive ion etch rates of SiGe alloy and elemental Si and Ge films as a function of the Ge content of the layer. The etch rate of the SiGe alloys is seen to increase as the percentage of Ge in the alloy is raised. The etch rate increase is consistent with the -3 times greater etch rate of elemental Ge as compared to that of Si. The inset shows the SiGe/Si etch rate ratio which is achieved under these conditions as a function of Ge con- tent, It is seen that for 15% Ge in the SiGe alloy the etch rate is -40% greater than that of Si. Etching of SiGe alloys under these conditions is highly directional, and near-vertical sidewalls are formed. This is demonstrated by the scanning electron micrograph shown in Fig. 2 which displays a shallow trench etched through a 800-nm-thick Sis,GetS alloy film into the Si substrate. An oxide mask (which was removed prior to taking the scanning electron microscopy (SEM) image) and a gas mixture of 98 seem HBr/2 seem CF4 were employed for this experiment. A near-vertical etch profile all the way from the top of the SiGe into the Si substrate is obtained, and no difference in directionality between the SissGe,, alloy and the Si sub- strate is noticeable. The width of the SiGe base layer of heterojunction bipolar transistors is of the order of only a few tens of nanometers and a precisely controlled Ge profile is re- quired for optimum electrical performance.' If, as a result of reactive ion etching significant enrichment or depletion of Ge at the surface occurred, the device performance could be degraded. The bromides of Ge are less volatile than those of Silo and therefore a Ge-rich reaction layer on the surface of the dry-etched SiGe surface may be ex- pected. The composition and extent of the surface reaction layer formed as a result of HBr RIE was examined using x-ray photoelectron spectroscopy (XPS ) . Blanket (unpat- terned) SiGe films etched for 5 min were used for the XPS studies. Immediately after etching they were transferred from the etching chamber to the analysis chamber (at a pressure of about 10 - lo Torr) using a vacuum transfer mechanism. XPS measurements were made using a Vac- uum Generators ESCALAB MkII system, with a non- monochromatized Al Ko x-ray source. Measurements were taken at two different electron emission angles: Nor

rates of single;crystal Si and Gem are shown. The etch rates of the SiGe alloys increase with the Ge content for each etching gas which is qualitatively consistent with the greater etch rate of Ge as compared to Si. The dashed lines which connect the SiGe alloy nata with the elemental Ge etch rate data are intended to help in the comparison of SiGe and Ge-etch rates, but it is not implied that the actual etch rates would show a depen-m dence on Ge content given by the dashed lines. The SiGe etch rate shows the strong@ increase with Ge content for the CF, plasma. Analysis of profiles etched in photoresist- or oxidemasked SiGe films has shown that mask undercutting takes place for fluorine-rich discharges, e.g., SF,. No mask undercutting was observed with CF&& and HBr plasmas. These etch profile characteristics indicate that fijr Si, -t;GeX alloys F-based etching is ion assisted with a strong chemical etching component; but that Cl- and Brbased etching require ion bombardment. The behavior of Sil -XGeX alloys, xQl.20, is therefore in this respect the same as that of Si.’ A more detailed discussion of the processing parameters and-etch profile results will

Surface modifications of silicon induced by electron cyclotron resonance (ECR) etching with CF4 is studied using x‐ray photoemission spectroscopy and Raman scattering techniques. It is shown that a silicon sample etched by ECR exhibits a thinner surface residual layer compared to those exposed to reactive ion etching (RIE) or hybrid ECR/RIE. Evidence of plasma‐induced structural disorder in the silicon surface was only observed in the RIE‐etched sample.

Reactive ion etching of Si and SiO2 in SF6 plasmas in which the samples are mounted on a liquid‐nitrogen‐cooled electrode has been studied. At this temperature SF6 condenses on the electrode surface, but it is possible to maintain a plasma. Si etch anisotropy has been demonstrated at low temperature, in agreement with previous studies. Mass spectrometry and optical emission spectroscopy indicate that fluorine is the dominant species in the plasma because SF6 and SFx species are removed from the gas phase by condensation.

We describe a plasma‐based dry etching procedure which permits selective etching of Si over Ge with a Si/Ge etch rate ratio of over 70 and negligible etching of the Ge underlayer. This is achieved in a SF6/H2/CF4 gas mixture by the formation of a thin (≂3 nm) involatile etch stop layer on the Ge surface which consists of Ge‐sulfide and carbonaceous material.

We describe controlled, self-limited etching of a polystyrene polymer using a composite etching cycle consisting of sequential deposition of a thin reactive layer from precursors produced from a polymer-coated electrode within the etching chamber, modification using O2 exposure, and subsequent low-pressure Ar plasma etching, which removes the oxygen-modified deposited reactive layer along with ≈0.1 nm unmodified polymer. Deposition prevents net etching of the unmodified polymer during the etching step and enables self-limited etch rates of 0.1 nm/cycle.

We investigate elastic buckling as a driving force for roughening of polystyrene, a model resist, during plasma etching. Force curve measurements of the effective modulus of etched polystyrene films, along with modeling which accounts for adhesive forces and surface corrugation, show that an extremely stiff modified layer results from Ar-ion etching, with the modulus increasing with incident energy. The results are in good quantitative agreement with analysis based on the observed dominant corrugation wavelength, and buckling theory providing the corrugation is taken into account.

Roughening of nanoscale polymer masks during plasma etching (PE) limits feature critical dimensions in current and future lithographic technologies. Roughness formation of 193 nm photoresist (PR) is mechanistically explained by plasma-induced changes in mechanical properties introduced at the PR surface (∼2 nm) by ions and in parallel in the material bulk (∼200 nm) by ultraviolet (UV) plasma radiation. Synergistic roughening of polymer masks can be prevented by pretreating PR patterns with a high dose of He plasma UV exposure to saturate bulk material modifications. During subsequent PE, PR patterns are stabilized and exhibit improved etch resistance and reduced surface/line-edge roughness.

The roles of ultraviolet/vacuum ultraviolet (UV/VUV) photons, Ar+ ion bombardment and heating in the roughening of 193nm photoresist have been investigated. Atomic force microscopy measurements show minimal surface roughness after UV/VUV-only or ion-only exposures at any temperature. Simultaneous UV/VUV, ion bombardment, and heating to surface temperatures of 60–100°C result in increased surface roughness, and is comparable to argon plasma-exposed samples. Ion bombardment creates a modified near-surface layer while UV/VUV radiation results in loss of carbon-oxygen bonds up to a depth of ∼100nm. Enhanced roughness is only observed in the presence of all three effects.

Results are presented from molecular dynamics (MD) simulations of 100eV Ar+ bombardment of a model polystyrene (PS) surface. The simulations show that the system transitions from an initially high sputter yield (SY) for the virgin polymer to a drastically lower SY as steady state is approached. This is consistent with corresponding ion beam experiments. The MD indicates that this drop in SY is due to the formation of a heavily cross-linked, dehydrogenated damaged layer. The thickness and structure of this layer are also consistent with ellipsometry and x-ray photoelectron spectroscopy measurements of Ar plasma-exposed PS samples.

Raman spectroscopy has been used to characterize the surface modifications introduced into Si(100) by reactive ion etching with a CF4/H2 plasma. Both Raman scattering due to the destruction of crystalline long range order by lattice damage and the stretching modes of protons bonded to Si atoms have been observed without any special sample preparation. Our results are consistent with backscattering and nuclear reaction profiling studies of similarly prepared samples and Raman studies of Si(100) bombarded by low‐energy protons and Ar ions.

Temperature profiles for rapid thermal annealing of ion implanted material are analyzed using asymptotic methods. Although only rapid thermal annealing is discussed, these methods are also applicable to many other annealing processes. Formulas for effective diffusion distance and effective annealing time are given which correct for the deviations of actual annealing profiles from ideal annealing profiles. The three major deviations considered are (1) ramp terms, (2) rising plateaus, and (3) plateau overshoots. The analysis shows that even relatively small deviations can have sizable effects.

Arsenic dopant profile motion in ion implanted Si samples annealed for a few seconds at 1100 °C is adequately described by a model involving concentration enhanced diffusion. There is no evidence of an initial rapid diffusive transient. Diffusion in samples preannealed at 550 °C is consistent with this result.

The effects of SiO2 reactive ion etching (RIE) in CClF3/H2 on the surface properties of the underlying Si substrate have been studied by photoemission and He ion scattering/channeling techniques. We find that RIE introduces a F, C, and Cl containing layer on the Si surface. Furthermore, displacement damage is introduced in the Si near‐surface region during RIE processing. The efficacy of O2 plasma or rapid thermal annealing RIE post‐treatments for removal of contamination and/or displacement damage has been investigated.

Silicon surfaces reactive ion etched in CF4/H2 plasmas have been examined using medium energy ion scattering and core level photoemission. Surfaces analyzed in ultrahigh vacuum have a significantly higher fluorine content than surfaces that have been exposed to air prior to analysis. In addition, an unusually large cross section exists for ion beam desorption of some, but not all, of the fluorine. Based on core level shifts, we demonstrate that fluorine desorption arises from both the fluorocarbon film and the underlying fluorosilyl layer.

Changes in the infrared absorption bands of different carbon‐oxygen complexes which exist in as‐grown Czochralski crystals occur concurrently with the precipitation of the interstitial oxygen during high‐temperature heat treatment. The rate of change in the interstitial oxygen concentration and in the infrared absorption due to the carbon‐oxygen complexes can be affected both in the same way through changes in heat treatment parameters such as the temperature and ambient gas. These results support the view that the different carbon‐oxygen complexes act as heterogeneous nuclei for oxygen precipitation in Czochralski silicon.

The structures of Si near‐surface damage induced after the removal of a thick SiO2 layer on Si using reactive ion etching with various etching gases were studied by transmission electron microscopy. Cross‐sectional micrographs showed the presence of a fluorocarbon film on the Si surface after the SiO2 layer is etched away. No extended defects were observed in Si etched utilizing pure CF4 or CF4/20% H2 etching gas, even after a 25‐min overetch into Si. For a CF4/40% H2 etching gas, no extended lattice defects were evident for overetch times of up to 5 min. However, extensive damage was found in the Si surface layer after a 10‐min or longer overetch. This extensive damage consists of {111} planar defects distributing underneath the Si surface to a depth of 300 A. The planar defects are highly decorated by impurities, likely H and possibly C, F. A pure H2 etching gas was found to introduce a heavily damaged layer and a high density of extended defects near the Si surface after only a 5‐min overetch. These ob...

Silicon surfaces which had been exposed to a CF4 /H2 plasma have been characterized by x‐ray photoelectron spectroscopy, ellipsometry, He ion channeling, and H profiling techniques. Plasma exposure leads to the deposition of a thin (∼30 A thick) C,F‐polymeric layer. Hydrogen and/or damage (displaced Si atoms) can be detected in the near‐surface region up to a depth in excess of 400 A from the Si surface.

The etching of silicon in CF3Br/O2 plasmas has been examined. In situ x‐ray photoelectron spectroscopy shows that silicon surfaces etched in CF3Br/O2 plasmas with a proportion of 30% O2 or less are covered with a reaction layer that is mainly due to bromine bonded to silicon. As the proportion of oxygen is increased above 30% the reaction layer becomes thicker and contains mainly fluorine and oxygen, and the silicon etch rate decreases simultaneously.

We solve analytically the problem of dopant redistribution from an arbitrary initial implanted profile. The diffusivity, assumed concentration independent, can be an otherwise arbitrary function of temperature. Next, we derive a closed‐form expression for the annealed concentration distribution in the special case of an initial truncated Gaussian. Our solution is valid over a much wider range of experimental conditions than is Seidel and MacRae’s approximation [Trans. AIME 245, 491 (1969)]. We offer a simple, yet precise, criterion for the validity of this approximation, and we guard against its indiscriminate use. Last, we fit short‐time annealed P profiles in implanted Si to get an average diffusivity D≂3×10−12 cm2/s. We describe simple and accurate one‐parameter data‐fitting procedures.

Silicon surfaces etched in CF4/O2 plasma have been characterized with the use of in situ x‐ray photoemisson spectroscopy. A SiFxOy layer on elemental silicon is formed under all conditions. For oxygen percentages greater than 5% in the feed gas, the oxygen content of the film and the film thickness increase, whereas the fluorine content of the film decreases. The Si etch rate decreases also and appears to be controlled primarily by the thickness of the SiFxOy layer, rather than by the F atom concentration in the gas phase, which increases up to 15% O2 addition.

The surface morphology of silicon (100) wafers has been measured by scanning tunneling microscopy. Samples which were bombarded with low‐energy argon ions are found to have an average root‐mean‐square roughness of 4.0 A, and the surfaces are covered with characteristic 50‐A‐diam hillocks. The roughness of nonbombarded (control) samples is observed to be 1.8 A, and this roughness arises in part from disorder at the interface between a native oxide and the silicon substrate.

The diffusion of ion‐implanted phosphorus in silicon during rapid thermal annealing has been studied. Depending on the implant dose, we find two distinct kinds of diffusion behavior. For low dose (1×1014 cm−2) P+‐implanted Si, a profile redistribution is observed which becomes observable at 900 °C for annealing times of 10 s, but which is temperature independent in the range 800–1150 °C. This initial redistribution is much more rapid than conventional diffusion coefficient data would predict. For high dose (2×1015 cm−2) P+ implanted and short time (10 s) annealed Si, the dopant profile broadening is strongly temperature dependent. The experimental profiles are in this case in agreement with concentration enhanced diffusion profiles.

The nature of extended defects in silicon introduced by hydrogen containing plasmas has been studied by transmission electron microscopy for a variety of technological processes. Depending on the doping level of the substrate, the substrate temperature and the presence/absence of simultaneous energetic ion bombardment, {111} planar defects, gas bubbles, and a heavily damaged near‐surface region have been observed.

Silicon specimens which had been reactive ion etched in CF4/x% H2 (0≤x≤40) have been characterized by x‐ray photoelectron emission spectroscopy. Angular rotation was used to study films deposited by the plasma process onto the Si surface. In agreement with previous studies it is found that plasma exposure of Si specimens leads to the deposition of a fluorocarbon film. An intriguing new finding was the discovery of a silicon‐carbide layer localized near the fluorocarbon‐film/Si interface. The existence of this carbide layer was found to be independent of gas composition from 0–40% H2 for a 1‐min plasma exposure.

It is demonstrated with the use of in situ x‐ray photoemission spectroscopy, secondary ion mass spectrometry, and transmission electron microscopy, that the commonly practiced in situ oxygen plasma/hydrofluoric acid dip treatment of reactive ion damaged silicon surfaces is insufficient in removing all reactive ion etching (RIE) related contaminants and damage. For CHF3/CO2 RIE the residual modifications are shown to be fluorine and carbon contamination and deeper lying modifications, e.g., hydrogen‐induced extended Si defects. An enhanced silicon oxidation rate during air exposure has been observed for post‐RIE‐treated silicon, which correlates to the amount of residual fluorine.