Nanotopography Impact in Shallow Trench Isolation Chemical Mechanical Polishing-Dependence on Slurry

JOURNAL OF RARE EARTHSVol. 22, Suppl., Aug. 2004, p.001Nanotopography Impact in Shallow Trench Isolation Chemical MechanicalPolishing Dependence on Slurry CharacteristicsJea-Gun Park'" , Takeo Katoh' , Ungyu Paik2( 1. Nano-S0I Process Laboratory, Hanyang University, 17 Haengdang _Dong，Seoungdong-Gu, Seoul 133-791 ,Korea; 2. Department of Ceramic Engineering， Hanyang University, 17 Haengdang-Dong ，Seoungdong-Gu,Seoul 133-791, Korea )Abstract: The nanotopography of the surface of silicon wafers has become an important issue in ULSI device manufactur-ing since it affects the post-chemical mechanical polishing (post-CMP) uniformity of the thickness deviation of dielectricflms. In this study , the nanotopography impact was investigated in terns of its dependence on the characteristics of ceria-based slurries, such as the abrasive size , the grain size of the polycrystalline abrasive and the surfactant added to the slur-ry. It was found that the magnitude of the post-CMP oxide thickness deviation due to nanotpography increased with thesurfactant concentration in the case of smaller abrasives but was almost independent of the concentration in the case oflarger abrasives . The grain size of the polyrystalline abrasive did not affect the nanotopography impact .Key words: CMP; sluny; ceria; abrasive; surfactant ; nanotopographyCLC number: TG175Document code: A Article D: 1002 - 0721(2004) -0001 -04The nanotopographyl12 of the surface of siliconthesize ceria powder . The grain size of the polyecrystal-wafers has become an issue of concern in the shallow-line ceria abrasives was controlled by employing a cal-trench isolation (STI) process used to fabricate ULSIcination process at various temperatures (400， 600,devices, since it affects the post-chemical mechanical700, 800 and 900 C) for 4 h. The abrasive particlepolishing ( post-CMP) uniformity of dielectric filmsize was controlled by a mechanical milling process,thicknesses . Nanotopography is defined as the compo-through which a longer milling time produced a smallernent of the height changes of the wafer surface in theabrasive. The milling time was varied, with durationsabsence of backside chucking, and its range of lateralof8, 15, 28 and 40 h. We prepared nine differentwavelength is generally from 0.2 to 20 mm:. A num-kinds of abrasives by combining the calcination tem-ber of studies have attempted to quantify the impact ofperature and milling time as shown in Table 1. Thethe nanotopography height (NH) on the post-CMP ox-ceria abrasives were dispersed in deionized water andide thickness deviation (OTD)[4-14). Ceria slury withstabilized by adding 100 ppm of a commerially avail-an ionic surfactant is widely used in the STI-CMP pro-able dispersant ( poly-meta-acrylate acid). We alsocess, because it can reduce nitride erosion and widenadded an anionic organic surfactant ( poly-acrylic ac-the processing margin by providing higher oxide-to-ni-id) at a concentration of 0 to 0.8 wt%. Each slurrytride selectivity in the polishing remnoval rate than sili-ca sluris5-19. It is important to study the role ofTable 1Combinations of milling time and calcinationtemperature for nine kinds of ceria abrasivesthe slurry, as one of the most critical consumables, incontrolling the nanotopography impact. We have thuscalcinationmilling timecaldnationmiling timeinvestigated the efects of the abrasive size, the graintemperaturatemperaturen]ro_ rC)_size of the polycrystalline abrasive and the surfactanta)|[@) 400in ceria slurry on the nanotopography impact on post-()5o 600CMP OTD.(C)800158001 Experimental[([)Cerium carbonate was used as a precursor to syn-中国煤化工MHCNMHGReceived date: 2004 - 06 - 03; revised date: 2004- 06 - 23.Foundation item: Project supported by the Korea Ministry of Science & Technology through the National Research Laboratory (NRL) program,Hynix Sermiconductor Ine. and Sumitomo Mitsubishi Silicon Corp* Correspoding author (E mail: parkjgL @ hanyang. ac. kr)2JOURNAL OF RARE EARTHS, Vol. 22 ,Suppl.，Aug. 2004was diluted with deionized water to produce a final ce-cm'min' - I . These polishing conditions were optimizedria abrasive concentration of 1 wt% . Each slurry's pHfor ceria slurry. To extract the specifc efet of thewas adjusted to the range of 7.0~ 8.0 by adding ansurfactant, the polishing time was adjusted to archivealkaline agent. The morphology of the abrasives wasthe same target removal depth (300 nm).observed with a high-resolution transmission electronmicroscope ( HRTEM; JEOL JEM-2010). The sec-2 Results and Discussionondary particle size in each slurry was measured withHigh-resolution TEM images of selected abrasivean AcoustoSizer II (Colloidal Dynamics).samples are shown in Fig.1. As these images show,We prepared prime -grade 8-inch silicon wafers.increasing the mechanical milling time caused minimalThe nanotopography height (NH) of the wafers werechange in the grain size and reduced the abrasive par-measured with a Nano-mapper (ADE Phase Shift). .ticle size (Figs.1(a), (c) and (d)). On the otherOxide films with a thickness of 700 nm were grown onhand, the grain size and morphology in the polyerys-the wafers by the plasma enhanced tetra -ethyl-ortho-talline abrasives changed significantly depending onsilicate (PETEOS) method . The oxide thickness devi-the calcination temperature (Figs. 1(e)，(g) andation (OTD) was measured with a spectroscopic ellip-(i)). The size of the grains was analyzed staistically.someter, the ES4G (Sopra). For the NH and OTD(30 grains for each condition) and then the averagedprofiles shown in the figures in this article, a double-value was calculated.Gaussian type high- pass filter (cutoff length = 20 mm)However, the abrasive size was difficult to ana-was applied to remove the long-wavelength compo-lyze from the TEM photographs due to the small areasnents，such as wafer bow，warp， and wafer-scaleof the images, so we used the secondary particle sizesCMP non-uniformity. The the oxide films were pol-measured with the AcoustoSizer II instead. These re-ished on a Strasbaugh model 6EC laboratory planari-sults are shown in Fig. 2. With increasing mechanicalzer. Break-in with a diamond dresser was caried oumilling time， the secondary particle size decreasedbefore each polishing. An IC100/Suba IV stackedfrom 290 to 70 nm，while the grain size was main-pad (Rodel) was used for polishing. In conductingtained at approximately 40~ 50 nm, as shown in Fig.CMP experiments with the test slurries, the actual pol-2(a). This occured because the ceria powder wasishing pressure was 4 psi = 27.4 kPa. The rotationcontinuously fractured during the milling process 20speeds of the head and the table were both 70 rpm,With increasing calcination temperature, the grain sizeand the relative velocity between the pad and waferincreased up to 40 ~ 50 nm，while the secondarywas 0. 758 m.s~'. The slurry flow rate was 100particle size was maintained at approximately 1 30 ~8h40h400C700900 g300 nm间)(C(d(e(9)()Fig.1 TEM photographs of the abrasives in selected sluries processed at various milling time and calcination temperature350(a){ + secondary particle size50[ (bt卜secondary paride size.30oo个grain size (TEM)< grain size (TEM)250200o15150■----100二中国煤化工.10203040MYHCNM H G°Milling time/hLalcinauon temperanure/ iFig.2 Average measured secondary particle sire of the abraives and averaged grain size with repect to (a) mlling time and (b) cal-cination temperatureJea-Gun Park et al .Nanotopography Impact in Shallow Trench lsolation Chemical Mechanical Polishing140 nm, as shown in Fig. 2(b). This phenomenonSlurmry (a) corresponds to the largest abrasive and slur-was due to grain growth during the calcination processry (d) corresponds to the smallest abrasive. The sur-as atomic diffusion caused wetting of the grain bound-factant concentration was varied from 0 to 0.8 wt%.aries of adjacent grains. The variation shown in Fig.2For all cases shown in Fig.3, the peak and valley po-demonstrates that both the grain size and the abrasivesitions of the NH and post-CMP 0TD coincide wellparticle size were well contolled independently .with each other. Therefore, the fluctuations in theUsing slurries (a) to (d) listed in Table 1, thepost-CMP OTD can be atributed to the wafer nanoto-abrasive size and the surfactant concentration effectspography . While the magnitude of the OTD for slurryon nanotopography impact were examined. Fig. 3(a) is similar with or without surfactant, that for slurryshows examples of the correlation between the lateral(d) increased with the surfactant concentration.profles of the NH and post-CMP OTD on a wafer.twwTrf."432180162460科移四1801832444妈318011344Distance from center/mmslurry (a), 0 wt%slurry (a), 0.8 wt%slurry (d). 0 wt%slurry (d), 0.8 wt%Fig.3 Crrelation between the NH and the post-CMP OTDThe correlation between the abrasive size, thenipulating the slurry characteristics. Note that even ifsurfactant concentration, and the standard deviation ofthe ceria slury generally used in STI-CMP has highthe OTD is shown in Fig.4. For slurries (a) and (b)，oxide-to-nitride removal selectivity, it basically avoidsthe standard deviation of the OTD is not infuenced byonly excessive thinning of the nitride. It does notthe surfactant concentration. For slurry (c), however,.avoid incomplete clearing of the oxide as an influencethe standard deviation increases gradually with in-on the nanotopography.creasing surfactant concentration. For slury (d), itFig.5 shows the RMS the post-CMP OTD for theincreases even more strongly with increasing surfactantslurries including abrasives calcined at various temper-concentration. Thus, when a slurry with smaller abra-ature, which correspond to slurries (e) to (i) listed insives is used, the magnitude of the OTD increases withTable 1. The surfactant concentration was fixed to beincreasing surfactant concentration. That is, the sur-0.5 wt%. As shown in the figure, the calcinationfactant more strongly influences the nanotopographytemperature did not influence the nantopography im-impact on:OTD after CMP in the case of smaller abra-pact unlike milling time. Therefore, it can be saidsives . Though other factors, such as the selectivity ofthat the polycrystalline abrasive particle size is morethe removal rate between oxide and nitride films，mustimportant than the grain size as a factor in contrllingbe taken into account in discussing the influence orthe nanotopography impact and planarization eficiencythe actual STI-CMP process, these findings show thatof a ceria slury .the nanotopography impact can be controlled by ma-35， 605C40ξa中国煤化工72(D)82(C)48(B)YHCNM H G0o9000.20 0.80” 290(A) abrasive size0.80calcination temperature rC]Fig.5 Rms of the post.CMP orxide thicknes deviain (OTD)as a function of calcination temperatureFig.4 Standard deviation (Rms) of the post-CMP OTD4JOURNAL OF RARE EARTHS, Vol. 22 , Suppl., Aug. 2004[9] ParkJG, KatohT, Yoo HC, etal. Spectral analyses3 Conclusionof the impact of nanotpography of silion wafers on oxidechemical mechanical polishing [J]. Jpn. J. Appl.The nanotopography impact was investigated inPhys., 2001, 40: L857.terms of its dependence on the abrasive size, the grain[10] ParkJG, KatohT, YooHC, etal. Spectral analysessize of the polycrystalline abrasive and the surfactanton pad dependency of nanotopography impact on oxideof ceria slurries. It was found that the magnitude ofthe post-CMP oxide thickness deviation due to nanoto-Phys., 2002, 41: L17.pography increased with the surfactant concentration in[11] Park JG, KatohT, YooHC, et al. Spectral analysisthe case of smaller abrasives but was almost indepen-method of nanotopography impact on pad and removaldent of the concentration in the case of larger abra-depth dependency in oxide CMP [A]. Proc. 5th Int.Symp. Chemical Mchanical Polishing, 201st Meeing ofsives . The grain size of the polyerystalline abrasive didthe Electrochemical Society [C]. Philadelphia, PAnot affect the nanotopography impact.2002 (The Electrochemical Society, Pennington, 2002)，Acknowledgement: We are indebted to Mr. Jin-Hyung Park,PV 2002. 202.Mr. Hyun-Goo Kang, Mr. Sung Jun Kim, Mr. Min-Seok Kim[12] Katoh T, ParkJC, Lee W M, etal. The nanotopogra-and Mr. Myung- Yoon Lee for assisting us in performing the ex-phy efct of improved single-side-polished wafer on oxideperiments and analysis.Phys., 2002, 41:L443.References :[13] Schmolke R, Deters R, ThiemeP, et al. On the im-pact of nanotopography of silicon wafers on post-chemical[1] XuCS, Zhao E, Jairath R, et al. Efcts of siliconmechanical polished oxide layers [J]. J. Elctrochem .front surface topography on silicon oxide chemical me-Soc..2002, 149:G257.chanical planarization [J]. J. Electrochem. & Solid-[14] Muller T, Kumpe R, Gerber H A, et al. TechniquesState Lett., 1998, 1: 181.for analysing nanotopography on polished silicon wafers[2] Ravi K V. Wafer flatness requirements for future techn-[J]. Microelectronic Engineering, 2001, 56:123.ologies [J]. Intel Corp.， Future Fab. International,[15] Nojo H, Kodera M, Nakata R. Slury engineering for1999, 7:207.self- sopping, dishing free SiO2-CMP [A]. Proc. IEEE[3] SEMI Document M43-0301, Guide for Reporing Waferidem [C]. San Francisco, CA, 1996 The Institute ofNanotopography .Electrical and Electronics Engineers, Piscataway, 1996.[4] Lee B, Gan T, Boning D, et al. Nanotopography ef-349.fects on chemical mechanical polishing for shallow trench[16] Hirai K, Ohtsuki H, Ashizawa T, et al. High perfor-isolation [ A]. Advanced Semiconductor Manufacturingmance CMP sluy for STI [J]. Hitachi Chemical Tech.Conf. [C], Boston, MA, 2000 (unpublished) .Report (in Jap.), 2000, 35: 17.[5] Lee B. Modeling of Chemical Mechanical Polishing for[17] Philipossian A, Hanazono M. Tribology and Fluid Dy.Shallow Trench Isolation [D]. Dept. of Elctrical Engi-namics Characterization of Cerium Oxide Slurries, www.neering and Computer Science, Massachusetts Institute ofinnovative- planarization . com, 2001.Technology , Massachusetts, 2002.6] Boning D, Lee B. Nanotopography isues in shallow[18] Tateyama Y, Hirano T, Ono T, et al. Study on ceria-based slurry for STI planarization [A]. Proc. Int. Symp.trench isolation CMP [ J] . Materials Research SocietyChemical Mechanical Planarization IV [C]. Phoenix,Bulletin, 2002, 27: 761.The Electrochemical Society, Penninglon, 2000. 297.[7] Fukuda T, Shimizu Y, Yoshise M, et al. The impact[19] Homma Y, Furusawa T, Kusukawa K, et al. Selectiveof nanotopography on oxide polish perforimed by polishCMP of organic S0G for low parasitic capacitance quar-pads with diferent pad hardness [A]. Proce. 3rd Int.ter-micron mutilevel iterconnectins [A]. Proc. CMP-Symp. Advanced Science & Technology of Silicon Mate-MIC [C]. Santa Clara, CA, Institute for Microelectron-rials [C]. Korea, Hai, 2000 (The Japan Society forics Inter-Connection, Tampa, 1996. 67.the Promotion of Science) , Tokyo, 2000.382.[8] Katoh T, KoBG, ParkJH, et al. Effects of flm type[20] Reed JS. Principles of Ceramics Processing [M]. NewYork: Wiley Interscience, 2nd ed. Chap. 1995， 17:and nanotopography of wafers on oxide CMP characteris-323.tics [J]. J. Korean Phys. Soc., 2002, 40:180.中国煤化工MYHCNMHG