Journal of The Electrochemical Society, 154 ͑7͒ H631-H635 ͑2007͒
H631
0013-4651/2007/154͑7͒/H631/5/$20.00 © The Electrochemical Society
Selective Chemical Mechanical Polishing Using Surfactants
Kyoung-Ho Bu and Brij M. Moudgilz
Particle Engineering Research Center and Department of Materials Science and Engineering,
University of Florida, Gainesville, Florida 32611, USA
Device fabrication using high density, small pattern size shallow trench isolation ͑STI͒ processes requires material removal
selectivity during chemical mechanical polishing ͑CMP͒ steps for optimum product processing and quality control. To improve the
selectivity of STI CMP processes, surfactants were applied to selectively polish silica as opposed to silicon nitrides surfaces. A
ten-fold increase in selectivity over conventional colloidal silica slurry was achieved by the addition of sodium dodecyl sulfate
͑SDS͒ and pH adjustment. Adsorption characteristics of SDS on silica and silicon nitride were measured as a function of slurry pH
and concentration of SDS. As indicated by streaming potential measurements and solution depletion adsorption experiments under
acidic pH conditions, SDS adsorption on silicon nitride was significantly higher than silica primarily due to the electrostatic
interactions. It was concluded that the preferential adsorption of SDS on silicon nitride results in the formation of a material-
selective self-assembled passivation ͑lubrication͒ layer leading to selective polishing. Effects of different alkyl chain length of
surfactants were tested. Various mixed surfactant systems were tested and it is believed that the addition of second surfactants
promotes the adsorption on silica diminishing selectivity. The material-targeted boundary layer lubrication concept may be used to
develop selective CMP polishing slurries.
© 2007 The Electrochemical Society. ͓DOI: 10.1149/1.2734802͔ All rights reserved.
Manuscript submitted December 15, 2006; revised manuscript received February 14, 2007. Available electronically May 17, 2007.
The fabrication of next-generation devices will require advances
in shallow trench isolation ͑STI͒ technology.1-3 STI chemical me-
chanical polishing ͑CMP͒ is particularly challenging due to the
highly integrated and variable pattern density as well as smaller
pattern sizes. Less than optimal CMP system can cause several de-
fects such as dishing, nitride erosion, and failure to clear oxide that
are detrimental to global planarization. To minimize such defects,
current STI CMP processes comprised multistep or raw structure
modification such as reverse mask, dummy active area, and addi-
tional active area.1 For better productivity and process simplicity, a
minimum number of processing steps is highly desirable. Accord-
ingly, “high selectivity single-step” slurry designs are being widely
investigated.4,5 The term “selectivity” in this paper is defined as the
ratio of material removal rate ͑MRR͒ of silica to that of silicon
nitride as described below. It is the slurry property as to how much
more it can polish silica than silicon nitride
determined to act as antipolishing agents. Subsequently, Vakarelski
and co-workers showed that formation of an intervening film of
surfactants on solid surfaces caused a significant reduction in fric-
tion between the wafer and the abrasive particles.10 In the proposed
strategy, it was envisioned that a selective coating of surfactant on
silicon nitride can act as an antipolishing barrier without signifi-
cantly affecting the MRR of silica. Considerations that the selected
surfactant needs to be easily removed at the end of the CMP process
preclude surfactants that chemisorb onto any of the CMP substrates.
Experimental
CMP experiments were conducted using Klebosol 1501-50 col-
loidal silica slurry obtained from Rodel Co. after diluting the origi-
nal slurry at 30 to 12 wt %. The slurry pH was measured to be
around pH 10.4 after dilution. The original slurry pH was around pH
10.8. Both silica and silicon nitride thin films were purchased from
Silicon Quest International. Silica thin films were fabricated by the
plasma-enhanced chemical vapor deposition ͑PECVD͒ method us-
ing tetra ortho silicate ͑TEOS͒ as the precursor. Silicon nitride films
were fabricated with low-pressure chemical vapor deposition
͑LPCVD͒ by using dichlorosilane ͑SiCl2͒ and ammonia ͑NH4͒.
Hardness values were measured by the Nanoindentation method us-
ing Hysitron Triboindenter purchased from Hysitron Co. All the
CMP experiments were performed under a 7 psi load using 1
ϫ 1 in. silica and silicon nitride blanket wafers. Relative errors in
MRR from CMP experiments were around 5%, and several points
are randomly repeated for reproducibility check. IC 1000/Suba IV
stacked pads supplied by Rodel Inc. were utilized as CMP pads. The
TegraPol-35 with TegraForce-5 from Struers Co. tabletop polisher
was used for polishing experiments. The rotation speed was con-
trolled at 150 rpm both for the pad and the wafer. MRR was mea-
sured by the decrease in thickness after CMP by ellipsometry ͑Wool-
lam EC110 Ellipsometer͒. For measurement of surface roughness
after CMP, a Digital Instruments Nanoscope III atomic force micro-
scope was used. Zeta potential of silica and silicon nitride wafers
was measured by the streaming potential technique ͑Paar Physica
Electro Kinetic Analyzer͒. Cleaning was done with 99% acetone,
ethyl alcohol, and nano purity water produced by a Millipore filtra-
tion system, which have internal specific resistance of more than
18.2 M⍀. No further chemicals were used on the wafers. For ad-
sorption studies, silica from Geltech Co. and silicon nitride from
Ube Co. ͑SN-E10͒ were used to simulate silica and silicon nitride
substrates. The particle size of the silica was measured to be around
0.53 m by Coulter and that of silicon nitride, which was measured
by centrifugal sedimentation, was reported to be around 0.5 m by
the manufacturer. Their specific surface areas were measured to be
Material Removal Rate of Silica
Selectivity =
͓1͔
Material Removal Rate of Silicon Nitride
In general, conventional silica-abrasive-based STI CMP slurries
exhibit selectivity in the range of 3–4.5 Besides its influence on
planarization, high selectivity slurries are known to provide more
reliable endpoint detection capability. Generally, if the oxide to ni-
tride selectivity is greater than 15, monitoring wafer carrier motor
current can be utilized for efficient endpoint detection.6
In recent years, ceria-based abrasives have shown potential for
high selectivity and have attracted attention in developing advanced
selectively polishing systems. In ceria CMP, maximum polishing
rate and selectivity are achieved at around pH 8, the isoelectric point
͑IEP͒ of ceria. This, however, leads to agglomeration of ceria abra-
sive particles yielding poor surface morphologies with scratches and
higher roughness.6 Alternatively, polymer additives have been used
to improve selectivity, however, the overall material removal rate
͑MRR͒ decreased significantly upon polymer addition.7 In this
study, a different approach to increase STI CMP selectivity by in-
troducing a selective passivation ͑lubrication͒ layer using surfactants
is investigated.
Previous research at the University of Florida8-10 involving sur-
factant mediated lubrication effects in CMP is well documented.
Basim and co-workers showed that the addition of long-chain cat-
ionic surfactant produced better defect-free surface morphology but
the polishing rate was extremely small due to the lubrication effect
of the added surfactant.8 In other words, long-chain surfactants were
z E-mail: bmoudgil@erc.ufl.edu
Downloaded on 2015-05-30 to IP 128.210.126.199 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).