`INNOLUX CORP. v. PATENT OF SEMICONDUCTOR ENERGY
`LABORATORY CO., LTD.
`
`IPR2013-00064
`
`
`
`Anisotropic Si deep beam etching with profile control using SF6/O2 Plasma
`H. Zou
`
`Microsystem Technologies 10 (2004) 603–607 Ó Springer-Verlag 2004
`DOI 10.1007/s00542-003-0338-3
`
`603
`
`In our investigation, SF6 and O2 are chosen for RIE
`etching of silicon. In a plasma containing SF6 and O2, each
`gas has its own specific function, so by altering the flow
`rate of one of the gases, changes to the etch profile can be
`achieved. The SF6 and O2 generate F and O free-radicals
`under the influence of strong electric fields generated in
`the reactive ion etcher. The F radicals initiate a chemical
`reaction with silicon, producing the highly volatile by-
`product SiF4. The O radicals act to passivate the silicon
`surface by forming SiOxFy (siliconoxyfluoride). In addi-
`tion to generating F radicals, SF6 is also the source of SFþ
`x
`ions, which act to remove the oxyfluoride layer. Alternate
`formation and removal of the oxyfluoride layer is the en-
`abling mechanism behind anisotropic etching of silicon
`[9–12].
`Substrate damage and mask erosion can be minimised
`through appropriate lowering of ion energy. Etch profile is
`also more controllable at lower ion energies. Ion energy is
`governed by the dark space voltage, Dv, which is the po-
`tential developed between the plasma and the powered
`electrode. The O2 creates high voltages, whereas SF6 results
`in lower voltages. The Dv voltage also increases with
`increasing input power, RRF and decreasing system pres-
`sure, p. The relationship can be expressed qualitatively as
`[13]:
`PRF F½O2
`p F½SF6
`where F[gas] refers to the gas flow rate. The dark space
`voltage is not easily measured. In practice DC bias is
`measured instead, as this is known to increase with
`increasing dark space voltage [9]. In the etching of silicon
`using the SF6=O2 mixture, there is a continual competition
`between etching and passivating reactions caused by F
`and O free-radicals, respectively. When RF power, system
`pressure and gas flow rates are at the right levels, etch
`features with vertical side walls result.
`
`Dv a
`
`ð1Þ
`
`Abstract This paper presents the results of dry plasma
`etching of single crystal silicon using SF6 and O2 as process
`gases in a traditional Reactive Ion Etcher. The highly
`anisotropic profiles are achieved for a deep beam feature
`with depths in excess of 100 lm. The effect of O2 con-
`centration on both etch rate and etch profile is investigated
`across a range of chamber pressures. Etch profile anisot-
`ropy can be controlled through appropriate variations in
`O2 and SF6 flow rate and SEM images are provided to show
`this effect over a range of chamber pressures and RF
`powers. Our results indicate O2 concentration to be the
`primary factor influencing etch profile, while system
`pressure is shown to have a strong influence over etch rate.
`Shadowing effect also has been discussed for the possible
`application of releasing the freestanding beams. These
`results aided in the formulation of a suitable process for
`fabricating long-travel electrothermally actuated beam
`structures with the depth and width of 100 lm and 20 lm.
`The ratio of beam depth to the mask-undercut is 10:1. This
`etching technique is resulting in the successful fabrication
`of thermoelectrically-driven long-travel beam structures.
`
`1I
`
`ntroduction
`An important factor in any silicon etch process is the
`ability to control the profile of the etch cavity. In the
`fabrication of MEMS sensors and actuators, vertical walls
`are especially desirable [1–4]. Reactive ion etching (RIE) is
`a popular choice for etching silicon; here, fluoride-based
`plasmas are used because of their ability to spontaneously
`react with silicon at room temperature to create isotropic
`etch features. Anisotropic etching and high etch rates have
`been demonstrated in RIE plasmas containing SF6 and O2
`[5–8].
`
`2E
`
`xperimental
`Etching trials were conducted using an Oxford Instru-
`ments PlasmaLab 80+ reactive ion etcher. Figure 1 shows
`the key components featured in the etcher.
`The PlasmaLab 80+ consists of a pair of parallel plate
`electrodes connected to a 13.56 MHz RF generator, capa-
`ble of automatic matching. The base-plate electrode is
`20 cm in diameter and is water-cooled, allowing an
`adjustable temperature range of 5–20 °C. Base-plate
`temperature was maintained at 20 °C for all experiments.
`
`Received: 17 April 2003 / Accepted: 9 September 2003
`
`H. Zou
`Optical and Semiconductor Devices Group,
`Department of Electrical and Electronic Engineering,
`Imperial College London, Exhibition Road,
`London SW7 2BT, UK
`e-mail: h.zou@imperial.ac.uk
`
`The author would like to thank Professor R. R. A. Syms and
`Mr Michael Larsson for encouraging and useful discussions
`during this work. This project is supported by the EPSRC under
`grant GR/R07844/01.
`
`
`
`photoresist, which is itself patterned using standard pho-
`tolithography processes. Sample sizes were approximately
`3.5 cm2 in area, and a dummy 4-inch wafer was employed
`to maintain loading consistency between trials.
`
`3R
`
`esults and discussions
`
`Oxygen concentration
`The effect of oxygen concentration on etch rate and profile
`has been investigated. Keeping the SF6 flow rate and
`substrate electrode temperature constant, the system
`pressure is set at 30 and 200 mTorr, respectively. The ef-
`fect of increments in oxygen concentration on etch rate
`and etch profile is demonstrated in Figs. 2 and 3, respec-
`tively.
`Figure 2 indicates an increase in silicon etch rate up to
`a maximum, followed by a decrease, with increasing
`oxygen flow rate. The maximum etch rate with system
`pressure set to 200 mTorr is approximately twice that of
`the minimum value. In addition, the etch rate is greater
`when system pressure is at 200 mTorr rather than
`30 mTorr.
`At low oxygen concentrations, further increments of
`oxygen have the effect of facilitating the conversion of
`SF6 to F radicals, as the oxygen reacts with fluorosul-
`phur radicals, thereby hindering their reaction with F
`radicals and the subsequent re-formation of sulphur
`hexafluoride (SF6). The result is a net increase in the
`concentration of fluoride radicals, leading to an increase
`in the silicon etch rate. At high oxygen concentrations,
`however, there is competition between the F and O
`radicals for reaction with silicon; the former resulting in
`etching, the latter resulting in surface passivation. At
`very high oxygen concentrations, the silicon etch rate is
`retarded due to polymerisation on the side walls and
`bases of etch trenches. The polymers protect the silicon
`from reaction with F radicals, thereby, causing a de-
`crease in silicon etch rate with further increases in oxy-
`gen concentration [9, 12].
`The effect of oxygen concentration on silicon etch
`profile is shown in Fig. 3. The SF6 flow rate and system
`pressure are kept constant at 12 sccm and 200 mTorr,
`respectively. At an oxygen flow rate of 4 sccm (Fig. 3a), the
`silicon etch profile is largely isotropic (i.e. non-direc-
`tional). At a flow rate of 6 sccm (Fig. 3b), the degree of
`
`604
`
`Fig. 1. RIE PlasmaLab 80+ component layout
`
`Fig. 2. Si etch rate dependency on O2 flow rate: 12 sccm SF6 at
`160 W RF power, with DC bias varying from 360 to 387 V at
`30 mTorr and from 117 to 175 V at 200 mTorr
`
`Gas-flows are controlled by a standard mass-flow regulator
`and mixed prior to entering the chamber via the gas inlet.
`Etch profile is investigated by sectioning samples and
`viewing in a scanning electron microscope (SEM). Mea-
`surements of etch depth and lateral etch of the vertical face
`beneath the mask layer are made using the SEM.
`All samples used originate from 4-inch <100> oriented,
`phosphorous-doped single crystal silicon wafers, with
`resistivity within the range 1–10 Xcm. The etch mask used
`is 2000 ˚A thick chromium, applied via sputtering.
`Patterning of the chromium layer is achieved using a
`1.3 lm layer of Shipley S1813 positive-working
`
`Fig. 3. SEM cross section micrographs of 30 lm wide trenches etched at different O2 flow rates of a 4 sccm, b 6 sccm, c 9 sccm,
`d 12 sccm respectively. Other conditions: SF6: 12 sccm; system pressure: 200 mTorr, RF power: 160 W
`
`
`
`isotropy is reduced, and the side-wall profile can best be
`described as following a negative taper. At an oxygen flow
`rate of 9 sccm (Fig. 3c), side-walls exhibit a dual positive-
`negative taper, with the overall result close to 90° with
`respect to the horizontal. Finally, at an oxygen flow rate of
`12 sccm, the side-walls exhibit a positive taper and the
`base of the etch trench becomes rounded (Fig. 3d).
`
`increasing system pressure until a pressure of 180 mTorr,
`after which the silicon etch rate continues to increase, but
`at a lower rate. A maximum etch rate of approximately
`880 nm/min is reached at a system pressure of 200 mTorr.
`The increase in etch rate with system pressure is directly
`attributable to the increase in plasma ion density resulting
`from increasing system pressure.
`
`System pressure
`Silicon etch rate is found to be strongly dependent upon
`system pressure; this is shown in Fig. 4. The SF6 and O2
`flow rates are kept constant at 12 sccm and 7 sccm,
`respectively and RF power maintained at 200 W. System
`pressure is varied between 150 mTorr and 200 mTorr. A
`monotonic increase in silicon etch rate is observed with
`
`Fig. 4. Si etch rate as a function of system pressure: 2 sccm SF6,
`7 sccm O2 at 200 W RF input power, with DC bias varying from
`223 to 167 V
`
`Fig. 5. Layout of electrothermally driven, long-travel beam res-
`onator
`
`605
`
`Anisotropic etching of the long-travel beam resonator
`The layout of an electrothermally-driven, long-travel beam
`resonator [3] is shown in Fig. 5. The intention is to fab-
`ricate the device using deep anisotropic reactive ion
`etching. The design in question has dimensions
`L ¼ 14 mm, W ¼ 25 lm and H ¼ 100 lm.
`The device is fabricated from a <100> oriented single
`crystal silicon wafer. A 2000 ˚A thick chromium etch-mask
`was applied via sputtering. Release of the long-travel
`beams is to be achieved via a final isotropic etch step.
`Figure 6 shows SEM images of beam-tips to illustrate
`the variation in anisotropy achieved at various oxygen
`flow rates, and the details of the processing parameters are
`summarized in Table 1. Figure 6a, b and c indicate the
`influence of an increase in oxygen flow rate on the etched
`profiles with the fixed SF6 flow rate, RF power and system
`pressure. An anisotropic profile can be obtained at an
`oxygen flow rate of 8 sccm (Fig. 6b). However oxygen flow
`rates of 5 sccm and 10 sccm, either side of this critical
`point, lead to worsening of beam profile; with either neg-
`ative taper (Fig. 6a) or positive taper (Fig. 6c), respec-
`tively.
`Comparison of Fig. 6d and 6e show the effect of a de-
`crease in system pressure from 200 mTorr to 160 mTorr
`on etch profile. It can be seen that the decrease in system
`pressure (all other process parameters remaining con-
`stant) results in a decrease in lateral etch in the region of
`side-wall directly beneath the etch mask; the reduction is
`from 11 lm to 8 lm. A reason for this observed decrease
`stems from the fact that at lower system pressures, fewer
`reactive ions exist within the chamber at any one time,
`resulting in fewer collisions and fewer chances for poorly
`directed ions to cause lateral etching.
`Figure 7 shows an isometric view of a long-travel beam
`resonator, formed via anisotropic reactive ion etching;
`etching parameters being 12 sccm SF6, 7 sccm O2, 200 W RF
`power and 160 mTorr’s of system pressure. The depth and
`
`Fig. 6. Variations in etch profile with various O2 flow rates a
`5 sccm, b 8 sccm, c 10 sccm; other parameters, SF6 12 sccm, RF
`power 200 W, system pressure 160 mTorr. SEM images showing
`
`anisotropic etching with different system pressure, d 200 mTorr,
`e 160 mTorr. Other parameters; SF6 12 sccm, RF power 200 W,
`O2 7 sccm
`
`
`
`Table 1. Processing para-
`meters for anisotropic etching
`of long, narrow and deep beam
`
`Samples
`
`SF6 flow
`(sccm)
`
`O2 flow
`(sccm)
`
`RF Power
`(W)
`
`Pressure
`(mTorr)
`
`Beam depth
`(lm)
`
`Mask-undercut
`(lm)
`
`(a)
`(b)
`(c)
`(d)
`(e)
`
`12
`12
`12
`12
`12
`
`5
`8
`10
`7
`7
`
`200
`200
`200
`200
`200
`
`160
`160
`160
`200
`160
`
`100
`95
`115
`110
`105
`
`10
`8
`9
`11
`8
`
`606
`
`Free-standing features, such as single lines or walls,
`which are subject to ion bombardment at angles-of-inci-
`dence deviating from the vertical, will eventually assume a
`negative taper. Such an effect can be used to under etch
`free-standing beams with negatively sloping side walls that
`eventually meet on the underside (Fig. 8).
`
`4C
`
`onclusions
`The anisotropic etching of single crystal silicon using SF6
`and O2 process gas mixtures has been demonstrated using
`a PlasmaLab 80+ RIE system. Our results suggest O2
`concentration to be the principal factor influencing etch
`profile, while etch rate is strongly affected by system
`pressure. Based on these findings, process recipes were
`developed to achieve anisotropic profiles for fabricating
`deep beams with the depth and width of 100 lm and
`20 lm. The ratio of beam depth to the mask-undercut is
`10:1. This etching technique allowed the successful fabri-
`cation of thermoelectrically-driven long-travel beam
`structures.
`
`References
`1. Syms RRA; Moore DF (2002) Optical MEMS for telecommu-
`nications. Materials Today July/August Issue: 26–35
`2. Syms RRA; Lohmann A (2002) A Tuning mechanism for
`a MEMS external cavity laser. Proceedings of IEEE/LEOS
`International Conference on Optical MEMS and their
`Applications, Lugano, Switzerland, 20–23 August:
`paper ThD4
`3. Syms RRA (2002) Long-travel electrothermally-driven reso-
`nant cantilever microactuators. J Micromech Microeng 12:
`211–218
`4. Jansen HV; de Boer MJ (1996) High aspect ratio trench
`etching for MEMS applications. Proceedings of IEEE The
`Ninth Annual International workshop Micro Electrical
`Mechanical System, San Diego, USA, pp. 250–257
`
`Fig. 7. Isometric view of long-beam resonator structure after
`isotropic etching, showing beam profiles a, c and d and electro-
`thermal actuator b. Etch conditions: SF6 12 sccm, O2 7 sccm, RF
`power 200 W, system pressure 160 mTorr
`
`Fig. 8. Free-standing beam with negatively sloped sidewalls
`etched by ions at angles-of-incidence deviating from the vertical
`direction
`
`width of the beam are 105 lm and 20 lm, respectively. The
`ratio of beam depth to the mask-undercut is 10:1. The
`cantilever beams were released using SCREAM processing
`[14]. After anisotropic etching the freestanding beams were
`released by isotropic etching of the SF6 plasma.
`
`Shadowing effect
`Ions with an angle-of-incidence substantially different to
`the vertical are prevented from proceeding along their
`intended path by the sidewalls of narrow trenches and / or
`nearby free-standing features; the effect is known as
`‘shadowing’. Shadowing of ions by beam or trench side-
`walls is one reason for the decrease in etch-rate that is
`observed with increasing feature aspect ratio, known as
`RIE-lag [15–17], if the etch rate is substantially governed
`by ion assistance.
`
`
`
`5. D’Emie CPD; Chan KK; Blum J (1992) Deep trench plasma
`etching of single crystal silicon using SF6/O2 gas mixtures.
`J Vac Scl Technol B 10(3): 1105–1110
`6. Pomot C; Mahi B; Petit B; Arnal Y; Pelletier J (1986)
`Anisotropic etching of silicon using an SF6/Ar microwave
`multipolar plasma. J Vac Scl Technol B 4(1): 1–5
`7. Francou M; Danel JS; Peccoud L (1995) Deep and fast plasma
`etching for silicon micromachining. Sens Actuators Phys A
`46–47: 17–21
`8. Darbyshire DA; Pitt CW; Stride AA (1987) Fabrication of
`large scale optical components in silicon by reactive ion
`etching. J Vac Sci Technol B 5(2): 575–578
`9. Manos DM; Flamm DL (1989) Plasma Etching: An Intro-
`duction. Academic Press, Boston, London
`10. Bosch Gmbh RB (1994) U.S. Pat. 4855017, U. S. Pat. 4784720,
`and Germany Pat. 4241045C
`11. Syau T; Baliga BJ; Hamaker RW (1991) RIE of Si trenches
`using SF6/O2 gas mixtures. J Electrochem Soc 138(10): 3076–
`3081
`12. Legtenberg R; Jansen H; Deboer M; Elwenspoek M (1995)
`Anisotropic RIE of Si using SF6/O2/CHF3 gas mixtures.
`J Electrochem Soc 142(6): 2020–2028
`
`13. Jansen HV; Deboer MJ; Legtenberg R; Elwenspoek MC (1995)
`The black silicon method: A universal method for determin-
`ing the parameter setting of a fluorine based reactive ion
`etcher in deep silicon trench etching with profile control.
`J Micromech Microeng 5: 115–120
`14. Shaw KA; Zhang ZL; MacDonald NC (1994) Scream I: a single
`mask, single-crystal silicon, reactive ion etching process for
`microelectromechanical structures. Sensors and Actuators A
`40: 63–70
`15. Volland BE; Heerlein H; Kostic I; Rangelow IW (2001) The
`application of secondary effects in high aspect ratio dry
`etching for the fabrication of MEMS. Microele Eng 57–58:
`641–650
`16. Cardinaud C; Peignon MC; Tessier PY (2000) Plasma etching:
`principles, mechanisms, application to micro- and nano-
`technologies. Appl Surf Sci 164: 72–83
`17. Arnold JC; Gray DC; Sawin HH (1993) Influence of reactant
`transport of fluorine reactive of ion etching of deep trenches
`in silicon. J Vac Sci Technol B 11(6): 2071–2080
`
`607
`
`