Plasma–surface interactions in fluorocarbon etching of silicon dioxide
`J. W. Butterbaugh, D. C. Gray, and H. H. Sawin
`
`Citation: Journal of Vacuum Science & Technology B: Microelectronics and Nanometer Structures Processing,
`Measurement, and Phenomena 9, 1461 (1991); doi: 10.1116/1.585451
`View online: https://doi.org/10.1116/1.585451
`View Table of Contents: http://avs.scitation.org/toc/jvn/9/3
`Published by the American Institute of Physics
`
`Articles you may be interested in
`Role of steady state fluorocarbon films in the etching of silicon dioxide using CHF in an inductively coupled
`plasma reactor
`Journal of Vacuum Science & Technology A: Vacuum, Surfaces, and Films 15, 1881 (1997); 10.1116/1.580655
`
`Study of the
` etch selectivity mechanism in inductively coupled fluorocarbon plasmas and a
`-to-
`-to-Si mechanism
`comparison with the
`Journal of Vacuum Science & Technology A: Vacuum, Surfaces, and Films 17, 26 (1999); 10.1116/1.582108
`
`Plasma etching of Si and SiO2—The effect of oxygen additions to CF4 plasmas
`Journal of Applied Physics 49, 3796 (1978); 10.1063/1.325382
`
`Fluorocarbon high-density plasmas. II. Silicon dioxide and silicon etching using CF4 and CHF3
`Journal of Vacuum Science & Technology A: Vacuum, Surfaces, and Films 12, 333 (1994); 10.1116/1.578877
`
`Fluorocarbon high-density plasmas. I. Fluorocarbon film deposition and etching using CF4 and CHF3
`Journal of Vacuum Science & Technology A: Vacuum, Surfaces, and Films 12, 323 (1994); 10.1116/1.578876
`
`High density fluorocarbon etching of silicon in an inductively coupled plasma: Mechanism of etching through a
`thick steady state fluorocarbon layer
`Journal of Vacuum Science & Technology A: Vacuum, Surfaces, and Films 16, 239 (1998); 10.1116/1.580978
`
`IP Bridge Exhibit 2225
`TSMC v. Godo Kaisha IP Bridge 1
`IPR2017-01843
`
`

`

`Plasma-surface interactions in fluorocarbon etching of silicon dioxide
`
`J. W. Butterbaugh
`IBM General Technology Division, Essex Junction, Vermont 05452
`
`D. C. Grayand H.H. Sawin
`Massachusetts Institute ofTechnologfi Department ofChemical Engineering, Cambridge,
`Massachusetts 02139
`
`(Received 1 January 1991; accepted 26 February 199 l )
`
`in a fluorocarbon plasma environment were simulated and
`The major species present
`independently controlled using radical and ion beams in an ultrahigh-vacuum apparatus. The
`beams used in this study were chosen to determine the importance of CFX radicals in a CF4
`plasma; the beams included F and CFg, with a beam of Ar“ to simulate energetic ion
`bombardment. Both CF2 and F enhance the etching yield of SiO2 under energetic Ar *‘
`bombardment; however, the enhancement with F is twice that seen with CF2 at similar fluxes.
`When CF2 and F fluxes are used simultaneously, F dominates and the CF2 flux has little effect on
`the overall etching yield. Combined with previous work on Si substrates,
`these results are
`consistent with qualitative theories for SiOz/Si selectivity in fluorocarbon plasmas. Possible
`elementary steps in the ion-enhanced etching process are proposed and reduced to a two-
`parameter model which represents the process as ion—enhanced neutral adsorption followed by
`ion-induced reaction to form volatile products.
`
`l. lNTRODUCTION
`
`During plasma—etching processes, the substrate surface is
`subjected to fluxes of several different reactive species. In
`fluorocarbon plasmas, the species striking the substrate sur-
`face include CFy (y = 0—3: ), F, and energetic CF):+ (.x z 0.,
`3). Knowledge of the relative importance of these species in
`the plasma etching of Sit)2 is necessary to understand selec-
`tivity and anisotropy in SiO2 etching processes. Selectivity of
`SiOZ etching over Si etching in fluorocarbon plasmas is at‘
`tributed to carbonaceous film formation on the Si surface,
`while the highly anisotropic nature of SiO2 etching is attrib—
`uted to the strongly ion-enhanced nature of the etching pro—
`cess. Previous research into these mechanisms has included
`
`post-process surface analysis with x-ray photoelectron spec—
`troscopy (XPS), Auger electron spectroscopy (ABS), and
`ellipsometry. Other research efforts have focused on the rel—
`ative importance of the radical and ionic species by isolating
`these species outside the plasma environment.
`The general conclusion of the post—process surface studies
`is that a carbonaceous film forms on the surface of Si while
`
`little, if any, carbonaceous film forms on the surface of SiO2 .
`After exposing Si and SK); surfaces to a C2 F6 plasma,
`Oshimal found a 20— to 30-A—thick fluorocarbon film on the
`Si surfaces and little or no fluorocarbon film on the SiO2
`surfaces. Coburn2 found similar results with AES analysis of
`Si and so, surfaces after exposure to CE, + H2 plasmas.
`Coburn found that the C signal on Si surfaces increases and
`the Si signal decreases as the amount of H2 in the plasma is
`increased; the C signal is not detectable on SiO2 surfaces
`until the level of H2 addition is above 17%. Jaso and Ochr—
`lein,3 using XPS and ellipsometry, also noted the absence of
`any carbonaceous film on SiO2 surfaces, although some Si—=
`C bonding was detected. Havering er al.‘ studied the SiC)2
`surface in a CF4 plasma with in situ ellipsomctry and again
`found a. carbonaceous layer only on the Si after the SiO2 was
`
`etched away. The probable explanation for the absence of a
`carbon layer on the SiO2 surface is the availability of oxygen
`to form volatile carbon—fluorineuoxygen species. Ab initio
`calculations5 show COFZ to be very weakly bound on SiO2
`surfaces, making it a likely product of the etching process.
`Several studies in the literature have isolated radical spe-
`cies present in the discharge and have measured their reacti-
`vity towards SiO2 .6 "3 Flamm et (11.6 measured the reactivity
`of atomic fluorine with SiO2 and found that F will spontan-
`eously etch SiO2 at a rate similar to that found in CF4 barrel
`etchers, where the SiO2 surface is subjected to little ion bom-
`bardment. Etching rates in parallel-plate or reactive ion
`etching (RIE) reactors, where the SiO2 receives consider-
`able energetic ion bombardment, are much higher than can
`be accounted for by spontaneous reaction with F.
`The reactivity of CF2 with "5&02 has been studied by sever-
`al researchc;:rs.7'10 Selamoglu et al.9 reported that 3102 does
`not etch in the presence of CF2 without simultaneous sur—
`face irradiation. Langan et 0].“) performed XPS analysis of
`SiO2 surfaces exposed to CF2 radicals and found that CF2
`adsorbs on the SiO2 surface, but does not dissociate or react
`with the surface. After bombarding the surface with energeb
`ic ions, the CF2 adsorption probability increased, but there
`was still negligible dissociation. These studies clearly Show
`that CF2 does not spontaneously etch SiOZ; CF2 will only
`etch SiO2 when the surface is under simultaneous energy
`irradiation (cg, ion bombardment).
`The reactivity of CF3 towards SiO2 has also been studied
`by several researchers.“1 1 "3 These studies reported that the
`reactivity of CF, toward SiO2 is below experimental detec“
`tion limits, placing the reaction probability at less than
`10 ’ 6. Joyce er a]. ’2 found that CF3 will adsorb on SiOz, but
`will undergo very limited dissociation. XPS analysis, after
`exposure to CF; , showed no evidence of O-F bonding and
`some evidence ofboth C—«Si and (Zn-O bonding.12 McFeely et
`(21.13 used XPS to examine SiOz surfaces exposed to CF, and
`
`1461
`
`J. Vac. Sci. Technol. 3 9 (3), May/Jun 1991
`
`0735-21 1 X/9‘i/D‘31 461 -1 0301 .00
`
`(<2) 1991 American Vacuum Society
`
`1461
`
`
`
`

`

`1462
`
`Butterbaugh, Gray, and Sawin: Plasma—surface interactions in Fc2 etching of SIC,
`
`1462
`
`also determined that CF3 adsorbs, but not dissociatively. In
`addition, McFeely concluded that adsorbed CF3 is mainly
`bound to O on the surface through C—O bonding. These
`studies Show that CF}, like CFz, does not spontaneously
`etch $0,.
`In addition to the flux of neutral radical species, there is
`significant ion bombardment of the SiO2 surface in a paral—
`lel—plate plasma reactor. Ions in a CF4 plasma are mainly
`CF73“, CF; , and CF 'F. Several studies”18 demonstrate
`that these molecular ions, which supply chemical reactants
`to the surface as well as energy, remove a larger number of
`SiOZ molecules than a noble gas ion like Ar ‘1. The etching
`yield (number of SiO2 molecules removed per incoming
`ion) has been measured at 0.2 for 500 eV Ar * and 0.5 for
`
`500 eV CF; .18 It has also been shown that the energetic ion '
`bombardment can induce the reaction of species already ad-
`sorbed on the surface. The SiO2 etching yield of Ar” is
`enhanced when the surface is simultaneously exposed to a
`neutral source of F, such as XeF2 .‘5'19 Because XeF2 disso—
`ciatively chemisorbs leaving a monolayer of F on the SiO2
`surface, ‘9 XeF2 has been used to simulate the flux of atomic
`fluorine in the plasma environment. Loudiana er al. 19 found
`an enhancement of the 500 eV Ar + etching yield of about
`3.6 X with simultaneous Xe}?2 exposure.
`It has also been suggested that the etching yield of SiO2 is
`enhanced by the adsorption of neutral CFy species. In stud-
`ies using CFJ', the SiO2 is also exposed to CFy species,
`which are produced along with the ions. Mayer and Bark—
`er20 attributed the pressure dependence of the etching rate in
`reactive ion beam etching to the adsorption of neutral spe—
`cies. The pressure dependence flattened out at higher pres—
`sure, indicating a saturation or a shift from an adsorption—
`limited to an ion flux limited etching process.
`In the present work, the fluxes to a SiO2 surface in a flu-
`orocarbon plasma environment are simulated with three
`controlled beams in an ultrahigh-vacuum (UHV) appara-
`tus. In the literature, the primary etchant in a CF4 plasma
`has been identified as atomic fluorine, although it. has been
`suggested that CF2 and CF3 may be active participants in
`the etching reaction?"22 a beam of atomic fluorine (F) was
`used as the primary etchant in the UHV apparatus. There
`are several fluorocarbon radical species present in the CF4
`plasma, including CF3, CFZ, and CF; a beam of CF2 was
`used to represent the effect of fluorocarbon radicals. The
`ionic species in a CR, plasma are mainly CFf , in addition to
`less fiuorinated ions. Due to difficulties in making a clean,
`high-flux beam of CF; , a beam of Ar * ions was used in the
`UHV apparatus as the source of energetic ions. This substi—
`tution has been made in past studies ofiorrenhanced etching
`of SiO2 with XeF2 to decouple the effects of ion inertia and
`ion chemistry.””‘9’23'24 Tu et (11.15 studied SiO2 sputtering
`with CF; ions and Ar "
`ions in the presence of XeF2 and
`found that the CF3+ ions were twice as effective as Ar 4
`in
`etching SiOZ, but that the effectiveness of both ions showed
`the same dependence on the flux of XeFZ. Winters” found
`that the products from the sputtering of SiO2 in the presence
`of XeF2 were relatively independent of the type of ion (Ar ‘*
`or CF: ) used to bombard the surface.
`
`J. Vac. Sci. Technol. B, Vol. 9, No. 3, May/Jun 1991
`
`This paper describes an investigation of the individual and
`combined effects of CF2 and F on the Ar 4'
`ion sputtering
`yield of SiO2 . The flux levels in these experiments are repre-
`sentative of the flux levels in the plasma environment. Sim-
`ple models involving the ion-enhanced adsorption of CF2
`and F were derived from a much larger set of possible surface
`reactions and ion enhancement mechanisms. These simple,
`two-parameter models capture the major features of the ion—
`enhanced etching of SiOZ.
`
`II. EXPERIMENTAL
`
`The experimental apparatus used for these studies has
`been described in detail elsewhere.25 A schematic of the sys—
`tem is provided in Fig. l. The SiO2 sample can be simulta-
`neously exposed to three separately controlled beams. The
`beams used in this work include an Ar + ion beam generated
`in a Kaufman ion source, a difluorocarbene (CF2 ) radical
`beam generated by pyrolysis of hexafluoropropylene oxide
`(HFPO), and an atomic fluorine (F) beam generated by a
`microwave discharge of F2 and Xe. The sample temperature
`is measured and controlled with a thermocouple and stain-
`less-steel heating filament. The etching rate of SiO2 films is
`measured by laser interferometry at a 45° angle with the sur-
`face [Fig. 1 (a) ]. The sample is mounted to a transfer assem-
`bly, which insures alignment of the sample to the beams and
`
`
`
` Photodiode
`
`“Si/5m,
`
`
`Film
`
`,
`. He-Ne Laser
`
`Detector
`
`Neutral Flux
`Probe
`
`Pchamber
`
`Manometer
`
`Diiierential
`Capacitance
`
`FIG. 1. The experimental apparatus. The profile View shows the laser inter»
`ferometry and line-of—sight mass spectrometer. The beam plane View shows
`the beams and the two flux probes {retarding grid analyzer and pitot tube)
`which can be rotated into the sample position.
`
`

`

`1 463
`
`Butterbaugh, Gray, and Sawin: Plasma—surface interactions in Ft;2 etching of Sit)2
`
`1463
`
`CF31L , with the commonly used ionizer electron energy of 70
`eV. Since CF2+ is a product of the electron impact of HFPO
`and CF3 CFO, as well as CF2, at high ionizer energies, it is
`advantageous to choose an electron energy which ionizes the
`CF2 radical, but which produces as little dissociation as pos-
`sible of higher molecular weight species. Therefore, the ion“
`izer electron energy was held at 20 eV (the minimum for the
`mass spectrometer) for the HFPO pyrolysis characteriza-
`tion. The appearance potential for the process
`
`CFZ +e‘ ==—>CF2+ +2e"
`
`(2)
`
`has been estimated at 13.3 eV.27 Appearance potentials for
`the dissociative processes are unknown, but are expected to
`be several eV larger.
`Figure 2 is a plot of the ratio of the CF2 peak to the CF3
`peak versus temperature. The ratio was plotted to eliminate
`noise due to variations in the flow rate and base pressure
`during the series of readings. The ratio starts increasing at
`about 500 K, which corresponds to the minimum tempera-
`ture at which Knickelbein et (21.26 could detect CF2 radicals
`by LIF. The ratio of CF2 to C173 has a sigmoidal shape with a
`knee at about 750 K. This saturation is at a significantly
`higher temperature than that detected by Knickelbein et al.
`(600 K) .26 attributable to a different hot‘zone residence
`time and to the qualitative nature of the mass spectrometric
`analysis. The pyrolysis dependence on the HFPO flow rate
`was also investigated at 730 K with no significant change in
`the CFZ to CF3 signal ratio over a range of 0.2—0.6 seem.
`The qualitative results of the mass spectrometric charac-
`terization of the pyrolysis ofHFPO in this apparatus indicat-
`ed that pyrolysis was occurring with an efficiency close to
`100%, based on the work of Knickelbein et (£1.26 The tem:
`perature of the HFPO pyrolysis tube was held at 740~760 K
`
`provides electrical connection to the thermocouple and
`stainless-steel heating filament.
`A Faraday cup with retarding grids measures the ion flux
`and energy distribution at the sample position. With, the
`sample removed, the cup is rotated into place on a rotary
`feedthrough, as shown in Fig. l(b). The flux of neutral spe—
`cies is measured with a pitot tube that is also moved into the
`sample position on a rotary feedthrough. The pressure dif-
`ferential in the pitot tube is measured with a difi‘erential har-
`atron.
`
`A quadrupole mass spectrometer is mounted on the upper
`flange and can be used for analysis of the beams or for analy-
`sis of etching products from the surface of the sample. The
`ionizer is equipped with a line-of—sight shield such that the
`signal from background gas is minimized.
`The chamber is pumped with an 8 in. cryogenic pump and
`the load lock is pumped with a turbomolecular pump. The
`base pressure in the chamber is typically around 5X 10‘ 8
`Torr after 5—10 h of pumping; the chamber is frequently
`brought. up to atmospheric pressure to replace the filaments
`in the ion source. The gases used are research purity
`( > 99.9995%) Ar, research purity ( > 99.99%) Xe, techni-
`cal grade ( > 97.0%) F2, and HFPO (98.83%). Gas flow
`rates are set with needle valves, based on pitot tube measure-
`ments. The SiO2 samples are Z-Iu-thick sputtered quartz
`films on Si substrates.
`
`The Ar + and F sources were previously described in de-
`tail.25 A monoenergetic beam of Ar i‘ ions is created with a
`Kaufman ion source with a 1 cm2 nominal beam cross sec—
`
`tion. The F beam is created by the microwave discharge dis:
`sociation of F2 in an alumina tube. The F2 is mixed with
`about 2 seem of Xe to maintain stability of the discharge.
`Mass spectrometric analysis indicates that dissociation of F2
`in this source is close to 100%.
`
`The CF2 beam was created from the pyrolysis of hexa—
`fiuoropropylene oxide (HFPO)26 in a resistively heated Py-
`rex tube. Knickelbein at all.” provided evidence that HFPO
`is completely dissociated at temperatures above 590 K in a
`pyrolysis tube with a residence time of 10 ms according to
`the reaction:
`'
`
`CF} —CF—CF2 _. CF3 —CF = o + CFZ.
`\ /
`0
`
`(1)
`
`Knickelbein et at. monitored the extent of this reaction via
`
`laser-induced fluorescence (LIF) detection of CF), Since
`the pyrolysis tube used in this work is of different dimensions
`(straight, 3 mm inner diameter) than that used by Knickel'
`bein at at”. (nozzle geometry ), it was important to character-
`ize the cracking of HFPO as a function of tube temperature.
`The pyrolysis of HFPO was characterized by monitoring the
`mass spectrum of the products flowing out of the tube. The
`chamber pressure, due to HFPO flow, was 3>< 10"" " Torr
`(chamber base pressure was less than 8 X 10 ’ 8 Torr), which
`corresponded to an HFPO flow rate of about 0.2 sccm and a
`10 ms residence time in the hot zone of the pyrolysis tube.
`Quantitative mass spectrometric characterization of
`HFPO pyrolysis would require much effort due to its com-
`plicated cracking pattern, which includes CF *, CF2* , and
`
`J. Vac. Sci. Technol. 8, Vol. 9, No. 3, May/Jun 1991
`
`(A,
`
`intensityRatioor;10F;(ArbitraryUnits) N
`
`
`
`
`
`
`
`Operating Region
`
`450
`
`500
`
`550
`
`600
`
`650
`
`700
`
`750
`
`800
`
`Temperature (K)
`
`FlG. 2. Mass spectrometric intensity ratio ofCth to CF; during the pyro-
`lysis of HFPO. The mass spectrometer ionizer energy was 20 eV.
`
`

`

`1464
`
`Butterbaugh, Gray, and Sawin: Plasma—surface interactions in FC2 etching of Sio2
`
`1464
`
`for all studies in which a CF2 flux was used.
`One must also be concerned with the effect of the
`
`CF3 CFO and any unreacted HFPO on the etching reaction.
`In the case of HFPO, the etching of SiO2 by Ar + ions was
`monitored in the presence of various flux levels of room-
`temperature HFPO (Fig. 3). The physical sputtering rate
`was unaffected by the impingement of HFPO, so it was rea-
`sonable to assume that the HFPO had a negligible sticking
`probability on SiOz. This was not surprising as Winters28
`determined that the sticking probability of molecular species
`like CF4 , CF3 H, and CF3 Cl is less than 10 “ 7. It was expect-
`ed that the sticking probability of CF3 CFO was also negligi-
`ble since it is a stable molecular species. Thus, it was assumed
`that the only species in the pyrolyzed HFPO beam that can
`affect the etching of the $02 was the CF2 radical.
`CF2 fluxes of up to 5 X low/cm2 s were obtainable; at
`higher fluxes back-mixing of fluorocarbon species into the
`Kaufman ion source affected the quality of the data. This
`flux level compares very well with the typical CF2 flux strik-
`ing the surface in a CF4 plasma. LIF measurements” indi-
`cate that the CF2 concentration in a CF4 plasma is on the
`order of lO‘z/cm3, which corresponds to a surface flux of
`about ION/cm2 s at 300 K. F fluxes of up to 2 X 10‘7/cm2 3
`(corresponding to 1.4)(10‘3/cm3 in the plasma environ-
`ment) and Ar 4’
`fluxes of up to 3X10‘5/cmzs at 250 eV
`(corresponding to 0.5 mA/cm2 in the plasma environment)
`were obtainable.
`
`rill. RESULTS AND DISCUSSION
`
`In the first series of experiments, the ion-enhanced etching
`ofSiO2 in the presence of CF2 was examined. The results are
`shown in Fig. 4, where the etching yield (the number of SiO2
`
`0.3
`
`
`
`Yield(SiOZJAr+) O
`
` Etching
`
`1.5
`o" 0.51"
`HFPO Flux (1016icm2-s)
`
`7'2 H 2.5
`
`FIG. 3. Dependence of the etching yield on the flux of room-temperature
`HFPO showing that HFPO has negligible interaction with the ion-bom-
`barded surface.
`
`J. Vac. Sci. Technol. B, Vol. 9, No. 3, May/Jun 1991
`
`0.6
`
`
`
`
`
`EtchingYield(SiozlAm
`
` 0.5
`
`0.3 .
`
`0.2 '
`
`o
`
`5
`
`
`"o "
`15
`20
`"
`25
`Flux Ratio (CleAr+)
`
`FIG. 4. The etching of SiO2 with Ar+ ions in the presence of CF2 at ion
`energies of 150, 250, and 350 eV. The error bars indicate estimated uncer-
`tainty in instrument readings. Filled symbols indicate baseline sputtering in
`a “dirty” system. The solid lines are derived from the two-parameter model
`oqu. (9).
`
`molecules leaving the surface per incident Ar + ion) is plot-
`ted against the CF2 —to-Ar + flux ratio for ion energies of 150,
`250, and 350 eV. The yields with a clean system and no CF2
`flux are in reasonable agreement with previously published
`values.”18 After the system was exposed to HFPO, the ap—
`parent baseline physical sputtering rates were higher than in
`a well-cleaned system due to residual fluorocarbon conta-
`mination in the ion gun (filled symbols in Fig. 4). Etching of
`the SiOZ was not observed with a CF2 flux in the absence of
`ion bombardment. Previous studies have shown that CF2
`adsorbs to some extent on the surface of both virgin and
`sputtered SiO2 , but does not etch the SiO2 surface10 (see
`discussion above). We saw no indications of polymer film
`deposition on the surface by the CFZ. Evidence for the lack
`of a polymer film was that the SiO2 etching rate immediately
`returned to its baseline level after the CF2 flux was terminat-
`ed.
`
`It is evident in Fig. 4 that the Ar+ etching yield is en—
`hanced by the flux of CF2 on the surface; at an ion energy of
`150 eV, the saturated yield with CF2 is over three times the
`physical sputtering yield. The yields tend toward an asymp-
`totic value as the ratio of CF2 to Ar + increases; i.e., increas—
`ing amounts of CF2 have less efl‘ect on the yield. It appears
`that saturation of the yield occurs at lower flux ratios for
`higher energy ions. This last feature is counter-intuitive in
`that one would expect higher energy ions to be more eflicient
`at removing products from the surface, causing saturation to
`occur at higher flux ratios. An explanation of this behavior is
`that the CF2 sticking probability increases with ion energy.
`The more damaged surface, expected with higher energy ion
`bombardment, may offer more adsorption sites and increase
`
`
`
`

`

`1465
`
`Butterbaugh, Gray, and Sawln: Plasma—surface interactions in F62 etching of Si02
`
`1 $65
`
`the sticking probability of the CF2 radical.
`The CF2 radical interaction with SiO2 vs Si surfaces (Fig.
`5) illustrates the importance of fluorocarbon radicals in de-
`termining SiO2 :Si selectivity. CF2 enhances the Ar + sput-
`tering yield of SiO2 while suppressing that of Si, due to the
`formation of fluorocarbon surface films. Therefore, the se-
`lectivity of SiO2 sputtering to Si sputtering increases with
`increasing CF2 flux. This selectivity effect is quantified in
`Fig. 5 for an Ar + ion energy level of 250 eV and substrate
`temperature of 80 °C, and is likely to become more pro-
`nounced with decreasing ion energy and substrate tempera“
`ture. Formation of passivating fluorocarbon films on silicon
`is also enhanced through the use of carbonaceous, rather
`than inert ions.
`
`Figure 4 assumes that the etching yield is dependent only
`on the neutraHo-ion flux ratio and not on the absolute neu—
`
`tral or ion flux level; surface kinetic modeling presented be—
`low also predicts this dependence. This assumption was in
`vestigated by varying the fluxes of CF’Z and Ar + while
`holding the ratio constant (Fig. 6). It was found that any
`dependence of the yield on the absolute flux level is within
`the experimental uncertainty of these data over a 2.5 X
`change in absolute flux, validating the use of this ratio as a
`dimensionless system parameter. It was also found that the
`yield is independent of the substrate temperature over a
`range of 350—475 K, as expected, since thermal etching of
`SiO2 by CF2 has not been detected.
`In the next series of experiments, the ion—enhanced etch-
`ing of SiO2 in the presence of F was examined. The results of
`these experiments are shown in Fig. 7 in which the etching
`yield is plotted against the Faro-Ar"r flux ratio. At low ion
`energy ( 150 eV), the yield, in the presence of F, shows much
`the same dependence on the flux ratio as the yield in the
`presence of CE. There is an immediate enhancement of the
`
`(SiOzlAr'P)
`
`
`
`EtchingYield
`
`1
`
`2
`
`3
`
`4
`
`ion Flux (1 01 51cm2-s)
`
`FIG. 6. Dependence of the etching yield on the absolute flux level at a con-
`stant flux ratio. Circles indicate data taken for F/Ar * at a flux ratio of 7
`and an ion energy of 250 eV. Equares indicate data taken for CF: /Ar * at a
`flux ratio of 3 and an ion energy of 250 eV. The error bars indicate estimated
`uncertainty in instrument readings. i
`
`yield as the F flux is increased from zero; however, this en-
`hancement is almost twice that seen with CFZ. The begin—
`ning of saturation is clearly seen for 150 eV ion bombard-
`ment, but experimental limitations precluded the attainment
`ofsufliciently high flux ratios to see the onset of saturation at
`250 and 350 eV. It has been shown that SiO2 is etched spon-
`
`(SiOzlAF’)
`
`
`
`EtchingYield
`
`o "
`
`20
`
`"
`
`"
`
`" "
`
`"100
`
`14"
`
`Flux Ratio (Flam
`
`(SiOQISi)
`
`Selectivity
`
`
`
`
`
`
`
`EtchingYield(Si/Air+orSiOzlAr+l
`
`Flux Ratio (CleAr+)
`
`FIG. 5. Comparison ofthe efieet ofCF2 on the etching of SiO2 and Si. As the
`CF2 flux increases, the selectivity ofSAG2 over Si increases. Si data are taken
`from Ref. 25.
`
`ions in the presence of F at ion
`FIG. 7. The etching of Sit)2 with Ar ‘
`energies of 150, 250, and 350 eV. The error bars indicate estimated uncer-
`tainty in instrument readings. Filled symbols indicate baseline sputtering in
`a “dirty" system. The solid lines are fits to the two-parameter model of Eq.
`(8). Only data with R > 15 were used in fitting the model.
`
`J. Vac. Sci. Technol. B, Vol. 9, No. 3, May/Jun 1991
`
`

`

`1466
`
`Butterbaugh, Gray, and Sawin: Plasma—suriace interactions in Ft;Z etching of SiOz
`
`,
`
`1466
`
`taneously by F,6 but the contribution to the yield at these
`fluxes is less than 0.001.
`
`As with CF2 , it is found that any dependence of the yield
`on the absolute flux level ofF or Ar + is within the uncertain-
`
`ty of the experimental measurements (Fig. 6). The ion-en—
`hanced reaction of F is also independent of substrate tem-
`perature over the range of 350—475 K. The activation energy
`for the spontaneous etching of SiO2 by F has been measured
`at about 0.16 er’ but the contribution of spontaneous etch-
`ing to the ion—enhanced etching rate is less than 1% at the F
`fluxes used.
`
`The next step in this series of experiments was to examine
`the combined efl‘ects of CF2 and F fluxes on the etching
`yield. The experimental results are shown in Fig. 8 in the
`form of a three-dimensional surface with the yield plotted as
`a function of the CFz—to-Ar 4‘ flux ratio and the F~tc-Ar +
`flux ratio. At the F flux ratios studiedhere ( > 5), the CF2
`had little effect on the yield. It appears that the ion-enhanced
`etching yield of F with SiO2 is unaffected by the flux of CF2 .
`This is consistent with plasma process observations, that
`’SiO2 etching in F containing discharges is relatively unaf-
`fected by increases in fluorocarbon radicals.30 These results
`also indicate that CF2 is not a limiting reagent to the etching
`of SiO2 and is not necessary for the removal of oxygen. This
`is consistent with the experimental work of Donnelly et 01.3 1
`in which SiO2 films were etched in both CF4 + 02 and
`NF3 + Ar plasmas. Donnelly found that the SiO2 etching
`rate correlated with the F concentration and was insensitive
`
`to the source gas (CF4 or NF3 ). The work of McFeely er
`al.” suggests that adsorbed CFX species may preferentially
`
`
`
`EtchingYield
`
`(SiOzlAr+)
`
`ions in the presence of CF; and F at
`Fit}. 8. The etching of SiO2 with Ar "
`an ion energy of 250 eV. The data points indicate experimental data used to
`construct this surface.
`
`J. Vac. Sci. Technol. B, Vol. 9, No.3, May/Jun 1991
`
`bond to oxygen at the surface which would keep it from
`interfering with F adsorption on Si adsorption sites. The
`reactivity of CFx species to the surface oxygen probably
`modifies the distribution of O-containing etch products.
`The major difference between these beam simulation stud-
`ies and a real plasma environment is that the ions in the real
`plasma are molecular, mainly CFf. Tu et (11.‘5 studied the
`chemical etching of SiO2 with Ar+ and CF? ions in the
`presence of XeF2 . They found that the CF3+ ion has an etch-
`ing yield about twice that of Ar + , but that the dependence
`on XeFl flux was the same. It is expected that the trends in
`the yield with (SF2 and F under Ar * bombardment will be
`similar under CF; bombardment, but. with a 2 X enhance-
`ment, putting the measured yields of the beam experiments
`in reasonable agreement with yields measured in our labora—
`tory single-wafer plasma etcher.
`
`IV. MODELING OF ION-ENHANCED SURFACE
`KINETICS
`
`Comprehensive models of the ion-enhanced surface kinet-
`ics described above require complete knowledge of product
`distributions, surface bonding states, and radical sticking co-
`efficients in the presence of energetic ion bombardment. The
`continuous mixing of the surface adsorbates into the near—
`surface bulk region complicates the kinetic treatment of this
`problem beyond the scope of classical treatises on gas—solid
`reactions at well—defined surfaces. A rigorous consideration
`of ion‘induced surface reaction pathways suggests applica-
`tion of Monte Carlo type models to this problem to capture
`the random nature of this process,32 since it is unlikely that a
`simple analytical theory ofchemical sputtering will be found
`to parallel Sigmund’s classical treatment of physical sputter—
`ing.33 Nevertheless, a few researchers have attempted to
`model ion~enhanced surface kinetics on the basis of steady-
`state yield data as a function of neutral-to—ion flux ratios.
`Gerlach~Meyer34 first modeled ion-enhanced etching of sili~
`con by XeF2 and C12 with noble gas ions of various masses,
`using yield information and neutral—todon flux ratios as we
`present here. Although many physical interpretations given
`to this model have since been invalidated, we follow a similar
`mathematical development that reveals great similarity be—
`tween the eifect of increasing ion energy and the effect of
`increasing ion mass modeled by Gerlach-Meyer.
`Winters and Coburn35 state that the basic steps for a
`chemical etching process parallel those in many heterogen-
`eous catalysis mechanisms. These steps include (1) adsorp-
`tion of reactive gas species on the surface, (2) chemical reac-
`tion to produce a product molecule, and (3) desorption of
`the product molecule into the gas phase, liberating a clean
`adsorption site. Depending on the specific etchant-substrate
`system, ion bombardment may enhance (or impede) any or
`all of these steps. At steady state, all steps proceed at the
`same rate, but one is rate-limiting; our data indicate both
`neutral adsorption limited and ion flux limited etching re-
`gimes as a function of the neutral-to—ion flux ratio and the
`ion energy. The ion-enhanced etching of Si and SiO2 by
`XeF2 has been studied extensively; several proposed ion-
`enhanced etching models based on these studies have been
`
`
`
`

`

`1457
`
`Butterbaugh, Gray, and Sawin: Plasma—surface interactions in FC2 etching of Sit)2
`
`1467
`
`summarized by Coburn and Winters,36 who present data
`which discredit a number of them. Serious controversy has
`evolved concerning the suggestion that ion bombardment
`enhances the adsorption of fluorine, which is subsequently
`involved in a rapid thermally driven reaction with the da-
`maged lattice.22 Winters37 has shown that Xer does not
`spontaneously etch an ion-damaged lattice at significantly
`increased rates and that continuous ion mixing of the surface
`adsorbates into the lattice is necessary to accelerate the reac-
`tion.
`
`We have constructed a simplified mechanistic model that
`decouples radical adsorption/recombination effects from
`the ion-driven reaction and mixing steps. The exact ion—en-
`hanced chemical pathway is unknown and involves multiple
`surface layers, where 5in -type species become fully fluorin-
`ated as the near-surface region is “mixed” by the ballistic
`effect of the ions. We have represented this process with glow
`bal reaction steps which attempt to capture the correct over-
`all stoichiometry, but should not be taken as elementary suru
`faceareaction steps. The correct functional dependence of
`the steady—state etching yields on neutral-to—ion flux ratios is
`well represented when the ion‘enhanced reaction rates are
`assumed to be first order in ion flux and fluorine adsorbate
`
`concentration,