`
`A ULSI shallow trench isolation process through the integration of
`multilayered dielectric process and chemical–mechanical planarization
`
`C.-F. Lina, b, W.-T. Tsengc,*, M.-S. Fenga, Y.-L. Wangd
`
`aInstitute of Materials Science and Engineering, National Chiao Tung University, No. 1001 Ta-Hsueh Road, Hsinchu 300, Taiwan
`bWinbond Electronics Corp., 4 Creation Rd. 3, Science-Based Industrial Park, Hsinchu 300, Taiwan
`cDepartment of Materials Science and Engineering, National Cheng Kung University, No. 1 Ta-Hsueh Road, Tainan 701, Taiwan
`dTaiwan Semiconductor Manufacturing Co., Ltd., Hsinchu 300, Taiwan
`
`Received 11 August 1998; received in revised form 6 January 1999; accepted 6 January 1999
`
`Abstract
`
`A multilayer thin film structure (SiO2/SiNx/SiO2) with a one-step chemical–mechanical polish (CMP) process is developed for shallow
`trench filling and planarization for ULSI devices in the quarter micron regime. By fine-tuning the plasma-enhanced chemical vapor
`deposition (PECVD) conditions we successfully modified the stoichiometry and other characteristics of the as-deposited SiNx and oxide
`films. As these film characteristics are changed, the CMP removal rate selectivity between sacrificial oxide and nitride stopper layer can also
`be adjusted. Correspondingly, the CMP process latitude for shallow trench isolation in 0.25 mm memory devices can be broadened by
`sequentially depositing multilayered oxide and nitride films with adjustable characteristics and CMP removal rates. Dishing-free wide trench
`areas with excellent planarity can be achieved through the integration of the proposed multilayered PECVD oxide/nitride scheme and the
`one-step CMP process. q 1999 Elsevier Science S.A. All rights reserved.
`
`Keywords: Multilayer thin film; Chemical–mechanical polish; Dielectrics; Shallow trench isolation; Silicon local oxidation
`
`1. Introduction
`
`For a long time, local oxidation of Si (LOCOS) was the
`standard technology to provide electrical isolation between
`active devices for integrated circuits (IC). As demands for
`smaller geometry and higher circuit density become stron-
`ger, even more stringent requirements are being placed upon
`the isolation performance, and problems with LOCOS begin
`to surface. For example, the lateral encroachment of field
`oxide into device active areas and the poor planarity set the
`limitations for the utilization of this technology beyond 0.25
`mm generation. In light of these limitations, an alternative
`process called shallow trench isolation (STI) has been
`pursued actively by IC manufacturers as the substitute to
`LOCOS for device isolation [1]. STI allows for higher chip
`density and thus promotes more efficient use of Si wafers. A
`typical STI process involves etching a trench pattern
`through a nitride layer, a thin pad oxide layer, and into the
`silicon. Subsequently, a chemical vapor deposited (CVD)
`oxide is laid on the entire wafer, filling into the trench
`area and overlying the nitride-protective active region. A
`chemical–mechanical polish (CMP) step follows then to
`
`* Corresponding author.
`
`topography from previous deposition
`the
`planarize
`processes and stops finally on the nitride layer. The remain-
`ing nitride is removed subsequently by wet chemistry or
`reactive ion etching (RIE). The planarization issue is
`perceived as the most critical step to the success of STI
`[2,3]. Specifically, due to pattern density effects, nitride
`thickness after CMP can vary, creating step height variation
`and, in some cases, promoting overpolish of nitride layer,
`leading to deleterious damage of the active region. Along
`with the overpolish and pattern density issues is the dishing
`into the oxide-filled trench region that may degrade the
`isolation performance. Extensive efforts have been invested
`to resolve these problems. These include adjustments in
`thickness of trench-fill
`layer [2] and its characteristics
`(e.g., Poly-Si, HDP-CVD, SOG, etc.) [3,4], modifying
`slurry chemistries to achieve high oxide/nitride polish selec-
`tivity [5,6], implementation of a post-CMP RIE step to
`assist planarization [7,8], and layering the trench oxide
`with an additional nitride [9].
`From a process integration point of view, a feasible and
`manufacturable STI process should take into account simul-
`taneously the thickness, characteristics and CMP perfor-
`mances of trench-fill layer and nitride stop layer, so as to
`optimize the process latitude while minimizing overpolish
`
`0040-6090/99/$ - see front matter q 1999 Elsevier Science S.A. All rights reserved.
`PII: S0 0 4 0 - 6 0 9 0 ( 9 9 ) 0 0 0 2 8 - 0
`
`IPR2014-00898
`Exhibit MX027II-1007, p. 1
`
`
`
`C.F. Lin et al. / Thin Solid Films 347 (1999) 248–252
`
`249
`
`Fig. 1. Cross-sectional view of the multilayer thin film structure for shal-
`low-trench-isolation process.
`
`and dishing. In addition, to reduce process complexity, use
`of an additional stop or sacrificial layer or an extra RIE step
`to help achieve planarity should be avoided. Based on the
`perspectives above, in this investigation, we modify the
`stoichiometry and other characteristics of the plasma-
`enhanced chemical vapor deposition (PECVD) SiNx stop
`layer and oxide trench fill layer, in order to optimize and
`improve the CMP process performances such as planarity,
`dishing, selectivity. Such modifications in thin film charac-
`teristics would help enhance the throughput of CMP process
`itself for the implementation of shallow trench isolation into
`0.25 mm device manufacturing.
`
`2. Experimental
`
`In this study, the multilayer thin film structure for the STI
`process is manufactured by the combination of various thin
`film deposition/growth techniques. The process sequence is
`summarized as follows:
`
`1. PECVD nitride stacks (nitride 1 and nitride 2) are grown
`on 150 mm Si wafers as the polish stopping layers. The
`NH3/SiH4 gas flow rate ratio is adjusted to modify nitride
`film properties. The thickness of the nitride stacks is
`varied to evaluate the CMP process latitude.
`2. I-line lithography and trench etch.
`3. Sidewall oxidation is performed to grow a thin oxide
`liner to protect the active region from subsequent proces-
`sing and to achieve better process control and device
`performance.
`chemical vapor deposited
`4. Bilayer
`subatmospheric
`(SACVD) oxides are filled into the 450 nm deep trench
`
`Table 1
`PECVD TEOS oxide characteristics
`
`TEOS
`
`TEOS1
`
`TEOS2
`
`TEOS3
`
`TEOS4
`
`TEOS5
`
`Stress (MPa)
`Density (g/cm3)
`R/R (nm/min)
`E/R (nm/min)
`
`386
`2.27
`185.6
`273.3
`
`330
`2.27
`218.7
`301.1
`
`205
`2.26
`245.6
`332.1
`
`38
`2.23
`261.7
`433.9
`
`244
`2.16
`333.1
`898.9
`
`Fig. 2. Effects of TEOS/N2O gas flow ratio on film stress, NH4F-buffered
`HF chemical wet etch rate and film density of oxides.
`
`at 4008C with TEOS (tetraethoxysilane)/O3 (ozone) gas
`mixtures. The ozone concentration is adjusted so that a
`low-ozone oxide (100 nm) and a high-ozone oxide (500
`nm) can be deposited subsequently.
`5. Dual-frequency PECVD TEOS oxides are deposited as
`CMP sacrificial layers. In this process, the N2O gas flow
`rate is maintained at 4.8 l/min, while the flow rate of
`TEOS gas and low/high-frequency power ratio are both
`varied to modify the stoichiometry and characteristics of
`the deposited oxides. The TEOS/N2O gas flow ratio is
`increased from TEOS1 to TEOS5, representing different
`oxide film characteristics. The cross-sectional view of
`the entire thin film structure is shown in Fig. 1.
`
`CMP experiments were carried out in an IPEC 472 Avanti
`polisher with IC1000/Suba IV pad and SC-112 slurry. The
`down force, back pressure, platen rpm, and carrier rpm used
`were maintained at 7 psi, 2 psi, 30 rpm and 20 rpm, respec-
`tively.
`After film deposition and CMP, physical and chemical
`characterization was performed. Standard wafer curvature
`measurement was used to determine the film stress based on
`Stoney’s equation. Fourier transform infrared (FTIR) spec-
`troscopy was employed to determine the bonding nature of
`the oxide and nitride films. Wet chemical etching rate was
`determined using 10:1 buffered oxide etchant (BOE). CMP
`
`Fig. 3. Correlation between the wet etch rate and CMP removal rate of PE-
`TEOS oxide.
`
`IPR2014-00898
`Exhibit MX027II-1007, p. 2
`
`
`
`250
`
`C.F. Lin et al. / Thin Solid Films 347 (1999) 248–252
`
`higher local tensile stress in the film leads to a higher diffu-
`sion coefficient of water or H1 in silica. As a result, films
`with higher tensile stress should have higher CMP removal
`rates through enhanced water/slurry diffusion and the stron-
`ger reactivity with the alkaline agent in slurry. However, the
`TEOS/N2O vapor reactions in plasma environment produce
`moisture or hydroxyls in the oxide, whose concentration
`increase with the amount of TEOS [12]. A direct conse-
`quence of this is the reduction in film density because the
`existence of hydroxyl groups, in the form of Si–OH bonds,
`results in a more open structure in the oxide. Meanwhile, the
`hydrophilicity of the oxide is enhanced due to the formation
`of Si–OH bonds. Both of the scenarios above, i.e. reduction
`in film density and enhancement in hydrophilicity, will
`enlarge the contact area and hence accelerate wet etch and
`polish rates, as found in the current study.
`The correlation between hydrogen bonding concentration
`in PECVD silicon nitride films and their respective film
`density is exhibited in Fig. 4. As shown, both the numbers
`of Si–H and N–H bonds change the film density signifi-
`cantly. The results are somewhat similar to previous studies
`on PECVD SiNx films [13,14]. It has been reported that a
`higher Si–H/N–H bond density ratio corresponds to a lower
`SiNx film density and higher chemical etch rate in aqueous
`fluoride or chloride media [5,6]. Therefore, it can be postu-
`lated that the existence of more Si–H bonds increases the
`porosity of the PECVD SiNx, leading to increases in both the
`chemical etch rate and the CMP removal rate. Increasing the
`porosity of the dielectrics also increases the specific area
`immersed in the slurry during CMP process. During the
`polishing process, the large specific area on the surface
`will provide larger area for the abrasives in the slurry to
`scrub the weakly bonded and porous SiNx film. All of
`these effects contribute to the enhancement in chemical
`wet etch rate and CMP polish rate dramatically as shown
`in Figs. 5 and 6. Table 2 summarizes the material character-
`istics of PECVD silicon nitride and their respective CMP
`polish rates. Further improvement in CMP performance can
`be realized by modifying the stoichiometry of the intermedi-
`ate nitride stopping layer. This is achieved by modifying
`bias power while reducing SiH4/NH3 gas flow ratio during
`deposition of PECVD nitride films. The FTIR spectrum
`shows that the resulting PECVD nitride films have different
`Si–H content. Concurrently,
`these as-deposited nitrides
`exhibit different BOE wet etch rates and compressive stres-
`ses, suggesting a significant change in nitride film structure
`and hence different CMP removal rates. With these changes
`in film characteristics, a multilayer PECVD SiNx film
`deposition process is implemented with the intention to
`induce high/low CMP removal rates. The process goals
`are to completely remove the remaining oxide on nitride
`stop layer, and to prevent CMP dishing effect resulting
`from over polish. This point will be elaborated later.
`To incorporate the above oxide/nitride thin film stacks
`into the STI process, the thickness and deposition conditions
`are adjusted in order to broaden the process latitude while
`
`Fig. 4. Effects of hydrogen-related bond in PECVD silicon nitride films on
`film density.
`
`removal rate (CMP R/R) was calculated from difference in
`film thickness measured at 49 points across the wafer.
`
`3. Results and discussion
`
`Table 1 summarizes the experimental results of PECVD
`TEOS oxide characteristics and their CMP results. These
`data are reproduced in Fig. 2, which shows that as the
`TEOS/N2O gas flow ratio increases from TEOS1 to
`TEOS5, the as-deposited film stress changes from high
`tensile to low compressive, while the film density decreases
`slightly. This decrease in tensile stress is accompanied by
`the increase in NH4F-buffered HF wet etch rate. Such an
`increase in wet etch rate also correlates well with the rise in
`CMP polish rate, as shown in Fig. 3.
`However, the above findings seem contradictory to the
`work by Osenbach and Knolle [10]. They postulated that the
`presence of tensile stress in the silica structure would
`increase the electrochemical potential and hence chemical
`wet etch rate of the oxide film, when the film is immersed
`into the corrosive aqueous media. Another study [11]
`pointed out that the water diffusion rate in silica glass is a
`function of local stress. In that study, it was found that a
`
`Fig. 5. Effects of stress and film density on chemical wet etch rate of
`PECVD nitride thin films.
`
`IPR2014-00898
`Exhibit MX027II-1007, p. 3
`
`
`
`C.F. Lin et al. / Thin Solid Films 347 (1999) 248–252
`
`251
`
`Fig. 6. Effects of chemical wet etch rate and film density on CMP removal
`rate of PECVD nitrides.
`
`minimizing dishing effects during polishing. A 200-nm
`thick nitride layer was deposited first on the Si wafer. A
`600-nm thick SACVD oxide was then filled into the 450-
`nm deep trench, whose width varies from 300 nm to 2000
`nm. Subsequently, PECVD TEOS oxide layers of varying
`thickness were deposited and their CMP planarization
`performances were evaluated in order to optimize the
`process latitude.
`For optimal planarization results, a 400-nm thick TEOS5
`(CMP R=R (cid:136) 333 nm/min) was deposited first followed by a
`200-nm TEOS1 (CMP R=R (cid:136) 186 nm/min). The nitride
`polish stop layers were formed by sequentially depositing
`a SIN1 (nitride 1, CMP R=R (cid:136) 212 nm/min) layer of 130-
`nm thick followed by a SIN5 layer (nitride 2, CMP R=R (cid:136)
`80 nm/min). Such a oxide/nitride combination with high/
`low polish rates effectively enhances the throughput for
`the STI process. Initially, a TEOS5 oxide layer with a
`high R/R was adopted for bulk removal. The removal rate
`was reduced then by inserting a TEOS1 layer underneath, as
`the polish proceeds towards the oxide/nitride interface. Such
`a deceleration of polishing action effectively reduces the
`risk of overpolish and brings the dishing effect, if any, to
`a controllable extent. A nitride with a slightly higher
`removal rate (SIN1) than TEOS1 was utilized as the top
`polish stop layer (i.e. nitride 1) in order to compensate the
`drop in TEOS1 oxide removal rate due to the existence of
`the nitride substrate. Such a substrate effect [15] results
`when the vertical displacement during polishing is reduced
`as the abrasives indent into a film on top of a mechanically
`harder substrate. The removal rate (R/R) can then be
`expressed as a function of Young’s modulus, the applied
`pressure, and relative velocity:
`
`R=R (cid:136) k(cid:133)1=Ep 1 1=Eu(cid:134) £ P £ V
`
`(cid:133)1(cid:134)
`
`where k is a parameter related to slurry concentration and
`chemical erosion effects; Ep and Eu are the Young’s moduli
`of
`the film being polished and the underlying film
`(substrate); P is the applied down pressure; and V is the
`relative velocity. The PE-TEOS oxide directly above the
`
`(a) The multilayer STI structure before CMP and TEOS oxide
`Fig. 7.
`deposition. (b) STI structure during an intermediate stage of CMP showing
`nitride remaining on the active region. The SEM sample is decorated to
`reveal the trench region. (c) The multilayered STI structure after CMP.
`Global planarization and minimum amount of dishing are achieved for
`wide trench patterns.
`
`harder nitride layer would exhibit a lower removal rate
`due to this substrate effect, while that above the trench
`area would polish at the same rate. Therefore, to maintain
`a planarized surface, the top nitride layer should be removed
`at a slightly higher rate than the TEOS1 layer, in order to
`compensate the dishing into the trench area. Finally, the
`
`IPR2014-00898
`Exhibit MX027II-1007, p. 4
`
`
`
`252
`
`C.F. Lin et al. / Thin Solid Films 347 (1999) 248–252
`
`Table 2
`PECVD nitride characteristics
`
`RI
`E/R (nm/min)
`CMP R/R (nm/min)
`Density (g cm23)
`Si–H (E22 cm23)
`N–H (E22 cm23)
`Bonding ratio (Si–H/N–H)
`Total-H (E22 cm23)
`Stress (MPa)
`
`SIN1
`
`2.28
`400.4
`211.9
`1.94
`1.48
`0.27
`5.48
`1.76
`490
`
`SIN2
`
`2.06
`239.4
`165.8
`2.45
`1.39
`0.39
`3.58
`1.78
`64
`
`SIN3
`
`1.96
`232.8
`119.7
`2.52
`0.87
`0.66
`1.32
`1.54
`2290
`
`SIN4
`
`1.90
`230.4
`99.7
`2.73
`0.05
`1.21
`4.05
`1.36
`2940
`
`SIN5
`
`1.89
`188.4
`80.0
`3.04
`0.00
`1.43
`1.54
`1.43
`2980
`
`nitride with the lowest removal rate (SIN5) was laid under-
`neath as nitride 2 to effectively stop the polish action at the
`interface. At this time, all the PECVD TEOS oxide layer
`was removed, exposing the underlying SACVD trench
`oxide and the residual SIN5 layer, which was removed
`subsequently by wet H3PO4 dip.
`The above findings serve as useful guidelines to design a
`multilayer PECVD oxide/nitride dielectric scheme for shal-
`low trench isolation process. Specifically, by introducing
`oxide/nitride layers with tunable CMP polish rates, the
`CMP dishing effect can be minimized easily without any
`modification in slurry chemistry [8] or the implementation
`of complicated two-step polishing processes [9]. Another
`advantage of the current process is that
`the CVD and
`CMP throughput and process latitude can be expanded
`simultaneously. Fig. 7a is a SEM micrograph showing the
`cross-sectional view of a typical thin film structure for STI
`before CMP. Trenches with different widths can be seen
`clearly. The STI structure during an intermediate CMP
`stage is exhibited in Fig. 7b. The SACVD low-ozone
`oxide was removed to decorate the trench. As shown in
`the picture, there is some residual nitride remaining on the
`active region. The final STI structure after CMP is displayed
`in Fig. 7c. As the picture shows, a wide (4.5 mm) trench with
`a fully planarized surface is attained and the STI structure is
`delineated successfully by the CVD/CMP process scheme
`proposed in this study.
`
`4. Conclusion
`
`By modifying the PECVD deposition conditions we can
`manipulate the stoichiometry and other characteristics of the
`as-deposited SiNx and oxide films for application in the
`shallow trench isolation process. The CMP removal rates
`of the PECVD films are fine-tuned and a multilayer thin film
`stack is proposed in order to allow for a one-step CMP
`process for the definition of the shallow trench patterns.
`The implementation of the this multilayer PECVD dielectric
`process leads to enhancements in the CMP remove rate
`
`selectivity and the efficiency of polish stop layer. The inte-
`gration of the proposed CVD and CMP processes produces a
`STI structure with a well-planarized surface and minimum
`amount of dishing.
`
`Acknowledgements
`
`Thanks are due to the kind support from thin film module
`in Winbond Electronics Corp., for the wafer process. This
`project is supported in part by the National Science Council
`of Taiwan under contract No. 88-2216-E-009-015.
`
`References
`
`[1] P.C. Fazan, V.K. Mathews, Int. Electron Device Meeting (IEDM)
`Digest, IEEE, New York, 1993, p. 157.
`[2] S.S. Cooperman, A.I. Nasr, G.J. Grula, J. Electrochem. Soc. 142
`(1995) 3180.
`[3] I. Ali, M. Rodder, S.R. Roy, G. Shinn, M.I. Raja, J. Electrochem. Soc.
`142 (1995) 3088.
`[4] S. Nag, A. Chatterjee, Solid State Technol. 40 (9) (1997) 129.
`[5] C.R. Mills, G.S. Grover, B.L. Mueller, J.S. Steckenrider, S. Ganesh-
`kumar, G.W. Leach, C.K. Huang, J. Grillaert, Proc. 2nd CMP for
`USLI Multilevel Interconnection Conf., CMP-MIC, Tampa, FL,
`1997, p. 179.
`[6] T. Detzel, S. Hosali, A. Sethuraman, J.-F. Wang, L. Cook, J. Grillaert,
`Proc. 2nd Int. CMP for ULSI Multilevel Interconnection Conf., CMP-
`MIC, Tampa, FL, USA, 1997, p. 202.
`[7] K. Smekalin, Solid State Technol. 40 (7) (1997) 187.
`[8] J.-Y. Cheng, T.F. Lei, T.S. Chao, D.L.W. Yen, B.J. Jin, C.J. Lin, J.
`Electrochem. Soc. 144 (1997) 315.
`[9] J.M. Boyd, J.P. Ellul, J. Electrochem. Soc. 143 (1996) 3718.
`[10] J.W. Osenbach, W.R. Knolle, J. Electrochem. Soc. 139 (1992) 3346.
`[11] M. Nogami, M. Tomozawa, J. Am. Ceram. Soc. 67 (1984) 151.
`[12] Y.-L. Wang, W.-T. Tseng, M.-S. Feng, Jpn. J. Appl. Phys. 36A (1997)
`5492.
`[13] I. Kobayashi, T. Ogawa, S. Hotta, Jpn. J. Appl. Phys. 2A (1992) 336.
`[14] B.T. Dai, W.F. Wu, C.H. Liu, C.W. Liu, C.F. Yeh, Proc. 12th VSLI
`Multilevel Interconnection Conf., CMP-MIC, Tampa, FL, 1995, p.
`485.
`[15] Y.-L. Wang, C. Liu, M.-S. Feng, J. Dun, K.-S. Chou, Thin Solid Films
`308–309 (1997) 543.
`
`IPR2014-00898
`Exhibit MX027II-1007, p. 5
`
`