Influence of reactor wall conditions on etch processes in inductively
`coupled fluorocarbon plasmas
`M. Schaepkens, R. C. M. Bosch,a) T. E. F. M. Standaert, and G. S. Oehrleinb)
`Department of Physics, University at Albany, State University of New York, Albany, New York 12222
`J. M. Cook
`Lam Research Corporation, Fremont, California 94538-6470
`共Received 25 June 1997; accepted 30 January 1998兲
`
`The influence of reactor wall conditions on the characteristics of high density fluorocarbon plasma
`etch processes has been studied. Results obtained during the etching of oxide, nitride, and silicon in
`an inductively coupled plasma source fed with various feedgases, such as CHF3 , C3F6 , and
`C3F6/H2 , indicate that the reactor wall temperature is an important parameter in the etch process.
`Adequate temperature control can increase oxide etch selectivity over nitride and silicon. The loss
`of fluorocarbon species from the plasma to the walls is reduced as the wall temperature increased.
`The fluorocarbon deposition on a cooled substrate surface increases concomitantly, resulting in a
`more efficient suppression of silicon and nitride etch rates, whereas oxide etch rates remain nearly
`constant. © 1998 American Vacuum Society. 关S0734-2101共98兲03504-0兴
`
`I. INTRODUCTION
`Etching of trenches or contact holes into silicon dioxide is
`an indispensable process in modern integrated circuit fabri-
`cation technology. A high oxide etch rate and etch selectivity
`of oxide to silicon and nitride are important requirements
`that need to be met in order for etch processes to be appli-
`cable in industrial manufacturing. Etch processes employing
`low pressure high density fluorocarbon discharges are ex-
`pected to meet
`these demands, and have been studied
`extensively.1–6
`Low pressure high density fluorocarbon plasma processes,
`however, have been found to suffer from process drifts.7 At
`low pressure, plasma-wall interactions are of significant im-
`portance in determining the discharge chemistry. Process
`drifts have been attributed to changes in reactor wall
`conditions.8,9 In order to limit changes in reactor wall condi-
`tions and to create a stable and reproducible process environ-
`ment, it is common in semiconductor processing to ‘‘sea-
`son’’ a reactor. A fundamental understanding of
`the
`important mechanisms operating in the conditioning proce-
`dure is lacking, however.
`This article presents results of a study of the influence of
`reactor wall conditions on the stability of etch processes.
`Results obtained during the etching of oxide, nitride, and
`silicon in an inductively coupled plasma source fed with
`various feedgases, such as CHF3 , C3F6 , and C3F6/H2 , will
`be reported. Explanations of the observed effects will be pre-
`sented.
`
`II. EXPERIMENTAL SETUP
`The high-density plasma source used in this work is a
`radio frequency inductively coupled plasma 共ICP兲 source of
`planar coil design. This plasma source has also been referred
`
`a兲Permanent address: Eindhoven University of Technology, P.O. Box 513,
`5600 MB Eindhoven, The Netherlands.
`b兲Electronic mail: oehrlein@cnsibm.albany.edu
`
`to in literature as transformer coupled plasma 共TCP兲 and ra-
`dio frequency induction 共RFI兲 source. A schematic outline of
`the ICP reactor used is shown in Fig. 1. It is similar to the
`one described by Keller et al.10
`The apparatus consists of an ultrahigh vacuum 共UHV兲
`compatible processing chamber in which the plasma source
`and a wafer holding electrostatic chuck are located. The cen-
`ter part of the ICP source is a planar, 160 mm diameter
`induction coil that is separated from the process chamber by
`a 19.6 mm thick, 230 mm diameter quartz window. The coil
`is powered through a matching network by a 13.56 MHz,
`0–2000 W power supply. The results presented here were all
`obtained at an inductive power level of 1400 W.
`A plasma confinement ring that can hold multipole mag-
`nets is located below the quartz window. For this work, no
`magnets were placed into the confinement ring, but the ring
`itself confines the plasma. The confinement ring has poor
`thermal contact with other parts of the reactor and its tem-
`perature cannot be controlled independently. In the rest of
`this article, the confinement ring will be referred to as the
`reactor wall 共since this is the surface that the plasma is in
`contact with兲. The inner diameter of the confinement ring is
`20 cm.
`Wafers with diameters of 125 mm are placed on a bipolar
`electrostatic chuck during processing. The chuck is located 7
`cm downstream from the ICP source and allows the wafer to
`be RF biased and cooled during processing. A helium pres-
`sure of 5 Torr is applied to the backside of the wafer during
`the experiment to achieve a good thermal conduction be-
`tween the wafer and the chuck.11 The electrostatic chuck is
`cooled by a refrigerator. A variable frequency RF power sup-
`ply 共500 kHz–40 MHz, 0–300 W兲 is used to bias the wafer
`for etching experiments. The experiments reported in this
`work were all performed at 3.4 MHz.
`Substrates placed at the center of a wafer can be moni-
`tored by in situ ellipsometry. Plasma diagnostics like a re-
`tractable Langmuir probe and optical emission spectroscopy
`
`2099
`
`J. Vac. Sci. Technol. A 16(cid:132)4(cid:133), Jul/Aug 1998
`
`0734-2101/98/16(cid:132)4(cid:133)/2099/9/$15.00
`
`©1998 American Vacuum Society
`
`2099
`
`IP Bridge Exhibit 2014
`TSMC v. IP Bridge
`IPR2016-01376
`Page 0001
`
`

`

`2100
`
`Schaepkens etal.: Influence of reactor wall conditions
`
`2100
`
`FIG. 1. Schematic of ICP source setup.
`
`FIG. 2. Nitride etch rate and oxide etch rate as a function of the time that the
`plasma was switched on 共1400 W inductive power, 6 mTorr operating pres-
`sure, ⫺100 V self-bias兲.
`
`共OES兲 can be used for gas-phase characterization. With the
`retractable Langmuir probe it is possible to make a scan of
`the ion current density over 70% of the wafer. The values
`reported here were measured in the center of the reactor, 2
`cm above the wafer surface.
`The process chamber is pumped using a 450 l /s turbo-
`molecular pump backed by a Roots blower and a mechanical
`pump. The process gases are admitted into the reactor
`through a gas inlet ring located just under the quartz window.
`The pressure is measured with a capacitance manometer.
`Pressure control is achieved by an automatic throttle valve in
`the pump line.
`The ICP chamber is connected to a wafer handling cluster
`system that allows all sample transport to occur under ultra-
`high vacuum conditions. Processed samples can be trans-
`ported from the ICP reactor to a surface analysis chamber for
`X-ray photoelectron spectroscopy 共XPS兲 without exposure to
`air.
`
`III. RESULTS AND DISCUSSION
`
`A. Investigation of process drift
`
`It was found in an earlier study by this group that oxide
`could be etched at a high rate and at the same time selec-
`tively to nitride in a C3F6/H2 共40 sccm/15 sccm兲 discharge at
`high inductive power 共1400 W兲 and low operating pressure
`共6 mTorr兲 when a sufficiently high RF bias power 共200 W
`corresponding to ⫺100 V self-bias兲 is applied to the wafer.12
`In that study, however, the samples were etched for only a
`relatively small time period. When etching for longer times,
`it was observed that etch rates of both oxide and nitride were
`dependent on the time that the plasma was switched on; see
`Fig. 2. The nitride etch rate decreases significantly as a func-
`tion of processing time, whereas the oxide etch rate remains
`at a near constant value. This results in an increase of oxide-
`to-nitride selectivity with etching time.
`Two time-dependent processes were initially suggested to
`be responsible for the changes in etch characteristics as a
`function of time, namely 共1兲 fluorocarbon contamination of
`the reactor walls during processing and 共2兲 increasing tem-
`
`J. Vac. Sci. Technol. A, Vol. 16, No. 4, Jul/Aug 1998
`
`perature of the reactor. In order to distinguish between these
`processes,
`the following experiment was performed; see
`Fig. 3.
`Blanket nitride samples were etched at the same condi-
`tions and the rates are plotted versus processing time. Results
`for three consecutive etch experiments are shown. The first
`experiment was performed starting with a reactor that has
`been cleaned with an oxygen plasma prior to the experiment.
`The reactor is initially at room temperature. In between ex-
`periments the reactor was allowed to cool down, but no oxy-
`gen plasma cleaning was performed.
`It can be seen that during each run there is a significant
`decrease of the nitride etch rate. The initial value of the etch
`rate in each run stays at a constant level. This experiment
`
`FIG. 3. Nitride etch rate as a function of processing time for three consecu-
`tive etch experiments. The first experiment was performed starting with a
`reactor that has no fluorocarbon material deposited at the walls 共reactor was
`cleaned with an oxygen plasma兲. Between the etch experiments the reactor
`was allowed to cool to room temperature, but no oxygen plasma cleaning
`was performed 共1400 W inductive power, 6 mTorr operating pressure, ⫺100
`V self-bias兲.
`
`IPR2016-01376 Page 0002
`
`

`

`2101
`
`Schaepkens etal.: Influence of reactor wall conditions
`
`2101
`
`FIG. 4. Wall temperature as a function of processing time. The triangles are
`experimental data; the curves show the model.
`
`shows that the heating of the reactor is the process that is
`dominant in producing the observed effect.
`
`B. Time dependence of the reactor wall temperature
`
`The time dependence of the wall temperature has been
`measured by attaching a thermocouple to the confinement
`ring. The resulting data for both a 600 and 1400 W C3F6/H2
`共40 sccm/15 sccm兲 plasma at 6 mTorr operating pressure is
`plotted in Fig. 4. Similar experiments were obtained using
`different fluorocarbon gases, such as CHF3 and C3F6 . The
`time dependence of the reactor wall temperature changed
`little with feedgas chemistry.
`A simple model that describes the time dependence of the
`wall temperature T has been developed. In the model it is
`assumed that the power dissipated in the wall results from
`collisions of ions with the wall. The ions have an energy
`equal to the sheath voltage between the floating wall and the
`plasma. The energy loss factor for the wall is assumed to be
`Stefan-Boltzmann radiation. The resulting differential equa-
`tion is
`m(cid:149)c(cid:149)dT/dt⫽ICD(cid:149) A(cid:149)V sh⫺e(cid:149)␴(cid:149)A(cid:149)共T 4⫺T RT
`4 兲,
`where m is the mass of the confinement ring, c is the heat
`capacitance of the anodized aluminum confinement ring,
`ICD is the ion current density, A is the area of reactor wall,
`V sh is the sheath voltage between the floating wall and the
`plasma, e is the emissivity of the anodized aluminum con-
`finement ring, ␴ is the Stefan-Boltzmann constant, and T RT
`is the room temperature.
`The ion current density was measured by Langmuir probe
`measurement. The sheath voltage was calculated to be 15 V
`by using an estimated value of 3 eV for the electron tempera-
`ture. Values between 0.4 and 0.5 for the emissivity were
`determined from the cooling rate of the wall after the plasma
`was shut off. The differential equation was solved numeri-
`cally, resulting in the curves plotted in Fig. 4. The agreement
`between the model and the experimental data is satisfactory.
`
`共1兲
`
`JVST A - Vacuum, Surfaces, and Films
`
`FIG. 5. 共A兲Passive deposition rates are measured as a function of time on a
`non-cooled wafer. Three fluorocarbon feedgas chemistries are compared,
`i.e., CHF3 共40 sccm兲, C3F6 共40 sccm兲, and C3F6/H2 共40 sccm/15 sccm兲 at
`1400 W inductive power and 6 mTorr operating pressure. The processing
`time is proportional to the wafer temperature. For C3F6 the 600 W data are
`also included. The temperature/time correspondence is different in this case.
`共B兲 The etch rates of fluorocarbon films deposited at the same conditions at
`which they are etched are measured as a function of RF bias power at 1400
`W inductive power, 6 mTorr operating pressure plasmas fed with various
`fluorocarbon gases, i.e., CHF3 共40 sccm兲, C3F6 共40 sccm兲, and C3F6 H2 共40
`sccm/15 sccm兲.
`
`C. Temperature dependence of plasma–wall
`interactions
`
`In order to study the temperature dependence of the pro-
`cesses occurring at a non-biased wall that has poor thermal
`contact, an experiment was performed on a non-biased wafer
`that is lifted above the chuck by three small Teflon studs.
`The thermal conductivity between the cooled electrostatic
`chuck and the wafer is poor in this case. The only energy
`loss factor is the Stefan-Boltzmann radiation from the wafer
`to the chuck. Apart from a dissimilarity due to differences in
`heat capacitance of the wafer and the confinement ring, the
`behavior of wafer and wall will be similar. The temperature
`of the wafer will increase with time until it reaches a maxi-
`mum after which it will no longer change.
`Figure 5共A兲 shows the results of this experiment per-
`formed in discharges fed with CHF3 共40 sccm兲, C3F6 共40
`sccm兲, and C3F6/H2 共40 sccm/15 sccm兲 using 1400 W induc-
`tive power and 6 mTorr operating pressure. The fluorocarbon
`passive deposition rates are plotted as a function of process-
`ing time 共i.e., increasing substrate temperature兲. The initial
`deposition rates are similar to the deposition rates obtained
`with a clamped wafer when a helium backside pressure is
`applied; see Fig. 5共B兲. As the processing continues, and thus
`
`IPR2016-01376 Page 0003
`
`

`

`2102
`
`Schaepkens etal.: Influence of reactor wall conditions
`
`2102
`
`function of RF bias power is plotted in Fig. 5共B兲. It shows
`that C3F6 has a lower threshold RF bias power for etching
`than C3F6/H2 even though the passive deposition rates 共i.e., 0
`W RF bias power兲 are similar. It also shows that CHF3 has a
`higher threshold RF bias power than C3F6 , even though the
`passive deposition rate is significantly lower. From this com-
`parison it can be suggested that, in fluorocarbon etching, the
`RF bias power and the substrate temperature play a similar
`role, i.e., an energy source for enabling chemical reactions. It
`further shows that the transition from deposition to etching is
`strongly dependent on the feedgas chemistry. The reason
`why CHF3 and C3F6/H2 deposited fluorocarbon films require
`a higher threshold RF bias power or a higher substrate tem-
`perature is due to differences in chemical composition. The
`F/C ratios of the CHF3 and C3F6/H2 fluorocarbon films are
`lower than the F/C ratio of C3F6 fluorocarbon material due to
`the fluorine scavenging effect of hydrogen.12,14
`
`D. Temperature dependence of the gas phase
`chemistry
`
`The results from Sec. III C indicate that fluorocarbon spe-
`cies that are precursors for deposition on low temperature
`surfaces do not deposit on hot surfaces and remain in the
`plasma. The 共low temperature surface兲 deposition precursor
`density thus increases. This is consistent with results from
`Chinzei et al.,15 who reported that in a C4F8 ICP no polymer
`deposition occurs on surfaces at temperatures higher than
`200 °C. The density of CF and CF2 radicals, as measured by
`mass spectrometry, is one order of magnitude higher in a
`200 °C heated reactor in comparison to a reactor at 30 °C.
`Hikosaka et al.16 also reported mass spectrometric results
`showing that, as the temperature of their quartz wall in-
`creased, the CF3 radical density increased due to a reduced
`surface loss of these species. They additionally found an in-
`creased CO density due to enhanced etching of the quartz
`wall at higher temperatures and a decrease in F density
`which they ascribed to the enhanced F absorption on the
`quartz wall. Fluorine absorption by an aluminum wall has
`also been reported by Chinzei et al. In mass spectrometry
`measurements Maruyama et al.17 observed that for CF4 plas-
`mas an increase in the CFx radical density occurs as the wall
`temperature goes up. O’Neill and Singh9 who performed ul-
`traviolet 共UV兲-absorption spectroscopy in order to examine
`the CF2 radical density in a C2F6/CF4 plasma also reported a
`steady increase of this density as processing proceeds. They
`ascribed this effect to increasing polymer deposition on the
`wall, but also mentioned that the reactor wall temperature
`may play a significant role.
`Figure 7 shows optical emission spectra taken on a 1400
`W, 6 mTorr, 40 sccm C3F6 discharge in both a relatively cold
`共60 °C兲 and a hot 共300 °C兲 reactor. Significant intensity dif-
`ferences can be observed in the emission from various car-
`bon containing species. In Fig. 8 the peak intensities of CF2 ,
`C2 , CO and also those of atomic fluorine and oxygen are
`plotted as a function of reactor wall temperature. The inten-
`sities are normalized to their value in a hot reactor.
`It shows that the initial value, i.e., in a cold reactor, of the
`
`FIG. 6. 共A兲Simulated correlation between the ellipsometric angle ⌿ mea-
`sured on a crystalline silicon wafer and the real part of the refractive index
`of the silicon. 共B兲 Ellipsometric angle ⌿ measured as a function of time on
`a non-cooled crystalline silicon wafer in a C3F6 plasma at 1400 W inductive
`power and 6 mTorr operating pressure.
`
`the temperature of the substrate increases, the deposition rate
`decreases until a point in time where no net deposition takes
`place. Beyond this point the deposited fluorocarbon material
`will be removed until no fluorocarbon is left on the c-Si
`wafer. It thus shows that on a surface with a temperature
`above a certain value no net deposition will take place.
`The temperature rise of the non-cooled wafer is quantified
`by measuring the ellipsometric angle ⌿ on the hot, clean
`c-Si wafer after the C3F6 experiment from Fig. 5. The ⌿
`angle is strongly related to the real part of the refractive
`index, Re共n兲, of the silicon.13 The correlation between ⌿ and
`Re(n) is calculated using an ellipsometry simulation routine,
`and is plotted in Fig. 6共A兲. Using the temperature coefficient
`of Re共n兲 of silicon at 632.8 nm, ␦n/␦T⫽4.52⫻10⫺4 K⫺1, it
`can be calculated from the data in Fig. 6共B兲 that the tempera-
`ture of the wafer, after all fluorocarbon material has been
`removed, is at a value around 360 °C. This temperature cor-
`responds well with the stable wall temperature in a 1400 W,
`6 mTorr plasma; see Fig. 4.
`In Fig. 5 it is shown that the point where the net deposi-
`tion rate is zero occurs earlier in time 共i.e., at a lower tem-
`perature兲 for C3F6 共40 sccm兲 than for either CHF3 共40 sccm兲,
`which has a much lower initial deposition rate, or C3F6 共40
`sccm兲 with 15 sccm H2 added to it, which has a similar initial
`deposition rate.
`The results from this experiment allow an interesting
`analogy to be made between substrate temperature and RF
`bias power when etching fluorocarbon material deposited on
`a cooled wafer. The etch rate of fluorocarbon material 共de-
`posited at the same condition as etching is performed兲 as a
`
`J. Vac. Sci. Technol. A, Vol. 16, No. 4, Jul/Aug 1998
`
`IPR2016-01376 Page 0004
`
`

`

`2103
`
`Schaepkens etal.: Influence of reactor wall conditions
`
`2103
`
`FIG. 7. Comparison of optical emission spectra measured on a C3F6 plasma
`at 1400 W inductive power and 6 mTorr operating pressure. The left panel
`shows the spectrum measured in a cold 共60 °C兲 reactor and the right panel
`shows the spectrum measured in a hot 共300 °C兲 reactor.
`
`FIG. 9. Ion current density as a function of reactor wall temperature 共1400
`W inductive power, 6 mTorr operating pressure兲.
`
`intensity of carbon containing species is around 50%–60%
`of the final, hot reactor value. The fluorine intensity is ini-
`tially 10% higher than its final intensity. This indicates that
`the concentration ratio of carbon containing species, i.e.,
`deposition precursors, over atomic fluorine, i.e., etch precur-
`sor, increases as the reactor temperature increases.
`The increase in the CO intensity coincides with a decrease
`of the atomic oxygen signal. The CO formation can therefore
`be explained by an increased carbon density in the plasma
`reacting with atomic oxygen.
`At a constant pressure, the particle density decreases as
`the reactor temperature increases. Therefore, the electron
`mean free path and thus the electron temperature will in-
`crease. The constant emission intensity at
`temperatures
`higher than 200 °C may therefore correspond to a decreasing
`radical density. A change in the electron temperature, how-
`ever, should not influence the relative intensity of F emission
`
`FIG. 8. Emission intensities measured in a C3F6 plasma at 1400 W inductive
`power and 6 mTorr operating pressure. The intensities of the various species
`are normalized to their value in a hot reactor.
`
`JVST A - Vacuum, Surfaces, and Films
`
`to CF2 and C2 , i.e., the gas phase F/C ratio, assuming ap-
`proximately similar excitation cross sections for these spe-
`cies.
`
`E. Temperature dependence of the ion current
`density
`
`The ion current density has been measured as a function
`of reactor wall temperature in CHF3 , C3F6 , and C3F6/H2
`discharges; see Fig. 9. The ion current density decreases as
`the wall temperature increases. The ion density may decrease
`more significantly as a function of reactor wall temperature
`than the ion current density does, assuming that the electron
`temperature increases. A higher electron temperature leads to
`an increase in the ion acoustic velocity, which results in a
`higher ion current density for a given ion density.
`The decrease in the ion density can partially be explained
`by the fact that the overall particle density goes down as the
`temperature of the reactor walls increases.
`
`F. Temperature dependence of plasma-surface
`interactions
`1.Fluorocarbondeposition
`The results previously presented suggest a time depen-
`dence of the fluorocarbon deposition on cold surfaces, such
`as cooled substrates. Figure 10 shows the passive fluorocar-
`bon deposition measured as a function of reactor wall tem-
`perature in a C3F6 共40 sccm兲 and C3F6/H2 共40 sccm/12 sccm兲
`plasma at 1400 W inductive power and 6 mTorr operating
`pressure. The deposition rate shows a maximum as a func-
`tion of wall temperature. This can be explained as follows.
`Initially passive deposition will take place on the walls at
`the same rate as on the wafer. As the walls heat up, the
`passive deposition rate on the walls will decrease and ulti-
`mately fluorocarbon removal will occur. The deposition pre-
`cursor density in the gas phase increases at the same time,
`raising the passive deposition on the cooled wafer. After all
`the fluorocarbon material has been removed from the walls,
`the deposition precursor density in the gas phase and thus the
`
`IPR2016-01376 Page 0005
`
`

`

`2104
`
`Schaepkens etal.: Influence of reactor wall conditions
`
`2104
`
`FIG. 10. 共A兲 Threshold RF bias power for etching, 共B兲 passive fluorocarbon
`deposition rate, and 共C兲 the fluorocarbon refractive index as a function of
`reactor wall temperature in a C3F6 共40 sccm兲 and a C3F6 /H2 共40 sccm/12
`sccm兲 plasma at 1400 W inductive power and 6 mTorr operating pressure.
`
`passive deposition rate at the wafer will decrease again 共but
`still be higher than the initial rate兲. This is consistent with
`results reported by Tsukada et al.,18 who reported that poly-
`mer deposition on the coldest surface is faster than on other
`surfaces. The reason that the maximum in the deposition rate
`occurs at a higher temperature in C3F6/H2 than in C3F6 is
`consistent with the data in Fig. 5, which show that a higher
`temperature is needed to remove fluorocarbon deposited in
`C3F6/H2 .
`The refractive index of the deposited fluorocarbon mate-
`rial shown in Fig. 10 has been shown to be inversely propor-
`tional to the F/C ratio.1 The maximum in deposition rate
`coincides with a maximum in the refractive index. Figure 10
`also shows the temperature dependence of the threshold RF
`bias power, defined as the RF bias power at which net fluo-
`rocarbon deposition is prevented by ion bombardment, but
`no net substrate etching occurs. The data show that the etch-
`ing of the fluorocarbon material is increasingly difficult as
`the fluorocarbon deposition rate increases and as the F/C
`ratio decreases. A maximum in the RF threshold bias power
`is observed at the same temperature as the maximum in the
`deposition rate and refractive index occurs.
`
`2.Oxideetching
`Figure 11共A兲 shows the oxide etch rates monitored for
`longer time periods than in Fig. 2 for C3F6/H2 discharges at
`1400 W inductive power, 6 mTorr operating pressure at
`⫺100 V self-bias voltage, and different H2 additions. In the
`temperature regime up to 200 °C, the oxide etch rate is fairly
`temperature independent.
`It
`is also independent of
`the
`feedgas chemistry. The slight decrease in the etch rate can be
`
`J. Vac. Sci. Technol. A, Vol. 16, No. 4, Jul/Aug 1998
`
`FIG. 11. 共A兲 Oxide etch rates and 共B兲 nitride etch rates as functions of wall
`temperature measured in C3F6 共40 sccm兲, C3F6/H2 共40 sccm/12 sccm兲, and
`C3F6/H2 共40 sccm/15 sccm兲 discharges at 1400 W inductive power and 6
`mTorr operating pressure at a RF bias power of 200 W 共i.e., ⫺100 V
`self-bias兲.
`
`explained by the decreasing ion current density since the
`oxide etch mechanism has been shown to be mainly ion
`driven.6,19 For higher temperatures a strong dependence on
`feedgas chemistry is observed, which cannot be explained by
`direct reactive ion etching of the oxide. When 15 sccm of
`H2 is added to 40 sccm of C3F6 , the etch process goes into a
`deposition mode at a temperature of around 250 °C. For a
`H2 addition of 12–40 sccm C3F6 the etch rate displays a
`minimum. For pure C3F6 , the oxide etch rate is temperature
`independent.
`The minimum in the etch rate at 12 sccm H2 addition is
`consistent with the maximum in the passive deposition rate
`as a function of temperature. It is also consistent with the
`maximum in the fluorocarbon refractive index 共minimum in
`F/C ratio兲. The minimum in the etch rate occurs also at 15
`sccm H2 addition, but then as a maximum in the deposition
`rate 共under biased conditions兲.
`The fact that for pure C3F6 no minimum in the oxide etch
`rate is seen, although there is a maximum in the passive
`deposition rate, indicates that the oxide etch mechanism is
`more complicated than just reactive ion sputtering.
`An explanation for this behavior can be found by observ-
`ing that the oxide etch rate as a function of self-bias voltage
`exhibits: 共1兲 fluorocarbon deposition, 共2兲 fluorocarbon sup-
`pression, and 共3兲 oxide sputtering.6,12,14 In the fluorocarbon
`deposition regime net fluorocarbon deposition takes place.
`As the RF bias power is increased and a larger self-bias
`
`IPR2016-01376 Page 0006
`
`

`

`2105
`
`Schaepkens etal.: Influence of reactor wall conditions
`
`2105
`
`voltage develops, the fluorocarbon suppression regime is en-
`tered and oxide etching occurs. The sputtering mechanism is
`however suppressed by the presence of a relatively thick
`steady-state fluorocarbon film. As the RF bias power is fur-
`ther increased, the oxide surface becomes free of fluorocar-
`bon material and chemical sputtering can occur. The etch
`rate is then proportional to the square root of the ion energy.
`The dependence of the oxide etch rate on RF bias was
`examined for different fluorocarbon feedgas chemistries. The
`passive deposition rate is similar for both C3F6 and C3F6/H2 ,
`but the RF bias threshold power for etching and the RF
`power at which the sputtering regime is reached are higher
`for C3F6/H2 . This is consistent with the etching of fluorocar-
`bon material, as shown in Fig. 5. For sufficiently high RF
`bias powers the oxide etch rates in C3F6 and C3F6/H2 are
`similar. At that point the sputtering regime is reached and the
`etching is mainly dependent on the ion flux to the surface.
`Figure 9 showed that the ion current densities for C3F6 and
`C3F6/H2 are fairly similar.
`In Fig. 10 the temperature dependence of the RF threshold
`power required to induce oxide etching in a C3F6 共40 sccm兲
`and a C3F6/H2 共40 sccm/12 sccm兲 discharge is shown. It
`shows a maximum, similar to the fluorocarbon deposition
`rate. This can be understood since at the etch threshold point
`the deposition and etching processes balance each other. If
`the deposition rate increases, one needs to compensate for
`this by increasing the bias. The opposite holds for a decrease
`in the threshold value.
`
`3.Nitrideandsiliconetching
`Figure 11共B兲 shows the nitride etch rate as a function of
`reactor wall temperature for C3F6/H2 discharges at 1400 W
`inductive power, 6 mTorr operating pressure at ⫺100 V self-
`bias voltage, and different H2 additions. Figure 11共B兲 is con-
`sistent with Fig. 2. It shows that the nitride etch rate depen-
`dence on the reactor wall temperature is different from the
`oxide etch rate dependence. This can be explained by taking
`in account results obtained when nitride samples were etched
`as function of the RF bias power for C3F6 and C3F6/H2 . For
`RF bias powers where oxide was being etched by a sputter-
`ing mechanism, the nitride etching was suppressed, and the
`degree of suppression depended strongly on the feedgas
`chemistry. Similar results were obtained for silicon sub-
`strates with the difference that these were suppressed to an
`even lower level.12,14,20
`
`4.Model
`
`In Figure 12 all the above information is combined. It
`shows schematically the influence of increasing wall tem-
`perature on the oxide etch rate as a function of self-bias
`voltage in both the C3F6 and the C3F6/H2 cases. As the pas-
`sive deposition rate increases, the RF bias threshold shifts to
`a higher value. For the C3F6 case etching at 100 V self-bias
`共⫽200 W RF bias power兲 still occurs through the mechanism
`of oxide sputtering. No temperature dependence can there-
`fore be observed; see Fig. 12共A兲. For C3F6/H2 , however,
`
`JVST A - Vacuum, Surfaces, and Films
`
`FIG. 12. Schematic view of the oxide and nitride etch characteristics for
`various wall temperatures for 共A兲 C3F6 processing and 共B兲 C3F6 /H2 pro-
`cessing. The thick solid curves represent the oxide etching, and the dotted
`lines the nitride etching. The sputtering curve 共thin solid line兲 is also in-
`cluded.
`
`etching has shifted into the fluorocarbon suppression regime,
`resulting in a temperature dependent etch behavior; see Fig.
`12共B兲.
`Joubert et al.21 used oxide etch rate versus self-bias volt-
`age curves similar to the ones presented in Fig. 12 in order to
`explain reactive ion etching 共RIE兲 lag effects. Since differ-
`ential charging at the bottom of etched features occurs, ions
`entering the feature are retarded, resulting in a decreased etch
`yield or even deposition. It is expected that the reactor wall
`temperature effect will enhance the RIE lag effect signifi-
`cantly.
`The wall temperature dependence of the nitride etch rate
`can also be explained using the information presented above.
`Figure 12 also schematically shows the effect of increasing
`wall temperature for both C3F6 and C3F6/H2 on the nitride
`etch rate. As the passive deposition rate increases, the RF
`bias threshold shifts to a higher value, and the etch rate val-
`ues at a certain self-bias voltage are being suppressed. This is
`the case for both C3F6 , Fig. 12共A兲, and C3F6/H2 etching, Fig.
`12共B兲, at 100 V self-bias 共⫽200 W RF bias power兲, resulting
`in a temperature dependent etch behavior. The same expla-
`nation holds for the silicon etching data.
`It can thus be suggested that the increased reactor wall
`temperature results in a more efficient suppression of the
`substrate etch rate due to enhanced fluorocarbon deposition
`which leads to a thicker steady-state fluorocarbon film. The
`
`IPR2016-01376 Page 0007
`
`

`

`2106
`
`Schaepkens etal.: Influence of reactor wall conditions
`
`2106
`
`FIG. 13. The silicon etch rate and the thickness of the steady-state fluoro-
`carbon film measured as a function of reactor wall temperature. As the film
`thickness increases, the silicon etch rate decreases 共1400 W, 6 mTorr, ⫺100
`V self-bias兲.
`
`suggested mechanism is supported by the data in Fig. 13,
`which show the silicon etch rate and the steady-state fluoro-
`carbon film thickness 共determined by ellipsometry兲 as a
`function of reactor wall temperature in a C3F6/H2 plasma at
`1400 W inductive power and 6 mTorr operating pressure.
`The silicon etch rate decreases as the fluorocarbon film thick-
`ness increases until the process moves from an etching mode
`into a deposition mode. This is consistent with results from
`other studies that report that the etch rate of a substrate is
`inversely
`proportional
`to
`the
`fluorocarbon
`film
`thickness.12,14,20
`
`G. Temperature dependence of process uniformity
`
`In the above it has been shown that the fluorocarbon
`deposition rate is a key parameter in the selective oxide etch-
`ing process. The temperature dependence of the process uni-
`formity is now investigated by measuring the fluorocarbon
`deposition rate as a function of position on the wafer 共using
`ex situ spatially resolved ellipsometry兲. In Fig. 14 the fluo-
`rocarbon deposition rate profile obtained in a C3F6/H2 共40
`sccm/15 sccm兲 discharge at 1400 W inductive power and 6
`mTorr operating pressure in a reactor with a wall tempera-
`ture of 100 °C is compared to the profile obtained at 300 °C.
`The fluorocarbon deposition rate in the 300 ° C