Chemomechanical polishing of silicon: Surface termination and mechanism of
`removal
`G. J. Pietsch, G. S. Higashi, and Y. J. Chabal
`
`Citation: Applied Physics Letters 64, 3115 (1994); doi: 10.1063/1.111365
`View online: http://dx.doi.org/10.1063/1.111365
`View Table of Contents: http://scitation.aip.org/content/aip/journal/apl/64/23?ver=pdfcov
`Published by the AIP Publishing
`
`Articles you may be interested in
`Abrasive rolling effects on material removal and surface finish in chemical mechanical polishing analyzed
`by molecular dynamics simulation
`J. Appl. Phys. 109, 084335 (2011); 10.1063/1.3575177
`
`Chemomechanical Production of Functional Structure on Silicon Surfaces
`Chin. J. Chem. Phys. 20, 655 (2007); 10.1088/1674-0068/20/06/655-659
`
`Mechanisms of copper removal during chemical mechanical polishing
`J. Vac. Sci. Technol. B 13, 2215 (1995); 10.1116/1.588106
`
`Impurity decoration of defects in float zone and polycrystalline silicon via chemomechanical polishing
`Appl. Phys. Lett. 62, 2519 (1993); 10.1063/1.109308
`
`Ultrafast diffusion of a defect in indium‐doped silicon introduced by chemomechanical polishing
`Appl. Phys. Lett. 53, 1426 (1988); 10.1063/1.99962
`
`
`
` Reuse of AIP Publishing content is subject to the terms at: https://publishing.aip.org/authors/rights-and-permissions. IP: 104.129.194.129 On: Tue, 01 Mar 2016
`23:09:57
`
`Raytheon2034-0001
`
`Sony Corp. v. Raytheon Co.
`IPR2015-01201
`
`

`
`termination
`
`polishing of silicon: Surface
`Chemomechanical
`and mechanism of removal
`G. J. Pietsch,‘) G. S. Higashi, and Y. J. Chabal
`AT&T Bell Laboratories, Murray Hill, New Jersey 07974
`(Received 20 October 1993; accepted for publication 15 March 1994)
`Infrared spectroscopy of Si(ll1)
`samples immediately after chemomechanical planarization with
`silica slurry (“siton polishing”) shows that the surfaces are predominantly
`terminated by hydrogen.
`This hydrogen termination is responsible for the observed strong hydrophobicity peak at a slurrypH
`of 11, at which point a monohydride
`termination prevails. At higher or lower pH, silanol groups
`replace some of the hydrogen species causing an increase in surface hydrophilicity. A removal
`mechanism
`is proposed which
`involves
`the interplay of oxidation by OH- and subsequent
`termination by H.
`
`feed; -lo4 Pa polishing pres-
`ing speed; -0.2 mI/s slurry
`sure); the polishing
`tlrst removes the thin oxide layer and
`then chemomechanically etches the Si itself. Finally, they are
`quickly removed from the rotating polishing pad and rinsed
`with de-ionized water for -10 s to remove excess polishing
`slurry. Insertion into the nitrogen-purged FI’IR bench is usu-
`ally done within 3 min. For the multiple
`internal reflection
`FTIR experiments,
`the rectangular plates
`(38X 15 X0.45
`mm3) are beveled at the ends (45”) to couple the IR radiation
`in and out.
`Figure 1 shows the dependence of the stock removal rate
`on the slurry pH. The inset demonstrates the dependence on
`both the silica and buffer concentrations. Figure 2 shows the
`degree of hydrophobicity
`(high contact angle) and hydrophi-
`licity
`(low contact angle) as a function of slurry pH. The
`hydrophobicity and the stock removal rate both peak at the
`same pH
`(pH-11.5).
`Figure 3 shows the IR spectra recorded on a sample
`chemomechanically polished at apH of 12.5, that is close to
`where maximum material removal occurs. The most impor-
`tant feature is the strong absorption at 2082.5 cm-‘, assigned
`to the Si-H stretch vibrationa
`It is evidence that, after ch-
`emomechanical polishing,
`the silicon surface is predomi-
`nantly covered with H bonded in an ideal termination of the
`(111) plane. The surface is therefore mostly hydrophobic.
`The left-hand
`inset in Fig. 3 compares the Si-H absorption
`
`Chemomechanical polishing (CMP) is a key step of sili-
`con processing in semiconductor
`technology. Since the late
`6Os, preferential stock removal by CMP is achieved by pol-
`ishing with a soft, planar pad in an alkaline suspension of
`colloidal silica (“Siton”).r
`It is the last and thus most crucial
`step of a series of surface shaping processes (slice cutting,
`lapping, etching)
`to produce a defect-free,
`flat surface for
`further processing of microelectronic devices? Shrinking de-
`vice dimensions and increasing wafer diameters demand bet-
`ter surface planarity with respect to local and total thickness
`variations of the wafer
`to meet the strict requirements of
`photolithographic patterning processes. Recently, CMP has
`become even more important as a method to remove excess
`interlevel dielectric SiO, over the connection studs in very
`large scale integrated multilayer chips after device patterning
`to allow
`interlayer connection and to produce a planar sur-
`face for the next level of wiriug.3 These technological needs
`and this latter low-cost alternative
`to plasma etchbacks re-
`quire a control of CMP on the atomic scale and, therefore, an
`understanding of the surface chemistry and atomic processes
`during material removal.
`We address here the origin of the well-known hydro-
`phibicity of Si surfaces right after chemomechanical polish-
`ing. Although Schnegg et al. have suggested that hydrogen
`termination must be responsible
`for
`the hydrophobicity,4
`their hypothesis has not yet been verified using chemical
`analysis of Si surfaces immediately after CMP. We use here
`Fourier-transform
`infrared spectroscopy (FTIR),5 which is a
`powerful
`tool to detect surface hydrogen and other relevant
`species such as hydroxyl groups.6 These results, along with
`stock removal rate and contact angle measurements, reveal
`that the hydrophobicity
`is indeed due to hydrogen termina-
`tion and that the removal rate strongly depends on the OH
`concentration.
`The samples are n-type (111) silicon (-150 n cm) with
`exact orientation
`(low
`initial step density). They
`first un-
`an RCA-SC1
`standard
`clean7
`(10 min
`in
`dergo
`NH40H:Hz02:H20=1:1:5
`at 80 “C); this clean forms a thin
`oxide layer. They are then polished on both sides for a few
`minutes while being held by a vacuum chuck (resilient, po-
`romeric polishing pad, -100 cm’
`large; -100 cm/s polish-
`
`‘)Permanent address: Philipps-Universiti Marburg, FB Physik, Renthof 5,
`D-35032 Marbug, Germany.
`
`I
`
`I
`
`I
`
`,
`
`5
`
`I
`
`I
`
`I
`
`BUFFER CQ~ENTR.4,&
`0%
`
`pH=ll
`
`fPn-%-
`
`B
`
`~~?.res..*
`
`%
`SILICA Ci%ENG$
`
`I
`
`100
`
`50
`
`rz
`E
`-2
`.5
`
`F
`d
`
`d
`
`6
`
`iz
`
`%o
`
`:i
`po
`
`-*so
`
`_ P
`-3;
`
`.
`
`‘0
`
`10 12 14 16
`8
`6
`4
`2
`SLURRY
`pH
`POLISHING
`PIG. 1. Stock removal rate during CMP at various pH values. Inset: re-
`moval rate vs K&O,
`concentration (forming a KOH:K&O,
`buffer; solid
`curve and upper scale) and silica concentration (dashed, pH=11.5, and dot-
`ted, pH=7,
`curve and lower scale) in the slurry.
`
`Appl. Phys. Lett. 64 (23), 6 June 1994
`
`0003-6951/94/64(23)/115/3/$6.00
`
`0 1994 American
`
`Institute of Physics
`
`3115
`
` Reuse of AIP Publishing content is subject to the terms at: https://publishing.aip.org/authors/rights-and-permissions. IP: 104.129.194.129 On: Tue, 01 Mar 2016
`23:09:57
`
`Raytheon2034-0002
`
`

`
`9o”
`80°
`
`z
`
`.
`
`I,
`
`e,
`
`8.
`NH4Ff
`
`.I.,.
`
`I
`
`.
`
`00’
`0
`
`’
`. ’
`-
`6
`4
`2
`POLISHING
`
`RCA-SC1 I
`. ’
`. 5 ’ ’
`8
`IO
`12
`SLURRY
`pH
`
`. ’ ’
`14
`
`FIG. 2. Contact angle measurements of Si(ll1) surfaces after CMP in silica
`slurry at various pH values. The reference points are the perfectly hydro-
`phobic Si(ll1) surface etched in ammonium Auoride (WF)
`and the hydro-
`philic surface obtained by a wet-chemical oxide (RCA-Xl)
`standard clean.
`
`flat (111)
`features of CMP silicon and ideally H-terminated,
`silicon etched in Nl&F. Comparison of the spectral areas of
`the SGH bands indicates that the CMP sample is only cov-
`ered by half a H monolayer. The lower
`frequency of the
`mode (2082.5 cm-‘)
`compared
`to
`that of
`the reference
`sample (2083.5 cm-‘)
`is consistent with
`the existence of
`small
`(111) domains (-8 atoms of more if
`ideally H-terminated
`OH adsorbed at steps also
`lowers
`the observed Si-H
`frequency).’ The width of the Si-H stretch mode is charac-
`teristic of inhomogeneity
`in either
`the terrace size or the
`chemical termination of the steps.
`cm-’
`The additional defect mode observed at -2071
`(see upper arrow in Fig. 3) has previously been assigned to
`coupled monohydrides and is consistent with the presence of
`(Oil) steps (miscut in the (211) direction).” Since the modes
`usually associated with higher hydrides, such as uncon-
`
`I
`
`I”.
`
`I ”
`
`8 ’
`
`I ‘.
`
`*
`
`*
`
`I.
`
`’ ”
`
`-4 ‘1’
`
`pFZfTfZi0~
`
`I ’
`
`t
`
`2080
`*.
`
`2060
`I*
`* 1.
`2000
`
`I L
`
`I..
`
`2100
`. # *
`2100
`2050
`WAVENUMBER
`
`3000
`.
`.
`
`4,
`
`2150
`[cm-’
`]
`
`sample q,, ,,*
`id’ m
`
`4S-bwekl
`
`w
`
`Ppo
`
`:,3
`
`I
`
`4000
`*
`
`.,
`
`(2135 cm-‘) dihy-
`cm-‘) or constrained
`strained (-2110
`drides’ or trihydrides”
`(2140 cm-‘) are not observed, the
`only possible defects on these otherwise atomically
`flat sur-
`faces are triangular pits, with monohydride-terminated
`steps
`(step edges predominantly
`running the (011) directions).”
`The right-hand
`inset of Fig. 3 gives a clue as to the
`nature of the species terminating
`the rest of the (111) surface.
`is assigned to the O-H stretch
`The absorption at 3250 cm-’
`region
`mode. That there is no absorption in the 1600 cm-l
`rules out the presence of molecular water. The position and
`width of the O-H stretch mode, characteristic of hydrogen
`bonded species, indicate that the hydroxyl groups are close
`together. The relative intensities in the two polarizations are
`particularly
`interesting:
`If OH were adsorbed on flat (111)
`ferraces, then the OH bond would be inclined with respect to
`the surface normal by roughly 65”, yielding a much stronger
`absorption
`in s polarization
`than in p polarization
`(due to
`strong screening of modes normal
`to
`the surface).6 The
`stronger absorption in p polarization
`indicates that the O-H
`dipoles are oriented close to the surface normal, This situa-
`tion is possible if the silanol groups are predominantly ad-
`sorbed at steps with a one dangling bond termination.
`In this
`configuration, steric interaction with
`the bottom
`terrace and
`neighboring Si-O-H
`would contribute
`to the O-H
`bond
`pointing away from the surface plane. The hydroxyl groups
`account for the lower hydrophobicity of the CMP samples
`compared to the completely H-terminated samples12 (com-
`after CMP vs >85”
`after
`pare max contact angle -70”
`“ideal” hydride termination by NH4F etching in Fig. 2).
`The atomic processes taking place during CMP (after the
`removal of the thin oxide
`layer) are now considered. The
`stock removal
`rate (Fig. 1) increases with slurry pH and
`peaks at the alkaline value of pH-11.5.
`This increase with
`pH suggests that OH-
`is instrumental
`in silicon chemical
`etching. Further, OH- molecules must be consumed during
`the removal process since the pH of the off-flowing
`slurry is
`reduced during polishing. Consequently,
`stabilizing
`the
`slurry pH by buffering
`the slurry significantly
`increases the
`removal rate (solid line, inset Fig. 1) because it maintains the
`OH-
`concentration.
`In addition,
`the polishing
`rate varies
`with the silica concentration (dashed and dotted curves, inset
`of Fig. l), indicating
`that the presence of colloidal silicon
`is
`also necessary in chemomechanical polishing.
`Away
`from
`the optimum pH value of 11.5, the surface
`becomes hydrophikc
`(Fig. 1). At very high slurry pH>13,
`the infrared absorption spectrum is dominated by the O-H
`stretch absorption, with a much decreased Si-H absorption
`band. The O-H absorption is now characterized by a strong
`absorption in s polarization, pointing
`to the formation of a
`thin hydrohilic oxide containing OH groups with
`random
`orientation.13 At lower pH (-lo),
`a drop of the removal rate
`is also observed together with an increase in hydrophilicity.
`Here, however, the oxide growth
`is due to the well-known
`slowly oxidizing effect of (oxygen-containing)14 water (re-
`moval
`rate minimum at pH-7)
`rather than due to direct
`the highly alkaline slurry at very high
`attack by OH-
`from
`pH. At pH<6,
`however, the colloidal suspension becomes
`unstable,r5 silica particles precipitate, and
`the polishing
`mechanism
`thus becomes entirely abrasive, leading
`to the
`
`(slurry pH=12.5). The lower arrow
`FIG. 3. FfTR spectra of CMP Si(ll1)
`termination. The upper arrow shows
`shows the predominant monohydride
`the presence of coupled-monohydride steps (defects). The diagram defines
`the p and s polarizations. The IR reflectivity AR/R
`is given per
`infernal
`reflection (-60). Left-hand
`inset: comparison of ChfP Si(ll1)
`(solid line) to
`N&etched
`Si(ll1)
`(dashed line). The frequencies and widths are all given
`in cm -*. Right-hand inset: O-H
`stretch region of CMP Si(ll1);
`the hydro-
`carbon contamination
`level (CT&) is very low (-% monolayer range).
`
`3116
`
`Appl. Phys. Lett., Vol. 64, No. 23, 6 June 1994
`
`Pietsch, Higashi, and Chabal
`
` Reuse of AIP Publishing content is subject to the terms at: https://publishing.aip.org/authors/rights-and-permissions. IP: 104.129.194.129 On: Tue, 01 Mar 2016
`23:09:57
`
`Raytheon2034-0003
`
`

`
`in Fig. 1.
`slight increase of the removal rate at very low pH
`In summary, these observations lead to the following mi-
`croscopic model
`for
`the removal mechanism during
`later
`stage of chemomechanical polishing
`(after
`the thin oxide
`layer has been removed); OH- of
`the polishing slurry
`is
`supplied at very high density by
`the surfaces”
`of
`the
`polyanionic OH--carrying
`silica particles and transferred
`through the boundary layer to the silicon surface by the me-
`chanical action of the polishing process. Locally
`the surface
`may be terminated either by OH or H. If the surface is H
`
`that it becomes
`then there is some probability
`terminated,
`partially oxidized by
`forming surface hydroxyl groups,14
`=Si-OH, particularly at sterically
`favored defect sites like
`coupled monohydrides at atomic steps.* These hydroxyl
`groups weaken the Si-Si
`backbonds of the surface silanol
`species by strong polarizationr4 and facilitate a further attack
`by HzO+H++OH-.
`(An attack by OH-
`ions directly is un-
`likely due to the electrostatic
`repulsion by the negatively
`charged, hydroxylated OH”-Si&+=surface
`sites). This re-
`action can be depicted as
`
`H
`I
`0
`I +=&$
`(H%errk=%e,
`--t 2Ha Si,,,+(HO),=Si,,,
`I
`II
`
`H
`I
`0
`I
`
`I
`
`H
`0
`I
`f
`+H20
`-+ 2H2 Siterr+ (HO),= Si,,,, + Sit,,,
`I *
`II
`OH
`
`.
`
`I
`
`(1)
`
`The silicic acid species, Si(OH)4, formed in this manner are
`removed frictionally by the slurry particles. The new surface
`generated by this row-by-row
`removal mechanism is again
`predominantly H terminated and will subsequently be at-
`tacked by OH-
`in
`the same way. The concentration of
`OH--hence
`the electrochemical potential for this oxidation
`reaction of silicon during polishing
`(Nemst equation)-
`controls
`the size of
`the silicic
`acid species
`removed,
`Si(OH)4,
`(OH),Si-0-Si(OH),,...
`and the resulting stock re-
`moval rate. At very high pH,
`the competing process of com-
`plete oxidation by
`immediate
`condensation of adjacent
`si-OH
`groups,
`
`ai-OH+HO-S&+-\i-0-S%+H20,
`
`(2)
`
`prevails, and a stable hydrophilic oxide is formed which fur-
`ther slows polishing removal rate.14’16
`leads to sub-
`In summary, chemomechanical polishing
`stantial hydrogen termination of Si(lll),
`after the initial
`thin
`oxide layer is removed. This hydrogen termination consists
`predominantly of “ideal” monohydride species, present in
`small triangular
`terraces enclosed by double-layer steps. Hy-
`droxylization
`of
`these H-terminated,
`sterically accessible
`step silicon atoms by OH-
`from the alkaline slurry initiates
`of the now polarized Si4i
`backbonds,
`subsequent crackin
`(Sis+)3&is++ OH P
`--, by attack of water molecules. Even-
`tually an oxidized silica species is removed, leaving the sur-
`face H-terminated behind. The microscopic surface morphol-
`ogy, the overall removal rate, and the surface hydrophobic@
`(contact angle) are controlled by this competition of oxida-
`tion by OH- and termination by H. The flatness achieved
`and the simple chemical model extracted for the Si(ll1)
`sur-
`faces may be useful
`to optimize CMP and wet chemical
`preparation of atomically smooth and homogeneous Si(100)
`surfaces, an important requirement of the semiconductor
`in-
`dustry.
`
`The authors would like to thank S. B. Christman and T.
`Boone for valuable technical assistance. G.J.P. gratefully ac-
`knowledges
`financial
`support
`by
`the Alexander
`von
`Humboldt-Stiftung.
`
`‘E. Mendel, Solid State Technol. 10, 27 (1967).
`‘A. J. Pidduck and V. Nayar, Appl. Phys. A 53, 557 (1991); T. Abe, E. F.
`Steigmeier, W. Hagleitner, and A J. Pidduck, Jpn. J. Appl. Phys. 31, 722
`(1992); S. Hahn, K. Kugimiya, K. Vojtechovsky, M. Sifalda, M.
`Yamashita, P. R. Blaustein, and K. Takahashi, Semicond. Sci. Technol. 7,
`A80
`(1992) J. Wamock, J. Electrochem. Sot. 138, 2398
`(1991); S.
`Kimura, J. Mizuki, J. Matsui, and T. Ishikawa, Appl. Phys. L&t. 60, 2604
`(1992).
`3W. J. Patrick, W. L. Guthrie, C. L. Standley, and P. M. S&able, J. Elec-
`trochem. Sot. 138, 1778 (1991).
`4A. Schnegg, I. Lampert, and H. Jacob, Electrochem. Sot. Extended Ab-
`stracts 85-1, 394 (1985).
`‘N. J. Harrick, Internal Reflection Spectroscopy (Wiley, New York, 1967).
`% J. Chabal, Surf. Sci. Rep. 8, 211 (1988).
`7W. Kern and D. A. Puotinen, RCA Rev. 31, 187 (1970).
`‘P. Jakob and Y. J. Chabal, J. Chem. Phys. 95, 2897 (1991).
`‘P. Jakob, Y. J. Chabal, and K. Raghavachari, Chem. Phys. Lett. 187, 325
`(1991); Y. J. Chabal, J. Vat. Sci. Technol. A 3, 1448 (1985).
`‘OP. Dumas, Y. J. Chabal, and P. Jakob, Surf. Sci. 269/270, 867 (1992).
`‘r G. J. Pietsch, U. Kijhler, and M. Henzler, J. Appl. Phys. 73,4797
`(1993).
`rZArtiiacts from water physisorbed on sample and mirror surfaces in the IR
`bench were ruled out by taking time-dependent spectra of the OH-stretch
`region (adsorption vs desorption; no change observed) and referencing to
`WF
`spectra obtained under identical conditions in Fig. 2.
`13A V Kislev and V. I. Lygin, . .
`
`
`in Infrared Spectra of Adsorbed Species,
`edited by L. H. Little (Academic, New York, 1966), pp. 213 and 228; G.
`S. Higashi and Y. J. Chabal, in Handbook of Semiconductor Water Clean-
`ing Technology: Science, Technology, andApplications, edited by W. Kern
`(Noyes, Park Ridge, NJ, 1993), p. 433.
`I’D. Griif, M. Gnmdner, and R. Schulz, J. Vat. Sci. Technol. A 7, 808
`(1989); M. Morita, T. Ohmi, E. Hasegawa, M. Kawakami, and M.
`Ohwada, J. Auul. Phvs. 68. 1272 (1990):
`in Semiconductor Silicon
`“L. Fengiwei, C. Guo-then; and %. G&g-yu,
`1986, edited by H. Huff, T. Abe and H. Kolbesen [Electrochem. Sot. Proc.
`86-4, 183 (1986)].
`16G. J. Pietsch, U. Kiihler, and M. Henzler, Chem. Phys. Lett. 197, 346
`(1992).
`
`Appl. Phys. Lett., Vol. 64, No. 23, 6 June 1994
`
`Pietsch, Higashi, and Chabal
`
`3117
`
` Reuse of AIP Publishing content is subject to the terms at: https://publishing.aip.org/authors/rights-and-permissions. IP: 104.129.194.129 On: Tue, 01 Mar 2016
`23:09:57
`
`Raytheon2034-0004