Keynote Papers

Technological Advances in Fine Abrasive Processes

R. Komanduri (11, D.A. Lucca (21, and Y. Tani* (2)

Mechanical b Aerospace Engineering, Oklahoma State University, Stillwater, OK, USA
*Institute of Industrial Science, University of Tokyo, Tokyo, Japan

Abstract

Fine abrasives, be they loose or fixed, are invariably used in the production of components of the highest
quality in terms of form and finish accuracy, and surface integrity. While optical, mechanical, and
electronic applications of advanced ceramics, glasses, and semiconductors may require high form and
finish accuracies (e.g., roughnesses in the angstrom range), the depth and nature of the subsurface
damage resultant from processing may be equally important. For economic manufacture and for improved
reliability of brittle materials, an understanding of the mechanisms of material removal in fine abrasive
processes, as well as the nature of damage imparted, are essential prerequisites. Knowledge of the
removal mechanisms and nature of damage can enable process improvement and minimize, if not
altogether eliminate, surface and subsurface damage. This paper focuses on. fine abrasive processes
with emphasis on material removal in brittle workmaterials. Generally, indentation models are used to
simulate abrasion and polishing. An attempt is made to rationalize various models by linking conventional
machining, grinding, ultraprecision machining, and indentation sliding as a cognate transition for material
removal operations, Whereas chemical mechanical polishing (CMP) has become the process of choice for
finishing semiconductors, the various models developed thus far, although very significant, have
addressed isolated aspects of the process and/or neglected certain issues. To use analytical models as
predictive tools for finishing of brittle materials, it is necessary to integrate existing understanding into a
comprehensive model of the process. This paper reports on some significant technological advances in
fine abrasive processes which have been made.

Kevwords: grinding, polishing, finishing, abrasives, ceramics, glass, silicon

"The smaller the particles of the substance (abrasives), the smaller will be the
scratches by which they continually fret and wear away the glass until it is polished..."

............ Sir Isaac Newton in Opticks (1695)

1. INTRODUCTION 1964). Consequently, the mechanisms of material removal

Fine abrasives have been in use for polishing since
antiquity as evidenced by innumerable marvelous stone
monuments around the world as well as the natural erosion
of them over centuries due to rain, wind, and sand erosion,
and the chemical interactions associated with the
environment (pollution) (Komanduri and Lucca, 1997).
Fine abrasives are generally obtained by comminution, a
process by which brittle materials are reduced in size,
e.g., from large size abrasives to fine size abrasives.
Consequently, there has been a limit on the fine size that
can be obtained (- a few pm) and the associated defect
structure inherited during the size reduction process.
Today, new advances in materials syntheses have
enabled production of ultrafine abrasives in the nanometric
range without the need for comminution. They are used in a
variety of forms including loose abrasives (polishing,
lapping), bonded abrasives (grinding wheels), and coated
abrasives for producing components of various shapes,
sizes, accuracy, finish, and surface integrity. The
requirements range from high resolution, high reliability,
high response, high throughput, and high integration for
electronic components (VLSI); to high durability, high
strength, and high reliability for structural components; to
high reflectivity for optical components; to high magnetic
permeability for magnetic materials.
The initial technological uses of fine abrasives have been
predominantly for optical applications such as
microscopes and telescopes, and for polishing of surfaces
for metallurgical investigation of microstructures (Holland,
in polishing have attracted the attention of such stalwarts
as Sir Isaac Newton (1695), Lord Rayleigh (1901), Sir
George Beilby (1921), recently by Late Professor Bowden
and Professor Tabor of Cambridge University (1950), and
more recently by one of our ClRP colleagues, Len Samuels
of Australia (1971), to mention a few. It may be noted that
surfaces produced on ground glass for optical applications
need to be smooth only to a certain degree (a fraction of
the wavelength, h of the incident light) for it to specularly
reflect light. Thus, a surface which negligibly scatters light
need not be smooth on an atomic scale. With high
emphasis on precision in Japan, many universities there
had for quite some time (some to this day) Departments of
Precision Engineering. They also have a professional
society dedicated to this area, namely, Japan Society for
Precision Engineering (JSPE) since 1947. The journal
Precision Enaineerinqwas launched with the help of many
of our ClRP colleagues in 1979, and is now maintained by
the American Society for Precision Engineering (ASPE)
which was started in 1986.
Since material removal in fine abrasive processes is
extremely small, on the order of submicrometers or less,
our distinguishedcolleague, Norio Taniguchi (1974) coined
the term "Nanotechnology" as a target of ultraprecision
processing. In 1983, at the ClRP General Assembly in
Harrogate a keynote paper was presented on the "Current
Status and Future Trends in Ultra Precision Machining and
Ultrafine Materials Processing," by Professor Taniguchi
(1983). An elegant introduction to Professor Taniguchi's

Annals of the CIRP Vol. 46/2/1997 545

 Achievable
Machmmg
accuracy
Fig. 1 The development of achievable machining accuracy (after Taniguchi, 1983)
presentation was made by our distinguishedcolleague and
former President of CIRP, Professor Pat McKeown of the
Cranfield Unit for Precision Engineering (CUPE) of the UK.
Taniguchi, in this paper, reviewed the historical progress
of machining accuracy achieved during this century under
the general classification of conventional machining (CM),
precision machining (PM), and ultraprecision machining
(UPM) and also included various inspection equipment
used for the measurement of accuracy [Fig. I]. By
continuously incorporating the advances made in UPM in
the previous years to P.M first, and ultimately to CM the
achievable accuracies continue to increase. For example,
Taniguchi predicted that by 2000 A. D., machining
accuracies in conventional machining would reach 1 pm,
while in precision machining (PM) and in ultra precision
machining (UPM) would reach 0.01 pm and 0.001 pm,
respectively. It is amazing that his predictions made in
1983 are in fact in line with the advances taking place
currently in the machine tool industry. Taniguchi also
addressed problems to be solved in the meantime towards
-
achieving true accuracies of 0.1 pm 0.05 pm for finishing
electronic, optical, and magnetic materials. They include
leveling up the inherent precision of all elements/sub-
systems, such as straightness, flatness, and squareness
of the tool/work motions; adaptive accuracy control of
machine tool motions; and application of quality control
techniques and systems at the highest levels of accuracy.
The development of deterministic prediction capability,
which our distinguished colleague Jim Bryan (1984) at the
Lawrence Livermore National Laboratory (LLNL) had been
advocating from early on to achieve machine accuracy
capability of 50 nm, was also pursued.
Traditionally, si.ngle or multiple point tools with
geometrically defined edges, such as turning and milling
tools have been considered for roughing and semi-
finishing, and abrasives either in loose or bonded form
whose geometry varies continuously in an unpredictable
manner during the process are considered for finishing.
However, this myth was laid to rest by the development of
rigid, ultraprecision machine tools for mirror finishing of
nonferrous materials using a single crystal diamond tool.
546
Machina tools Measurina instruments
(pccessuq ecuipmnt) (inspection squimnt)
Surface roughness on the order of a 10 A rms is obtained
today relatively easily on a range of surface geometries,
such as planar, round or, parabolic (concave or convex).
Subsequently, Professor McKeown of CUPE developed an
extremely rigid, very precise state-of-the art grinding
machine (Nanocentre 600) (1990) incorporating Professor
Nakagawa's electrolytic in-process dressing (ELID)
(Ohmori and Nakagawa, 1990,1995; Ohmori and
Takahashi, 1994) for grinding a range of hardened steels,
ceramics, and glasses.
It may be pointed out that the terms polishing, finishing,
lapping, and honing are all used for precision finishing with
differences in the emphasis on the requirements. For
example, traditionally, polishing was meant to imply best
finish without regard for shape or form accuracy, e.g., in
the preparation of metallurgical samples where the
objective is to remove the scratches so that
microstructure can be observed after etching. It is also
used for removing the damage, such as microcracks,
voids, etc., caused by the previous manufacturing
operation, such as grinding. Conventional lapping in
contrast is used not for finish but for form accuracy, such
as flatness in the case of flat objects or sphericity in the
case of balls. The term honing is used, similar to lapping,
for form and shape accuracy and for generating the
topography to 'trap' the lubricant. Finishing is a general
term used to describe any or all of these processes.
Today, in view of our ability to generate extremely fine
finish and form accuracy by a range of processes, these
terms are used somewhat interchangeably and may not
adhere strictly to these definitions by researchers as well
as practitioners.
Several Keynote papers which have been presented at the
CIRP General Assemblies in the past by our distinguished
colleagues, have related to the technology of finishing with
fine abrasives. In view of their special importance, some of
those references are incorporated in the keynote paper:
Taniguchi, N., 1983, Current Status in, and Future
Trends of Ultraprecision Machining and Ultrafine
Materials Processing, Annals of CIRP, 32/2: 573-582.
 Keynote Papers

Machining.

k
by bonded
I 1 UltraprecisionGrinding
abrasives 2 Honing
3 Magnetic Field Assisted Polishing
4 Coated Abrasive Machining
Abrasive
machining
I- 5 Barrelling

Mechanical
processes
Machining
by loose
abrasives

Machining by
4
free abrasives I
6 Lapping
7 Ultrasonic Machining
8 Polishing
9 Viscoelastic Flow Machining
10 ChemicaCMechanical Polishing (CMP)
Machining by 11 Single Point Diamond Turning
single point tool
12 Etching
Chemical processes
[ - ,
13 Chemical Polishing
Photo-chemical processes 14 Photo Etching
Electro-chemical processes 15 (Electrolytic Polishing)
r 16 Electric Discharge Machining
t Electric
processes
17 Electron Beam Machining
18 Ion Beam Machining

L Optical
processes
19 Plasma Machining
20 Laser Beam Machining

Fig. 2 Classification of material removal processes in terms of the energy source used (after Nogawa, 1988)
Konig, W., Tonshoff, H.K., Fromlowitz, J. and Dennis,
P., 1986, Belt Grinding, Annals of CIRP, 35/2: 487-494.
Inasaki, 1.,1987, Grinding of Hard and Brittle Materials,
Annals of CIRP, 36/2: 463-471.
McKeown, P.A.,1987, The Role of Precision Engineering
in Manufacturing of the Future, Annals of CIRP, 3612:
495-502.
Salje, E., and Paulmann, R., 1988, Relations Between
Abrasive Processes, Annals of CIRP, 37/2: 641-648.
Tbnshoff, H.K., v. Schmieden, W., Inasaki, I., Konig,
W., and Spur, G., 1990, Abrasive Machining of Silicon,
Annals of CIRP, 39/2: 621-635.
Ikawa, N., Donaldson, R.R., Komanduri, R., Konig, W.,
McKeown, P.A., Moriwaki, T., and Stowers, I.F., 1991,
Ultraprecision Metal Cutting , the Past, the Present, and
the Future, Annals of CIRP, 4012: 587-594. .
T6nshoff, H.K., Peters, J., Inasaki, I., and Paul, T.,
1992, Modelling and Simulation of Grinding Processes,
Annals of CIRP, 41/2: 677-688.
Venkatesh, V.C., Inasaki, I., Tbnshoff, H.K.,
Nakagawa, T., and Marinescu, ID., 1995, Observations
on Polishing and Ultra Precision Machining of
Semiconductor Substrate Materials, Annals of CIRP,
44/2: 611-618.
Malkin, S., and Hwang, T.W., 1996, Grinding
Mechanisms for Ceramics, Annals of CIRP, 45/2: 569-
580.
Also, at least three monographs were written on this topic
by our distinguished colleagues. They are "Principles of
Machining by Cutting, Abrasion, and Erosion," by Jan
Kaczmarek (1976); "Principles of Abrasive Processing,"
by Milton Shaw (1996); and "Grinding Technology," by S.
Malkin (1989). In addition, many technical papers were
published in the Annals of CIRP over the years on various
topics related to the subject matter, many of which relating
to ultraprecision finishing with fine abrasives are
incorporated here as references. However, the authors
apologize for any inadvertent omission of papers by our
esteemed colleagues.
Figure 2 gives a classification of various material removal
processes in terms of the energy source used (after
Nogawa, 1988). They involve both mechanical material
removal, such as single point and multiple point cutting
tools as well as with abrasives (bonded, coated, and
loose) as well as non-mechanical techniques. The
techniques involving non-mechanical material removal for
fine finishing include, several types of beam machining
processes (laser beam, electron beam, ion beam,
molecular beam), electrical discharge machining, electro-
chemical machining, chemical machining, etc. They are
considered as outside the scope of finishing with fine
abrasives, and hence are not covered in this article.
Instead, the readers are referred to the excellent keynote
papers by Taniguchi (1983, 1993) and other specialized
articles on this subject (e.g. Taniguchi, 1989).
There are many advances taking place in the finishing of
materials with fine abrasives, including, the processes,
the abrasives and their bonding, the machine tools
(rigidity, precision, speed, power, feed back controls,
etc.), and the environment. It would be difficult, if not
impossible, to cover all the advances in these areas in one
paper. Hence, an attempt is made to give an appreciation
of some of the rapid advances taking place in this field with
a few illustrations rather than attempting to be all inclusive.
Generally, indentation models are used to simulate
abrasion and polishing. An attempt will be made in this
paper to link conventional machining, grinding,
ultraprecision machining, and indentation as a cognate
transition for material removal operations as one varies the
scale of cut depth and tool geometry.
Chemo-mechanical polishing (CMP) is becoming an
indispensable process for finishing hard, brittle materials
for optical, electronic, and structural applications. While
some valuable models were developed towards

547
 r 1 I Hiah 1
High-function work materials for + integration
optical devices (mirrors, lenses),
Ultra-precision machine tools electronic, electromagnetic Ultra- + High
(CNC control, rigid, high and structural applications precision reliability
precision, motion accuracy Functional
and resolution, high loop + surface
properties High parts
stiffness, feed back control) (finish, form, response (electronic

Ultra-Precision -
nd accuracy.
+ High
sensitivity
-
parts,
sensors, etc.)

Finishing Surface
(fine abrasives of various
materials, uniformity of grain
- integrty
(absenceof +
Minimal
sub-surface - Ferro-
damaqe
size, grinding wheel (type and with Fine defects like magnetic

I
bond), polishing compounds, +microcracks, High Pa-
carrier fluids) Abrasives metallurgical -+ permeability

U
changes,
t
I I stresses,
? residuaI High
reflectivity
Ultra-precision sub-surface Optical

P
High-level clean environment measurement damage) High
(temperature, humidity, vibration, (Ta IySurf, Ultra- ccuracy, form,
and dust control) TalyRond, precision and finish
TalyStep, finishing
ZYGO, WYKO, properties properties High
X-ray, p-Raman, evaluation specific to reliability Structural
SEM, TEM, the material
AFM, AES, etc.) strength and
toughness I 1

M:;-a;;g + Machining control

and assessment + Machining
+- Required
performance
of product
-+ 4 product

Fig. 3 Conceptual diagram of ultraprecision finishing with fine abrasives (after Nogawa, 1988)

understanding the process, most of them dealt with
isolated aspects of the process and ignored the chemical
aspects of CMP. Also, erosion of the individual colloidal
silica particles on the silica layer (formed during chemical
action) has to be considered instead of, or, in addition to
abrasion. More efforts are needed in the direction of
developing a comprehensive model taking into account
various aspects of CMP to take advantage of this unique
process. Also, chemo-mechanical polishing is becoming
the process of choice for the finishing stage of many brittle
materials with stringent requirements of finish, accuracy,
and surface integrity. Hence, research efforts should be
directed in this important field of activity in addition to the
mechanical aspects.
2. ULTRAPRECISIONMACHINING SYSTEMS
In view of the peculiarities of the workmaterials (advanced
ceramics, optical glasses, semiconductor materials, etc.),
the stringent requirements of these materials in practice,
the need to remove very small amounts of material (atomic
bits) to minimize or eliminate brittle fracture, the very high
surface finish and form accuracy requirements, and the
required surface integrity (i.e., minimal or no surface or
near surface damage including metallurgical changes,
residual stresses, defects such as microcracks), it is
necessary to consider the entire ultraprecision machining
process as a system and not as individual components. It
is this approach that enabled Jim Bryan and his colleagues
at LLNL to conceive, design, and build submicron diamond
turning machines and develop this technology to finish
surfaces that surpassed precision grinding technology of
the time by orders of magnitude (Bryan and Loewen, 1974;
Bryan, 1979; Donaldson, 1979). A similar systems
approach is needed for all precision machining systems as
well as sub-systems using fine abrasives. Although
specific requirements may vary from process to process,
for ultraprecision finishing with fine abrasives, we are
typically interested in an integrated system with the
following considerations:
548
1. High precision, vibration-free, rigid machine tool
2. High loop stiffness between the tool and the
workpiece
3. High resolution motion control
4. Submicron grain size abrasive of uniform grain size’
bonded into a wheel or in the f o m of a coated
abrasive or as free abrasive
5. In-process dressing in the case of grinding, and new
dressing techniques for different bonding systems,
e.g., EDM, ECM, microblasting, wire brushing, laser
dressing, chemical etching
6. Thermal stability
7. Feedback control
8. Ultraprecision metrology system integrated into the
machine tool but isolated from the response of the
machine tool during machining
9. High level of clean environment including
temperature, vibration, humidity, and dust control
Figure 3 is a conceptual diagram of ultraprecision
machining (UPM) using fine abrasives, adopted from
Nogawa (1988). The central block is the ultraprecision
machining process with fine abrasives, the central theme
of this paper. To its left are arrows coming into the block
indicating the type of machine tools (grinding, polishing,
lapping, etc.), including their characteristics, such as
stiffness and damping characteristics, precision, motion
accuracy and resolution, speed and power capability, feed
back control system, on machine metrology frame; the
types of abrasives (material, size and method of bonding)
or tools for cutting; the metrology frame in the machine
tool; and the environment (clean room, temperature
control, vibration control, dust free). The arrow from the
top into the main block is the condition of the workmaterial
(material type, size, shape, initial condition, method of
processing, etc.), the arrows from the bottom into the
block are the ultraprecision measurement and evaluation
of properties, and the arrows on the right pointing away
from the block are the surface and subsurface features
(roughness, residual stresses, geometry, accuracy,
metallurgical damage, cracks, etc.) and the ultimate
properties required (or retained) for various functional
applications (structural, optical, magnetic, electronic,
etc.).
In the following, some aspects of ultraprecison machining
systems, namely, the machine tools, the abrasives and
the workmaterials will be briefly reviewed as examples. For
a complete review, one may refer to appropriate papers
published in the Annals of ClRP and other relevant journals
dealing in detail with the various aspects of the system
mentioned above.
2.1 Machine Tools
Pioneering work on the design and construction of single
point diamond turning (SPDT) machines was initiated in the
1970's at LLNL. A variety of SPDT machines for finishing
laser optics used in defense applications were built.
Following this, many machine tool builders in the U.S.,
U.K., Japan, and Europe developed specialized
ultraprecision machine tools for a range of optical,
electronic, and magnetic applications. Details on the
evolution of single point diamond turning and grinding
technologies were covered in an earlier keynote paper by
lkawa et at. (1991). Here, representative machine tools
built by LLNL (LODTM), Moore Special Tool Co., Toshiba,
CUPE, two machines built by the researchers at the
University of Osaka (a grinding machine and a float
polishing machine), and Canon's super smooth polishing
machine (CSSP) are briefly reviewed as examples.
In 1980, Dr. R.R. Donaldson and his colleagues at LLNL
designed and built the vertical spindle Large Optics
Diamond Turning Machine (LODTM) to manufacture very
precise, large diameter reflective optics for advanced
laser systems (1.5 m in diameter) by direct machining
without the need for polishing [Fig. 41. This machine
Fig. 4 Schematic of the vertical spindle Large Optics
DiamondTurning Machine (LODTM) at LLNL (after
Donaldson, 1979)
tool is essentially a lathe with a work zone 61 cm high and
1.63 m diameter, a load capacity of 1.4 tons, and has an
rms positioning accuracy of 28 nm. The metrology loop
consists of high-stability passive components made of
superinvar. laser interferometers, and hiah Drecision
capacitance gages. To avoid quantization irrors in the
digital control system, measurements were made with a
resolution of 0.6 nm. Much of the repeatability of this
machine comes from the fact that the mechanical support
structure is separated from the metrology loop. The
metrology frame is an unloaded structure that is
kinematically mounted to the main machine frame. In this
way, tool positioning information is not affected by the
changes in strain due to movement of the machine slides.
The entire machine tool is located in a temperature
controlled (<0.005 K variation) room and operated
remotely. The most critical components of the machine
tool are cooled with a constant temperature water system.
Moore built a diamond tuming and grinding machine [Fig. 5)
for the production of aspheric optical and infrared
components. It uses an oil shower to control the
+
temperature to within 0.006 K. Similarly, in Japan, Ueda
et al. (1991) at Toshiba developed an ultraprecision CNC
machine tool for the manufacture of large aspheric mirrors
up to 650 mm in diameter and 250 mm in axial length [Fig.
61 . The spindle is supported by externally pressurized air
bearings. The spindle speed range is from 3,000 to 30,000
rpm. The spindle stiffness is 10 N/pm in the radial direction
and 20 N/pm in the normal direction. A laser interferometer
feedback system with 2.5 nm resolution is incorporated
both in the X- and Z- slide motions. The machine tool is
installed in a temperature controlled room (+ 1 K) with the
machine tool chamber maintained at ?s 0.1 K. Using this
machine tool, a concave mirror made of electroless nickel
plated aluminum was finished to a form accuracy of < 0.1
pm PV and a 2 nm R, surface roughness. A 500 mm
diameter glass-ceramic mirror with a 1000 mm radius of
curvature was also finished using a fine grain (4 pm),
resinoid bonded diamond grinding wheel (200 mm dia) at a
wheel speed of 10,000 rpm, with a work spindle speed of
10 rpm and depth of cut of 1 pm. A surface roughness of
0.6 nm Ra and a form accuracy of 0.3 pm PV were
obtained.
Fig. 5 Moore's M-18 Oil Shower Aspheric Generator
Fig. 6 Toshiba's ultra precision CNC machine (after Ueda
et at., 1991)
McKeown of CUPE (1990) developed a state-of-the-art
ultraprecision grinding machine (Nanocentre) with a
Cuproc CNC 3000 control system [Fig. 71. This is a 3-axis
diamond turning, grinding, polishing, and measuring
machine which can finish complex profiles through 1.25 nm
resolution interferometry, high traction friction drives, and
549
stiff hydrostatic bearings. It can handle workpieces up to
600 mm in diameter and 260 mm in length. It has a T- base
constructed from granite-epoxy to give excellent
dimensional stability, high stiffness, and high internal
damping. It incorporates ELID dressing, ultraprecision
truing and has a high loop stiffness between the tool and
the workmaterial. It is operated in a temperature controlled
environment. In addition, the machine tool has the
capability for oil shower temperature control on critical
machine tool elements.
Namba et al. (1987) developed an environmentally
controlled, ultraprecision float polishing machine for
finishing large sized wafers (up to 180 mm) of electronic
materials to a high finish with minimal subsurface damage
[Fig. 81. This machine has been widely used in Japan for
finishing semiconductor materials in the electronic Fig. 7 CUPE's ultraprecision grinding machine
industry. Namba and Tsuwa recognized that the key to (Nanocentre) (after McKeown et al., 1990)
making flat parts by polishing is the flatness of the lap
itself which in turn is transferred on to the worksurface study to grind various types of glasses with cup-shaped
during the polishing process. They accomplished this by resinoid bonded diamond grinding wheels (up to #3000 or
ultraprecision diamond turning the tin lap (Namba and 2-6 pm grain size) to produce extremely smooth surfaces.
Yonezawa, 1982). The structure of the machine tool is a without the need for subsequent polishing. They reported a
double-column vertical lathe. A tin lap of 460 mm OD.and surface roughness of 0.142 nm rms and 1.71 nm (PV) on a
120 mm I.D. is set on a horizontal faceplate of the main ground BK7 glass using a finer grain diamond (#3000) and
spindle supported by a rigid, high accuracy, and vibration a lower feed (5 pm/rev) at a wheel speed of 1200 m/min.
absorbing hydrostatic oil bearing. The spindle drive motor They concluded that super-smooth surfaces (0.2 nm rms,
(2.2 kW) is mounted off the machine to reduce vibration. and 2 nm Rmax which are better than required for optical
The main spindle (125 mm in diameter) rotates surfaces can be obtained by ultraprecision grinding
continuously up to 500 rpm. The tin lap consists of a tin without the need for any subsequent polishing.
plate glued onto a stainless steel plate with an epoxy. The
rail head having a tool post is driven horizontally by an oil
cylinder on a rigid cross rail which is also provided with an
upper spindle head for rotating a small holder. A 1 pm radial
straightness and axial deviation on the tin lap surface is
obtained by machining spiral grooves of 2 mm pitch in the
tin lap with a sharp diamond tool. The width and depth of
the grooves are 1 mm and 0.5-1.2 mm. The machine is
enclosed in a superclean unit (Class 100) to prevent dust
from collecting and interacting with the workmaterial. The
temperature of the oil supply to the hydrostatic bearing
and the polishing fluid is controlled to within 0.05 K. The
polishing fluid consists of 20% of fumed silica (7 nm) in
filtered, deionized or distilled water. During polishing, both
the main and upper spindles rotate at almost the same
revolution but in the opposite directions. The polishing
pressure is applied by the weight of the sample holder. The
samples are floated in the polishing fluid by the
hydrodynamic behavior of the fluid between the sample
and the grooved tin lap. Fig. 8 Environmentally controlled ultraprecision float
polishing machine for finishing large sized wafers of
Namba et al. (1989) also developed an ultraprecision electronic materials (after Namba et al., 1987)
surface grinder with a glass-ceramic spindle (125 mm
diameter) of zero-thermal expansion material capable of Also in Japan, Ando et at. of Canon Inc. (1992, 1995)
rotational speeds of 3500 rpm [Figs. 9(a) and 9(b)]. The developed a super-smooth polisher (CSSP) [Fig. 1O(a)] for
precision grinder has two vertical spindles with hydrostatic the finishing of aspherical surfaces used in excimer
bearings of high precision and rigidity. Extremely fine lasers, soft X-ray, and other short wavelength light
depths of cut (0.1 pm) are numerically controlled using applications. The CSSP polisher has a high resolution, on-
optical scales. Submicron flatness and nanometer surface machine coordinate measuring system that utilizes a
roughness (5 nm Rmax) were reported on optical (NBFI unique contact probe structure by which both inclination
optical glass), magnetic (Mn-Zn ferrite), and electronic and motion errors are simultaneously compensated.
materials using diamond abrasive grinding wheels. Namba
et al. (1992) using this grinder ground Mn-Zn ferrite single Fourteen axes interferometers were used to compensate
crystals at 1200 m/min using various grain size diamond for the mechanical motion errors. To improve the surface
grinding wheels (SD 200-3000 or 4 to 100 pm mean grit quality they used a local pitch polisher with a flexible tool
size) and down feeds (1-100 pm). They classified the laminated with an elastic sheet which adopts to various
material removal process into three types: 1.) brittle mode, contours. For higher contour accuracy, a super-smooth
2.) "ductile" (here it is denoted as damage-free) mode, and polishing process was developed which polishes 500 mm
3.) ductile-fracture modes. They found damage free diameter optical elements with a contour accuracy of 78
surfaces depend on the down feed and the grain size of nm and a surface roughness of 0.13 nm rms on fused
the abrasive used with lower feed and finer grain size silica, CaF2, and CVD-SIC toroidal mirrors. They developed
favoring damage-free surfaces. They developed diagrams a contour tracking tool which employs an elastic layer for
showing the conditions required for damage free grinding contour tracking performance. A polychloroprene elastic
of Mn-Zn ferrites. Namba and Abe (1993) extended this layer is sandwiched between a metal shank and the pitch

550

Fig. 9(a) Ultraprecision surface grinder with a glass-
ceramic spindle (125 mm dia. ) of zero-thermal
expansion material (after Namba et al., 1989)
c)
Fig. 9(b) Section of the ultraprecision surface grinder
showing details (after Namba et al., 1989)
layer [Fig. 10(b)] using a pitch surface molding press. In
addition to the machines discussed above, another
example is the ultraprecision grinding machine developed
at Toyoda for the manufacture of non-axisymmetric mirrors
recently reported by Suzuki and Murakami (1995).
2.2. Workmaterials
A wide range of workmaterials including metals and their
alloys, ceramics, glasses, and semiconductors are
finished to a given geometry, finish, accuracy, and
surface integrity to meet service requirements. In view of
the wide range of metals and their alloys in various
metallurgical structures, chemical compositions, etc., and
their extensive coverage in various publications, they will
not be covered in any detail here. Instead, the emphasis
will be on brittle materials mostly ceramics, glasses, and
semiconductors which are finished by fine abrasives. For
example, ceramics are used for mechanical,
electromagnetic, optical, thermal, biochemical, and energy
oriented applications. Figures 11(a) to (d) give details of
the functions and types of ceramics used which are
finished by fine abrasives (after Nogawa, 1988). As
Keynote Papers
COUN7ZR EALANClNC - M WSLIDE
2 AXIS AIR
COORDINATE U R E MFOR
ENT
sy- \
/
V~EFWT~ONisounoN
S Y r n \ AKnSElSMlC STOPPER
Fig. lO(a) Schematic of Canon's super-smooth polisher
,
(CSSP) (after Ando et al., 1992)
1 ,/
I Shank
yElasdc layer
1 I
' Pitch
Fig. 10(b) Elastic contour tracking tool used in Canon's
super-smoothpolisher (after Ando et at., 1995)
can be seen, the range of workmaterials and the
requirements for finishing, including size, shape,
geometry, finish, accuracy, and surface integrity are much
broader than in the past where concentration was mainly
on some metals and glasses for mostly optical
applications (i.e., mirrors and lenses).
A qualitative comparison of ductile metals with nominally
brittle non-metals (ceramics, glasses, semiconductors)
indicates the differences and difficulties in finishing of the
later with fine abrasives compared to metals and their
alloys [Table 11. Metals, in general, have many desirable
features from the point of. material removal by fine
abrasives, such as metallic, non-directional bonding, high
symmetry in crystal structure, high thermal conductivity,
low density, no porosity, high purity, high fracture
toughness, large strain to fracture, and high impact
energy. In contrast, the non-metals are characterized by
covalent (directional), or ionic bonding, low symmetry,
limited or inadequate slip systems for plastic deformation,
low thermal conductivity, and low fracture toughness and
low breaking energy.
The absence of plasticity in crystalline ceramics can also
be traced to the nature of their bonding. The mechanism of
plastic flow in materials is due to the movement of
dislocations. In metals, because of non-directional
metallic bonding, extensive movement of dislocations can
take place even at low stresses. In covalent ceramics, the
directional bond is between specific atoms, leading to a
narrow dislocation and a high resistance to movement. In
terms of microscopic parameters, it can be shown that the
dislocation width is given by a function of the ratio of G/K
where G is the shear modulus and K the bulk modulus
551
 brides. such as ZrO,. MoS,,
LaCrO,. Sic. Ti. Zr

CdSCu,S ZnO-Bi,O,-Sic

Game rcria ferrite

QUMZ, LiNbO,
Pb(Zr. T i p , rue eanh prnct

Fig. 11(a) Electromagnetic functions and types of ceramics (after Nogawa, 1988)

Property Metals Non-metals

Tic, TiN. Type of atomic bond metallic ionidcovalent bond
Sic. no directionalitv directional
Al,O,-Zr,O,.
BJ. BN
I crystal structure 1. high symmetry I low symmetry I
I Thermalexpansion I high I low I
Fig. 1l(b) Mechanical functions and types of ceramics Thermal conductivity high low
(after Nogawa, 1988) Density hsh low
Mode of deformation ductile brittle
I Microstructure I
I
TIN, TIC,
CaF,, COO Heat-resistant Photo
Photo- I Intergranular
structure
Porosity I relative simple

1I
~ m ~ l y n o genedypores
complex

remain because of
the consolidation
1I
Dmesses used
glass
porcelain,
I I
high purity can be hi h punty is
obtained diacult I
LEDs, I Heat resistance 1 low to moderate I moderate to high I
Cap, GaAs,
GaAsP
Rare earth Toughness 210 (carbonsteel) 5.3 (S,N
i ,)
compounds: (MN/m2) 34 (A1 alloys)
EufiI,O, : Cr, Strain at fracture 50/. 02 %
Glass : optical
Nd, Y communication Weibull index P 5-20
1 Failure mechanism 1 plastic deformation I brittle fracture I
Glass fibers
Fig. 1l(c) Optical functions and types of ceramics (after
I Breakingenergy
(J/cm2) 1
I lo
I 10-2
I
Nogawa, 1988) Thermalshock high low
resistance

552

(Wyatt and Dew-Hughes, 1974). This ratio is given by G/K
= 3(1- 2v)/2(1+ 2v), where v is the Poisson's ratio. For
metals v is about 0.3 and G/K = 0.37, whereas for
ceramics v = 0.1 and G/K = 1. Consequently, the yield
stress at which dislocations move is much higher for
ceramics than for metals, and approaches the fracture
stress. Thus, covalent ceramics are characterized by their
brittleness in the single and polycrystalline states. This
should be the case, for example, with Si3N4 ceramic which
is predominantly covalent.
In polycrystalline ionic materials, the adjacent grains are
compelled to change their shape in the same manner, if
void formation is to be avoided at the boundaries. von
Mises showed that this requires five independent slip
systems. A slip system is the combination of a slip plane
and a slip direction, and an independent system is one that
produces a deformation that cannot be produced by a
combination of the others. Whereas metals have many
slip systems, ionic crystals are only able to slip on a
limited number of slip planes, due to the restriction that
similarly charged ions must not be forced into near-
neighbor positions such that cohesion would be lost.
Consequently, polycrystalline ionic materials are brittle
with cracks forming at the grain boundaries instead of
plastic deformation.
From the above, it is clear that metals should behave
differently than ceramics during finishing (Komanduri,
1996). The former fails predominantly by plastic
deformation while the latter by microchipping or
microcleavage, grain boundary cracking, and grain
dislodgments, etc. The presence of grain boundary glassy
phase in some ceramics, as in the case of Si3N4 with MgO
as a sintering aid, should not be overlooked in formulating
the mechanism of failure. Similarly, glasses, though brittle
like ceramics, should behave entirely different than either
metals or ceramics as the structure in the former case is
amorphous. At room temperature, failure should occur
predominantly by brittle fracture and above the glass
transition temperature by viscous flow. Thus many of the
models advanced on polishing of glass and ceramics
based on the application of polishing of metals should be
viewed with caution. Also, chemo-mechanical effects play
an important role in many of these processes., The key to
the containment of the cracks due to brittle fracture with
these materials is to chose "gentle" conditions during
processing. Since metals pose less difficulty to finish to
the requirements for specific applications compared to
non-metals, this paper will focus on the finishing of brittle
materials.
a. Advanced Cerarnics
Advanced ceramics for structural and wear applications of
interest include alumina (Al2O3), silicon nitride (Si3N4),
silicon carbide (Sic), zirconia (ZrOq), and SiAION. The
nature of some of these materials will be briefly discussed
in the following.
The nature of atomic bonding determines the hardness as
well as the Young's modulus of the material. Figure 12 is
the variation of Vickers hardness with modulus of
elasticity of a range of materials (after Gilman, 1973). The
ratio of VH, for the metallic bonded materials which are
ductile is about 250, while that for the covalent bonded
materials which are brittle is about 20. The ratio may vary
between them for ionic materials. Thus, this ratio may play
an important role when considering the micro-deformation
fields in the initial stages of indentation at light loads
involving irreversible deformation to the onset of median
vents. Low density and low mobility of dislocations are the
reasons for the high hardness of some of the brittle
materials. In the following, a brief description of the
synthesis of one ceramic, namely silicon nitride is given as
an example.
Keynote Papers
r
GaAs I 0 ' vcc$MbC
G a S b l N /(si
Ice TIC
0
I 0
/
/ 0 O lr
2 loo Ionic bondin<
PbS
1lLiF, 0
)ONi
'Rb
m
j'l
I / uP0Pt
a NaCl
Y I,@??Pb Metalic banding
V
.- ./At A9
lo-' P b o 04KcL
KBr
I
L I I
10' lo2 10'
Modulus o f elasticity G Pa
Fig. 12 Variation of Vickers hardness with modulus of
elasticity for a range of metallic, ionic, and
covalent bonded materials (after Gilman, 1973)
Silicon nitride (Si3N4), is considered as the candidate
material for high-strength, high-temperature applications
such as gas turbine components and for balls and rollers in
hybrid bearings. Si3N4 is predominantly a covalent (75 %)
material built up of SigNq-tetrahedra joined in a three
dimensional network by sharing corners P-Si3N4. Si3N4 can
be processed by sintering, reaction bonding, hot-pressing,
and HIP'ing. YqO3 or MgO are the common sintering aids.
During the high-temperature hot-pressing of Si3N4 with
small amounts of MgO addition, a complex glassy phase is
formed at the grain boundaries. It is primarily a magnesium
silicate modified by Ca, Fe, Al, and other impurities initially
present in Si3N4. At temperatures -1 100°C grain boundary
sliding occurs under loading. Addition of yttria to Si3N4,
generally leads to a crystallization in the glassy phase at
the grain boundaries instead of pure glassy phase with
MgO. However, the oxidation resistance of this material
was found to be inferior to those with MgO additions
(Lange, 1974). For this reason, MgO added to Si3N4 is
increasingly used in high temperature applications.
Si3N4 is produced either by conventional uniaxial hot-
pressing or HIP'ing. One starts with an a-SigN4 powder to
which sintering aids, such as MgO, or Y2O3 are added.
Under a pressure of -14 MPa and temperatures in the
range of 1650 OC to 175OoC, some of the a-SigN4 reacts
with the additive and a thin layer of SiO2 coats each
particle of Si3N4, producing a liquid silicate in which the
remaining a-SigN4 dissolves and re-precipitates as
elongated P-Si3N4 grains. On completion of the a to P
transformation, the elongated P-grains are surrounded with
residual silicate oxynitride grain boundary phase. The
elongated nature of these grains, that are typically 0.5 to 4
pm, gives hot-pressed Si3N4 its high strength.
The apparent limitations of the hot-pressed Si3N4 are due
to the nature of the grain boundary phase and not intrinsic
to the Si3N4 (Leatherman and Katz, 1989). Attention has
therefore been focused on controlled modification of the
grain boundary phase, such as grain boundary
crystallization in hot-pressed Si3N4 with Y2O3 additions.
While such a modification was found to provide higher
strength at both room and elevated temperatures
(-14OO0C) as well as better creep and oxidation
resistance, this material was found to suffer from an
intermediate temperature oxidation (-1 OOO°C). This
material is also found to be difficult to finish by grinding
553
and polishing. Consequently, complex parts made of this usually alkali silicate glasses and flints are lead alkali
material are rather expensive, once again shifting the silicates. Figure 13 shows the variation of index of
emphasis to Si3N4 with MgO additions. refraction (Na line) with dispersion for a variety of glasses
according to Schott nomenclature (Boyd and Thompson,
Relatively pure Si3N4 material was found to be elastic up to 1980).
fracture exhibiting practically no plasticity. This is in view
of the predominantly non-ionic or covalent nature of Table 2 Physical and chemical properties of some silicate
bonding. Impurities in the material seem to enter into the and borate optical glasses (after Izumitani, 1979)
grain boundary glassy phase and lower its viscosity. Thus Actd- Water-
the high-temperature mechanical properties of hot- Glass Softening Vickers resistance resistance
pressed Si3N4 appear to be controlled by the viscous grain tvpe pant hardness weight loss weight loss
boundary phase, the viscosity at a particular temperature ("C) (kg/mm2) ("/I ("/I
being controlled by impurities, and more than likely, the Silicate Glasses
hot-pressing aid. Grain boundary sliding is the suggested SF6 470 41 3 1.3 0.03
mechanism for subcritical crack growth, plasticity, and KF2 490 627 0.07 0.07
creep. FK1 475 666 1.9
b. meal Glasses BK7 61 5 707 0.08 0.13
SK2 700 707 0.70 0.05
Glass is an inorganic material supercooled from the molten SKI 6 680 689 3.3 0.58
state to a rigid condition without crystallizing. The
viscosity of glass changes drastically above the glass Borate Glasses
transition temperature, Tg. Consequently, at temperatures LaK12 670 743 1.7 0.35
above Tg, glass can flow viscoplastically, i.e., stress is LaLK3 650 762 1.9 0.70
proportional to strain rate rather than strain. The glass' LaLF2 675 803 1.3 0.25
transition temperature is not particularly high (only a few LaKl 0 670 803 1.2 0.02
hundred deg. C). Hence, it would not be too surprising to NbF1 650 a24 1 .o 0.01
find viscoplastically deformed chips in the grinding of NbSF3 650 803 0.76 0.01
glass. At room temperature, the viscosity is > 1020 MPa TaF2 685 847 0.74 0.01
sec. The modes of failure from brittle to viscoplastic can
take place at relatively low temperatures. Hence, some In three dimensions, a glass has microscopically irregular
researchers prefer to use the term viscoplastic flow or form. Under strain this irregular structure insures that the
deformation of glass instead of simply plastic deformation links are equally loaded. The void space permits very high
(Fiedler, 1988).
The basic structural unit of silicate glasses is the silicon-
oxygen tetrahedron in which a Si atom is tetrahedrally
c'oordinated to four surrounding oxygen atoms. Oxygens
shared between two tetrahedrons are called bridging
oxygens. Most glasses fall into the category of silicates
containing modifiers and intermediates. Addition of
modifiers, such as sodium oxide, Na20 to the silica
network alters the structure by cleaving the Si-0-3 bonds
to form Si-0-Na linkages. Separating the silica tetrahedra
from each other makes the glass more fluid and amenable
to melting and forming. Modifiers (or fluxes) also cause a
decrease in resistivity, an increase in thermal expansion,
and generally lower chemical stability. Various types of
glasses include soda-lime glasses, leaded glasses,
borosilicate glasses, aluminosilicate glasses, and various
refractory glasses, such as fused silica. Soda-lime glass
accounts for some 90% of, all the glasses produced. In
addition to Na20, CaO, and Si02, these glasses may
contain MgO, Al2O3, BaO, etc. Alumina increases v Value
durability and MgO prevents devitrification. Replacement Fig. 13 Variation of the index of refraction with dispersion
of alkali by boric oxide in a glass network gives a lower for a variety of glasses. The shaded area
expansion glass. Aluminosilicate glasses can be indicates the region of glass formation. (after
chemically strengthened and can withstand high Boyd and Thompson, 1980)
temperatures. Leaded glasses have a high refractive
index and can effectively shield from high energy hydrostatic compression as overloaded links buckle and
radiation. Table 2 gives physical and chemical properties permanent compressive densification occurs. The majority
of some silicate and borate glasses and Table 3 gives the of plastic deformation is of this type, since the irregular
type and chief chemical constituents for some structure permits little shear or plastic deformation by
representative optical glasses. dislocation motion.
Optical glasses are usually described by their refractive Flow of glass is directly related to the proportion of the
index at the sodium 0 line (589.3 nm) and their v value (or network modifiers which introduce non-bridging atoms into
Abbe number) which is a measure of the dispersion or an otherwise complete network. Thus flow is facilitated by
variation of index with wavelength, i.e., v = (nD-l)/(nF-nc) having a large number of non-bridging atoms which are
in which nF is the refractive index at the hydrogen F line 'loose ends' in the structure and permit flow in a manner
(486.1 nm) and n is the refractive index at the hydrogen C somewhat analogous to the half planes which terminate on
line (656.3 nm). Ttese data are incorporated by Schott into dislocations.in crystals (Marsh, 1964). It may be noted
a six-digit numbering system used to identify optical that the conventional concept of plastic flow due to
glasses. 'Glasses with index nD < 1.6 and a v value of 55 dislocation motion in silicate glasses has to be
or above are defined as crown glasses; those with a v abandoned. Plastic flow requires deformation in shear
value below - 50 are defined as flint glasses. Crowns are which necessarily involves the breaking of inter atomic

554
 Keynote Papers

bonds and the formation of new bonds with new partners. exploded rapidly into the computer industry. The ability of
semiconductor manufacturers to incorporate extremely
Table 3 Designation and chief chemical constituents of complex electronic circuits on a single chip of Si has
some silicate and borate optical glasses (after revolutionized the design and manufacture of countless
Izumitani. 19791 products. What makes a semiconductor the basis for the
electronic devices is the capability to alter, at will, the
SF dense flint electrical properties by adding controlled amounts of
F flint Si02-Pb0-R20 impurity atoms (dopants) into its crystal structure.
LF light flint
LLF extra light flint A semiconductor is generally, a nonmetallic crystal held
KF crown flint together by ionic/covalent bonds with no free electrons
available for conduction. It is a solid in which the highest
K crown Si02-RO-R20 occupied energy band (i.e., valance band) is completely
BaSF dense barium flint 1 Si02-PbO-BaO-R20 filled and the next band (i.e., conduction band) is
BaF barium flint
-
BaLF liaht barium flint
1. Si0;-B203-Pb0-Ba0
completely devoid of electrons at 0 K. The valence
electrons are, however, not very tightly bound to their
atoms (energy gap is -0.66 eV for Ge and -1.1 eV for Si)
BaK barium crown Si02-Ba0-R20 and may be readily excited by thermal, electromagnetic,
BaLK light barium photonic, or other energy sources. Semiconductor
crown behavior is determined by the number of electrons in the
SSK extra dense barium
crown
1 Si02-B203-Ba0 conduction band and the number of holes in the valance
band. The electrons or holes may be created either by
SK dense barium crown 1 energy excitation as just outlined or by the introduction of
suitable impurities in the lattice. Semiconductor materials
LaF lanthanum flint (Si02)-B203-La20- whose conductivities do not depend upon impurities are
PbO-AIpOn called intrinsic semiconductors and those which depend
on the presence of impurity atoms (e.g. P in Si in the n-
LaLF light lanthanum flint (Si02)-B203 -Lap03 type which has one extra valence electron, or Al in Si in the
-PbO-RO p-type which has one electron less or has a hole) are
LaLK light lanthanum (Si02)-B203-La203 called extrinsic semiconductors.
crown -ZrOp-RO Figure 14 shows the semiconductor elements in the
LaK lanthanum crown Periodic Table (Wyatt and Dew-Hughes, 1974). Of these,
TaSF dense tantalum) B2O3 -La203-Th02- silicon (Si) and germanium (Ge) are the most frequently
flint Ta205 used semiconductor materials. Further, Ge is finding
TaF tantalum flint application in infrared optics. The latter is due to its
excellent transmission characteristics in the infrared
TaK . tantalum crown B203.La203-Th02-RO spectral region of 2-15 pm. Compound semiconductors
BK borosilicate SiOZB2O3-R20-Ba0 also exist. They include the Ill-V compounds, e.g., AIP,
crown AIAs, Gap, AISb, GaAs, InP, GaSb, InAs, and InSb, the II-
PK DhOSDhate crown VI compounds, e.g., CdS, ZnS, ZnSe, CdTe, ZnO, and
~

other compounds, e.g., CdSb, ZnSb, PbS, PbTe, PbSe,

PSK dense phosphate crown AI2O3-P2O5-RO
FK fluorcrown . Si02-B03-R20-RF
This process of bond formation and breakage is a
recognized phenomena in metals and certain ionic
crystals. There is no precedent for it in covalently bonded
materials even in the crystalline state, unless the material
is strained at a temperature from 1/2 to 2/3 of its absolute
melting temperature. The atomic mechanism of
densification strictly occurs under high hydrostatic
pressure which is typically the condition under the indenter
during indentation. In the case of glass, it involves simply
a collapse of the structure into a more close packed
arrangement by a process of minor bond rotations and
changes in bond angles. That is, densification is a
"displacive" transformation while plastic flow is
"reconstructive."
In conclusion, it may be pointed out that in the case of
glass, the very nature of the inelastic deformation remains
a highly contentious issue; it is no simple matter to
establish which of the two basic, competing processes,
namely, shear induced flow (viscoplastic above glass
transition temperature) or pressure induced densification
(phase change or compaction of an open structure)
dominates with the contact zone (Lawn and Wilshaw,
1975).
c. Semiconductor Materials
The invention of the transistor in the 1940'swhich is based
on semiconducting materials brought a major technological
revolution in the electronic industry and its applications
Bi2Teg, Sic, cubic BN, and Sic. Doped diamond is also
semiconducting. GaAs is seriously being considered as a
replacement for Si. Its major advantages include higher
speed, capability for light transmission, larger energy gap
(1.43 eV), and higher operating temperature (-200OC).Its
drawbacks are availability and cost, and greater
processing complications.
Semiconductor materials have electrical conductivities
between conductors on one end and insulators on the
other. Figure 15 shows the range of electrical conductivity
for insulators, semiconductors, and conductors (Van
Vlack, 1982). Conductors have resistivity in the range of
10-5-1 0-6 a-cm; semiconductors from 10-2 R-cm for
extrinsic to 10 - lo5 R-cm for intrinsic semiconductors. A
good insulator may have resistivity as high as 1015 R-cm.
Thus one can obtain some 21 orders of magnitude.change
in the resistivity as one goes from insulators to highly
conducting materials.
For electronic applications Si and Ge must be of high purity
and melted in an atmosphere of hydrogen or helium to
prevent oxidation. They should also be dislocation free.
Such high purities can be obtained only by zone refining
and zone melting techniques. A widely used technique for
growing single crystals is the Czochralski method of
crystal pulling.
Si and Ge have strong directional covalent bonds with a
diamond structure. A diamond structure is composed of
two fcc lattices displaced from each other by 1/4 of the
body diagonal. They crystallize in the diamond cubic (fcc)
structure and the slip system is c110> {l 11). Pure edge

555

lllB I VB VB VIB VlIR
C they considered various abrasive processes such as ID
GiI
In
TI Pb Bi Po
Fig. 14 Various semiconductor elements in the Periodic
Table (after Wyatt and Dew-Hughes, 1974)
Paraffin
P.T.F.E.
5 P.E.
Mica
Quartz
PhenoI - Resistivity at 300 K, R cm
I 0-9
I 03
1 Graphite
Fig. 15 Range of electrical conductivity for insulators,
semiconductors, and conductors (after Van
Vlack, 1982)
dislocations do not occur and only pure screw dislocations
with a Burgers vector at 60° to the dislocation line occur.
At room temperature the dislocations are reluctant to move
and hence the materials are brittle but ductility may be
imparted at higher temperature, or by irradiating with light,
applying an electric field, or doping with donor impurities.
Table 4 lists properties of Si and Ge. Although the earliest
electronic devices were fabricated with Ge, Si has became
the industry standard. The abundance of Si makes its
economical. Also, Si-based devices can 'operate at higher
temperatures (-150°C for Si compared to -100 OC for Ge)
without breakdown. Silicon's important processing
advantage is that its oxide, SiO is an excellent insulator
and can be used for isolation an$ passivation purposes. In
contrast, germanium oxide is water soluble and unsuitable
for electronic devices. Doped with arsenic, Ge is used as a
transistor element in many electronic applications. The
use of Ge for IR applications has surpassed its use in
electronics. It is also finding wider applications as a
detector in 'y-ray spectroscopy and in fiber optics. Si
doped with B, Ge, P, or As, is used in transistors, solar
556
cells, rectifiers, and solid state devices. Tonshoff et al.
(1990) presented a keynote paper on "Abrasive Machining
of Silicon" at the ClRP General Assembly in 1990 where
cut-off grinding, surface grinding, lapping, polishing, and
dicing for machining Si and should be referred to for
details.
Table 4 Pro
Si Ge
Lattice structure Diamond Cubic Diamond Cubic
BondType Covalent Covalent
~atticeconstant A 5.4307 5.6569
Density, 4/cm3 2328 5.323
Melting Point, "C 1410 937.4
Young's MO~UIUS, N/mm2 188,400 155,800
Poisson's Ratio, v 0.28 0278
Hardness, Hv, kg/mm2 950
Mohs Hardness 6.5 6.3
FractureToughness, 15
MPa rnm1l2
2.3 x lo5 46
Energy Gap, eV 1.1 0.67
2.3. Abrasives
Both conventional abrasives, such as A120 , SIC, ZrO,,
B4C as well as superabrasives, such as diamond and
cubic boron nitride are used extensively in finishing with
fine abrasives. In addition, for promoting chemo-
mechanical action, soft abrasives, such as MgO, Ce02
and fumed silica (- 5-7 nm in diameter) are used.
Innovations in this area include new methods of synthesis
of the abrasives in the nanograin size and their bonding for
different applications. Production of fine grain size
abrasives by synthesis (chemical routes) instead of
comminution has enabled the production of nanograin size
materials. The sol-gel process is used for the precipitation
of fine powders. Either amorphous or crystalline material
can result depending on the composition, the precursors,
the handling, and heat treatments. A sol is a colloidal
suspension of particles in a solvent which is obtained by
hydrolysis of an inorganic (aluminum sulfate) or organic
precursor (tetraethyl-orthosilicate) solution. The
hydrolysis can be controlled and/or polycondensation
process can be initiated to yield inorganic oxide networks
composed of -M-0-M- linkages or hydro-gels. For the
preparation of ultra-fine abrasive powders, the hydro-gels
are dried, calcined at required temperatures, and ground to
fine particle sizes. The precursor solutions commonly
employed in sol-gel process are numerous and vary from
simple inorganic sulfates, chlorides and nitrates to
complex organic alkoxide solutions such as isopropoxide,
butoxide, etc. Abrasive particles in the range of a few
nanometers can be produced depending on the nature of
the precursor and processing conditions. Significant work
was carried out on the development of alumina based
abrasives produced by sol-gel technique by 3M Co. and
Norton Co. and is covered by a number of U.S.patents.
Similarly, fumed silica, used extensively in the chemical
mechanical polishing of Si wafers, is made by a similar
technique. Nanocrystalline diamond is produced by shock
synthesis, for example, by duPont: Magnetic abrasives
used in the magnetic float polishing (MFP) of ceramic balls
(Komanduri et al., 1996a) utilizes nanosize (-10 nm
diameter) Fe O4 particles in a water or kerosene medium to
which suitabb fine abrasives are added prior to polishing.
The fine iron oxide particles are prevented from
agglomeration by coating them with a suitable surfactant.
 Keynote Papers

In addition to conventional metal, resin, and glass continuous layer of silica is deposited on the blade. Using
bonding, new bonding media and methods have been this technique, the authors were able to demonstrate
developed. For example, Professor Nakagawa developed chipping free dicing of Si wafers. Semba et al. (1996)
a cast iron bonded diamond grinding wheel (Nakagawa et developed a malamine bonded diamond wheel (of fine grain
al., 1986) which can be continuously dressed using the size) with high porosity for mirror finishing of die materials
electrolytic in-process dressing (ELID) technique again -
(cemented tungsten carbide, HA 90.5). They developed a
developed by him and his colleagues. Using this technique highly foamed structure to increase the porosity to > 75 %.
Nakagawa et al. (1988), demonstrated grinding of hard This increased the flexibility of the wheel, reduced swarf
material as well as grinding of mirror finished surfaces for loading, and enabled self dressing of the grinding wheel.
optical, electronic, and magnetic applications. New
machine tools with higher stiffness, accuracy, speed .Blade cover. cathode
capability, and power were developed to meet the needs
for these applications. Nakagawa's technique of CI
bonded diamond grinding wheels and ELID dressing was
also incorporated in the ultraprecision grinding machines
developed by McKeown and his colleagues at CUPE. IAsi3' , lade'
&(B , anode
lkeno et al. (1990) developed an electrophoretic
deposition technique for the production of abrasive pellets
(10 mm diameter) containing ultrafine colloidal silica
particles in the range of 10-20 nm in a sodium alginate
-,
/ S
il
ia layer
c

,yp ...... . ..
-
,Sectionof
a workpie

binder (- 4 wtY0). In this technique, ultrafine silica

particles are negatively charged in an alkaline solution
which causes homogeneous dispersion. When an electric
field is applied, the charged abrasives move to the anode
and adhere to it [Fig. 161. They used these pellets to
t - Chuck
Fig. 17 Apparatus for chipping-free dicing applying
1
accomplish nanometric grinding of brittle materials, such electrophoretic deposition (after lkeno et al.,
as Si wafers and sapphire without the need for an 1991)
ultraprecision grinding machine for producing extremely
smooth surfaces (c 10 nm PV). The low bond strength of
the binder (softer wheel) enabled higher removal rate Optical Polishing for Glass Lens
without loading. The surface finish obtained was found to .-2 (Mirror Finish)
be better and the removal rate higher than with xerogel r
P
pellets.
0
m
Ern ct
DC poser supply

h - ' h
sorp.
I.I

r
0
(Matte Finish)
Lapping for Various Materials
Fig. 18 Schemati,c showing various combinations of
backing material (soft or hard tools) and abrasive
(coarse or fine) for different applications resulting
in different finishes (after Kasai et al., 1990)

3. PRECISION MATERIAL REMOVAL PROCESSES WITH

Fig. 16 Synthesis of ultra-fine colloidal silica (10-20 nm)
abrasive pellets (10 mm diameter) in a sodium
alginate binder (-4 wt%) by an electrophoretic
deposition technique (after lkeno et al., 1990)
lkeno et al. (1991) extended this study to develop a similar
concept for dicing of Si wafers without chipping [Fig. 171.
The technique employs the use of electrophoretic
deposition of ultrafine abrasives (collodial silica, 5-7 nm
dia.) onto a conductive diamond saw blade in an electric
field. A bronze bonded diamond saw blade which forms
the anode is mounted on a high-speed air spindle and the
blade cover forms the cathode. A groove is formed on the
blade cover through which colloidal silica solution passes
continuously. When an electric field is applied, a
FINE ABRASIVES
Kasai et al. (1990) classified lapping and polishing on the
basis of the backing material (or lap material) and
abrasives used. Figure 18 is a schematic showing various
combinations of backing material (soft or hard tools) and
abrasive (coarse or fine) for different applications resulting
in different finishes. For example, in the finishing of Si
wafers or glass lenses, a soft backing plate and a fine
abrasive are used while in conventional lapping giving a
matte finish, a hard backing plate (e.g., cast iron) and a
coarse abrasive are used. Similarly, for finishing of
sapphire a hard backing plate and a fine abrasive are used
while for finishing of die materials, a soft backing plate and
a coarse abrasive are used.
Figure 19 is a schematic showing the material removal
mechanisms and various polishing methods after Kasai et
al. (1990). The two main material removal mechanisms are

557
mechanical and chemical. The mechanical action can be
abrasion, or erosion involving plastic deformation or
fracture. Tribological conditions may often initiate
mechanical or chemical change due to the generation of
frictional heat without the need for hard abrasive particles
(tribo-chemical). The chemical action can result in
chemical reaction products or dissolution of the material.
Often the processes can be a complex mixture of
mechanical and chemical depending on the polishing
conditions as well as the abrasive-workmaterial-
environment.
Chemical and Mechanical Polishing
E i m ~ E m nydeamr
p-Mw
POWlii
P oField
h n iAssislsd
i
Chemo-MechanicalPolishing
Fig. 19 Schematic showing the material removal
mechanisms and various polishing methods (after
Kasai et al., 1990)
The following is a brief review of various precision material
removal processes using fine abrasives (Imanaka and
Okutomi, 1979; Stowers et al., 1988). All these processes
are predominantly mechanical in nature, as material is
removed by the fine abrasives by mechanical action.
However, by proper choice of the abrasive-workmaterial-
environment, it is possible to induce additional chemical
action to promote chemo-mechanical finishing. Various
processes under consideration include:
1. Ultraprecision Grinding
2. Conventional FixedLoose Abrasive Polishing
3. Computer Numerical Control (CNC) Polishing
4. Float Polishing
5. Elastic Emission Machining (EEM)
6. Ultrasonic MachinindGrinding
7. Erosive Jets
8. Chemical Hydrodynamic Polishing
9. Magnetic Field Assisted Finishing
10. Electrolytic Abrasive Mirror Finishing
11. Honing
12. Precision Sawing
13. Single Point Diamond Turning (SPDT)
14. Microcrack-freeor Damage-free Grinding
Single point diamond turning (SPDT) uses a single crystal
diamond as the tool instead of fine abrasives as in
ultraprecision grinding. Consequently, it does not strictly
qualify under "finishing with fine abrasives." However,
these two processes are somewhat similar in the sense
that (a) the size of the chips are small and of the same
order and (b) both processes yield mirror finished surfaces
using specially designed high precision, rigid machine
tools generally in a controlled environment (Shaw, 1996).
558
Since these two processes are used interchangeably for
some applications, e.g., finishing of semiconductor
materials, SPDT is briefly reviewed for completeness.
. .
1. Ultraprecision Gnndinq
The technology of grinding brittle materials and
ultraprecision machining of both ductile and brittle
materials have advanced significantly in the past two
decades. All aspects of grinding systems have advanced,
including new abrasives, new bonding methods, new
lubricants and their supply to the grinding zone, new
machine tools, new controls, and new processes. Special
machine tools were built to address creep-feed grinding of
aerospace materials such as nickel-base super alloys and
hardened steels, high-removal rate machining using high
speed, high power machine tools, and ultraprecision
grinding machines for generating the required size, form,
finish, and surface integrity. Extremely rigid, high power,
high speed, high performance machine tools were built to
handle machining of difficult-to-machine materials.
Nakagawa et al. (1986, 1988) used a modified machining
center to grind 3-D profiles on ceramics (AI2O3, Zr02 and
Si3N ) and hard metals (cemented carbide) using their cast
iron Iiber bonded diamond grinding wheels. The high
rigidity of the machine tool and the toughness of the
grinding wheel enabled very high removal rates 42,000
mm3/min in deep grinding of alumina and 2000 mm3! /min in
!he case of high feed rate grinding of Si3N4). The grinding
ratio with the cast iron fiber bonded diamond grinding wheel
was about 1600 in the case of Si3N4 and 2200 in the case
of hot pressed Zr02 workmaterials.
Ohmori and Nakagawa (1990, 1995) developed a highly
efficient grinding technique for mirror finished surface
grinding of Si wafers utilizing electrolytic in-process
dressing (ELID) of micro-grain diamond abrasive in a cast
iron fiber bonded grinding wheel. ELID dressing with the
new grinding wheel enabled mirror finishing on
conventional grinding machines without the need for
subsequent polishing or use of conventional lapping. The
metal binder of the grinding wheel is preferentially removed
by the ELID process to expose new grits (Fig. 20). The
wheel becomes the positive pole by a brush smoothly
contacting its surface. The electrode fixed below the
wheel surface is the negative pole. In the clearance of -
0.1 mm between the two poles, electrolysis occurs by the
supply of the electrically conducting fluid. By continuously
dressing at a slow rate, the protrusion of the grits is
ensured. They obtained a surface finish of 48 nm Rmax
and 7 nm Ra using a 4 pm diamond abrasive in the grinding
wheel. By using submicron diamond abrasive they could
obtain a surface finish of 8.92 nm R
,, and 1.21 nm Ra.
Chemical-~olution-type Nozzle
grinding fluid
'-'
A. System construction B. Electrode detail
Fig. 20 Schematic illustrating the principle of electrolytic
in-process dressing (ELID). Ttie metal binder of
the grinding wheel is preferentially removed by
electrical discharge to expose new grits (after
Ohmori and Nakagawa, 1995)
Ohmori and Nakagawa (1995) extended this work using a
constant in-feed pressure to other workmaterials including
monocrystalline silicon, glasses and ceramics. They
 Keynote Papers

characterized the surfaces using SEM, AFM, and

subsurface damage using X-ray diffraction. Good surface
finish (2.8 nm Ra and 18 nm ,),R
, high flatness h/4 (A = Table I. specifications of the machine.
632.8 nm) for Si of 60 mm in diameter, and low subsurface
Maximum workpiece dimensions 500 mm x 500 mm
damage (< 1 pm depth) were achieved. Using constant X axis travel 1100 mm
pressure grinding and ultrafine diamond grinding wheel Y axis travel 700 mm
(#120000 and 3000000) they obtained an Ra of 3.29 A Ra Z axis travd 280 mm
and an,,R of 23.4 A for Si and an Ra of 3.25 A and an A axis rotational angle flQ
Rm, of 23.6 A for glass. 6 axis rotational angle 115'
Wheel spindle speed W O O 0 min
Zhong (1992) introduced a new method for grinding toroidal Rapid feed rate ( X . Y,2 axis) 300 mm min-I
(CVD-Sic on graphite, size: 50 x 15 x 6 mm), elliptic (SIC, Rapid rotational speed (A, B axis) 18' min-'
size: 180 x 180 x 20 mm), and circular cylindrical mirrors Axis positional resolution ( X , Y. 2 axis) 0.01 pm
(Sic, size: 180 x 180 x 20 mm) with a large radius of Angular resolution (A, 8 axis) 0.00002"
curvature (150 mm). He used a micro-displacementtable
with piezoelectric actuators to shape the grinding wheel Fig. 21(a) Toyoda's 5-axis ultraprecision grinding
and then to precisely control the relative position of the machine: machine specifications (after Suzuki
ground surface and the grinding wheel. The surface finish and Murakami, 1995)
of the mirrors ground by this method was 20-40 nm Rma,
and 2-5 nm Ra
Suzuki and Murakami of Toyoda Machine Works (1995) ~~

developed a large 5-axis (X-, Y-, Z-, and A-, B- rotational

axis on the work table) ultraprecision grinding machine for
non-axisymmetric aspheric mirrors [Figs. 21(a)-(d)]. The
overall dimensions of the machine are 3.2 m x 2.7 m x 4.4
m and the overall weight is 34 tons. The resolution of this
system is 10 nm and the rotational positioning is controlled
by a laser rotary encoder with a resolution of 0.00002°.
The machine is equipped with an on-machine measurement
and feedback control system. Fig. 21(a) gives the
specifications of the machine and Fig. 21 (b) shows
schematically the machine configuration. The maximum
workpiece size is 500 mm x 500 mm. Rigid and precise
hydrostatic oil bearings were used for the slideways and
rigid ball screws for the feed system. To achieve precise
motion of the table, the guide rails with hydrostatic Side View
bearings for vertical motion are supported by wedge type
leveling blocks [Fig. 21(c)]. By adjusting the height of
each block, the guide rails can be deformed to be straight.
In order to attain high accuracy, the required profiles are
generated by transferring the peripheral contour of the
grinding wheel to the workpiece. To avoid the influence of
grinding wheel contour error on profile accuracy, the point
of contact of the grinding wheel with the workpiece is kept
constant by five-axis control. Profile error from the
deformation of the workpiece and wear of the grinding
wheel is measured on the machine by the system shown in
Fig. 21(d) and compensated for by the NC control. Using
this machine tool, Suzuki and Murakami (1995) reported
grinding of a toroidal surface 500 mm x 100 mm made of
CVD Sic for which they obtained a profile accuracy of 0.5
pm and a surface roughness of < 5 nm rms.
librrliw irelalor/
3300 4
Front View
Nakao and Katamura (1996) developed a unique,
intelligent face grinding machine (weighing about 8.3 tons) Fig. 21(b) Toyoda's 5-axis ultraprecision grinding
to fabricate ultraflat surfaces of thin, brittle materials [Fig machine: machine configuration (after Suzuki
221. The vertical spindle machine was designed to be and Murakami, 1995)
highly rigid and thermally stable to reduce tool positioning Another interesting development deals with the
errors. Multi-layered piezo elements were used in the ultraprecision machining of 3-D shapes on brittle materials
actuators for fine positioning. Wax was used to hold the with high surface accuracy using a lathe-type
workmaterial. Thermal deformation sensors and a heater ultraprecision milling machine and pseudo ball end mills
on the column legs of the machine tool control the wheel (Takeuchi et al., 1996). The three control axes are
position in addition to spindle attitude control mechanism. mounted on a cast iron base with concrete, which is
An active knowledge based control system with sensors, supported by air and oil dampers not only to dampen
computers and actuators were used to further reduce the vibrations from the surroundings but also to absorb the
deviation. Details of the actual design process and the vibration of the machine tool during machining. They used
development of an intelligent machining center an air bearing spindle capable of 55,000 rpm. The
incorporating active compensation for thermal distortion translational axes are mounted on air bearing linear slides.
are given elsewhere (Shinno et al., 1992; Hatamura et al., The resolution of the two translational axes and the
1993). Using a #325 diamond grinding wheel (100 mm rotational axes are 1 nm and 0.0001 deg, respectively. The
diameter) at a spindle speed of 1500 rpm, feed rate of 200
machine tool is located in a temperature controlled clean
mm/min, and a depth of cut of 0.03 mm, they obtained a
room. They machined a glass mask of 1 mm diameter and
flatness of 0.5 pm over 200 x 200 mm. 30 pm in height with a surface roughness of 50 nm .),R
,(

559
Perhaps, this technology can lead to micro-electro- 2. Conventional Fixed/Loose Ab rasive Polishing
mechanical (MEM) devices of the future.
Abrasive polishing could be considered the oldest
manufacturing technology. Over two millennia of
Capacltance prove craftsmanship has resulted in local optimization of these
S t r h h t edge Ball screw processes. Individual opticians and organizations have
Upper rail developed preferred combinations for abrasives, laps,
r speeds and pressures on a proprietary basis. In the optics
industry, in particular, the underlying physical reasons for
the apparently superior performance of the polishing pad
are rarely studied. Less secrecy surrounds the
manufacture of large astronomical optics and Meinel
(1 982) reports that prior to 1980, large telescope mirrors
\Guide rail ulth f i n e were finished primarily to spherical shapes using rough
deformation mechaniqp cam-guided grinding followed by loose abrasive grinding
and finally by pitch polishing. Rough grinding was done
with a cylindrically-shaped fixed abrasive (Sic or diamond)
grinding wheels. The workpiece was rotated about a
vertical axis and the vertical position of the grinding wheel
was cam-controlled. Contour accuracies of 100 pm are
typically obtained in this step with a corresponding surface
roughness of about 100 pm rms. Grinding with loose
abrasives, such as Sic, A1203 or diamond is accomplished
~y using a spherically-shaped segmented metal lap (cast
iron or brass), ceramic, or other glass-like materials. A
typical procedure involves the use of progressively finer
r
Transmitter Automtic M: data Analyser of abrasives (60 pm, 40 pm, 30 pm 12 pm, and 3 pm) as the
of Hc Data +

programing device figure error process progresses. Contour accuracy at the end of this
step is reported to be on the order of 25 pm or better, and
the surface roughness in the range of 1 to 3 pm rms. The

-
final stage utilizes a polish on a soft pitch lap using fine
' Controller to abrasives such as iron oxide (FeZO,), aluminum oxide
Receiver of measure figure (AI2O3), rare earth oxide (CeOz), or diamond. Several
HC data accurasy finishing steps are generally utilized with progressively
decreasing abrasive particle sizes ranging from 2 pm down
Tr igger IUDconverter1 to 0.8 pm. Final contour accuracy for large mirrors (2 to 5
m) is about one wavelength of visible light or 633 nm with a
surface roughness of 1-2 nm rms.

I tra-precision grinding orchin In conventional pitch polishing, pressures (load divided by

Fig. 21 (d) Toyoda's 5-axis ultraprecision grinding
machine: on-machine error measurement and
compensation system (after Suzuki and
Murakami, 1995)
- ,Motor for vertical feedina
Fig. 22 Schematic overview of the intelligent face
grinding machine for fabrication of ultraflat
surfaces of thin, brittle materials (after Nakao and
Katamura , 1996)
Jeong et al. (1996) qombined a grinding process using an
ultraprecision grinding machine with electrolytic in-process
dressing (ELID) and #4000 cast iron bonded diamond
abrasive wheel and chemical mechanical polishing (CMP)
using ultrafine soft abrasives (SiOz). They used a hard
lapping pad made of polyvinyl chloride (PVC) instead of
softer pads commonly used. The slurry used was colloidal
silica stabilized with potassium hydroxide (21% SiO, of 10-
20 nm particle size) diluted with deionized water, pH was
kept at about 10.2. A flatness of 0.3 pm/l25 mm diameter
and a roughness of R , 4 nm on a Si wafer was produced.
apparent area) between the pad and the glass are usually
modest, in the range of 1-3 kPa and the maximum relative
velocity between the pad and the glass is held to about
0.05-0.25 d s to minimize frictional heating. Most modern
polishing machines utilize a combination of reciprocating
and circular motions, or synchronous circular motions of
the polishing pad and the workpiece. Because abrasive
particles become imbedded in the pitch, or synthetic
material of the polishing pad these methods are classified
as fixed abrasive polishing (or fixed abrasive grinding).
As the pitch pad moves across the glass surface, some of
the fine abrasive particles become imbedded or fixed in the
soft pitch. The smoothing or polishing action occurs by
means of a combination of scratching or wearing of
asperities on the surface of the glass, and a chemical
reaction between the water-base slurry and glassy
material. In recent years some synthetic materials, such
as polyurethane, foam, felt, and Teflon (in which there is
considerable elastic resiliency) have been successfully
used in place of pitch for polishing glasses and metals.
Brown (1 986) has reported that hard pitch laps are usually
chosen for critical optical component polishing. Hard laps
have produced the smoothest abraded surfaces,
approximately 0.2 nm rms and even as low as 0.08 nm rms
on small 0.6-0.8 mm gyro mirrors. This process can
routinely maintain surface contour accuracies of 25-50 nm
rms. Surface removal rates for glass ranges from 1.5 n d s
for polishing large optics (4 m diameter aspheric mirrors) to
20 nm/s when polishin 0.25 m diameter flats. Assuming a
9
pad size of 1 x lo4 mm , this leads to a volumetric removal
rate of 0.015 mm3/s.
Various polishing methods used in lapping of
semiconductor materials can be classified into three

560

categories (Kasai et al., 1988). 1) close contact lapping,
e.g., pitch polishing, wax polishing, metallurgical polishing,
bowl feed polishing; 2) semi-contact polishing, e.g.,
conventional mechanical chemical polishing of Si wafers;
and 3) non-contact polishing, e.g., float polishing, elastic
emission machining, hydroplane polishing. By changing
the nature of contact from close contact to non-contact
one can alter the mechanism of polishing from mechanical
to chemical.
A variety of lapping techniques using flexible laps have
also been reported. Synthetic fabric-faced laps with
diamond paste are used to post polish single point
diamond turned optics plated with electroless Ni (Parks
and Evans, 1994). The lap consists of a metal backing or
pin plate covered with a self-adhesive Pellon (a non-woven
synthetic fabric). In between the Pellon and backing plate,
a layer of self-adhesive, closed-cell foam tape is inserted
to provide a pliable lap surface. Using diamond paste
rubbed into the Pellon and a few drops of polyglycol-based
diamond paste extender, electroless Ni plated optics were
finished to a surface roughness of < 1 nm rms. Teflon laps
have also been used to superpolish a variety of optical
glasses to an rms surface roughness of c 0.1 nm
(Leistner, 1993). Examples include using a teflon lap with
A1203 abrasives for the finishing of BK7 glass, and with
SiO2 for single crystal Si. Recently, a lapping technique
based on a "rapidly renewable lap" has been reported
(Evans and Parks, 1995; Parks et al., 1997). The concept
involves a thin film used to carry abrasive and which
covers a stable lap structure machined to the required
figure. A surface roughness of 1 nm rms was obtained for
the final finishing of black filter glass, in which a Mylar film
was used with a foamed SIC substrate and a cerium oxide
slurry. Other examples on the use of flexible tools for
ultra-fine lapping include the use of a fluorocarbon foam
polisher with SiO2 powder which resulted in a surface
roughness of 0.3 nm on fused quartz (Kasai et al.,
1990). Another example is the development of the Canon
Super-Smooth Polisher (CSSP) and the use of a flexible
tool laminated with an elastic layer for the local pitch
polishing of aspherical surfaces (Ando et al., 1995).
Attained surface roughnesses on a toroidal mirror using
this approach include 0.13 nm rms on fused silica, 0.15 nm
rms on CVD-Sic, and 0.12 nm rms on a CaF2 aspherical
lens. Some of these are briefly discussed in this paper.
Watanabe and Suzuki (1981) developed a polishing
machine for high precision, damage free polishing of
semiconductor wafers. In this process, the wafer is
polished by hydrodynamic action without contact with the
polishing lap. The polisher is designed such that the lap
surface has alternate flat and sloping surfaces similar to a
hydrodynamic thrust bearing [Figs. 23(a) and (b)]. The
guide rings are provided for work rotation as well as for the
polishing lap. Using this machine, Si wafers were polished
to a flatness of 0.3 pm over 76.2 mm except for the outer 3
mm circumference and a surface roughness of 1 nm.
Kasai et al. (1988) developed a fully automatic polishing
machine for finishing mirror-like, damage-free, flat
surfaces of compound semiconductor wafers. This
technique is called grogressive Eechanical and Chemical
polishing (P-MAC polishing). In this method, the polishing
mechanism is altered progressive!y from mechanical to
chemical. GaAs (100) wafers (50 mm in dia.) were polished
to a flatness of <2 pm for 80% of the diameter with a
surface roughness of < 2 nm R m a using bromine methanol
solution and soft polishing laps (fluoro-carbon plastic).
3. Comouter Numerical Control ICNC) Polishing
A number of organizations have developed computer
controlled polishers (see for example Jones and Rupp,
1990; Becker and Beckstette, 1988) including Perkin-
Keynote Papers
Elmer (now Hughes Danbury Optical Systems) for the
Hubble Space Telescope. Computer controlled polishing
methods are more frequently used for aspheric, higher
cost optics than for flats and spheres. The long polishing
times put a premium on the accuracy of the preceding
process. Large telescopes mirrors are finished by
computer controlled (CNC) machining, grinding, and
polishing. Both the pitch polishing and CNC grinding and
polishing are illustrated in Fig. 24. One of the machines at
the Optical Sciences Center (OSC) of the University of
Arizona is the Large Optics Generator (LOG). This is a
large bridge-type vertical spindle grinder. The LOG can
machine single mirror glass blanks 8 m in diameter with
focal ratios as fast as f/0.8 and 2 m segments of 10-15 m
diameter mirrors as fast as W0.35. Grinding removal rate
with a 355 mm wheel operating at 1750 rpm was 28 mm3/s.
The contour error of the finished part was 4 pm and the
surface roughness about 2 pm rms (Parks et al., 1985).
Drive dear
Rinq locating dote CircuJar '0°'
Fig. 23(a) High precision polishing machine using a
hydrodynamic fluid film (after Watanabe and
Suzuki, 1981)
Tool length
Fig. 23(b) Tool-workpiece interface of high precision
polishing machine (after Watanabe and Suzuki,
1981)
Becker and Beckstette (1988) reported a CNC polisher and
a coordinate measuring machine with a 600 x 400 x 50 mm
measuring volume used to feed error information directly
back to the polisher to remove contour errors. In an
iterative process the system has made parts with a
residual error of 0.7 arc-sec (3.5 prad). The CNC polisher
has 9 NC axes of which 6 can operate simultaneously
under computer control. The polishing pad is
pneumatically applied to the surface and the contour error
data is used to determine polishing dwell time necessary to
remove the contour error. This set of machines allows the
construction of strongly aspheric components with less
than 1 arc sec of tangential error.
Jones et al. (1985) reported on a machine for computer
numerical controlled generation of [Computer Controlled
Optical Surfacing (CCOS)] of large aspheric optics. Like
many other non-deterministic processes, the machine
utilizes a small polishing pad (75-100 mm diameter) and
56 1
iteratively approaches the desired optical contour by
alternately polishing and then performing an
interferometric inspection. Final contour accuracy on a
large (nearly 4 m) optic was reduced from 24 to 1.9 pm.
Final surface roughness on a 1 m optic was 1.6 nm rms.
Abrasive pertick Motion Pitch or other
4 soRmaterIal 7
Fig.24 A schematic view of the conventional pitch
polishing and CNC grinding/polishing processes
(after Stowers et at., 1988)
For large optics, a group at Steward Observatory
(Anderson et at., 1991) reduced the problem of the lap
size limitation imposed by the requirement that the lap fit
the local curvature of the part by the use of an adaptive
lap, the basic shape of which is adjusted by edge moment
bending to fit the local part contour. Extension of the
technique to smaller optics has not yet been
demonstrated. An obvious inversion of this approach is to
bend the part during spherical polishing and let it "spring
back" to the required asphere. This approach was
developed by Nelson (1981) for the Keck telescope and
has been used subsequently for synchrotron optics.
4. Float Polishing
In 1977, Namba and Tsuwa (1977) first reported on a
technique, named "float polishing" for ultrafine finishing of
sapphire single crystals. They used free abrasives (a
combination of colloidal Si02 (4-7 nm grain size), Ce02,
and A1203) in deionized water on an ultraprecision,diamond
turned tin lap supported on a rigid support base [Fig. 251.
The concentration of the abrasive particles in water ranged
from 2-8% by weight. Surface roughnesses of I 1 nm rms
were routinely obtained. By assuming a 100 mm2 area, a
removal rate of 1 x- mm3/s was obtained. Unique
features of this method include: .
. A relatively soft material (tin) was used for the lap
instead of pitch or synthetics used in conventional
polishing;
A fluid gap (many times greater than the particle size)
was maintained between the lap and the surface to be
polished to develop a hydrodynamic film; and
Very high polishing speeds (15 2 . 5 m/s)between the
lap and workpiece were used.
This method was later extended by Namba and Tsuwa
(1978, 1980a) for polishing single crystal and poly-
crystalline Mn-Zn ferrites for magnetic heads for
computers and tape recorders, using diamond, SO2, y-
A1203, Fe203, Zr02, ZnO, Cr203, and MgO for abrasives.
The material removal rate was found to be a function of the
size of the abrasive particles, concentration of particles in
the polishing slurry, and applied load. They found very
good flatness (h/lO), excellent retention of the edge
geometry, and extremely low surface roughness 1 nm R,
for single crystal materials (and 5 nm R, for polycrystalline
562
materials). They proposed a mechanism of material
removal involving erosion of the surface atoms by the
bombardment of abrasive particles without the introduction
of dislocations. The removal rate thus depends on the
probability of bombardment of particles, kinetic energy of
the particles, impact angle, bonding energy distribution at
the topmost layers of atoms, etc., without involving the
hardness of the abrasives.
Vessel I
!7 Diamond-turned
I
:ron
ickne! S S
Fig. 25 Schematic of the float polishing apparatus using
a diamond tumed tin lap (after Namba et at., 1987)
Namba and Tsuwa (1980b) extended this free abrasive
float polishing method to finishing metals, such as type
304 stainless steel, iron, chromium and nickel. A surface
roughness of 2 nm Rmax was obtained in 304 stainless
steel with MgO powder. This was obtained by terminating
the damage-free polishing at a point prior to the
appearance of the grain boundary due to anisotropy in the
polycrystalline material.
Namba and Tsuwa (1980c), Namba (1982), Bennett et al.
(1987), and Namba et al. (1987) have continued the
development of float polishing technology for optical
materials, including borosilicate glasses, fused silica, and
low-expansion synthetic amorphous glasses such as ULE
(Corning Glass) and Zerodur (Schott Glass). Details of the
ultraprecision polishing machine were given earlier under
Machine Tool Systems, Sec 2.1. The specimens are
suspended in the polishing fluid by hydrodynamic action
due to the specially shaped diamond turned tin. lap used.
The temperature of the supply oil for hydrostatic bearing
and polishing fluid is controlled to within 0.05 K to maintain
the flatness of the tin lap surface. Similarly, the
temperature of the polishing fluid was maintained constant
to minimize evaporation of the water. Apparent pressures
between the lap and workpiece were in the range of 70-500
Pa, about an order of magnitude higher than pressures
normally used for conventional pitch polishing. Substrates
up to 180 mm diameter have been polished to a flatness of
h/20 = 0.03 p m and surface roughness of I-2A rms
measured with a stylus, optical and x-ray scattering
methods.
Touge and Matsuo (1996) investigated the effect of
various lapping conditions on the polishing rate of a
polycrystalline Mn-Zn ferrite. They used a tin lap made by
single point diamond turning and lapped with a 0.5 to 2 pm
diamond abrasive under a pressure of 42.7 and 90.5 kPa.
They reported that feed, depth of cut, and nose radius of
the diamond tool all affect the amount of stored area of
fluid. They classified the diamond abrasive in the lap into
those that are anchored in the lap and those moving freely
in the lapping fluid. They found the pitch size of the
grooves on the lapping plate generated by diamond turning
had a significant effect on the removal rate which is about
9 times with a pitch of 250 pm when compared to a pitch of
100 pm. They correlated this with the amount of stored
lapping fluidlnumber of free abrasives. The average
 Keynote Papers

material removal rate with fixed and loose abrasives was controlled to within 20 nm. The removal rate, however, was
about 2.7 higher than that with only with fixed abrasive, found to vary non-linearly with concentration of abrasive in
indicating that much of the workmaterial is removed by the the slurry, over the range of concentration of 3 to 35% by
loose abrasives. weight. The interaction of the slurry and workpiece is
probably similar to that taking place in float polishing.
5. Flastic Emission Machinina ) However, the relationship between the various process
Mori et at. (1976) of Osaka University investigated the
feasibility of removing material on the order of atomic size
(which is the ultimate unit of machining) by mechanical
methods to obtain a finished surface which is not only a
complete mirror but also which is crystallographically and
physically undisturbed. Conceptually, the material removal
in conventional machining is in part due to the deformation
or fracture based on migration or multiplication of pre-
existing dislocations, or by the enlargement of cracks
originating from the pre-existing microcracks.
Consequently, the limit of the surface finish is determined f Feed
by the distance between these defects. If however, the 'Fluid lubrication
material removal can occur on atomic size units, then the
finish generated can be close to the order of atomic Fig. 26(a) Rotating sphere/workpiece interface in elastic
dimensions. Using ultrafine particles to collide with the emission machining (EEM) (after Tsuwa et al.,
work surface, it may be possible to finish the surface by 1979)
the atomic scale elastic fracture without plastic
deformation. They termed this new process Elastic
Emission Machining (EEM). Using single crystals of Si and
Crossspring rl
A1 as workmaterials they removed material by EEM using
alumina particles of 2, 8, and 20 pm size with a vibratory
collision method. They calculated the material removal
rate and examined the replica of the surface in a TEM and
surface integrity of the samples by X-ray diffraction. They
compared the kinetic energy of the powder particles with
the binding energy and the stress induced by the collision
of particles with the theoretical strength of the material
(Mori et at., 1978). Mori et at. (1976) established
theoretically and experimentally that atomic scale fracture
can be induced elastically and the finished surface can be
undisturbed crystallographically and physically.
Mori et at. (1976) and Tsuwa et al. (1'979) extended this
concept for the purpose of computer numerically
controlled (CNC) shaping and finishing of semiconductors
and glass. The primary difference between this method
and float polishing is that a rotating sphere, made from a
relatively soft polyurethane material, is used for the lap, as
compared to a relatively harder tin disc used in float
polishing, [Figs. 26 (a) and (b)]. Similarities to float
polishing include:
A fluid is maintained between the spherical lap and the
workpiece. The thickness of this gap is much larger than Fig. 26(b) Schematic of the EEM assembly used on an NC
the size of the abrasive particles used, machine (after Tsuwa et al., 1979)
Relative velocities between the workpiece and the lap parameters, such as velocity, pressure, abrasive particle
are very high (-3 m/s), and size and concentration, material removal rate, and surface
roughness are still not well understood.
Pressures between the lap and workpiece are extremely
high at 500 kPa, about 3 orders of magnitude higher than Linear material removal rates varied from 0.6 nm/s when
with float polishing. polishing GaAs with Zr02 to 23 nmls when polishing single
In the EEM process a polyurethane ball, 56 mm in crystal Si with this same abrasive. Removal rates for float
diameter, is mounted on a shaft driven by a variable speed glass with Zr02 abrasive were as high as 75 nm/s.
motor. The axis of rotation is oriented at an angle of - 4 5 O Assuming a I mm2 contact area, the removal rates range
relative to the surface of the workpiece to be polished. The from 6 x lo-' to 7.5 x mm3/s. Surface roughnesses as
workpiece is submerged in a slurry made up of Zr02 or low as 0.5 nm rms have been reported on glass and a
A1203 abrasive particles and water. Pa,rticle sizes ranged surface roughness of 2 1 nm rms was obtained when
from 20 nm to 20 pm. However, in most cases a fluid gap polishing single crystal Si. Contour accuracy over short
of about 1 pm is maintained between the bottom of the distances has been demonstrated to be better than 100
sphere and the workpiece, and abrasive particle size nm.
ranges from 0.1-0.6 pm. The slurry is circulated through Mori et at. (1983, 1985a,b) have modeled the removal
the work volume by means of a diaphragm pump, and process and concluded that it could not be the result of a
maintained at constant temperature with a heat kinetic energy exchange with the surface but must be a
exchanger.
surface energy phenomenon in which each abrasive
The material removal rate for the workpiece was found to particle removes a number of atoms after coming into
be linear with dwell time at a particular location, allowing contact with the surface. The type of abrasive used has
the total amount (thickness) of material removed to be been found to be critical to the removal efficiency and a

563
CVD process (Mori et al., 1988) for producing the
abrasives has been developed. In this process fine a-
quartz particles are coated with a few atomic layers of
alumina or zirconia. Mori et al. (1987a,b) compared the
mechanism of atomic removal in EEM and chemical
etching. In EEM, chemical activity of the solid surface is
employed instead of an etchant as in chemical etching.
This enables an atomistically finished surface.
.. . .
6. Ultrasonic Machinin-
Suzuki et al. (1993) introduced a new technique of
ultrasonic assisted grinding of ceramics and other hard
materials using a non-rotational tool for producing holes or
cavities with sharp inner edges or comers. In this method
a fixed abrasive type grinding tool (electroplated diamond
tool #loo-#800 grit size) is vibrated by a complex
ultrasonic transducer longitudinally at 29.4 kHz frequency
and 6 pm in amplitude and torsionally at 19.4 kHz
frequency and 32 pm in amplitude at the same time. The
complex ultrasonic vibration keeps the grinding force low
and stable for a long time in grinding ceramics even without
rotating the tool. Workmaterials used are alumina, optical
glass, and sintered carbide.
7. Frosive Jets
Finnie (1958, 1960) and Hashish (1984) have reported on
the machining of metals, glasses and ceramic materials by
utilizing abrasives suspended in jets of water or air [Fig.
271. The removal mechanism is one of pure erosion
powered by the inertia of the abrasive particles and the
hydrodynamic forces applied by the moving fluid. Plastic
flow is reported even with brittle materials when fine
abrasives (1000 mesh) impinge on the surface at 70° from
the normal. Removal rates with glass were found to be 0.75
mm3/s when impacting the workpiece with 10 g/s of SIC
abrasive traveling at 150 m/s. The use of coarser
abrasives (120 mesh) resulted in brittle fracture of the
glass surface. The use of erosive jets for precision
surface figuring does not appear to have been reported in
the literature. This may be the due to the observations
Fig. 27 Material removal in erosive jet machining (after
Finnie, 1960 and Hashish, 1984)
that a rippled surface texture is nearly always produced in
the case of metals when cutting with erosive jets.
8. Chemical Hvdrodvnamic Polishing
A variation of float polishing has been used by Gormley et
al. (1981) and lves et at. (1988) to obtain smooth, damage-
free surfaces on semiconductor crystab of GaAs, InP,
and HgCdTe. These materials were etched in a mixture of
methanol, ethylene glycol, and bromine. No abrasive
particles were mixed with the etchant fluid. The smoothest
surfaces were obtained in InP and GaAs with a linear
material removal rate of 150 nm/s. Assuming a 100 mm2
contact area, this leads to a 1.5 x mm3/s removal
rate.
. .
9. Maanetic Field Assisted Finishing
In conventional finishing of ceramic balls for bearing
applications, the common manufacturing practice is to use
564
grinding followed by V-groove lapping (which is basically
the same method used for making the steel balls for
bearings) at low polishing speeds (50 rpm), high loads (10
N per ball), and using diamond abrasives. In practice, it
takes considerable time (some 6-16 weeks depending on
the material and size of the balls, final specifications of the
balls, and the manufacturing practices used) to finish a
batch of ceramic balls. The long processing time and use
of expensive diamond abrasives result in high processing
costs. Furthermore, the use of diamond abrasives at high
loads can result in surface damage, such as deep pits,
scratches, and microcracks and subsurface damage,
such as the large lateral and radiavmedian cracks. These
surface defects may bring about the catastrophic failure of
Si3N4 balls by the propagation of large brittle fracture.
To minimize the surface damage, “gentle”/”flexible”
polishing conditions are required, namely, a low level of
controlled force and abrasives not much harder than the
workmaterial. Higher removal rates and shorter polishing
cycles can be obtained by high polishing speeds. This is
accomplished by a new and effective process known as
magnetic float polishing (MFP). This process was
originally developed by Tani and Kawata (1984) and
improved significantly by adding a float to produce uniform
and higher polishing force by Umehara (1988) and
Umehara and Kato in Japan (1990, 1994), Childs et al. in
the UK (1992,1994a,b, 1995), and Komanduri et al. in the
USA (1996a; Bhagavatula and Komanduri, 1996; Umehara
and Komanduri, 1996; Raghunandan et al., 1997;
Raghunandan and Komanduri, 1997; Jiang and Komanduri,
1997a,b). The material removal rate by magnetic float
polishing is about 50 times higher than the conventional V-
groove lapping, owing to the use of higher polishing
speeds (up to 10,000 rpm) compared to lapping (50 rpm).
Further, the resulting surface and subsurface damage on
the finished ceramic balls can be minimized in MFP due to
the application of extremely low polishing load (-1 N/ball)
and damage-free chemo-mechanical polishing.
The salient features of the magnetic abrasive finishing
techniques are the following :
1. Very high finish and accuracy can be obtained,
2. Very little or no surface damage is imparted, such as
microcracks, to the ceramic parts, during the finishing
operation due to extremely low forces involved,
3. Finishing operations can be significantly faster than by
conventional techniques. This is due to higher spindle
speeds possible by this technique,
4. Smaller polishing batches can be handled. Very few
ceramic balls are needed in a batch for polishing, unlike
in conventional polishing where a large number is
required for alignment and accuracy requirements.
Number of balls depend on the size of the balls as the
float pad can be changed’to suit the size of the balls,
5. Fewer polishing steps are needed. The balls can be
processed from the rough to finished state in one
operation by varying the strength of the magnetic field
intensity, and the type and size of the abrasives used.
It is, therefore, not necessary to change the polishing
machines for roughing and finishing, and
6. The equipment can be easily adapted in the existing
equipment without the need for high capital investment.
Although the magnetic abrasive finishing process had
originated in the U.S. in the 194O’s, it was in the late 50’s
and 60s that much of the initial development took place, in
the former U.S.S.R. by Baron and his associates (Baron,
1975; Sakulevich and Kozuro, 1977, 1978; Konovalov and
Sulev, 1967; Konovalov and Sakulevich, 1974) and in
Bulgaria by Makedonski et al. (1974). These researchers

showed that the technique could be applied for the
finishing of a wide range of products mostly heavy
workpieces. In the late 198O's, Japanese researchers
(chiefly Professors Kato and Umehara of Tohoku
University, Professor Nakagawa and his associates of
University of Tokyo and Dr. Shinmura of Utsunomia
University) followed this work and conducted research for
finish polishing of value added products. For example,
Takazawa, Shinmura, and Hatano of Japan (Takazawa et
al., 1983, 1985) have conducted extensive research on
the magnetic abrasive finishing of rollers and the
associated equipment. They studied the principles of
operation, finishing characteristics under different
conditions, and various applications of magnetic abrasive
finishing. Shinmura et at. (1990) later extended these
studies and designed various equipment for internal
finishing of tubes, external finishing of rods, finishing of
flat surfaces etc. Childs et al. (1992) concentrated on the
design of the magnetic fluid grinding apparatus and on the
magnetic fluid grinding mechanics to predict the motion of
the balls as it is forced by the drive shaft through the
magnetic fluid in the cell and the onset of skidding (Childs
et al., 1994b). Komanduri et al. (1996a) concentrated on
the design of the MFP apparatus using FEM analysis; the
chemo-mechanical polishing effects (Gibbs free energy
analysis) of the various reaction products that can be
formed between the workmaterial, abrasive, and the
environment; and the overall technology for finishing hot-
pressed Si3N4 balls for hybrid bearing applications (Jiang
and Komanduri, 1997b).
The magnetic float polishing technique is based on the
magneto-hydrodynamic behavior of a magnetic fluid that
can levitate a non-magnetic float and abrasives
suspended in it by a magnetic field (Rosenweig, 1985). The
forces applied by the abrasive to the part are extremely
small (1 Nhall) and highly controllable. Raghunandan et al.
(1997a) investigated the MFP of Si3N4 balls. Figs. 28(a)
and (b) show an overall schematic and a photograph of the
magnetic float polishing apparatus used. A bank o!
permanent magnets (Nd-Fe-B) are arranged alternate N
and S below an aluminum chamber which is filled with the
required amount of magnetic fluid and appropriate abrasive
(5-10% by volume). The magnetic fluid is a colloidal
dispersion of extremely fine (100 to 150 A) subdomain
ferromagnetic particles, usually magnetite (Fe304), in a
carrier fluid, such as water or hydrocarbons (e.g.,
kerosene). Water during the polishing process not only
acts as a coolant but also participates in the chemical
reaction with the workmaterial; oil plays only the role of a
coolant with no chemical action; and in dry polishing, the
frictional heat generated accelerates the flow of the glassy
surface layer. It is made stable against particle
agglomeration by the addition of surfactants. In this
investigation a water base ferrofluid was used. Figure 29
shows an FEM simulation of the MFP process
(Raghunandan et at., 1997a), and Fig. 30 shows the
displacement of the abrasive particles in a magnetic fluid
under the action of a magnetic field (Umehara, 1994).
When a magnetic field is applied, the Fe304 particles in the
ferrofluid are attracted downward to the area of higher
magnetic field and an upward buoyant force is exerted on
all non-magnetic materials to push them to the area of
lower magnetic field. The abrasive grains, the ceramic
balls, and the acrylic float inside the chamber, all being
nonmagnetic materials, are levitated by the magnetic
buoyant force. The drive shaft is fed down to contact the
balls (3-point contact is obtained by contact of the balls
with the shaft at the top, float at the bottom and the ball
chamber on the side) and presses them down to reach the
desired force level. The balls are polished by the abrasive
grains under the action of the magnetic buoyancy force
when the spindle rotates. Damage-free surface on
Keynote Papers
ceramic balls are produced by the MFP technique because
a low buoyant force is applied via the flexible float. The
function of the acrylic float is to produce more uniform and
larger polishing pressure (the larger buoyant force near
the magnetic poles can be transmitted to the polishing
area by this float). An urethane rubber sheet is glued to
the inner guide ring to protect it from wear. The material of
the drive shaft is austenitic stainless steel (a non-
magnetic material).
Jiang and Komanduri (1997a) used MFP to finish 9.5 mm
diameter Si3N4 balls. They used a polishing shaft driven
by a high-speed, high-precision air bearing spindle with a
stepless speed regulation up to 10,000 rpm. The magnetic
field was measured by Gaussnesla meter. The pH value
of the polishing environment was measured by a
pH/Temperature meter. The polishing load was set up by
measuring the normal force with a Kistler piezoelectric
dynamometer connected to a charge amplifier and a
display. To calculate material removal rates, the weight
reduction in the balls was measured by noting the weight
before and after polishing at every stage using a precision
balance. The surface finish of the polished balls was
analyzed using a Form TalySurf 120L, Zygo laser
interference microscope, an atomic force microscope
(Digital Instruments Nanoscope Ill), and an ABT-32
scanning electron microscope. The sphericity of the balls
was measured using a TalyRond 250. Figures 31(a), (b)
and (c) are surface roughness data of the polished surface
taken with a stylus profilometer (Talysurf), laser
interferometric microscope (Zygo) and AFM. Figures 32(a)
and (b) show roundness traces of the as-received and
finished balls. It may be noted that the as-received balls
of Si3N4.had a ring of .-5mm wide x 200 pm deep band at
the parting line due to the HIP'ing process. The initial
diameter was 13.4 mm and had to be finished to 12.7 mm,
The total polishing time for finishing a batch of balls was
about 20 hours of polishing time. The surface finish
obtained was 4 nm Ra and 40 nm Rmax. The best
sphericity obtained of the Si3N4 balls was 0.15 to 0.2 pm.
Finished surfaces relatively free of scratches, pits, etc
were obtained.
TT-I Spindle
' Drive Shaft
Guide Ring
Magnetic Fluid
and Abrasives
Rubber Ring
Ceramic Ball
Float
Aluminum Base
Magnet
Steel Yoke
I
Fig. 28(a) Schematic of the magnetic float polishing
apparatus for finishing advanced ceramic balls
(after Komanduri, 1996b).
565

Fig. 28(b) Photograph of the magnetic float polishing
apparatus (after Komanduri, 1996b).
Fig. 29 Contour map of the magnetic field intensity of the
magnetic float polishing process simulated by
FEM (after Rzghunandan et al., 1997a)
Fig. 30 Displacement Of abrasive particles in a magnetic
under the action Of a magnetic field
(Umehara, 1994)
In the' magnetic abrasive finishing (MAF) process,
magnetic-abrasiveagglomerate is used where the finishing
566
pressure is exerted by the magnetic field (Rosenweig,
1985). Figure 33(a) is a schematic of the magnetic
abrasive finishing process. The magnetic abrasives are
linked to each other magnetically between the magnetic
poles N and S along the lines of magnetic force, forming
flexible magnetic abrasive brushes. A cylindrical
workpiece, such as a ceramic bearing roller, is clamped to
the chuck of the spindle providing a rotary motion. Axial
vibratory motion is introduced in the magnetic field by the
oscillating motion of the magnetic poles, surface and edge
finishing can be accomplished by the magnetic abrasive
brushes [Fig. 33(b)]. The rollers can be magnetic or non-
magnetic. Hence, this technique is equally applicable to
both steel and ceramic rollers. The process is highly
efficient and the removal rate and finish depend on the
workpiece circumferential speed, magnetic flux density,
working clearance, workpiece material, size of the
magnetic abrasive conglomerate including the type of
abrasive used, its grain size and volume fraction in the
conglomerate. The size of the magnetic abrasive
conglomerates is about 50 to 100 pm and the abrasives
are in the 1 to 10 pm range. Typical conditions used for
finishing of steel rollers are the following: roller speed 200
ft/min, magnetic flux density 1.2 T, working clearance 10.2
mm, vibrational frequency 15 Hz, vibrational amplitude
1.52 mm, mean diameter of magnetic abrasive .particles
100 pm. Under these conditions, stainless steel rollers
can be mirror finish ground from an initial R, of 0.22 pm to
final Ra of 7.6 nm in 30 sec. For finishing Si3N4 rollers B4C,
SIC, followed by Cr203 abrasives in a ferromagnetic
material conglomerate are used for the roughing, semi-
finishing, and finishing, respectively.
10. Electrolvtic Abrasive Mirror Finishing
Maehata et al. (1987) has reviewed electrolytic assisted
abrasive polishing of metal mirrors. They have reported
that large workpieces of stainless and carbon steel can be
easily mirror-finished in a short time and with high
efficiency. In the process the removal of metal is
performed mainly by electrolytic dissolution with the aid of
grain (abrasive) action and is proportional to the working .
current and time (i.e., subject to Faraday's law). The
generation of surface roughness is mainly the result of
grain action. The final roughness is determined by the
density of abrasives, their motion (feed rate), and their
edge angles rather than by their grain size. Removal rates
on the order of 0.1 mm3/s are reported and final surface
-
roughness values of 10 100 nm rms. The process had
not been applied to aspheric shapes, however, cylindrical
surfaces were polished with a contour accuracy of < 1 pm.
The EAMF apparatus is illustrated in Fig. 34.
11. Honing
Honing is a finishing operation using bonded abrasives in
the form of a stone to imDrove dimensional and form
accuracy in addition to finish of the parts (Kaczmarek,
1976). It is generally used for finishing of inside walls of
cylinders and considered as an alternative to internal
grinding of large diameter bores of I.C. engines or tong
hydraulic cylinders (Shaw, 1996). Honing is also used for
finishing external surfaces and flat surfaces. It is a low
velocity (speeds < 100 d m i n ) finishing process; 'much
lower than in grinding. As it is generally the last operation
performed on the surface of a part, honing generates
functional characteristics specified for a surface, such as
geometric form accuracy, dimensional accuracy, and
finish and involves corrections of errors carried over from
the previous operations. The motion of the tool involves
both rotary and reciprocating actions. Sometimes an
oscillatory motion is superimposed. Consequently, a
crosshatch lay pattern with an included anale between 20°
and 60° is generated. The valleys acts as reservoirs for
the lubricants in the case of cylinders used in
 Keynote Papers

reciprocating motion and the peaks carry the load (Dallas, a)

1976; Indge, 1991). Typically, several honing sticks or
stones are mounted on a honing head. Since the abrasives
are in continuous balanced contact with the workmaterial,
there is less deflection of the work and consequently
4
1
U.
. .
. b

higher accuracy can be achieved. Any type of abrasive, 4 b

A b

b 1

L 1

,
.r .n ..
-
In

I . a-

-. .
a. .2.
1 . ..
I 1
0.2 pm

-. .
a n
100 prn
1

-. ..
ma

a n

C a (5 pm) abrasive
R. 3.a ~n
RI.0.029 pm 1
4
11..
.
. c
-.-
I..

171.
I
I ..
Fig. 32 Roundness traces of the (a) as-received and (b)
finished balls (after Jiang and Komanduri, 1997a)

Magnetic Poles
Vibratory Motion ,/

Fig.31 Surface roughness data of the polished surface
of a Si3N4 ball taken with (a) stylus profilometer
(Talysurf), (b) laser interferometric microscope
(Zygo) and (c) AFM. (after Jiang and Komanduri,
1997a)
/ Magnetic Abrasives
Rotating Workpiece
Fig. 33(a) Schematic of the magnetic abrasive finishing
(MAF) process (after Fox et al., 1994)
including AO
,I , Sic, diamond, and cBN of any grain size
can be used. I%e, harder abrasives are used for finishing
of hard materials while coarse, softer abrasives are used
for roughing or semifinishing of softer workmaterials. The
length of the stroke is shorter than that of the work
surface, for during the honing operation, the hone should
not leave the workpiece. An important advantage of honing
is it enables very high dimensional accuracy almost
ndependent of the kinematic and dynamic properties of the

567

Fig. 33(b) Photograph of magnetic abrasive finishing
showing the magnetic abrasive brushes (after
Fox et al., 1994)
+
i a final surface roughness of 9 nm rms and contour
Fig. 34 Schematic of electrolytic abrasive mirror finishing
(EAMF) (after Maehata et al., 1987)
machine tool. As the number of grains involved is very
high, high removal rates are, achieved. Since very fine
chips are generated, the heat generated during the
process is minimal. Consequently, surface integrity is
maintained.
12. Precision Sawing .
Precision sawing is used for cut-off operations, such as
sawing of ceramics, glasses, and semiconductors either
using a fine wire or a thin disc. Examples of sawing include
slicing of silicon rods into wafers or cutting of wafers to
size. In wire sawing, a fine wire (steel = 125- 250 pm) with
a single layer of copper coated diamond grits over its
entire length is used. Fine wire sawing (with the wire
uncoated) is sometime performed in a fine abrasive slurry
(Fujisawa et al., 1983). As was previously discussed in
Sec. 2.3, lkeno et al. (1991) developed a chipping-free
dicing technology applying electrophoretic deposition of
ultrafine abrasives. They successfully diced Si wafers
without chipping and obtained optically smooth surfaces.
Thin discs are generally metal discs (brass or bronze) with
a single layer of copper coated diamond grits on its
periphery. Silicon rods are cut into wafers with an ID cut-
off wheel which is made of a thin spring steel (- 125 pm)
annular ring, the inner periphery of which is coated with a
single layer of copper coated diamond grits (Brinksmeier
and von Schmeiden, 1987). The blade should be as thin as
possible to reduce the kerf width but thick enough to retain
stiffness and stability during the operation. For details,
refer to Brinksmeier et al. (1987) and the keynote paper
"Abrasive Machining of Silicon" by Tonshoff et al. (1990).
The in-feed of the blade into brittle workmaterials, such as
Si wafers determines if median cracks will be formed and
propagated into the workmaterial. In addition, if lateral
cracks form following plastic deformation underneath the
568
abrasive due to large hydrostatic pressure, this can lead
to chipping of the workmaterial on either side of the
diamond disc. This may necessitate subsequent polishing
to remove the damage or to discarding the wafer as scrap,
both of which are costly.
13. Sinale Point Diamond Turnina (SPD v
Single point diamond turning is a well established
technology for the fabrication of metal mirrors and
aspheric optics (see for example, Saito, 1976, 1978a,b).
The design of the machine tool for this application was a
revolutionary advancement of the conventional machine
tool technology to nanometer precision. Air-bearing
spindles, pneumatic slides, high rigidity, high precision
machine tools, feed back control, and temperature
controlled environment are some of the features that
enabled this technology. The salient feature of the SPDT
process is the ability to directly control the contour as well
as the surface roughness through direct numerical control
(DNC). SPDT utilizes large single crystal diamond tools
with a fine edge radius (-20 nm). Donaldson (1979) reports
on the fabrication of 1.6 m diameter Cu aspheric optic with
accuracy of 25 nm. When machining non-ferrous
materials, such as Cu and Al, surface roughness values <
1 nm rms can be achieved. This technology is now being
extended for precision finishing of brittle materials. As we
are interested in the finishing of brittle materials, some of
the highlights of SPDT of these materials will be briefly
covered.
Gerchman and McLain (1988) investigated ultraprecison
machining of single crystal and polycrystalline Ge using a
Rank Pneumo MSG-325, two axis CNC lathe. The cutting
tool was a single point diamond tool with a -25O rake with a
nose radius of 0.63 mm. The spindle speed was 1500 rpm,
the feed was 2.54 mm/min, and the depth of cut was 0.75
pm. The surface roughness obtained was - 6 nm R .,
Nakasuji et al. (1990) investigated turning of Ge, Si, and
LiNbO, on an ultraprecision lathe with a single crystal
diamond tool of rake angle varying from 0 to -25O and nose
radius varying from 0.5 mm to 1 mm. In all.cases the
negative rake angle tools enabled higher cut depths before
the occurrence of brittle fracture. They obtained a surface
roughness, PV of 0.02 pm for Ge, 0.04 pm for Si and 0.03
to 0.1 pm for LiNbOg.
Shibata et al. (1994,1996) machined Si on an
ultraprecision lathe (Hitachi DPL-100) with a single crystal
diamond tool of nose radius 0.8 mm, a negative rake angle
of -40°. The large negative rake angle provides the
needed high hydrostatic pressure to cause plastic
deformation under the tool. The Si workpiece (11 mm in
diameter) was fastened on the chuck of the lathe. Turing
was carried out at a depth of cut of 100 and 500 nm at a
feed rate of 10 mrn/rev and a speed of 3.3 m/sec.
Kerosene was used as the cutting fluid. The
crystallographic orientation of the turned surface and
cutting directions were the (001) plane and [110] direction
respectively. At 100 nm cut depth, a mirror finished
surface of surface roughness 20 nm , , ,R measured with
a TalyStep was obtained. They reported that the sample
bent in a convex shape towards the turned surface due to
residual strain in the surface layer caused by turning.
Consequently, the flatness was about 1 pm PV. The chips
generated at 100 nm cut depth were reported to be
continuous [Fig. 351. Examination in the TEM indicated
that their structure is amorphous. TEM examination of the
machined surface indicated the damaged layer can be
divided into two parts, the upper layer (-150 nm) devoid of
crystallographic contrast (amorphous) which is
independent of cut depth and the lower layer with many
dislocations whose thickness was found to depend on the

cut depth. They concluded that a phase transformation
may have been involved in the observed phenomenon.
Fig. 35 SEM photograph indicating the appearance of
continuous chips obtained by turning Si at 100 nm
cutting depth (after Shibata et al. (1994)
Kunz et al. (1996) investigated single point diamond
machining of Si (100). They used a single crystal diamond
tool of -25O rake and 64 pm nose radius. The final feed rate
was 0.95 pmhev. They found subsurface damage of 1-3
pm due to the presence of dislocations networks even
though the surface roughness was 5 nm rms. Damage-free
material was obtained only after chemo-mechanical
polishing.
Venkatesh et al. (1997) investigated machining of single
crystal Si using a single crystal diamond tool of zero rake
and 0.75 mm nose radius. They reported a surface
roughness of 1 nm R, at a depth of cut of 1 pm, feed of 0.4
mdmin and a speed of 400 dmin.
14. Microcrack-free or Damaae-free Grinding
It has been observed by many researchers that the
tendency for subsurface microcracks to develop in the
case of brittle materials decreases with decrease in the
undeformed chip thickness and to almost disappear below
a critical value of cut depth. This has been thought to be
due to the material being less brittle below a certain value
and has therefore been termed "ductile" mode
grindingcutting. This is because at very shallow depths of
cut or down feed on an extremely rigid, high precision
machine tool, the stresses generated with most blunt tools
may not reach the values required to generate median
vents in brittle materials. The term "blunt" is used here not
to imply that the tool is worn but to point out that most tools
are blunt when the edge radius of the tool is larger than the
cut depth. The sharpest edge radius that can be produced
on tool is in the range of 20-70 nm. If the depth of cut is
less than this value, the material is removed with the
radius of the tool and not by the rake face. The material
under these conditions behaves in an elastic-plastic
manner without fracture. Alternately, it has been argued
that at very shallow depths of cut with blunt indenters, the
energy required to propagate cracks may be larger than
the energy required for plastic yielding, so plasticity may
become the dominant material removal mechanism (Tabor,
1970, 1986; Puttick et al., 1979, 1980).
In reporting on the diamond turning of brittle materials for
optical components, Nakasuji et al. (1990) discuss the
ductile-brittle transition as follows. In material removal of a
homogeneous material, the slip plane or cleavage plane
coincides with the plane of maximum shear or tensile
stress. Plastic deformation takes place when the resolved
shear stress in the easy slip direction exceeds a certain
critical value inherent to the workmaterial before cleavage
takes place. Similarly, cleavage fracture takes place when
the resolved tensile stress normal to the cleavage plane
exceeds a certain value prior to slip. The type of chip
formed is determined by the predominant criterion in a
particular cutting condition. Thus there can be a critical
Keynote Papers
condition of the stress field and the workmaterial
orientation that determines whether a chip is removed by
plastic deformation or cleavage fracture. The critical value
may depend on the density of the defects in the
workmaterial (microcracks vs. dislocations). As the
density of defects which act as nuclei for crack
propagation is not high for brittle materials, the critical
value of fracture depends on the size of the stress field. In
contrast, the density of dislocations may be sufficiently
high to cause plastic deformation even in a small stress
field. Consequently, the critical value for plastic
deformation may not show as significant a size effect as in
brittle fracture. Therefore, the stress state and the defect
distribution may have to be controlled in order to obtain
chip removal not by brittle fracture but involving plastic
deformation.
Nakasuji et al. (1990) further point out that plastic
deformation may be necessary to obtain mirror smooth
surfaces in brittle materials. However, it is not necessary
that chips be deformed in a ductile manner (Syn et al.,
1988). In fact, plastic deformation needs to take place
only in the region near the finished surface. Fortunately,
because of the radius of the tool and/or the use of large
negative rake tools, the material immediately beneath the
tool is under a significant hydrostatic pressure and
consequently is subjected to plastic deformation. As long
as the stresses are not high enough to cause median
vents, the surfaces generated can be quite smooth with
evidence of plastic deformation just underneath the
machined surface. In other words, evidence of a smooth
surface is no indication that chips are formed in a ductile
manner in brittle materials as in the case of metals.
Professor Shaw (1995, 1996) pointed out that it is a
misnomer to call the process "ductile-regime"cutting under
conditions (very small undeformed chip thickness) where
there may be no subsurface cracks and the surface is
mirror finished. It would, therefore, be beneficial to
delineate what is implied or understood by the term,
"ductile" machining of brittle materials. This way any
confusion regarding, or mystery surrounding the
terminology can be minimized, if not eliminated.
If the term means that below a cut depth (the so-called
"critical depth of cut" or, below a critical load), it would be
feasible to remove material without initiating residual
cracks at or near the surface that would compromise on
the mechanical properties of the material, then the use
of the term appears reasonable. Alternately, the material
may be removed, whatever manner is appropriate,
leaving essentially microcrack-free surfaces.
If the term means that plastic deformation, or better yet,
irreversible deformation, in brittle materials occurs
underneath the tool similar to an indenter at low loads, it
is also reasonable. This is because the indentation
process involves the use of a blunt tool (radius in the
case of a spherical indenter and large included angle in
the case of pyramidal indenter) that causes very large
component of hydrostatic pressure which inhibits brittle
failure and promotes irreversible deformation
immediately underneath the indenter. Thus when one
observes smooth surfaces in the grooves in brittle
materials, they may be due to plastic deformation
underneath the tool due to high hydrostatic pressures.
In the case of ultraprecision machining of brittle
materials, such as Si or optical glasses large negative
rake tools are invariably used andor the edge radius is
generally a significant fraction of the cut depth. Only
under these conditions would it be possible to machine
Si or various types of glasses without causing cracks.
However, in this mechanism, as Shaw (1996) rightly
pointed out, a relatively large volume of material has to
be brought to fully plastic state in order for a relatively
569
 small amount of material to escape as a chip. This is one
of the reasons attributed for the rapid increase in the
specific energy with uncut chip thickness at extremely
low cut depths.
If the term means that chips are removed from the brittle
material by plastic deformation ahead of the tool similar
to conventional machining of metals using positive rake
tools with a concentrated shear zone or plane, one has
to exercise extreme caution. Puttick et al. (1994)
reported that the debris of Si tumed by a diamond tool by
ultraprecision machining consisted of ragged platelets
and fine particles down to a few nanometers in size (not
long continuous chips).
It appears that surfaces apparently free of cracks or
near surfaces with dislocations are taken as evidence
for "ductile" machining of brittle materials. The former is
due to "gentle" conditions used and the later due to
indentation by a blunt indenter. Hence, caution again
should be exercised for this type of interpretation.
Based on this, it would be more appropriate to term this
process as microcrack-free grinding or damage-free
grinding.
Wills-Moren (1982, 1988, 1989) has reported damage-free
or micracrack-free grinding using a facing lathe with an air
bearing spindle of high axial stiffness (618 MN/m). It is
equipped with a piezoelectric driven tool holder capable of
cuts with 1 nm increments. The machine has been used to
machine spiral grooves (0.4 mm pitch) in BK7 glass with a
depth of cut varying sinusoidally from 0 to 400 nm. The
"critical depth of cut" was less than 100 nm with BK7 and
between 100-200 nm for Zerodur. A coolant and grinding
fluid are separately fed to the wheel to prevent thermal
growth and to lubricate the grinding zone, respectively. A
surface roughness value of 2.4 nm rms has been achieved
on TiC-AI2O3 ( H i = 1850) surface. Best results were
obtained with resinoid bonded diamond wheels when great
care was taken in truing and dressing.
Miyashita (1988) and Yoshioka et at., (1982, 1984, 1985)
report that historically grinding has used the brittle mode of
material removal. This results in considerable difficulty in
understanding and compensating for the wheel wear, and
predicting the material removal rate. Because of the
mutual wear of both the surface being ground and the
wheel itself, the entire process is difficult to control to high
accuracy. A transition from brittle mode to 'ductile' mode
grinding was reported by reduGing the down feed rate per
pass from 1.7 pm to 0.2 pm with quartz crystal.
Conventional single point dressing uses a depth of cut of
10-100 pm which results i n brittle mode dressing of the
grinding wheel. Current practice also utilizes self-
sharpening of the grinding wheel which means brittle mode
failure of the wheel. This also results in highly
unpredictable wheel wear. Miyashita suggests truing the
wheel by microtruing with a depth of cut of 0.2 pm which
utilizes a microtrued regulating wheel. A surface
roughness of 40 nm R, when using 800-1500 mesh
diamond abrasives was achieved. Utilizing the microtrued
wheel they obtained removal rates of 2 x mm3/s with
resultant surface roughness values of 100 nm rms and
contour accuracies of 200 nm. They have also ground
quartz crystal workpieces to a final surface roughness of 2
nm rms.
Fielder (1988) refers to the material removal process in the
case of glasses not involving brittle fracture as
viscoplastic grinding. He described the construction of an
Aspheric Lens Generator (ALG). The axisymmetric
workpiece was mounted to an air spindle on an air bearing
rotary table. The vertical axis of the grinding wheel was
mounted on an air bearing spindle attached to a double
slide. A coarse slide sets the radius of curvature of the
570
optic and a precision stage was used to adjust for the
asphericity. Cutting speeds of up to 60 m/s are
achievable; however a surface speed of 16 m/s is more
typical for grinding. Machine stiffness is > 100 MN/m in the
X direction (workpiece to grinding wheel carriage) and 18
MN/m in the Y direction(paralle1 to the grinding direction).
Experiments were conducted on fused silica (Herasil),
BK7, LaK8, SK2, SF5, SF10, and ceramic glass (Zerodur).
Slope errors of 10 arc-sec (50 prad) have been achieved.
Surface roughness values of 1.5 to 3.2 nm rms (parallel
and traverse to cutting direction) have been achieved.
Hashimoto (1988) presented a review of "ductile" regime
grinding at Nikon. Rough grinding with #400 grit diamond
has given roughness values of 10-20 pm R, x. Using
resinoid bonded wheels with Sic abrasives &ey have
achieved surface roughness values of 100 nm Rmax. A
cam-guided machine was used to make a 35-mm lens with
3 pm accuracy. They performed rough grinding with a
metal bonded diamond wheel followed by a resinoid bonded
diamond wheel. Flat-topped abrasive grains rather than the
randomly fractured shapes resulting from dressing a
grinding wheel with a single point diamond were
recommended. Grinding conditions for BK7 were as
follows: 50 rpm for the workpiece, 3000 rpm for the grinding
wheel, 40 pm/rev feed, 1 pm depth of cut. Using a #I3000
grit diamond in a metal bonded wheel, the author could
achieve a roughness of 50 nm R ,, and a profile accuracy
of 0.5 pm.
lzumitani (1979) conducted extensive experiments on
polishing, lapping, and diamond grinding of various optical
glasses. Table 2 in Sec. 2.2 gives physical and chemical
properties of some optical glasses he used whose general
designations (type and chemical constitution) are given in
Table 3. Figure 36 shows the variation of polishing rate
with micro-Vickers hardness for various types of silicate
-c
*g
0.5 >
0.45
Silicate glasses
-- 0 Borate glasses
SK,6.
-1 0.4 --
3 0.35--
al
a
rn
.C
c
0.3 --
n
In
0.25 --
0.2 I
I
I
I
I
I
I
I
I
I
350 450 550 650 750 850
Microvickers Hardness (kglmm')
Fig. 36 Variation of polishing rate with micro-Vickers
hardness for various silicate and borate optical
glasses (after Izumitani, 1979)
and beorate glasses adopted from lzumitani (1979).
lzumitani combined the data for both types of glasses and
concluded that there was no correlation between the two
properties. A careful reexamination (note the linear
variation) of the data after separating the silicate glasses
from borate glasses, and after considering the softening
points, the relationship appears more obvious (note the
best fit lines are added in this review). In the case of
silicate glasses (KF2, FKI, BK7 and SK16), the polishing
rate decreases with increasing hardness (see Table 2)
while for the borate glasses (LaK12, LaLK3, LaLF2,
LaK10, CdFI , TaF2), the polishing rate varies less steeply
even though the microhardness varies (from 650 to 803
kg/mm2). This is because the softening points of the
borate glasses are nearly the same and in the range of
650-670°C. As it is the hardness at the softening
 Keynote Papers

temperature that should be considered (and not room

temperature hardness), both hardness and softness have
to be considered in developing the relationship for the
polishing rate. Also, the chemical nature of the glass has
to be considered as the nature of bonding between the
glass and the water as well as the reaction with the
abrasive depends on this. Figure 37 shows the variation of
the polishing rate with the softening point for both silicate

-.-
I
0 Silicate glasses SK16
C
E
0.45 -- 0 Borate glasses 0 LaK12

-
'-i
I

Q)
0.4 I
% 0.35
K
0)
c 0.3
EUJ
0.25
n
0.2 ! I
1
I I
I
I
I
I
I

450 500 550 600 650 700

Softening Temperature( "C)
Fig. 37 Variation of polishing rate with softening
temperature for various silicate and borate optical
glasses (after Izumitani, 1979)
glasses and borate glasses. The author again states that
there is no correlation between them by combining the data
for both types of glasses. Actually, by separating the
silicate glasses from borate glasses again a clear relation
can be seen (the best fit lines are again added in this
revew). For example, for the silicate glasses the softening
point varies widely from 47OoC for SK6 to 7OO0C for SK2.
The polishing rate for these glasses can be seen to
decrease linearly with increase in the softening point
(except SK16). In the case of borate glasses, the
variation in the softening point is in a very narrow range
(from 650-685OC). Hence, one would expect only a slight
decrease in the polishing rate with softening point and this
in fact can be observed from the experimental results by
noting a steep line. In fact, many models of glass finishing
assume that there is no correlation of the polishing rate
with the hardness of the glass 0'. the softening
temperature based on this work. For example, Brown and
Fuchs (1989) point out that lzumitani had laid to rest the
theories of abrasion and thermally induced flow by showing
that there is no correlation between the polishing rate of
glasses and the hardness or softening point. In light of the
present interpretation, this may have to be reconsidered.
Puttick et al. (1989) extensively investigated the
machining of brittle materials including Si and various
glasses, such as soda-lime glass and fused quartz. They
emphasized the significance of stiffness and precision
requirements of the machine tool system used for the
ultraprecision machining of brittle materials at fine depths Fig. 38 The single point diamond machining of glass
of cut. They showed some fascinating photographs of the (Spectrosil): (a) general view of spiral swarf
machined surfaces of fused quartz (Spectrosil) where adhering to machined grooves (b) detail of
many areas contained an interesting type of swa.rf detachment of spiral ribbons from single pass
adhering to the sides of the grooves [Fig. 381. For machined grooves, (c) machined grooves
example, they showed a ribbon detaching from the side of showing detached ribbon outside original groove
a machined groove. They also showed numerous ribbons (Puttick et al., 1989)
detaching from one side of the grooves only, leaving a been formed not in front of the grooves but on the sides of
striped structure or tearing a ragged foil alongside the the grooves as Puttick et al. comment on some of these
original groove or two filaments of the workmaterial peeling strips. In situ machining studies on glass conducted
off in opposite directions. Finally, they showed many inside an environmental scanning electron microscope at
strips in a state of incipient detachment with loops fixed at
Oklahoma State University indicated that many of these
both ends. They actually considered that these filaments filaments may have been formed at the sides of the
are generated in the grooves ahead d the tool. Based on a
grooves due to side flow and subsequent motion of the tool
critical re-examination of the photographs of these ribbons in the direction of cutting. The delamination and
on the machined grooves, it appears that they may have subsequent separation of the filaments led Puttick et al. to

57 1
conclude that the material (on the sides of the machined
grooves) is in a state of longitudinal compressive residual
stress and subjected to buckling. Puttick et at. proposed a
model to explain this effect. The mechanism of material
removal in micromachining of glass, according to Puttick
et at. is accomplished by at least two distinct processes,
namely, groove formation and subsequent swarf removal,
though their relative importance may vary between
materials. The fact that the two are not concurrent is
particularly noteworthy. It appears that the groove
formation is due to indentation and the swarf formation is
due to side flow during indentation and subsequent sliding.
Pyramidal indenters such as Vickers or Knoop indenters
are typically classified as sharp, whereas spherical
indenters are characterized as blunt in the indentation
hardness testing field. If we consider sliding of these
indenters over a surface to simulate machining, grinding,
or abrasion then all these shapes are in fact blunt. For
example, a Vickers indenter with 136 deg. included angle
would be equivalent to a high negative‘rake angle of -68
deg. A Knoop indenter with an included angle of 172.5
deg. in one direction would be equivalent to a negative
rake angle of -86.25 deg. and an included angle of 130
deg. in the other direction would be equivalent to a
negative rake angle of -65 deg. A conical indenter with an
included angle of 120 deg. would be equivalent to a
negative rake tool of -10 deg. Finally, a spherical indenter
would be equivalent to a varying high negative rake angle
from close to -90 deg. at the tool-workmaterial interface to
the angle subtended at the depth of cut line. Thus, both
radius and the depth of cut of the spherical indenter are
important in determining the tool/workpiece interface
geometry.
4. MATERIAL REMOVAL MECHANISMS IN BRllTLE
MATERIALS
Grinding of hard and brittle materials have been studied
extensively (see for example Inasaki, 1986, 1987; Malkin
and Hwang, 1996). Recently, Jahanmir et at. (1 992) made
a concise assessment of ceramic machining practice.
These and other relevant papers should be referred to for
further details.
It has been reported that crack-free hardness indentation
is possible on single crystals of Ge and Si, at extremely
low loads (-0.05 N). Gerk and Tabor (1978) attributed this
in part t0.a structural transformation could occur under the
conditions of indentation or machining. Ge has a diamond
cubic structure with a four-fold coordination. Under high
hydrostatic pressures, -1 15-120 kbars, the crystal
structure of Ge is found to change to a six-fold
coordination, and become metallic. Puttick et at. (1 994)
proposed a mechanism of material removal in Si involving
extrusion of heavily deformed material ahead of a large
radius tool and/or high negative rake tool. This could be
the case with Si which under high hydrostatic pressure
may undergo a metallic phase transformation and exhibit
plastic behavior.
The indentation hardness of Ge is -80 kbars, which is
comparable to the pressure at which the transition is seen
to occur. Si also shows similar behavior. The
semiconductor-metal transition in this case occurs at a
pressure of 190 kbar; the indentation hardness is -120
kbar. Thus both the transition pressure and hardness
value are again .comparable and are -50 % greater for Si
than for Ge. Gerk and Tabor proposed that the material
around the indenter would become sufficiently ductile (due
to its in situ metallic transformation) to sustain plastic flow.
In other words, both Ge and Si may exhibit metallic
behavior including extensive ability to deform plastically. If
this is the case then it is unlikely that Si or Ge would
exhibit similar ductile behavior if positive rake tools or
sharp tools (edge radius << than cut depth, as in
572
conventional machining) are used. This is because the
hydrostatic pressure may not reach such high levels
required with positive rake tools and even if they do, it
would be limited to extremely shallow depths. This is
perhaps the reason why large negative rake tools are used
in practice in machining Si. It may be possible to check
this concept with molecular dynamics simulation of Si with
an atomically sharp and a radiused tool for comparison.
4.1 Factors Affectina Material Removat in Finishina of
Brittle Materials
In finishing of brittle materials, the material response is an
important factor that affects the quality of the surface
generated. The material response in turn depends on the
magnitude and size of the stress field as well as the
response of the workmaterial-abrasive-environment
combination under the conditions of the process.
Yoshikawa (1967) conducted pioneering studies on the
brittle-ductile behavior of crystal surfaces in finishing. He
considered the size of the stress field in terms of “working
units” and classified it into four domains as shown in Fig.
39. Each domain is characterized by the defect structure
which either pre-exists or has been generated by the
processing method.
Domain
t.Crack Void
Grain boundary layer
Dislocotion
-I Micr0-cr ack
I t e r s t i t i a l atom
7Vacancy
Atom
I I I 1 I I I
In Domain I, the material removal is considered only on
the order of a few atoms or molecules. Material removal
on this order could not possibly occur by pure
mechanical action. Chemical action enhanced by
mechanical stress and temperature may play an
important role in the material removal process in this
domain.
In Domain 11, where no dislocations or cracks exist, a
crystal should behave mechanically as an ideal crystal.
Theoretical considerations indicate that under a given
load the crSistal surface in this domain should give rise
to the generation of dislocations prior to brittle fracture.
After the generation of dislocations, the crystal is
assumed to behave as in Domain Ill.
In Domain Ill, with only dislocations present, plastic
deformation occurs first, some amount of which will allow
cracks to nucleate in the deformation zone.
In Domain IV, the defects due to cracks are the
dominant factors in material removal.
Of course, this classification is for an ideal single phase
material with no pre-existing defect structure, such as
point defects, dislocations, grain boundaries, presence of
a second phase, microcracks, or voids. The presence of

these defects will alter the mode of material removal in all
the domains. However, it appears reasonable to assume
that at extremely fine depths of removal (angstrom range),
elastic removal of the material on the surface by erosion or
chemo-mechanical action may play a dominant role in
"gentle" grinding and polishing, followed by plastic
deformation/microfracture depending on the conditions of
polishing.
4.2. indentation and lndentation SIidina
- of Brittle Materials
Perhaps the most widespread use of indentation testing in
the context of brittle materials, in addition to hardness
measurements, is the evaluation of fracture toughness
(Lawn and Wilshaw, 1975). Palmquist (1962) recognized
that cracks generated during indentation must indicate a
measure of the toughness of a material. He, therefore,
suggested that the average length of the cracks
emanating from the corners of a Vickers indenter might be
a measure of relative.toughness. Thus one seeks to relate
the fracture resistance of a ,material to the scale of the
crack pattern (critical crack size). For example, in the
case of pyramidal indenters, cracks would form on the
surface and at the corners of the indenter.
In the simulation of nanometric cutting, grinding, or
polishing etc., it is important to focus attention on the early
stages of indentation under sliding conditions, namely, the
elastic-plastic deformation under the indenter prior to the
onset of median vents that can potentially lead to median
cracks once the indenter moves forward. Due to the
inclined loading of the surface, there is a strong tensile
component behind the point of contact [Fig. 401. It is this
tensile component that can lead to material removal by
brittle fracture.
t "YF
Constant
shear stress
I\ '\,+Tension
Constant
maximum
shear stress
\.
Fig. 40 Resultant stress field due to inclined loading of a
surface showing both compressive and tensile
stress components (after Suh, 1986)
It is implicit in all continuum-based plasticity models that
the deformation process during indentation is volume
conserving, a consequence of which is that a state of
residual stress must exist around the indentation site.
Closer examination of the deformation region beneath the
contact area reveals some important departures from the
idealized picture as outlined below (Marshall and Lawn,
1986).
First, the deformation process is by no means uniformly
distributed within the plastic zone, but is manifested (at
least in part) as an accumulation of discrete shear events.
These events are akin to the dislocation slip processes
which occur on preferential glide planes in softer materials
but differ in two important respects: 1) they occur at stress
levels close to the theoretical shear strength of the
structure in the more covalent materials, and 2) the shear
surfaces are not necessarily crystallographic but are
Keynote Papers
determined more by stress trajectory patterns. One
therefore, has to be extremely cautious before using
classical dislocation concepts to describe the flow
properties of solids with intrinsically rigid bonding. The
departure from the ideal behavior is apparent in certain
materials which show a greater tendency toward
deformation in hydrostatic stress than in shear. Fused
silica, for example, undergoes densification when
subjected to confining pressures (Ernsberger, 1968).
Compaction modes can accommodate the volume of the
impression with relatively little stress mismatch at the
deformation zone boundary. Also, many crystalline solids
undergo pressure induced phase transformations, which
may be either expansive or compactive.
The shear events are critically important in acting as
embryonic nuclei for crack formation. It has been
observed that most of the crack development occurs not
on loading but on unloading of the indenter, indicating that
it is the irreversible component of the contact stress field
which provides the dominant driving force for fracture. In
cutting, abrasion, polishing and other similar processes,
the workmaterial is indented by the tool and the tool leaves
the indent soon after, thereby offering a potential for the
crack formation where median vents may have been
formed a few instants earlier. Measurement of the elastic
recovery after indentation is of interest since it can shed
light on the partition of input deformation energy into
reversible and irreversible components and for the
characterization of the elastic-plastic field in the
development of a fracture mechanics formulation. The
parameter of interest here is the ratio of hardness to
elastic modulus, H/E (rigidity).
Brittleness should also be a measure of the competing
actions between deformation and fracture processes, as
manifested for instance, in the ductile-brittle transition.
The ratio of hardness to toughness, H/K,, can be used as
an appropriate index of brittleness. This way, the relative
susceptibilities to flow and fracture in terms of indentation
threshold conditions can be determined. Marshall et al.
(1982) have shown that the critical indentation size, a, is
proportional to (E/H)(K,/H)*. Physically, this represents
the scale of damage above which the mechanical
response is essentially fracture dominated and below
which it is deformation dominated (Marshall and Lawn
(1986). For an understanding of the mechanism of material
removal in brittle materials with fine abrasives generating
microcrack-free surfaces, we are interested in the physics
of the process at a scale below a., As pointed out earlier,
this is referred to by some researchers as "ductile" regime
grinding.
4.3 Material Removal Mechanisms in Brittle Materials
Jnvolvina Microfracture
Generally, during roughing and semi-finishing of nominally
brittle materials be it by machining, grinding, 'or polishing,
much of the material is removed by brittle fracture. This
enables high removal rates and results in an energetically
more efficient process, provided these defects do not
extend below the finished surface and that there is
sufficient material left for finishing them to the desired
form, size, accuracy, and finish. An appreciation of the
mechanism of material removal by this mode can be
obtained by comparing this process with the indentation-
sliding analysis conducted by Lawn et al. (1977, 1980).
Consider indentation of a brittle material. The various
stages of indentation (loading and unloading) are as
follows [Fig. 411: (a) The material under the indenter is
initially subjected to elastic deformation. As indentation
continues, the material below is subjected to a state of
high hydrostatic pressure and hence a small inelastic
(irreversible or plastic) deformation zone is produced; (b)
At some threshold, a deformation-induced flaw suddenly
573
develops into a median vent (which can subsequently
develop into a median crack during unloading), on a plane
of symmetry containing the contact axis at the plastic-
elastic boundary; (c) An increase in load causes further,
stable growth of the median vent; (d) On unloading, the
median vent begins to close (but not heal): (e) In the
course of indenter removal, sideways extending cracks,
termed, lateral vents, begin to initiate near the base of the
INCREASING LOAD REMOVAL OF LOAD
I t
d
0
Plostic
i etorma t lo< 1/ Sharp poinl
indentor
Load omlied
Unloading crack ,
b t
Small median Laleral I
C i
Growlh of Larger lateral
cracks leading
Fig. 41 Various stages of indentation showing plastic
deformation underneath the indenter, median
vent formation during loading, growth of the
median vent into a median crack, and formation of
lateral cracks ultimately reaching the surface
resulting in material removal during unloading
(after Lawn and Wilshaw, 1975)
plastic deformation zone below the contact and spread out
laterally on a plane closely parallel to the specimen
surface. This is the consequence of a residual tensile
stress field due to the presence of a zone of permanent
deformation constrained by the surrounding field of elastic
strain. Residual stresses are identified as the primary
driving force for lateral cracks. Immediate reloading of the
indenter leads to closure of the lateral vents, and
reopening of the median vents; (f) Upon complete removal,
the lateral vents continue their extension towards the
specimen surface and may eventually lead to the removal
of material by chipping. The driving force for crack
formation is provided generally by the residual stress field
which results from a mismatch in the elastic-plastic
deformation (Marshall et al., 1982).
4.4 Material Removal Mechanisms in Brittle Materials
without Microfracture
It is well known from plasticity theory that while the yield
strength of a material is determined by the deviatoric
stress state, the magnitude of the hydrostatic stress
determines the extent of plastic deformation prior to
fracture. In other words, hydrostatic stress determines the
strain at fracture. Bridgman (1953) showed that many
brittle materials are capable of ductile behavior but only
under high hydrostatic pressures. Thus, a high value of
574
the hydrostatic pressure is a prerequisite for plastic flow to
occur in brittle materials at room temperature. Such
conditions generally exist at light loads under the indenter
in indentation testing. In the indentation problems an
assumption is made that spherical symmetry of the
"bottom half" of a spherical cavity is maintained in the
deformation zone. Immediately below the indenter the
material is considered to behave as a radially expanding
"core", exerting a uniform hydrostatic pressure on its
surroundings; encasing the core is an ideally "plastic
region" within which flow occurs according to some yield
criterion: beyond the plastic region lies the elastic matrix
[Fig. 421 (Johnson, 1970).
1
Fig. 42 Model for elastic-plastic indention of brittle
materials. Dark region denotes "hydrostatic
core", shaded region denotes "plastic zone", and
surrounding region denotes " elastic matrix "
(after Johnson, 1970)
Lawn et al. (1980) developed an elastic/plastic indentation
model for the evolution of medianhadial crack system. It is
based on a model in which the elastic/plastic field beneath
the indenter is resolved into elastic and residual
components. The elastic component being reversible,
assumes a secondary role in the fracture process;
although it does enhance the downward (median)
extension during the loading half-cycle, it, suppresses
surface (radial) extension to the extent that significant
growth continues during unloading. The residual
component (i.e., the plastic) provides the primary driving
force for the crack configuration in the final stages of
evolution, where the crack tends to a near-half penny
geometry. The origin of the irreversible field thus lies in the
accommodation of an expanding plastic hardness
impression by the surrounding elastic matrix. The
equilibrium crack growth is found to relate to the hardness-
to-modulus ratio as well as toughness. Lawn et al.
developed a relation between the relative plastic zone
radius to the hardness-to-modulus ratio and the indenter
angle.
Lawn and Evans (1977) also developed a model for the
crack initiation in the elastic/plastic indentation fields.
Using an approximation for the tensile stress distribution in
the elastic/plastic field, in conjunction with the principle of
geometric similarity, Lawn and Evans determined the
critical conditions for the growth of penny-like, "median
cracks" from subsurface flaws. This analysis provides a
functional relationship between the size of the critical flaw
and the indentation load necessary to make this flaw.
extend. The parameters of the workmaterial involved in the
analysis include the intrinsic deformation/fracture
parameters, hardness, and fracture toughness. Such a
relationship would establish the upper bound in the case of
microcrack free finishing and establish the process
parameters for finishing different workmaterials. Hagan

(1979) proposed an alternate mechanism for subsurface
flaws that does not depend on the existence of pre-
existing subsurface cracks. Instead, these flaws are
generated during the indentation process due to the
interaction of dislocations in the case of crystalline
materials, or inhomogeneous deformation in the case of
some glassy materials which play an important role in the
nucleation and growth of the subsurface cracks required
for the model of Lawn and Evans.
In ultraprecision machining, be it single point diamond
turning, grinding, or polishing, it is important to identify the
conditions that involve the full development of plasticity
underneath the tool or abrasive grain (indenter) prior to the
ensuing fracture by median cracks. Although load (critical
load) or depth of indentation or cut (critical depth of cut),
can be easily measured, it is the state of stress which
determines whether fracture will occur. The geometry of
the tool or abrasive grain, coupled with the depth of cut
and workpiece material properties determine the resulting
force system which acts at the surface and which
determines whether fracture will occur or not.
4.5 Material Removal Mechanisms from Conventional
Machinina to Indentation
In ultraprecision machining or grinding of brittle materials
at small depths of cut, most tools used have large edge
radii (relative to the depth of the layer removed) and large
negative rake angles. For example, in the single point
machining of Si, negative rake angles as high as -30' or
higher are typically used. It is the severe negative rake
angle which provides the necessary hydrostatic pressures
for enabling plastic deformation of the workmaterial
beneath the tool radius to occur. One should not liken this
type of plastic deformation with the concentrated plastic
deformation which occurs ahead of the tool in the case of
conventional machining of metals with positive rake tools.
Plastic deformation in the case of brittle materials with
large negative rake tools is energy intensive and
inefficient compared to machining metals with positive
rake tools. Figure 43 is a molecular dynamics simulation
Fig. 43 Molecular dynamics simulation of cutting with a
large negative rake angle tool (Komanduri et al.,
1997)
with a large negative rake tool (after Komanduri et al.,
1997). Examine the highly inefficient material removal.
Also, note the significant deformation underneath the tool
due to the indentation effect. If one were to use positive
rake tools or very sharp tools, i.e., an edge radius small
compared to the depth of cut, one may not induce
adequate hydrostatic pressure to cause any plastic
Keynote Papers
deformation and elastic deformation followed by the
formation of median cracks will result. This can result in
brittle microfracture. As it is not possible to generate an
atomically sharp tool, it should be difficult to test this
hypothesis. However, it may be possible to simulate by
MD (molecular dynamics) simulation of nanometric cutting
with tools with finite edge radii and tools which are
atomically sharp.
For the case of conventional machining, the tool geometry
is such that the rake angle is generally positive or close to
zero. With a positive rake angle tool, the cutting force is
generally about twice the thrust force. The deformation
ahead of the tool is either in a concentrated shear plane or
in a narrow zone [Fig. 44(a)]. In grinding, although one
cannot identify a definite rake angle as it is unknown and
varies continuously due to wear and self-sharpening
action (friability), it is generally agreed that the tool
presents a large negative rake. Also, the cutting force is
about half that of the thrust force [Fig. 44(b)]. Large
negative rake angle cutting tools have been used to
simulate grinding (Komanduri, 1971). In ultraprecision
machining at depths of cut smaller than the tool edge
radius, the tool presents a large negative rake angle to the
workmaterial and the radius of the tool edge acts as an
indenter [Fig. 44(c)]. The final progression in this series
which represents a rotation of the resultant force vector
downward toward the workpiece surface (Lucca et al.,
1991) is indentation sliding of a blunt indenter across the
workpiece surface. This situation could represent the
case when a tool is rigidly supported and cuts the
workpiece under a stress state such that no median vents
are generated but material underneath the tool is
plastically deformed due to large hydrostatic pressures
[Fig. 44(d)]. It is apparent when one considers the force
Fig. 44 The progression of tool/workpiece interactions
which result in a rotation of the force system
downward toward the workpiece surface: (a)
conventional cutting, (b) grinding, (c)
ultraprecision machining at small depths of cut,
and (d) indentation sliding
system which results over a range of toollworkpiece
interactions, there is an obvious progression from
conventional machining with positive rake tools, to
grinding, to ultraprecision machining, and finally to
indentation sliding. It appears that an indentation sliding
model would be appropriate when considering machining
575
brittle materials with tools of large edge radius relative to
the cut depth or large negative rake tools.
5. CHEMO-MECHANICAL POLISHING
It is interesting to note that there are several terms used to
describe the process of chemo-mechanicalpolishing. The
first one is called chemo-mechanical or mechano-
chemical polishing (Vora et al., 1982), the second one is
called tribo-chemical polishing (Heinicke, 1984; Fischer,
1988), and the third one is called chemical-mechanical
polishing (Nanz and Camilletti, 1995). These distinctions
are made by researchers as well as practitioners mainly to
place proper emphasis on a particular factor that plays a
major role in the material removal process. For example, in
chemo-mechanical polishing (CMP), the driving factor for
material removal is chemical action which is followed by
mechanical action for the removal of the reaction product.
In contrast, in mechano-chemical polishing, the emphasis
is on mechanical followed by chemical action. Tribo-
chemical polishing involves the formation of a reaction
layer due to friction without the need for specific abrasive
particles. In chemical-mechanical polishing, a specific
fluid is introduced to form a reaction with the workmaterial
(similar to tribo-chemical) and this reaction product is
subsequently removed by the mechanical action of the
abrasive. CMP is the most common operation used in the
finishing of Si wafers in semiconductor industry. The
difference between chemo-mechanical and chemical-
mechanical polishing is that in the former, a reaction
product is formed between the abrasive and the
workmaterial under the conditions of polishing and the
reaction product is subsequently removed by the
mechanical action of the abrasive particles. These
particles can be softer than the workmaterial, for the
material removal is not by abrasive action. In chemical-
mechanical polishing, the reaction is between the fluid and
the workmaterial and the abrasive removes the reaction
product by mechanical action. However, some
researchers use these terms interchangeably, and
therefore care should be exercised in identifying the exact
process.
The chemo-mechanical process initiates a chemical
reaction between a given abrasive and the workmaterial.
The environment used can facilitate the chemical action.
The reaction product thus formed is generally brittle and
subsequently removed by the abrasive action as wear
debris or polishing debris. This results in the production of
a smooth surface. However, the presence of defects,
such as pits, or grooves from an earlier semi-finishing
operation can still exist in the final operation unless all the
defects are removed by this operation. Alternately, these
defects should be gradually minimized as one progresses
from roughing, to semi-finishing, to finishing. Hence, the
history of polishing is a very important consideration in the
final finishing of brittle materials.
Chemo-mechanical action depends on the availability for a
short duration of a certain threshold pressure and
temperature at the contact zone to initiate a chemical
reaction by the interaction of a suitable abrasive,
workmaterial, and environment [Fig. 451 (Yasunaga et al.,
1978). The exact values depend on the type of reaction
product that can form. The process produces a weaker
reaction product compared to either the abrasive or the
workmaterial. In this way, a higher removal rate can be
accomplished without causing damage due to brittle
fracture to the brittle workmaterial. Also, abrasives softer
than the workmaterial can be used in cases where a
reaction product can be formed between the abrasive and
the workmaterial since hardness is of no consequence but
material removal depends on chemical affinity.
Consequently, no scratching or groove formation is
expected with these abrasives leading to a very smooth
576
surface finish. Hence, for finishing of brittle materials,
chemo-mechanical action is an attractive alternate
material removal mechanism especially during the final
stages of polishing.
Mechano-chemical polishing was first demonstrated by
Yasunaga et al. (1974, 1977, 1978, 1979) for polishing of
single crystals of sapphire, Si, and quartz using soft
abrasives. They used SiO2 for polishing sapphire; BaCO3
Ce02, and CaC03 for polishing Si single crystals; and
Fe304 and MgO for polishing quartz crystals, and
demonstrated that darnage-free, smooth surfaces can be
generated by this technique. Chemo-mechanical action
was used to advantage in polishing without causing any
detrimental effects to a range of workmaterials including
sapphire, Si, Sic, quartz, gallium arsenide, and various
ferrites (Namba and Tsuwa, 1977, 1978; Karaki-Doy, 1993;
Kashiwagura et al., 1983; Kawata and Tani, 1993; Kikuchi
-
et at., 1990, 1992).
Direction of Movement
Hard Workpiece
Fig. 45 Schematic of the chemo-mechanical action
between abrasive, workmaterial, and the
environment (Yasunaga et al, 1978)
Chemo-mechanical polishing is expected to overcome
many problems of surface damage associated with hard
abrasives, including pitting due to brittle fracture,
dislodgment of grains, scratching due to abrasion, etc.,
resulting in smooth, damage-free surfaces. However, if
chemo-mechanical reaction products remain on the
surface or diffuse into the workmaterial, they may affect
the performance and reliability of the product in
subsequent use. It is therefore necessary to ensure that
chemo-mechanical action generates the reaction product
but this reaction product is completely removed from the
workmaterial, i.e., at the surface as well as the
subsurface. In the following, two examples of chemo-
mechanical polishing, namely, finishing of Si3N4 and Si
wafers are given.
5.1 Chemo-Mechanical Polishina ICMPI of Si3N4
Chromium oxide (Cr 0 3 )is found to be an excellent
abrasive for polishing of Si3N4. As the hardness of Cr203
is nearly the same as that of Si3N4, it appears that the
mechanism of material removal would favor chemical
action over mechanical abrasion.
Vora et al. (1982, 1983) and Uematsu et at. (1993)
demonstrated the feasibility of polishing Si3N4 to a fine
finish by mechano-chemical polishing with iron oxides
(both Fe2O3 and FegO4). A chemical reaction between
iron oxide and Si3N4 at the areas of contact was assumed
which is promoted by a local mechanical effect associated
with nonequilibrium conditions, such as high temperature
and high pressure, although the nature of this reaction was
not investigated. Suga et al. (1989) also investigated
polishing of Si3N4 using several abrasives including
CaC03, MgO, Si02, Fe2O3, Fe304, and Cr2O3 and
concluded that Cr2O3 is a more suitable abrasive for the
mechano-chemical polishing of Si3N4. In a later
investigation, Kikuchi et at. (1990,1992) found Cr2O3 to be
a suitable abrasive for mechano-chemicalpolishing of Sic
as well. They proposed that Cr2O3 may be operating

catalytically thereby enhancing the ,surface oxidation
during polishing. It may also lower the activation energy
and/or the starting temperature for the reaction. The tribo-
chemical form of wear mentioned earlier is interpreted as
due the removal of material, molecule by molecule, instead
of the classical removal of entire particles by brittle
fracture (Fischer, 1988).
Singhal (1975) studied the effect of water vapor on the
oxidation of hot pressed Si3Nq and interpreted its
formation as due to the following reactions:
Si3N4 + 6 H20 + 3 SiO2 + 4 NH3
The dissolution of SiOp in water is given by :
Si02 + 2 H20 + Si(OH)4
Kanno et al. (1983) found the formation of NH3 in the milling
of SigN4 powder in water confirming Singhal's
interpretation. Fischer and Tomizawa (1 985) and
Tomizawa and Fischer (1 986) studied wear of Si3N4 in
water and concluded that wear is due to tribo-chemical
dissolution of the material via the formation of Si02 and the
dissolution of the Si02 in water to form silicic acid, i.e., no
solid wear particles are generated. Thus Si02 is dissolved
in water more rapidly than amorphous Si02 particles.
Fischer and Tomizawa also found the debris to be white in
dry sliding of Si3N4 (NBD 200, which is black in color)
indicating the formation of SiOp particles but found no wear
debris in sliding in water. However, when the water was
evaporated, particles of Si02 were found to be
precipitated.
Sugita et al. (1984) studied the mechanism of material
. removal during rubbing of Si3N4 in water. They found the
wear debris in water to be amorphous. They interpreted
the results as due to oxidation of Si3N4 which
subsequently changed to amorphous hydrate SiO2:xHpO
during rubbing in water. The amorphous hydrate was
removed from the rubbing interface as a result of sliding.
Akazawa and Kato (1988) and Akazawa et al. (1986)
investigated the wear properties of Si3N4 in rolling contact
and rolling-sliding contact. They identified one type of wear
debris to be glassy film with Si02 structure. Similar results
were reported by Jahanmir and Fischer (1987), Kapsa and
Kanemura (1988), Gee (1992), Fischer and Mullins (1993),
Jahanmir (1993).
Although it is widely believed that advanced ceramics can
be finished by chemo-mechanical action by proper choice
of the abrasive and the environment, the exact mechanism
of this action has not been well established. Some
plausible explanations have been offered including the
formation of silicates, spinals, oxidation of Si3N 4
with/without catalytic action of Cr2O3, and less chemical
stability of CqO3 compared to Al2O3.
The chemo-mechanical polishing of Si3N4 workmaterial
with Cr20g abrasive in MFP was studied both analytically
and experimentally (Bhagavatula and Komanduri, 1996).
The flash temperatures generated at the real areas of
contact during polishing can be relatively high ( -1 OOO°C)
(Hou and Komanduri, 1997 a,b). Also, the flash times
range from 1-15 p e c depending on the flash temperature
under consideration, which in turn depends on the contact
pressure, the sliding velocity, and the friction coefficient
between the abrasive and the workmaterial. The longer
flash times are associated with the lower temperatures and
vice versa. The reactions involved in chemo-mechanical
polishing are essentially the adsorption of hydroxyl ions to
the surface of Si3N4 and oxidation of the surface, which
forms initially a silica layer which further reacts with water
to form a hydrated silica layer. When the nitrogen ions pair
up with hydrogen ions present in water, NH3 results.
Generation of NH3 gas has been detected by previous
investigators (Kanno et al., 1983) thus lending support to
Keynote Papers
this hypothesis. Thus, it is reasonable to assume that the
oxidation of Si3N4 in water proceeds by the following
reaction:
Si3N4 + 6 H20 + 3 9 0 2 + 4 NH3
The action of Cr in Cr20g in its 3+ state alone on Si3N4 can
be represented by
Si3N4 + 2 Cr203 + 3 Si02 + 4 CrN
The combined action of CrpOg and water can be
represented by
Si3N4 + Cr203 + 3 H20 + 3 Si02 + 2 CrN + 2 NH3
Further, Cr203, with Cr in its 2' oxidation state can react
with silica in the presence of water to form chromium
silicate (Earnshaw and Harrington, 1973).
2 Cr203 + 2Si02 + 2Cr2 SiO4 + 0 2
This reaction also evolves oxygen, partly replenishing the
amount that was spent during the previous reactions.
Therefore, oxygen is made continuously available even in
water for oxidation. The reaction of water with silica is
represented by the following:
3 Si02+ 6 H20 + 3 Si(OH)4
This reaction produces a hydrated layer of silica forming
silicic acid (H2Si03) as a reaction product.
Bhagavatula and Komanduri (1996) investigated the
mechanism of chemo-mechanical polishing of Si3N4
workmaterial with Cr 0, abrasive in MFP and MAF. They
showed evidence c$ the chemo-mechanical action by
examining the wear debris generated in polishing using an
SEM with an X-ray microanalyzer and a low angle X-ray
diffraction apparatus. The analysis showed conclusively
that Cr2O3 does participate in the chemo-mechanical
polishing of Si3N4 forming chromium nitride and chromium
silicate, thus establishing the role of Cr2O3 not as a mere
catalyst as some researchers suggested but of one
actively taking part in the chemical reactions during
chemo-mechanical polishing [Figs. 46(a) and (b)]. They
proposed a new model of the chemo-mechanicalpolishing
of Si3N4 workmaterial in air and in water environments with
Cr2O3 abrasive. This model is based on the formation of
such reaction products as chromium silicate and chromium
nitrate in addition to the formation of silica layer on Si3N4
surface as well as the gaseous reaction products, such as
NH, (in water) and nitrogen (in air).
Chemo-mechanical action between a given workmaterial,
the abrasive, and the environment can be studied using
thermodynamic analysis (Gibbs free energy of formation).
For example, consider the case of polishing of Si3N4 with
Ce02 abrasive in water, i.e., the case of magnetic float
polishing. Figure 47(a) shows the variation of the mole
fraction of various species with temperature at equilibrium
based on the thermodynamic calculations of the chemical
reaction system consisting of 1 rnol of Si3N4 and 1 rnol of
H20. It can be seen that at low temperatures < 3OO0C,
ammonia [NH3(g)] formation is promoted while at higher
temperatures, H2(g) and N2(g) gases are evolved. The
3 0 2 mole fraction as well as Si3N4 mole fraction, however,
seems to be somewhat independent of temperature
indicating very little effect, if any, of temperature on the
removal rate under these conditions. When one considers
the variation of various chemical species with temperature
for 1 mol of Si3N4 and increasing amounts of H20, one
sees that with an increase in the mole fraction of H20,'the
amount of Si02 increases and the amount of Si3N4
correspondingly decreases both accounting for an
increase in the material removal due to chemo-mechanical
polishing. This shows the beneficial role of H20 in CMP. In
577
a similar manner, NH3(g), H2(g), and N2(g) gases also
increase with increases in the mole content of H20.
Fig. 46(a) SEM micrograph and EDX mapping of Cr and Si
of a typical particle from the wear debris in
polishing Si3N4 with Cr2O3 abrasive showing
simultaneous existence of Cr and Si
(suggesting the formation of CrSi). Compare
the morphology of the wear particle with that of
the Cr and Si mapping (after Bhagavatula and
Komanduri, 1996)
Fig. 46(b) X-Ray diffraction analysis of the wear debris
showing the presence of Cr2SiO4 and CrN (after
Bhagavatula and Komanduri, 1996)
Figure 47(b) shows the variation of the reaction products
with temperature at equilibrium based on the
thermodynamic calculations of the chemical reaction
system consisting of 1 mot of Si3N4 and 1 mol of Ce02.
CeO2 is not stable and with increase in temperature
converts to CeOl.83, CeOl.72, and Ce2O3. It can be seen
that the 3 0 2 mole fraction is independent of temperature
up to -3OOOC and increases gradually up to 1000°C.
However, as the mole fraction of Ce02 is increased, the
amount of SiOp increases and the amount of SigN4
decreases both accounting for the increase in the material
removal rate due to CMP (Jiang and Komanduri, 1997b).
5.2 Chemical Mechanical Polishina (CMP) of Si Wafers
CMP of Si wafers is a basic processing technology for the
production of flat, defect free, highly reflective surfaces.
Such wafers are vital as the sfarting material for
subsequent processing steps in integrated circuit
fabrication. As the device dimensions are scaled down to
the lower end of the submicron (0.1-0.5 p n ) range, local
and global planarization are assuming increasing
importance for device fabrication as well as for
interconnects (Warnock, 1991, 1994). At the same time,
the size of the incoming wafers to be finished are
578
mol
1.0
0.9
oa
0.7
0.6
05
0.4
03
02
0.1
0.0 T
0 400 600 800 loo0 c
Fig. 47(a) Variation of the species mole fraction with
temperature at equilibrium based on the
chemical reaction system of 1 rnol Si3N4 and 1
mol H20 in the chemo-mechanical polishing of
Si3N4 with Ce02, (after Jiang and Komanduri,
1997b)
0 200 400 600 800 I000 c
Fig. 47(b) Variation of reaction products with temperature
of the chemical system consisting of 1 mol
Si3N4 and 1 mol Ce02 (after Jiang aiid
Komanduri, 1997b)
increasing, close to 300 mm in diameter, to reduce
manufacturing costs. If planarization is not performed
metal circuit lines would become coated with an intermetal
insulative layer such that the upper surface of the
interlayer dielectric comes to exhibit non-planar "steps".
These step variations are undesirable for they limit the
depth of focus of the high numerical aperture lenses used
by optical projection lithography equipment, and also
because metal can thin over steep topographies. CMP has
been widely accepted as the manufacturing process of
choice that can achieve both local and global planarization
for ultra-large scale integrated circuits (ULSI). However,
for CMP application to modern ULSl fabricatipn,
cleanliness, planarity, and uniformity (the so-called CPU
factors) are essential, and the technical requirements for
these CPU factors are becoming increasingly stringent
with time (Hayashi et aL1996).
In CMP planarization,the wafers are pressed face down by
carriers and rotated against a polishing pad covered with a
layer of slurry. A schematic representation of the CMP
process is illustrated in Fig. 48. The pad is used to hold
the slurry particles, transmit load to the particlelwafer
interface, and conform precisely to the wafer being
polished. The most widely used technique for polishing is
based on the use of an aqueous colloidal silica suspension
(5-7 nm diameter) with a pH between 8.5-1 1. The polishing

process proceeds by a combination of initial chemical
reaction of Si with the aqueous solution to form a thin silica
layer and the subsequent mechanical removal of the
reaction products by the polishing slurry.
Waferchuck & d
Fig. 48 Schematic of a CMP process for planarization
(Hayashi et a1.,1996)
The role of hydrodynamic lubrication of the fluid film is
distribution of the slurry, lowering of the friction force, and
removal of the eroded material in the form of chips, and
removal of the heat generated by the CMP process. For
abrasion process to take place, the slurry particles should
be in direct contact with both the pad and the surface of
the wafer during the polishing process. Since the hardness
is a determining factor in how deeply a particle (slurry
carried particles in the case of CMP) penetrates into a
surface, the harder the deformed material, the less it is
penetrated under a given load. Thus, the depth of
penetration into the pad surface (made of polyurethane
which is softer) has to be greater than into the surface of
the wafer. During CMP, the slurry particles move between
the wafer and the pad and the kinetic energy of the
particles will erode the surface of the wafer.
CMP of Si wafers involves simultaneous and intimate
contact between the wafer surface and a pad charged with
a layer of colloidal silica slurry. The relative motion
between the wafer and the pad, combined with applied
pressure and chemical activity of the slurry results in the
erosion of wafer surface. Specifically, when a particle
embedded in the fibers of the pad has sufficient strength
and energy to overcome the fracture toughness of
polished material, the particles strike the wafer surface
resulting in the removal of material from the surface. It
thus, appears as an erosion process although most
modeling considers it as an abrasion process and uses
Preston's equation.
The variables in the CMP of the Si wafer polishing include
the following: size and number of wafers per load,
accuracy and finish required, pad speed, carrier speed,
pad (material, geometry, and stiffness), abrasive content
in the slurry, slurry flow rate, distance between the pad
and the wafer, size and material of the abrasive, pressure
of the slurry fluid, viscosity of the slurry, pad conditioning
method, density of the slurry fluid, etc.
Typical polishing conditions used for CMP of Si wafers are
given below:
Polishing media fumed silica (5-7 nm) suspension dis-
persed in aqueous potassium hydroxide
(pH - 10.3) distributed across the pad
Pad speed 20 rpm
Wafer size 200 mm
Pressure 27-76 kPa
Carrier speed 10-50 rpm
Pad material Polyurethane impregnated polyester
Particle content 20 %m fumed silica (5-7 nm) in the slurry
Slurry flow rate 50 mVmin
Viscosity 109~
Keynote Papers
Pad conditioning Rotating pad conditioner to clean the
pad of old slurry and lift the pad fibers for
further processing
Various models dealing with the mechanism of material
removal in CMP will be briefly discussed below.
As pointed out earlier, CMP relies on both chemical action
of the slurry with the surface of the wafer and the
subsequent mechanical action by the abrasives for
removing the reacted species. From a thermodynamic
point of 'view, Si is unstable in aqueous solutions and
should dissolve in water with the formation of H2 gas.
However, the formation of a coherent oxide layer prevents
further dissolution. For that reason, etching of Si can only
occur in a solution in which the oxidation products are
soluble. This requires the use of an acidic solution
containing HF, in which SiF:-,can be formed or relatively
strong alkaline media in which the oxidized Si forms
soluble silicates. In the pH range from 7-11, the etch rate
is negligible. However, when the oxide layer is removed
mechanically, Si can dissolve in pure water similar to the
etching process in KOH.
Both in pure water and in concentrated KOH solution, the
following overall reaction takes place for the chemical
dissolution reaction
Si + 2H20 + 20H + H2Si0z-+ H2
or, the water molecules hydrolyze the sizable bridging
bonds according to the rate dependent reaction.(H-O-H) +
(-Si-0-Si-) + (-Si-OH,HO-Si-) to form terminal hydroxyl
groups. Below a pH of 11, this reaction becomes
extremely slow as silicates form coherent oxide layers at
lower pH. When the silicates are mechanically removed (in
this case by the flow of polishing slurry containing fine
fumed silica particles), a better interaction between Si and
water is possible and the concentration of dissolution
intermediates becomes higher. Even though the chemical
action is very crucial, it is paradoxical that it is hardly
considered in the analytical modeling. In fact, the overall
mechanism of polishing is still not well understood and our
understanding of this technology rests essentially at an
empirical level.
Venkatesh et al., (1995) presented various aspects of
polishing and ultraprecision machining of semiconductor
materials including possible material removal
mechanisms. Nanz and Camilletti (1995) reviewed various
models developed for chemical-mechanical polishing
(CMP) of Si wafers. In brief, these models can be
classified into the following:
1. Abrasion model (Preston, 1927)
2. Chemical model (Cook, 1990)
3. Fluid mechanics/erosion model (Runnels, 1994)
4. Tribological model ( Runnels and Eyman, 1994)
The abrasion model originally developed for glass polishing
(Preston, 1927) is commonly applied to CMP of Si wafer
polishing, perhaps for lack of alternate models. It is an
empirical equation whose validity for CMP has never been
challenged (Tseng and Wang, 1997). The equation, now
commonly termed Preston's equation, incorporates the
applied pressure (hardness of the Si wafer) and the
relative sliding velocity of the wafer with respect to the pad
as the main contributors to the removal rate, RR, as
RR=Cp P V
The experimentally obtained RR is used in the equation to
obtain the value of the constant, Cp which is termed
Preston's constant. The work done in a given time t, is
given by pAPVt, where p is the coefficient of friction and A
is the area of contact. Here, pP is the specific traction
force of the pad on the wafer. It may be noted that this
model does not take into account the CMP process at all;
579
consequently it has no physical basis. Even though this bonding occurs between the hydroxylated molecules on
model is used extensively, the results obtained using this the two surfaces and these are carried away by the moving
model are often not satisfactory and the differences are silica particles. A fresh surface layer is then exposed
hidden in the proportionality constant. Also, the effect of ready for the reaction to proceed. The reaction rate
various processes involved in CMP are also shrouded in increases with temperature and the pH of the solution. The
this constant. pH is adjusted between 9.7 and 11 to obtain acceptable
removal rates. Chemical activity may be affected by the
If abrasion is the dominant material removal mechanism
(which may not be the case in CMP), the work done in
removing material is proportional to the tangential force
which results from the loaded pad on the wafer and the
sliding velocity. More precisely, it is the integral of the
P
products of the tangential forces exerted by the individual
abrasive particles on the wafer and the corresponding
sliding velocities of the particles. Since the tangential
force is typically not determined but the normal load is
known, Preston's equation incorporates the normal force
with the ratio of the tangential force to the normal force
assumed to be a constant and included in the proportional
constant. For simple sliding, one can take the coefficient
of friction p = F/N where F is the tangential force and N is
the normal force. In contrast, in cutting and abrasion,
Coulomb's conditions do not exist and the F/N ratio is
expected to change with the rake angle and the
processing conditions.
Runnels and Eyman (1994) took into account the normal
and shear stresses that are involved in polishing, and
modified Preston's equation as RR = C,,o?.tV. The normal
stress, on originates from the pressure imposed by the
polishing arm while the shear stress, z arises from the
slurry flow across the wafer surface resulting from the
relative velocity between the rotating pad and the carrier I
fluid.
Cook (1990) perceived the removal process as a traveling
indenter plowing across the wafer surface. The abrasive
particles transported in by the turbulent slurry encroach
upon and indent into the wafer surface. The indentation
process is considered Hertzian and the resulting stresses
are calculated from the theory of elasticity. A question
arises as to whether the material removal from the Si wafer
by the silica is by abrasion or by erosion. As the polishing
slurry flows, it appears that the most probable mechanism
is erosion and not abrasion. If this is the case, no Fig. 49(a) Mechanics of slurry particle/wafer surface
mechanism incorporating erosion has been developed for contact (top) and schematic representation of
CMP of Si wafer polishing. the contact between slurry particle of radius R
Liu et at. (1996) developed a model based on statistical and wafer surface from the engagement point
methods and elasticity theory to determine the removal (point c) to some depth of penetration (bottom)
rate in CMP. Figure 49(a) is a schematic of the slurry (after Liu et al., 1996)
particle/wafer surface contact and the contact between a hydrostatic pressure which in turn is related to the applied
slurry particle and the wafer surface. This model, similar to load and velocity. According to Cook (1990), the rate of
Preston's model takes into account the applied pressure surface removal during polishing is determined by the
and the relative velocity between the pad and the wafer on following: 1) the rate of molecular water diffusion into the
the removal rate. In addition, it is capable of delineating glass or silica surface in the case of Si wafer, 2)
the role of mechanical properties of the slurry particles and subsequent glass or silica dissolution under the load
the film to be polished. They found the removal rate to be imposed by the polishing particles, 3) the adsorption rate
dependent on the elastic moduli of slurry particles and the of dissolution products onto the surface of the polishing
polished film. grain, 4) the rate of silica deposition back onto the glass or
Cook (1990) considered the chemical processes involved silica surface, and 5) the aqueous corrosion rate between
in the polishing of glass and extended it to the polishing of particle impacts. Cook developed simple equations for the
Si wafers. He considered the central chemical process approximate calculation of the depth to which water
which occurs by the interaction of the glass surface or diffuses below a glass surface during the polishing of silica
silica layer in the case of Si wafer and the polishing particle by silica particles as well as Ce02 particles to compare the
with water and other chemicals in solution. He thus number of collisions needed to remove one molecule of
considered the aqueous corrosion chemistry involved in SiO, during polishing. These equations can be
the process. For example, the water in the slurry incorporated in the development of a comprehensive
chemically reacts with the Si02, resulting in model of CMP.

i Si-0-Si I + HO
, + 2 E Si - OH
The diffusivity of water in SiO, being low, only the
molecules on the surface of the wafer and the silica
particles in the slurry participate in the reaction. Hydrogen
The abrasion model (Preston equation), when applicable,
may be applied only to the cases where the pad (actually
the abrasive in the pad) is in direct contact with the wafer.
In an attempt to link the chemical effects modeled on the
particle scale to process conditions modeled at the wafer

580

scale, Runnels and Rentein (1993) and Runnels (1994)
approached the removal mechanism in CMP from the
principles of fluid mechanics based on two-dimensional
Navier-Stokes equation, i.e., flow of the polishing slurry at
the velocity of the fluid between the pad and the wafer. The
modeling geometry for the fluid mechanics/erosion
process and details of the erosion feature are given in Fig.
49(b). The hydrodynamic layer thickness, h is assumed to
be between 10-50 pm as found from the wafer-scale
modeling studies (Runnels and Eyman, 1994). The
stresses induced by the flowing slurry on feature surfaces
were computed and used in the erosion models. However,
they did not relate the fluid motion (silica particle impact) to
actual material breaking mechanisms from the principles of
solid mechanics. As the fluid layer thickness increases
with increase in velocity, the removal rate decreases
because the contact area is reduced and the modified
surface is not removed efficiently. Also, at very high
pressures, the removal rate may actually decrease
because of inadequate supply of slurry reaching the wafer
surface.
Union of Rd
r I 1 a gimble mechanism which allows the carrier to adjust
inlet
U2’0
Fig. 49(b) The modeling geometry for the fluid
mechanics/erosion process (top) and details of
the erosion feature(bottom) (after Runnels,
1994)
At high velocities the wafer may glide on the pad at an
angle, i.e., inclined pad, with a thin layer of slurry
separating the two solid surfaces. In this case, the load is
entirely supported by the fluid pressure built in this
converging fluid film. A tribological model, using
hydrodynamic lubrication of bearings has been developed
by Runnels (1994) which assumes that the pad and the
wafer are planar surfaces in ielative motion, separated by
a wedge shaped film of slurry, which is an incompressible
viscous fluid under laminar flow. The thickness of the fluid
layer, h, for a plane slider bearing is given by
h = Kh (pV/PA)’”
where p is the kinematic viscosity of the slurry, and Kh is a
constant. The fluid film thickness where polishing occurs
is estimated to be in the range of 60-100 pm.
The total pressure on the wafer surface is assumed to be
the sum of the asperity contact pressure and the
hydrodynamic pressure. When the wafer/pad separation
distance is large (low normal load, high velocity), the
pressure is mostly hydrodynamic and the rate is very low
(hydroplaning). At intermediate separation distance
Keynote Papers
(increased normal load, decreased velocity) partial
asperity contact occurs. At minimal separation distance,
rate is high as asperity contact dominates.
Solid-solid interface involving a fluid (polishing slurry)
depends on:
1. Direct contact where the load between the two
surfaces is supported almost entirely through solid
structure
2. Semi-direct contact where there remains some solid
contact but the fluid between the surfaces also
partially supports the load
3. Hydrodynamic lubrication where the load is supported
only by a fluid layer, and thus the nature of stress
transferal is markedly different than that of solid
contact.
Figure 49(c) is a schematic representation of the
tribological modeling of CMP after Runnels and Eyman
(1994). This model takes into account the curvature of the
wafer as well as its attack angle. The converging flow of
the slurry under the wafer supports the load of the wafer
similar to that in a bearing. The wafer carrier is mounted on
during polishing to prevent the wafer sagging on the pad.
The model focuses on the flow of the slurry in the gap.
Fig. 49(c) Schematic representation of the tribological
modeling of CMP (after Runnels and Eyman,
1994).
Yu et al., (1993) considered the statistical nature of pad
asperities. Since the pad and the wafer are rough on a
microlevel, the actual area of pad-wafer contact is small
relative to the apparent area. The actual area is taken by
considering the plastic deformation of the pad asperities.
The contact properties, such as contact load, area, and
effective Young’s modulus are determined from Hertz’s
equation. The statistical model assumes that each pad
asperity is spherical at its peak and the variations in
asperity height and radius are normally distributed. The
statistical nature of the asperities is used to characterize
the surface roughness of the pads and pad-wafer contact
properties are then calculated. Although this model
reflects the effects of pad properties such as roughness
and elasticity, it does not explain the non-uniformity in the
wafer scale.
It is clear from a review of the literature that very useful
CMP models were developed by researchers using
isolated phenomena. Consequently, several mechanisms
of CMP of Si wafers were proposed. The first one uses the
normal load and sliding velocity (the so-called Preston
equation). A similar model uses a statistical methods and
elasticity theory to describe the material removal. It
considers the effects of applied pressure and relative
velocity between the pad and the wafer and is capable of
delineating the role of the moduli of slurry particles and
58 1
polishing film. The second model uses the chemical
aspects of CMP, the third model uses the fluid flow
(polishing slurry) between the pad and the wafer surface
(fluid mechanics approach) and erosion, and the fourth
one uses the effects of polishing pad roughness and
dynamic interactions between the pad and wafer. A
statistical model of pad surface roughness is used.
With the need to finish large sized Si wafers (up to 300 mm)
to better flatness and finish economically, it is necessary
to seek better physical understanding of the process. The
model developed should be capable of not only to predict
the removal rates accurately during CMP but also to
monitor and control the process. Even though a chemical
model was developed it is not used in the development of
the analytical equations in other models. It is time to
consider a comprehensive model taking into account
individual phenomena. The chemical aspect and the
erosion aspect of CMP have to be considered in the
development of a plausible mechanism of material removal
in CMP of Si wafers. One feasible approach, in spite of, or
because of, many variables is dimensional analysis. Some
preliminary work in this direction at OSU indicates that a
model can be developed that would include the Reynolds
number, the Euler's number, and other dimensionless
parameters. The Reynolds .number should address the
fluid mechanics approach of the flow of the polishing slurry
while the Euler's number should address both the
mechanical abrasive action of the abrasive particles on
the wafer surface under normal load and the erosion of the
wafer surface under the impact of the silica particles in the
polishing slurry.
6.AlTAINABLE ACCURACIES AND SURFACE INTEGRITY
The ultimate performance of the fine abrasive processes
can be determined by their achievable surface finish, form
accuracy, and resulting surface integrity. These
requirements vary with the individual technological
application. For example, requirements for form and finish
may be most important for metal optics with subsurface
damage a lesser concern, whereas for the processing of
semiconductors, requirements for surface roughness may
be easier met than for flatness or subsurface damage.
Below, a sampling of various fine abrasive processes
including both fixed and loose abrasive processes have
been surveyed for reported performance with regards to
surface finish, form accuracy and resulting subsurface
damage. This compilation that follows is by no means
meant to be comprehensive, but rather to be a sampling of
various processes and their performances as reported in
the recent literature.
6.1. Surface Rouahness
The current performance of many of the fine abrasive
processes in terms of achievable surface roughness is
such that methods used for characterizing the topography
must be carefully considered when reporting achieved
results. There have been several seminal keynote papers
on the measurement of the roughness of smooth surfaces
including that by Vorburger (1 987) and recently by
Lonardo et al. (1996). Rapid advancements have taken
place in the past decade in our ability to measure smooth
surfaces. In particular, the introduction of the atomic
force microscope has allowed a convenient measure' of
microtopography unconstrained by limited lateral spatial
resolution. A brief summary of the lateral and vertical
resolution and range of some of the instruments used to
examine smooth surfaces is shown in Table 5.
The differences in measured topographies of very smooth
surfaces by using instruments with different lateral spatial
resolution has been demonstrated. Sakata et al. (1993)
evaluated the microroughness of mechano-chemically
polished (100) Si surfaces by x-ray Fresnel reflectivity.
582
They found surface roughnesses of 1 nm rms as evaluated
using the x-ray reflectivity method which has nanometer
scale lateral resolution. The same surfaces measured with
an interferometric microscope with a lateral spatial
resolution of 1 pm gave surface roughnesses of 0.3 nm
rms. Also, differences in measured roughness parameters
using a stylus instrument, optical profiler and AFM in the
measurement of polished hardened tool steel surfaces
have been reported by De Chiffre et al. (1996). Tables 6(a)
and 6(b) show the range of reported average surface
Table 5 Resolution and range of some instruments used
to examine smooth surfaces (after Lonardo et al.,
1996;Whitehouse, 1994)
Resolution (nm) Range Im)
Instrument Lateral Vertical Lateral Vertical
Stylus 100-250 0.3 >I00 1
lnterferometric 500 0.1 7 0.1
microscope
AFM' 2 <o. 1 0.1 0.005
STM 2.5 0.2 0.1 0.0001
Nomarski >500 - - -
microscope
SEM 10 2 - 0.002
- 0.0001
TEM 2 2,000
-
roughnesses, Ra, Rq, Rmax for a variety of fine abrasive
processes (both fixed and loose). Where possible, data
taken with instrumentation with the finest lateral spatial
resolution is used.
a. Fixed abrasive Drocesseg
Grinding with fine abrasives of both ductile and brittle
materials is currently being used to generate surface
roughnesses in the nanometer range. Nakagawa's fine
grinding machine with in-process dressing (ELID) and
micro-grain cast iron fiber bonded diamond grinding whee!
is an example (Ohmori and Nakagawa, 1990). Reported
results for Si3N4 include surface roughnesses as low as
6.2 nm Ra (1994),and 3.2 nm Ra (Bandyopadhyay et al.,
1996). In another study, single crystal Si was ground to an
Ra of 2.8 nm and BK7 glass to an R, of 2.4 nm using a
#40000 wheel, and when constant pressure grinding was
employed with ultrafine grit wheels (#3000000 metallic
bond wheel with 5 nm diamond grits), a roughness of 0.33
nm Ra for both Si and BK7 glass was achieved (Ohmori and
Nakagawa, 1995). The ELID technique has also been
used together with chemical mechanical grinding which
used an ultrafine grit wheel composed of 7 to 40 nm silica
abrasives. This integrated planarization technique enabled
a 6 inch Si wafer to be finished to a surface roughness of 4
nm Rmax (Jeong et al., 1996).
Another example is Namba's ultraprecision grinder with a
glass-ceramic spindle operated in a controlled
environment (Namba et al., 1989). Using this machine, a
variety of optical glasses were ground to a surface
roughness of < 0.2 nrn Ra, as measured with an AFM
(Namba and Abe, 1993). Results on Mn-Zn ferrite single
crystals produced by the grinder using a 2-6 pm grain size
diamond wheel in a single pass were 0.6 nm Ra, as
measured with an interferometric microscope, and 2.7 nm
Ra, as measured with an STM (Namba et al., 1992).
The fine grinding of Sic toroidal mirrors has produced
roughnesses in the range of 2-5 nm Rae Suzuki (1995)has
reported on obtaining a roughness of 4 nm Ra on a CVD-
Sic toroidal surface (500 mm x 100 mm) ground with metal
 Keynote Papers

Table 6(a) Reported surface roughness results for fixed abrasive processes

Process Workpiece I I (2)

I ; :I ; Reference

1
(:I)

Fine Grinding
ELlD with micro-grain cast iron 6.2
bonded wheels
3.2 Bandyopadhyay
(1 996)
single crystal Si 0.33-2.8 Ohmori (1995)
BK7 0.33-2.4
ELlD with chemomechanical grinding single crystal Si Jeong (1 996)
Ultraprecision grinder wlglass- optical glasses Namba (1993)
ceramic spindle
Mn-Zn ferrite Namba (1 992)
Grinding w/metal bonded wheels + Sic Suzuki (1 995)
diamond abrasives
Cast iron fiber bonded diamond Sic 2-5 Zhong (1 992,
wheels 19941
~~

Fine grain diamond wheels (1 -2 pm) A1203-TiC 30 Matsuo (1 997)

Mn-Zn ferrite . 20
Si3N4 4-9 lchida (1 993)
3-6 pm diamond wheels Si 11 Puttick (1 994)
Grinding w/electrophoretically Si c8 lkeno (1 990)
deposited abrasive pellets
sapphire 22
Double-sidedlapping w/resin bond high purity A1 alloy 20 Tomita (1 996)
stone
Coated abrasive tape texturing Al alloy whndercoat 14-17 Yanagi (1 992)
Grinding w/directionally aligned SIC hardened steel 4 Yamaguchi
wheel (1 995)
Superfinishinn w/cBN hardened steel c30 Onchi (1 995)
Superfinishing w/fused alumina hardened steel
I 6o I I Puthanangady
I(1995)

bonded wheels with fine grain diamond abrasives. Other
grinding studies on Sic toroidal mirrors using cast iron
fiber-bonded diamond abrasive wheels with in-process
dressing yielded values of 2-5 nm R a (Zhqng, 1992), and
using cast iron fiber bonded diamond cup wheels
roughness of 2-4 nm Ra were obtained (Zhong and
Venkatesh, 1994).
Studies on the grinding of ceramics with very fine grain
diamond abrasives have demonstrated the ability to
produce similar surface roughnesses. Matsuo et al.
(1997) reported on the effect of grain size on surface
roughness in the grinding of ceramics with superfine grain
diamond wheels. Surface roughnesses of -30 nm Rmax
-
were obtained on A1203-TiC and 20 nm Rmax on Mn-Zn
Ferrite using 1 pm mean grain diameter abrasives. Also,
the grinding of Si3N4 with a 1.47 pm average grain size
diamond wheel was shown to achieve a roughness of 4-9
nm Rmax (Ichida et al., 1993).
Examples of the fine grinding of Si include the study of
Puttick et al. (1994) in which a surface roughness on Si of
1 1 nm Ra was reported. lkeno et al. (1990) performed
nanometer grinding using abrasive pellets made by
electrophoretic deposition of 10-20 nm ultrafine silica
grains, and produced surface roughnesses of less than 8
nm PV on Si and 22 nm PV on sapphire.
Table 6(a) shows that while fine grinding of brittle materials
has demonstrated average surface roughnesses of
typically e l 0 nm, with sub-nanometer roughnesses
possible, the fixed abrasive fine finishing of metals yields
a broader range of roughness results. A study of the
ultraprecision grinding of a high-purity A1 alloy for use as a
magnetic disk substrate was performed on a double-sided
lapping machine using a grinding stone with a resin bond,
composed of a mixture of PVA, melamine and phenol
resins. Using a 3 pm Sic abrasive, the A1 was ground to
an R, of 20 nm (Tomita and Eda, 1996). An interesting
comparison can be made by considering the fixed abrasive
process used for the subsequent texturing of thin film
magnetic disk substrates. Using a coated abrasive tape
with a 3 pm diameter A1203 abrasive, Al alloy substrates
with undercoat were given a radial texture resulting in an
R, of 14-17 nm (Yanagi et al., 1992). Yamaguchi and
Horaguchi (1995) have reported on the use of a
directionally aligned Sic whisker wheel, in which hardened
die steel was ground to a surface roughness of 4 nm Ra. In
addition to grinding, results on the superfinishing of
hardened steel have also been reported. Hardened steel
bearing races were ground with the use of a cBN stone with
a porous structure to a surface roughness of less than 30
nm R a (Onchi et al., 1995), and using a 500 grit fused
alumina abrasive, a surface roughness of about 60 nm R a
on hardened steel was obtained (Puthanangady and
Malkin, 1995).
b. Loose abrasive orocesses
The fine abrasive processes using loose abrasive can be
roughly classified into three categories: 1) polishing, 2)
lapping, and 3) field-assisted processes. All three types
of processes have been shown to be capable of producing
fine finishes on both ductile and brittle materials.
Significant interest of late has been focused on magnetic
field assisted finishing for both ductile and brittle
materials. The finishing of Si3N4 balls and rollers for high

583
 Table 6(b) Reported surface roughness results for loose abrasive processes
~~

I Process Workpiece

Field-Assisted
Magnetic fluid grinding
w/SiC abrasives Si3N4 ball
alumina plates
1 Umehara (1994)

w/diamond abrasives SiaN4 balls Childs (1994b)

Magnetic float polishing
wlAI2O3 abrasives Si3N4 balls Komanduri

1 1
(1996b)
w/CqOg abrasives 9.1
w/CeO2 abrasives 4.0 40 Jiang (1997b)
w/SiC abrasives acrylic resin 40 Tani (1984)
Magnetic abrasive finishing Si3N4 rollers 40 Shinmura (1990)
w/diamond abrasives
Magnetic abrasive finishing stainless steel rods 7.6 Fox (1994)
brass tubes 40 Umehara (1995)
Lithium Niobate 20 lkeno (1991)
deDosition + diamond blade
Polishing using electrophoretic BK7 40 Suzuki (1994)
deposition + elastic tool
1apDindPolishinq
Lapping w/sintered cast iron plate tungsten carbide Hagiuda (1981)
witti cBN abrasives
Lapping w/tin lap with diamond poly Mn-Zn ferrite Touge (1996)
abrasives
Lapping w/tin lap + various fine single crystal sapphire Narnba (1978)
particles
single crystal Mn-Zn
poly Mn-Zn
Lapping wheel WAOW bonding single crystal Si Tani (1986)
strength
Kawata (1993)
BK7, lime glass, AIN
Synthetic fabric-faced lap w/diamond electroless Ni Parks (1994)
paste
Teflon lap w/A1203 abrasives BK7 Leistner (1993)
w/SiO;! sinale ctvstal Si
Lap wMylar film + foamed Sic black filter glass Parks (1997)
substrate

II Lapping w/fluorocarbon foam

polisher w/SiO:, powder
Polishing w/flexible tool w/elastic
fused silica

fused silica
Kasai (1990)

Ando (1995)

I
layer
CVD-SIC
CaF2
(100) Si-oxide surface Sakata (1993)
(100) Si/oxide interface
polySi Yasseen (1997)
SIC Yamoaka (1994)
CdS I
Vitali I1996)
Mechanical polishing w/0.33 pm Ni-Zn ferrite Xie (1996)
diamond abrasives
pure Cu

'm
Polishing w/cloth polisher + fine AlSl 304 stainless steel Namba (1980b)
particles
Float polishing fused silica, glass ceramic 0.1-0.2 Namba (1987)
natural quartz Soares (1994)
Float Dolishina w/fluorocarbontool GaAs Kasai 11988)

584

temperature bearing applications has received particular
attention (Komanduri, 1996a). Umehara (1994) has
employed magnetic fluid grinding with Sic abrasive to
finish Si3N4 balls to a surface roughness of 100 nm Rmax.
In a study aimed at identifying the material removal
mechanisms of the magnetic fluid grinding of Si3N4 balls,
Childs et al. (1994a) report the achievement of an Ra of 25
nm. In the magnetic float polishing of Si3N4 balls,
Komanduri (Raghunandan et al., 1997a) found that use of
Cr2O3 abrasive particles resulted in a significantly higher
removal rate and improved surface finish over that in which
A1203 abrasives were used. This was shown to be due to
chemo-mechanical action between the Si3N4 and Cr2O3.
Surface roughness values achieved were 35.8 nm Ra
using A1203 abrasives and 9.1 nm Ra using Cr2O3
abrasives. Jiang and Komanduri (1997b) have used
chemo-mechanical polishing as the last stage of magnetic
float polishing and obtained a surface finish of 4 nm Ra and
40 nm Rma. A study on the magnetic abrasive finishing of
Si3N4 rollers using diamond abrasives showed achievable
surface roughnesses of 40 nm Ra (Shinmura et al., 1990).
However, the same Si3N4 rollers without using diamond
abrasive but instead using conventional abrasives
followed by chemo-mechanical polishing with Cr2O3
resulted in 5.6 nm Ra (Komanduri, 1996a). The finishing of
alumina plates by magnetic fluid grinding resulted in a
surface roughness of 60 nm Ra (Umehara and Komanduri,
1996).
Work on the magnetic assisted finishing of ductile
materials has also been performed. In a study on the
magnetic abrasive finishing of non-magnetic stainless
steel rods to simulate non-magnetic Si3N4 a surface
roughness of 7.6 nm Ra was reported (Fox et al., 1994) and
the finishing of the internal surface of long brass tubes
yielded surface roughnesses of 40 nm Ra (Umehara et al.,
1995). Tani and Kawata (1984), in an early study on the
magnetic field assisted polishing of acrylic resin reported a
surface roughness of c 40 nm Rmap
Another field-assisted finishing technique, described
earlier, is the use of electrophoretic deposition of ultrafine
abrasives onto a conductive diamond saw blade in an
electric field (Ikeno et al., 1991) for chipping-free dicing.
Lithium niobate was cut using this technique resulting in a
surface roughness of 20 nm Rmax. Suzuki et al. (1994)
also employed electrophoretic deposition technique to
polish 3-surfaces in an electric field employing an
electrically conductive elastic tool. Using a tool consisting
of a semi-spherical elastic polymer wrapped with carbon
fabric, BK7 glass was finished to a surface roughness 40-
nm Ra.
Traditionally, lapping refers to a process which uses hard
tooling, relatively coarse abrasives and results in surface
roughnesses which are typically larger than polishing.
With the introduction of flexible tools used for ultra-fine
lapping and the combination, for example, of a tin lap with
the use of fine abrasive particles and a hydrodynamic film,
the distinction between lapping and polishing becomes
less clear. One approach to attempt to delineate the
various non-field-assisted loose abrasive processes is to
focus on the type of tool/workpiece contact in the removal
zone, where the cases of close contact (mechanical
loading), non-contact (fluid film separation), and a mixed
condition containing some of both loading conditions are
considered. Chemical effects between tool and workpiece
may or may not be present for all three situations.
One can first consider surfaces generated by lapping
where the dominate tooVworkpiece interaction is "close
contact". In the work of Hagiuda et al. (1981), a sintered
cast iron lapping plate was manufactured with fixed
abrasives (diamond or cBN) such that the abrasives were
premixed in the matrix cast iron powder. Using this
Keynote Papers
approach, lapping of cemented tungsten carbide with a
composite plate of cast iron and cBN abrasives resulted in
a surface roughness of 100 nm Rmax. The approach of
involving both fixed and loose grains to perform material
removal has also been reported. The high precision
-
lapping of polycrystalline Mn-Zn Ferrite with 0.5 2.0 Bm
diamond abrasives was performed using a tin lap (Touge
and Matsuo, 1996). It was observed that the surface
characteristics, as measured by AFM, differed
significantly depending on whether only fixed abrasives or
both fixed and loose abrasives were used. Final surfaces
generated by only fixed grains had surface roughnesses
as low as 0.9 nm Ra. In an early study on the ultra-fine
finishing of single and polycrystal Mn-Zn and single crystal
sapphire, Namba and Tsuwa (1978) reported on using a tin
lap and various fine particles to obtain a surface
roughness of less than 1 nm RZ on single crystal sapphire,
less than 1 nm R, for single crystal Mn-Zn, and 5 nm R, for
polycrystalline Mn-Zn.
The development of techniques which utilize a lapping
wheel with low bonding strength to aid in the self-dressing
of the wheel and provide a compliant backing for the
individual abrasive grains enabling uniformity of the grain
depth of cut have also been investigated. Tani and
Kawata (1986) reported on the use of a lapping process
using a "liquid-bonded wheel" which consisted of silicon
oxide or silicon carbide abrasive grains, 5 pm in size,
bonded with an alkaline solution of diethanolamine.
Surface roughness of 20 nm RZ was obtained on single
crystal Si. In another study, the lapping of Si and various
brittle materials including BK7 glass, soda-lime glass, and
aluminum nitride was performed using a lapping disc
possessing a high concentration of abrasives and low
bonding strength (Kawata and Tani, 1993). A surface
roughness of 2-3nm Ra was obtained for single crystal Si
wafers using a disc of silicon oxide, zirconium oxide, or
calcium carbonate abrasives and a colloidal silica lapping
fluid. It was hypothesized that the abrasive grains may
undergo a mechano-chemical action with the Si. Surface
roughnesses of c 10 nm Ra were obtained for other brittle
materials included in the study.
As discussed earlier, a variety of lapping techniques using
flexible laps have also been reported. Using diamond
paste rubbed into self-adhesive Pellon (a non-woven
synthetic fabric) and a few drops of polyglycol-based
diamond paste extender, the electroless Ni plated optics
were finished to a surface roughness of c 1 nm nns. Teflon
laps have also been used to superpolish a variety of
optical glasses to an rms surface roughness of less than
0.1 nm (Leistner, 1993). Examples include using a teflon
lap with A1203 abrasive for the finishing of BK7 glass, and
with Si02 for single crystal Si. Recently, a lapping
technique based on a "rapidly renewable lap" has been
reported (Parks et al., 1997). The concept involves a thin
film used to carry abrasive and which covers a stable lap
structure machined to the required figure. A surface
roughness of 1 nm rms was obtained for the final finishing
of black filter glass, in which a Mylar film was used with a
foamed SIC substrate and a cerium oxide slurry. Other
examples on the use of flexible tools for ultra-fine lapping
include the use of a fluorocarbon foam polisher with Si02
powder which resulted in a surface roughness of 0.3 nm
Rmax on fused quartz (Kasai et al., 1990). Another
example is the development of the Canon Super-Smooth
Polisher (CSSP) described earlier and the use of a flexible
tool laminated with an elastic layer for the local pitch
polishing of aspherical surfaces (Ando et al., 1995).
Attained surface roughnesses on a toroidal mirror using
this approach include 0.13 nm rms on fused silica, 0.15 nm
rms on CVD-Sic, and 0.12 nm rms on a CaF2 aspherical
lens.
585
Consideration of the fine abrasive processes (using loose
abrasives) in which there may be some combination of
close contact and fluid film contact can be focused on the
more conventional mechano-chemical or chemo-
mechanical polishing. These techniques have been used
to obtain very low roughness surfaces on both brittle and
ductile materials, a few examples of which follow. Using
standard mechano-chemical polishing with a suspension
of silica particles in an alkaline solution, (100) Si surfaces
were prepared and the resulting surface roughness of both
the oxide top surface and the Si/oxide interface were
measured by x-ray reflectivity measurements (Sakata et
-
al., 1993). Reported roughness results were 0.8 1.1 nm
rms for the oxide top surface, and 1.1-2.8 nm rms for the
Si/oxide interface. Polysilicon films are routinely polished
by chemo-mechanical polishing. Average surface
roughnesses of 1.7-1.9 nm have been reported (Yasseen,
1997).
Both hard and soft brittle materials can be routinely
polished by conventional techniques to low surface
roughness. Examples include a SIC X-ray mirror polished
to 0.47-0.6 nm rms (Yamoaka et al., 1994), and an average
surface roughness of 1.3 nm Ra obtained for the soft
semiconductor CdS by conventional low load chemo-
mechanical polishing (Vitali et al., 1996). Conventional
mechanical polishing using 0.33 pm diamond abrasives
yielded surface roughnesses of 6.19 nm rms on Ni-Zn
ferrite (Xie, 1996), and roughnesses on pure copper were
found to be similar, viz., 8.22 nm rms. Namba (1980b)
reported on the ultra-fine polishing of AlSl 304 stainless
steel with a cloth polisher with a polishing fluid with water
and 7 nm particles of Si02, Fe2O3 or MgO powder after
lapping with 3 pm diamond powder and tin lap with results
of 2 nm R,. Typical surface roughnesses of Ni-P plated
films on AI-Mg computer disk substrates after polishing are
1-2 nm R, (Dornfeld and Liu, 1993).
Float polishing or hydrodynamic film polishing belongs to
the category of fine loose abrasive processes where there
is little or no mechanical contact between tool and
workpiece. Low surface roughness results have been
reported for glasses and semiconductors. Namba et al.
-
(1987) report surfaces of 0.1 0.2 nm rms on fused silica
and glass ceramic, and natural quartz was float polished to
a roughness of 0.2 nm rms (Soares et al., 1994). Results
on semiconductors are similar. Kasai et al. (1988) report
on using a technique for progressive mechanical and
chemical polishing (P-MAC) with the final processing being
polishing with a fluorocarbon plastic polishing tool and
bromine methanol solution. Two inch diameter GaAs
wafers were finished to a surface roughness of less than 2
nm Rmax. Watanabe and Suzuki (1981) developed a
hydrodynamic polishing technique which resulted in an
achieved surface roughness on Si of 1 nm.
6.2. Form Accuracv
Fine abrasive processes (both fixed and loose) have been
shown to yield accuracies of form (figure) in the nanometer
range. In general, one can consider the achievable
flatness and the contour accuracy. A wide range of
reported values for achievable form accuracy exists in the
literature and is evidence that achievable accuracies are
clearly dependent on not only the type of process and
workpiece material, but also the specific machine. In the
case of ultraprecision aspheric optics, it may in fact be the
uncertainty in the metrological system that limits the
achievable accuracy. Some examples are given below for
the form accuracies which have been reported for fine
abrasive processes (both fixed and loose).
For the case of fine grinding with the combination of ELlD
and micro-grain cast iron fiber bonded wheels, reports of
achievable flatness in Si wafers include a flatness of 300
nm over a 6 inch wafer (Jeong et al., 1996), and a flatness
of 160 nm for a wafer 60 mm in diameter (Ohmori and
Nakagawa, 1995). Using the ultraprecision surface grinder
with a glass-ceramic spindle previously discussed, Namba
et al. (1989) report that NbFl optical glass was ground flat
to 200 nm over an 18 mm square. Contour accuracies for
grinding with metal bonded wheels have also been
reported. Suzuki (1995) used fine grain diamond wheels to
grind a CVD-SIC toroidal surface of 500 mm by 100 mm to a
profile accuracy of 0.5 pm.
Magnetic fluid grinding of Si3N4 balls has resulted in
reported range of sphericity from f 1 pm using diamond
abrasive (Childs et al., 1995) to 0.14 pm using Sic
abrasives (Umehara, 1994) and 0.15 p m using
conventional abrasives followed by chemo-mechanical
polishing (Jiang and Komanduri, 1997a). Magnetic fluid
grinding was also used to finish alumina plates to a
flatness of 500 nm over 25.4 mm (Umehara and
Komanduri, 1996).
Studies on ultra-fine lapping demonstrate an achieved
flatness of 30 nm over 7 mm in the early work of Namba on
single crystal sapphire (Namba and Tsuwa, 1978), and with
the use of a lapping process employing the "liquid-bonded
wheel" previously discussed, Tani achieved a surface
flatness of 500 nm over a 3 inch diameter single crystal Si
wafer (Tani and Kata, 1986). The Canon Super-Smooth
Polisher (CSSP) was used to pitch polish 500 mm diameter
toroidal mirrors to a contour accuracy of 78 nm PV on
fused silica and 60 nm PV on CVD-SIC. It also resulted in a
77 nrfi PV accuracy on a 200 mm diameter CaF2 aspherical
lens (Ando et al., 1995).
Flatnesses achieved from polishing a broad range of
materials have also been reported. Using the P-MAC
process with a fluorocarbon plastic polishing tool and
bromine methanol solution, 2 in. diameter GaAs wafers
were polished flat to less than 2 pm over 80% of the wafer
diameter (Kasai el al., 1988) and InP wafers were chemo-
mechanically polished to a flatness of 2 pm over 100 mm
diameter (Wang et al., 1994 ). A flatness of 150 nm over
100 mm achieved by polishing a Sic X-ray mirror has also
been reported (Yamoaka et al., 1994). Results for float
polishing include a flatness deviation of less than 1 pm
over a 3 inch diameter Si wafer (Watanabe and Suzuki,
1981), and a PV flatness of 30 nm on 100 mm diameter
6K7 and Zerodur (Namba et al., 1987)
Total thickness variation (TTV) is another important
measure of the achievable accuracy of the wafer polishing
process. In the chemo-mechanical polishing of CVD -
Si02, a TTV of +/- 20 nm over a 6 inch wafer was achieved
using a "hydro chuck consisting of a quartz plate with a
water supported surface (Hayashi et al., 1996).
6.3. Surface lntearity
Whereas some information on the resulting subsurface
damage due to fine abrasive finishing exists, when
compared to the data available on surface finish and form
accuracies, much less has been reported on the resulting
condition of the subsurface. In part, this is due to the fact
that the length scales of the damage (micrometers down to
nanometers), make measurements a formidable task. Still
however, significant progress on the assessment of both
the extent and type of damage caused by fine abrasive
finishing has been made.
Several ClRP keynote papers have addressed the
importance of the assessment of the surface integrity
which results after machining. Tcinshoff and Brinksmeier
(1980) presented a review of the mechanical and thermal
effects on machined surfaces by residual stress and
microhardness measurements. Brinksmeier (1989,
Brinksmeier et al., 1984) presented an extensive review of
non-destructive methods used .for evaluating surface
integrity. A report on a cooperative work conducted in the
STC S (Tonshoff et al., 1987) presents a review of different
measurement techniques for the detection of cracks. In a
seminal keynote paper on the abrasive machining of
silicon, Tonshoff et al. (1990) presented a listing of some
of the methods currently being used for the assessment of
subsurface damage. This keynote paper also presents a
careful review of five methods used for measuring
subsurface damage in semiconductor materials. In
particular, high resolution X-ray diffraction (HRXRD), p-
Raman spectroscopy, total reflection X-ray fluorescence
spectroscopy (TXRF), photothermal microscopy, and
bonds. ELID technique was used to dress the metal
bonded wheels. They found the mean depth of damage in
both the diamond turned and the ground specimens was in
the range of 100-400 nm, however in the ground
specimens, the damage depth was very variable. In the
ground specimens the damage was seen to consist of 1)
regions of well defined dislocation loops on several slip
systems, 2) regions with a very high density of irregular
dislocation arrays, 3) occasional microcracks up to 500
nm deep which did not always intersect the surface, and 4)
patches of amorphous silicon. Cross-sections of the
ground specimen are shown in Figs. 51(a) and (b). In

Fig. 50 Dislocations introduced by the grinding process

of a Si wafer clearly discerned by the use of a Fig. 51(a) Cross-section of a ground Si (111) crystal
"Yang"- base etch (after Tonshoff et al., 1994) showing spacing of feed marks (after Puttick et
al., 1994)
etching methods were reviewed. A typical technique for
assessing damage introduced into Si wafers is by etching.
As shown in Fig. 50, Tonshoff et al. (1994) demonstrated
that dislocations introduced by the grinding process of a Si
wafer could clearly be discerned by the use of a "Yang"-
base etch.
In addition to the techniques mentioned above, recent
studies have reported on the use of a broad variety of
techniques for measurement of subsurface damage, some
having nanometer scale depth resolution. As examples,
the use of cross-sectional transmission electron
microscopy (XTEM) has been reported for assessing
damage in nanomachined single crystal Si (Puttick et al.,
1994) and in mechanically polished CdTe (Nouruzi-
Khorasani, 1990). Atomic level resolution can be obtained
with the use of cross sectional high resolution
transmission electron microscopy (HRTEM) which was
used to examine crystalline damage introduced to (100) Si
by lapping and single point diamond machining (Kunz,
1996). Bowen (1993) presented a characterization
technique based on grazing incidence x-ray reflectivity
(GIXR) which can be used for the measurement of film Fig. 51(b) Surface crack and dislocations on a ground Si
thickness of semiconductors and magnetic multilayers (111) crystal (after Puttick et al., 1994)
and has sub-nanometer resolution. Additional techniques
including scanning photo-luminescence(PL) (Laczik et al., another investigation on the fine grinding of single crystal
1996a,b), photoacoustic spectroscopy (Soares, 1994), Si, Ohmori and Nakagawa (1995) evaluated the
ion channeling (Lucca et al., 1995), reflection high energy subsurface damage through X-ray topography, angle
electron diffraction (RHEED) (Vitali et al., 1996), and X-ray polishing and step etching. Using ultrafine grain metallic
topography (Ohmori and Nakagawa, 1995) are further bond wheels with ELID, single crystal Si was ground. X-
reported on below. ray topography revealed that cracks existed even in mirror
The subsurface damage resulting from the fine grinding of surfaces ground to several nanometers Ra. By using
single crystal Si has been reported by several angle polishing and etching, the depths of the resulting
investigators. Examples include the work of Puttick et al. cracks were measured. Crack depths of 1.3 pm and 0.4
(1994), and Ohmori and Nakagawa (1995). Puttick et al. p m were reported for #2000 and #8000 wheels
employed XTEM to surfaces which had been single point respectively. A total damage depth less than 1 pm was
diamond turned and precision ground with a diamond revealed with the use of step etching.
abrasive wheel. The grinding was performed on the CUPE A variety of approaches for the assessment of subsurface
seven axis grinder using cup diamond wheels with 3-6 pm damage in polished compound semiconductors have been
diamond abrasives bonded with either resin or metal reported. Several examples follow. Lucca and Maggiore
 (1997) have reported on the measurement of subsurface
lattice disorder in mechanically polished and chemo-
mechanically polished CdS and ZnSe. They employed ion
beam channeling using a combination of incident beam
energies and detector positions to obtain a quantitative
measure of lattice disorder. Damage was reported both as
the depth of an equivalent amorphous layer and an actual
physical damage depth. Damage depths due to
"obstruction-type'' defects (amorphous or amorphous-like
regions) were found to be 105 nm (CdS) and 377 nm (ZnSe)
for mechanical polishing with 0.25 mm diamond abrasives,
and 49 nm (CdS) and 135 nm (ZnSe) for chemo-mechanical
polishing with a slurry of sodium hypochloride and colloidal
silica. Differences in defect densities of the bulk CdS and
ZnSe crystals was proposed as a possible cause for the
substantial differences in damage depths for the two
materials polished under the same conditions. For 0.25 pm
mechanically polished ZnSe, indications were found that
"distortion-type'' defects (dislocations, dislocation loops,
stacking faults, etc.) extended as deep as 1.4 pm. These
results indicate that it is simply not sufficient to refer to a
"damage depth" since as was shown here, as well as in
other studies on subsurface damage, various "types" of
damage may have quite different spatial extent into the
subsurface. These results are summarized in Table 7
which also shows damage depths obtained for diamond
turned CdS surfaces. Vitali et al. (1996) have measured
the depth of the near-surface damaged layer of chemo-
mechanically polished CdS by RHEED. By varying the
glancing angle of the incident electron beam relative to the
surface, they were able to probe different depths, and
observe the angle (i.e., depth) at which the obtained
diffraction pattern indicated undamaged material. Using
this approach, the damaged layer was estimated to be
about 10-20 nm thick. They reported on the successful
solid state recrystallization of this damaged layer to single
crystal material by the use of low power pulsed laser
annealing.
Table 7 Subsurface Damage Depths for Fine Finishing of
CdS and ZnSe Measured by Ion Channeling
(after Lucca and Maggiore, 1997)
Eq. Amorphous Damage Depth
Process Laver Inm) (nm)
(0001) (100) (0001) (100)
CdS&&CdS&l&
1 pm mech. polish 21 53 147 424
1/4 pm mech. polish 13 46 105 377
chemo-mech. polish 2 4 49 135
machined along (1070) 50 217
depth of cut = 0.1 pm
machined along (1150) 25 120
depth of cut = 0.1 bm
Other studies on polished semiconductors include the
work of Laczik et al. (1996a,b) who utilized the
photoluminescence response of InP to obtain a measure
of damage introduced by polishing. Scanning PL together
with chemical angle polishing was used to investigate
subsurface damage in "commercially polished" InP wafers.
Three distinct regions below the surface were observed:
1.) a "good" or undamaged zone, extending about 10 nm
into the wafer from the surface, 2.) a damaged zone
extending from about 10 nm to 70 nm consisting of
dislocations and microcracks, and 3.) underlying bulk
material. The study was inconclusive as to whether the
"good" zone in the first 10 nm of the surface was truly
undamaged or whether damage was present which was not
revealed by PL. The subsurface damage introduced into
CdTe by polishing was examined by XTEM in another study
(Nouruzi-Khorasani, 1990). Results of five different
588
polishing procedures were reported. For conventional
lapping using a cast iron plate and a water/600 grade SIC
slurry, an intense banded deformation for the first 4 pm of
the subsurface was found, followed by a more isotropic
distribution of dislocations decreasing in density to a
depth of about 20 pm. For lapping with 3 pm diamond
paste, heavy damage was confined to a region less than 3
pm deep with the total extent of damage not extending
further than 4 pm. Polishing with 50 nm alumina mixed with
water yielded high damage less than 500 nm deep, with a
few dislocations extending beyond 1 pm. Polishing with 50
nm alumina mixed with ethanediol resulted in high damage
extending to less than 30 nm into the CdTe and beyond
500 nm no damage observed. Finally, when polishing was
performed with a SytonlHpOp slurry, the morphology of the
damage away from the surface was found to be different
from the preceding cases. A distribution of small
dislocation loops extending I pm into the surface was
observed.
Fig. 52 Cross-sectional HRTEM image of a Si(100)
surface polished using 50 nm diameter silica grit
(after Kunz et al., 1996)
The defects introduced into the subsurface of Si wafers by
polishing have been examined by HRTEM as reported by
Kunz et al. (Kunz et al., 1996). As shown in Fig. 52, they
found that when polishing was performed with 50 nm
diameter silica particles suspended in a pH 10 ammonium
hydroxide slurry no detectable damage could be observed.
Changes introduced to the near surface stoichiometry or
to the electrical or magnetic properties of the subsurface
can also be thought of as "damage" if the functionality of
the surface is affected by these changes. As an example,
Kinoshita et al. (1976) in an early work demonstrated that
although lapping of Mn-Zn resulted in a surface damage
layer not more than a few micrometers deep, assessment
of the surface revealed that the magnetic domain structure
change extended more than 100 times deeper.
7. CONCLUSIONS
1. In this paper an attempt was made to review various
technological aspects of fine abrasive processes. These
include the machine tools, abrasives, workmaterials
(chiefly, advanced ceramics, glasses, and semi-
conductors), and various processes developed to obtain
nanometric finish, sub-micrometric accuracy, and minimal
subsurface deformation. It may be pointed out that it is not
sufficient to advance one or two aspects of the finishing
process, e.g., abrasives, or machine tools. The total
process as an integrated system needs to be considered.

By incorporating various advancements in the system,
there is a synergistic benefit gained that is responsible for
extending the limits of accuracies, finish, and geometry,
and enabling the processes to be economically
implemented.
2. Techniques such as fine grinding using nanocrystalline
size abrasives, nanometric cutting using single crystal
diamond tools, float polishing using a diamond turned tin
lap, CNC polishing using loose abrasives, lapping using
nanometer-sized fumed silica and appropriate chemical
conditions for the polishing Si wafers, magnetic field
assisted polishing, and elastic emission machining are
some of the techniques recently developed to accomplish
this task.
3. In the finishing of brittle materials, features specific to
these materials have to be considered in developing
appropriate models for the material removal. For example,
glasses are amorphous, while most ceramics are
covalently or ionically bonded crystalline materials.
Hence they have limited slip systems and plastic
deformation by dislocations may not be the preferred mode
of deformation. However, under very high hydrostatic
pressures, most materials can deform plastically. Thus, in
addition to the brittle fracture mode commonly experienced
with brittle materials under heavier loads, plastic
deformation under the tooVabrasive (or indenter) due to a
variety of effects must be considered. These include: high
hydrostatic pressure, densification of the material due to
its amorphous (or lack of long range order) nature in the
case of glasses, structural transformations in the case of
some semiconductors (Si, Ge) subjected to high
hydrostatic pressure, and chemo-mechanicalaction under
a given workmaterial-abrasive combination in a given
environment and appropriate conditions of finishing need
to be considered.
4. TooVworkpiece interactions involved in machining and
grinding processes can be related by noting the continual
rotation of the resultant force system towards the
workpiece surface as one progresses from conventional
machining with positive rake angle tools, to grinding, to
. ultraprecision machining at depths of cut smaller than the
edge radius of the tool, to indentation sliding of a blunt
indenter along the workpiece surface. An indentation
sliding model may be appropriate when developing
mechanical models for the machining of brittle materials
with tools of large edge radius relative to the cut depth or
large negative rake tools.
5. The development of useful models for microcrack-free
finishing of brittle materials must consider nanoindentation
and sliding at loads low enough so as to not initiate median
cracks. The structure and properties of the workmaterials,
the geometry of the indenter, and the loading conditions all
play an important role in determining the state of stress.
The magnitude of hydrostatic pressure determines the
consequent ability of brittle materials to undergo strain
without initiating cracks.
6. While various important models have been developed
for chemical mechanical polishing (CMP), need exits for a
comprehensive model that takes into account the various
physical parameters involved in the process. Such a
model would enable prediction of the material removal rate
during CMP and control of the process. Dimensional
analysis may allow the number of variables involved to be
reduced and the physical phenomenbn to be more clearly
discerned.
7. The ultimate performance of fine abrasive processes
can be determined by their achievable surface finish, form
accuracy, and resulting surface integrity. These
requirements vary with the individual technological
application. Whereas for example, form and finish may be
Keynote Papers
most important for metal optics with subsurface damage a
lesser concern, in the processing of semiconductors,
requirements for surface roughness may be easier met
than for flatness or subsurface damage.
8. The current performance of many of the fine abrasive
processes in terms of achievable surface roughness is
such that methods used for characterizing the topography
must be carefully considered when reporting achieved
results.
9. Reported results on subsurface damage indicate that it
is not sufficient to refer to a "damage depth" since various
"types" of damage may have quite different spatial extent.
The extent of the damage depends on the local
tooVworkpiece contact, and may also be dependent on
workpiece material factors such as defect density.
Reported damage depths resulting from fine abrasive
processes are seen to span a wide spatial range.
10. It is a matter of pride and a time to recognize that many
of the significant developments in the ultraprecision
machining, grinding, and polishing of both ductile and
brittle materials are due to the efforts of many of our
distinguished colleagues of CIRP. Both technological and
scientific contributions were made by them over the years
that enabled Prof. Taniguchi's'predictions to be realized.
ACKNOWLEDGMENTS
This paper is respectfully dedicated to the memory of our
distinguished colleague and dear friend, Late Professor
Hideo Tsuwa of Osaka University who made seminal
contributions in the field of finishing with fine abrasives
Dedicated to Late Professor H. Tsuwa
This keynote paper is based on many fine contributions of
our CIRP colleagues and other researchers whose work
has been referred to in this paper. The authors would like
to acknowledge their valuable contributions. The authors
also thank Drs. C.J. Evans, I. Inasaki, N. Umehara, Y.
Namba, T. Shinmura, and Y. lchida for providing
information on their work. Special thanks are also due to
Prof. P. McKeown and Dr. I. Stowers for many valuable
discussions on this topic. Special thanks are extended to
Prof. E. Brinksmeier, President of the STC G for his
valuable contributions. The authors (RK and DAL) also
thank the National Science Foundation for support of this
work. In particular they extend thanks to Drs. B.M.
Kramer, M. Leu, and J. Lee of the Division of Design,
Manufacture, and Industrial Innovation (DMII). RK would
like to thank Dr. W. Coblenz of DARPA and Dr. J. Larsen
Basse of the Surface Engineering and Tribology Program
at NSF, and acknowledge the support of the NSFs US-
Japan Co-Operative Program and the DOD EPSCoR
Program on Finishing of Advanced Ceramics.
589
REFERENCES
Akazawa, M., Kato, K., Umeya, K., 1986, Wear
Properties of Silicon Nitride in Rolling Contact,
Wear, 110: 285-293.
Akazawa, M., Kato, K., 1988, Wear Properties of
Si3N4 in Rolling-Sliding Contact, Wear, 124: 123-
132.
Anderson, D.S., Angel. J.R.P., Burge, J.H.,
Davison, W.B., DeRigne, S.T., Hille, B.B.,
Kettlesen, D.A., Kittrell, C., Martin, H.M., Nagel,
R.H., Trebisky, T.J., West, S.C., Young, R.S.,
1991, Stressed-Lap Polishing of 3.5m f/1.5 and
1.8m f/1 .O Mirrors, Proc. SPIE, 1531 : 260-269.
Ando, M., Negishi, M., Takimoto, M., Deguchi, A.,
Nakamura, N., 1992, Super-smooth Polishing on
Aspherical Optics, SPIE, 1720: 22-33.
Ando, M., Negishi, M., Takimoto, M., Deguchi, A.,
Nakamura, N., 1995, Super-Smooth Polishing on
Aspherical Surfaces, Nanotechnology, 6: 1 11-1 20.
Bandyopadhyay, B.P., Ohmori, H., Takahashi, I.,
1996, Ductile Regime Mirror Finish Grinding of
Ceramics with Electrolytic In-process Dressing
(ELID) Grinding, Materials and Manufacturing
Processes, 11: 789-801.
Baron, J.M., 1975, Technology of Abrasive
Machining in a Magnetic Field, Masino-strojenije,
Leningrad (in Russian).
Becker, K., Beckstette, K.,1988, M400 and P400 -
A Pair of Machines for Computer-Controlled Fine
Correction of Optical Surfaces, Ultraprecision in
Manufacturing Engineering, Springer-Verlag, Berlin,
Germany, 212-223.
Beilby, Sir George ,1921, Aggregation and Flow of
Solids, Macmillan & Co., London, U. K.
Bennett, J.M., Shaffer, J.J., Shibano, Y., Namba,
Y., 1987, Float Polishing of Optical Materials,
Applied Optics, 26: 696-703.
Bhagavatula, S.R., Komanduri, R., 1996, On the
Chemo-Mechanical Polishing of Silicon Nitride with
Chromium Oxide Abrasive, Phil. Mag., 74: 1003-
101 7.
Bowden, F.P.,Tabor, D., 1950, The Friction and
Lubrication of Solids- Vol. I, The Clarendon Press,
Oxford, U. K.
Bowen, D.K., Tanner, B.K., 1993, Characterization
of Engineering Surfaces by Gazing-Incidence X-
Ray Reflectivity, Nanotechnology, 4: 175-182.
Boyd, D.C. Thompson, D.A., 1980, Glass,
Encyclopedia of Chemical Technology, John Wiley
& Sons, New York, 1 1 : 807-880.
Bridgman, P.W., 1953, Effects of Very High
Pressures on Glass, J of App. Phy., 24: 405-413.
Brinksmeier, E., Schneider, E., Theiner, W.A.,
Tdnshoff, H.K., 1984, Nondestructive Testing for
Evaluating Surface Integrity, Annals of CIRP, 33/2:
489-509.
Brinksmeier, E., Von Schmieden, W., 1987, ID-Cut-
Off Grinding of Brittle Materials., Annals of CIRP,
36/1: 21 9-222.
Brinksmeier, E., 1989, State-of-the-Art of Non-
Destructive Measurement of Sub-surface Material
Properties and Damages, Precision Engineering,
1 1 : 21 1-224.
Brown, N.J., 1986, Preparation of Ultrasmooth
Surfaces, Ann. Rev. of Material Science, 16: 371-
388.
Brown, N.J., Fuchs, B.A., 1989, Shear Mode
Grinding, ASPE Spring Conference Tuscon, AZ.
Bryan, J.B., Loewen, E.G., 1974, The Limit of
Accuracy of Machine Tools, Proc of the ICPE,
Tokyo, Japan, 24-32.
Bryan, J.B., 1979, Design and Construction of an
Ultraprecision 84 in. Diamond Turning Machine,
Precision Engineering, 1 : 13-18.
Bryan, J.B., 1984, The Power of Deterministic
Thinking in Machine Tool Accuracy, Prof. of the 1st
Int. Mach. Tool Engineers Conf., Tokyo, Japan, 39-
53.
Childs, T.H.C., Yoon, H.J., 1992, Magnetic Fluid
Grinding Cell, Annals of CIRP, 4Vl: 343-346.
Childs, T.H.C., Jones, D.A., Mahmood, S., Kato, K.,
Zhang, B., Umehara, N., 1994a, Magnetic Fluid
Grinding Mechanics, Wear, 175: 189-198.
Childs, T.H.C., Mahmood, S., Yoon, H.J., 1994b,
The Material Removal Mechanism in Magnetic Fluid
Grinding of Ceramic Ball Bearings, Proc. of the 1.
Mech. E., 208: 47-59.
Childs, T.H.C., Mahmood, S., Yoon, H.J., 1995,
Magnetic Fluid Grinding of Ceramic Balls, Tribology
International, 28: 341-348.
Cook, L.M., 1990, Chemical Processes in Glass
Polishing, J. Non-Crystalline Solids, 120: 152-171.
Dallas, D.B., 1976, Honing, in Tool and
Manufacturing Engineers Handbook, 3rd Edn.,
SME, Dearborn, MI.
DeChiffre, L., Hansen, H.N., Bronstein, A., 1996,
Investigation on the Surface Topography in
Polishing Using Atomic Force Microscopy, Annals
of CIRP, 45/1: 523-528.
Dornfeld, D., Liu, J.J.B., 1993, Abrasive Texturing
and Burnishing Process Monitoring Using Acoustic
Emission, Annals of CIRP, 42/1: 397-400.
Donaldson, R.D., 1979, Large Optics Diamond
Turning Machine, Lawrence Livermore National
Laboratory Report UCRL-52812 (Vol. 1).
Earnshaw, A., Harrington, T.J., 1973, The
Chemistry of Transition Metals, Oxford University
Press, Oxford, U.K.
Ernsberger, F.M., 1968, Role of Densification in
Deformation of Glasses Under Point Loading, J.
Amer. Ceram. SOC.,51: 545-547.
Evans, C.J., Parks, R.E., 1995, Rapidly Renewable
Polishing Lap, Proc. SPIE, 2554.
Fielder, K.H., 1988, Precision Grinding of Brittle
Materials, Ultraprecision in Manufacturing
Engineering, Springer-Verlag, Berlin, Germany, 72-
77.
Finnie, I., 1958, The Mechanism of Erosion of
Ductile Metals, Proc. of the 3rd National Congress
on Applied Mechanics, ASME, 527-532.
Finnie, I., 1960, Erosion of Surfaces by Solid
Particles, Wear, 3: 87-103.
Fischer, T.E., Tomizawa, H., 1985, Interaction of
Tribo-chemistry and Microfracture in the Friction
and Wear of Silicon Nitride, Wear, 105: 29-45.
Fischer, T.E., 1988, Tribochemistry, Ann. Rev.
Mater. Sci., 18: 303-323.
Fischer, T.E., Mullins, W.M., 1993, Relation
Between Surface Chemistry and Tribology of
Ceramics, in S. Jahanmir (ed.), Friction and Wear of
Ceramics, Marcel Dekker, New York.

Fox, M., Agrawal, K., Shinmura, T., Komanduri, R.,
1994, Magnetic Abrasive Finishing of Rollers,
Annals of CIRP, 4311: 181-184.
Fujisawa, M., Oku, T., Arakawa, N., Watanabe, M.,
1983, Precision Sawing with Wire Saw, Annals of
CIRP 32/1: 87-90.
Gee, M.G., 1992, The Formation of Glass in the
Wear of Reaction-Bonded Silicon Nitride, J. Phys.
D. App. Phy. 25: A1 82- A1 88.
Gee, M.G., Butterfield, D.B., 1993, Wear, 162-164:
234-245.
Gerchman, M.C., McLain, B.E., 1988, Infrared
Optical Materials VI,Proc. of the SPIE, 929: 94-98.
Gerk, A.P., Tabor, D., 1978, Indentation Hardness
and Semiconductor-metalTransistion of Germanium
and Silicon, Nature, 271: 732-733.
Gilman, J.J., 1973, Hardness- A Strength
Microprobe, in the Science of Hardness Testing and
Its Research Applications, ASM Int., Metals Park,
OH.
Gormley, J.V., Manfra, M.J., Calawa, A.R., 1981,
Hydroplane Polishing of Semiconductor Crystals,
Rev. Sci. Instruments, 52: 1256-1258.
Hagan, J.T., 1979, Micromechanics of Crack
Nucleation During Indentations, J. of Mat. Sci, 14:
2975-2980.
Hagiuda, Y., Karikomi, K., Nakagawa, T., 1981,
Manufacturingof a Sintered Cast Iron Lapping Plate
with Fixed Abrasives and Its Lapping Abilities,
Annals of CIRP, 30/1: 227-231.
Hashimoto, H., 1988, Machining of Aspherical
Surfaces, Ultraprecision in Manufacturing
Engineering, Springer-Verlag, Berlin, Germany,
184-186.
(53) Hashish, M., 1984, A Modeling Study of Metal
Cutting with Abrasive Waterjets, J. Eng. Mat. and
Tech., 106: 88-100.
(54) Hatamura, Y., Nagao, T., Mitsuishi, M., Kato, K.I.,
Taguchi, S . , Okumura, T., Nakagawa, G.,
Sugishita, H., 1993, Development of an Intelligent
Machining Center Incorporating Active
Compensation for Thermal Distortion, Annals of
CIRP, 42/1: 549-552.
(55) Hayashi, Y., Nakajima, T., Kunio, T., 1996,
Ultrauniform Chemical Mechanical Polishing (CMP)
Using a "Hydro Chuck, Featured by Wafer Mounting
on a Quartz Glass Plate with Fully Flat, Water-
Supported Surface, Jpn. J. Appl. Phys. 35: 1054-
1059.
Heinicke, G. 1984, Tribochemistry, Carl Hanswe
Verlag, Munich, Germany.
Holland, L., 1964, The Properties of Glass
Surfaces, John Wiley & Sons, New York.
Hou, H.B., Komanduri, R., 1997a, Magnetic Field
Assisted Polishing of Ceramics, Part I: Thermal
Model, ASME, J. of Tribology, submitted for
publication.
(59) Hou, H.B., Komanduri, R., 1997b, Magnetic Field
Assisted Polishing of Ceramics, Part II: Thermal
Aspects of magnetic Abrasive Finishing (MAF) of
Ceramic Balls, ASME, J. of Tribology, submitted for
publication.
Ichida, Y., Kishi, K., 1993, Nanotopography of
Ultraprecise Ground Surface of Fine Ceramics
Using Atomic Force Microscope, Annals of CIRP,
42/1: 647-650.
Keynote Papers
Ikawa, N., Donaldson, R.R., Komanduri, R., Konig,
W., McKeown, P.A., Moriwaki, T., Stowers, I.F.,
1991, Ultraprecision Metal Cutting-The Past, the
Present and the Future, Annals of CIRP, 4012: 587-
594.
Ikeno, J., Tani, Y., Sato, H., 1990, Nanometer
Grinding Using Ultrafine Abrasive Pellets-
Manufacture of Pellets Applying Electrophoretic
Deposition, Annals of CIRP, 3911: 341-344.
Ikeno, J., Tani, Y., Fukutani, A., 1991,
Development of Chipping-Free Dicing Technology
Applying Electrophoretic Deposition of Ultrafine
Abrasives, Annals of CIRP, 40/1: 351-354.
Imanaka, O., Okutomi, M., 1979, New Concepts on
Surface Finishing and Its Application to Ceramics--
Recent Progress in Ultrafine Finishing in Japan, The
Science of Ceramic Machining and Surface
Finishing, National Bureau of Standards, Special
Pub. No. 562, B.J. Hockey and R.W. Rice, Eds.,
Gaithersburg, MD, 157-169.
Inasaki, I., 1986, High-Efficiency Grinding of
Advanced Ceramics, Annals of CIRP, 3511: 21 1-
214.
Inasaki, I., 1987, Grinding of Hard and Brittle
Materials, Annals of CIRP, 3612: 463-471.
Indge, J.H., 1991, Lapping, Honing, and Polishing,
Engineering Materials Handbook, ASM Int., Metals
Park, OH, 4: 351-358.
'
Ives, N.A., Leung, M.S., 1988, Noncontact Laminar
Flow Polishing for GaAs, Rev. Sci. Instruments, 59:
172-175.
Izumitani, T., 1979, Polishing, Lapping, and
Diamond Grinding of Optical Glasses, Treatise on
Material Science and Technology, Academic Press
Inc., New York, 17: 116-149.
Jahanmir, S., Fischer, T.E., 1987, Friction and Wear
of Silicon Nitride Lubricated by Humid Air, Water,
Hexadecane and Hexadecane + 0.5 percent Stearic
Acid,Trans STLE, 31: 32-43.
Jahanmir, S., Ives, L.K., Ruff, W.A., and Peterson,
M.B., 1992, Ceramic Machining: Assessment of
Current Practice and Research Needs in the United
States, N E T Special Publication No. 834,
Gaithersburg, MD.
(72) Jahanmir, S., 1993, Friction and Wear of Ceramics,
Marcel Dekker, Inc., New York.
(73) Jeong, H., Ohmori, H., Doy, T.K., Nakagawa, T.,
1996, Integrated Planaritation Technique with
Consistency in Abrasive Machining for Advanced
Semiconductor Chip Fabrication, Annals of CIRP,
45/1: 311-314.
(74) Jiang, M., Komanduri, R., 1997a, Finishing of Si3N,
Balls for Bearing Applications, submitted for
publication.
(75) Jiang, M., Komanduri, R., 1997b, Mechanism of
Chemo-Mechanical Polishing of Silicon Nitride with
CeO, Abrasive, Submitted for publication.
(76) Johnson, K.L., 1970, The Correlation of Indentation
Experiments, J. Mech. Phy. Solids, 18: 115-126.
(77) Jones, R.A., Plante, R.L., 1985, Rapid Fabrication
of Large Aspheric Optics, Paper presented at the
Southwest Conference on Optics, Albuquerque,
NM.
Jones, R.A., Rupp, W.J., 1990, Rapid Optical
Fabricationwith CCOS, Proc. SPIE, 1333: 34-43.
59 1
 Kaczmarek, J., 1976, Principles of Machining by
Cutting, Abrasion, and Erosion, Peter Peregrinus,
Stevenage, Herts, U.K.
Kanno, Y., Suzuki, K., Kuwahara, Y., 1983, NH3
Formation Caused by the presence of H20 in the
Wet Grinding of Silicon Nitride Powder, Yog Yo-K
yokai-Shi, 91: 386.
Kapsa, P., Kanemura, Y., 1988, Sliding Damage on
Hot-Pressed and Sintered Silicon Nitride Caused by
a Diamond Tip Under Controlled Humidity, Wear,
127: 65-83.
Karaki-Doy, T., 1993, Study on Mechanism of
Mechano-chemical Polishing and its Application -
An Instance of Silicon Wafers, J. of Saitama Univ.,
Faculty of Education (Mathematical and Natural
Science), 42: 33-47.
(83) Kasai, T., Matumoto, F., Kobayashi, A., 1988,
Newly Developed Fully Automatic Polishing
Machines for Obtainable Super-Smooth Surfaces of
Compound Semiconductor Wafers, Annals of CIRP,
37/1: 537-540.
Kasai, T., Horio, K., Karaki-Doy, T., Kobayashi, A.,
1990, Improvement of Conventional Polishing
Conditions for Obtaining Super Smooth Surfaces of
Glass and Metal Works, Annals of CIRP 39/1: 321-
324.
Kashiwagura, N., Harada, J., Ogino, M., 1983,
Characterization of Mechanochemically Polished
(111) Surface of Silicon by Diffuse X-Ray
Scattering, J. Appl. Phys., 54: 2706-2710.
Kawata, K., Tani, Y., 1993, Development of High-
Concentration Lapping Discs with Low Bonding
Strength and Their Application to Mirror Finishing of
Brittle Materials, JSME Int. Journal, Series C, 36:
264-270.
Kikuchi, M., Takahashi, Y., Suga, T., Suzuki, S.,
and Yasunaga, N., 1990, Mechanochemical
Polishing of Silicon Carbide, Proc. Fall Meeting of
JSPE, 327-328 (in Japanese).
Kikuchi, M., Takahaski, Y., Suga, T., Suzuki, S.,
Bando, Y., 1992, Mechano-Chemical Polishing of
Silicon Carbide Single Crystal with Chromium (111)
Oxide Abrasive, J. Amer. Ceram. SOC.,75: 189-
195.
Kinoshita, M., Murayama, T., Hoshina, N.,
Kobayashi, A., Shimizu, R., Ikuta, T., 1976, The
Surface Damaged Layer Study of Mn-Zn Single
Crystal Ferrites Using Magnetic Domain
Observation Technique, Annals of CIRP, 291: 449-
454.
Komanduri, R., 1971, Some Aspects of Machining
with Negative Rake Tools Simulating Grinding, Int.
J. Mach. Tool. Des. Res, 11: 223-233.
Komanduri, R., Lange, J., Wickstead, J., and
Krasinski. J., 1996a, Ceramic Bearing Technology
Program, Final Report (WL-TR-96-4020) Wright
Laboratory, OH.
(92) Komanduri, R., 1996, On Material Removal
Mechanisms in Finishing of Advanced Ceramics and
Glasses, Annals of CIRP 45/1: 509-514.
(93) Komanduri, R., Umehara, N., Raghunandan, M.,
1996b, On the Possibility of Chemo-Mechanical
Polishing of Silicon Ntiride, ASME, J. of Tribology,
1 18: 721-727.
(94) Komanduri, R., Lucca, D.A., 1997, Technological
Advances in Fine Abrasive Processes, MAE
592
Research Report, Oklahoma State University,
Stillwater, OK.
(95) Komanduri, R., Chandrasekaran, N., Raff, L.M.,
1997, Some Aspects of Machining with Negative
Rake Tools Simulating Grinding: An MD Simulation
Approach, to be published.
(96) Konovalov, E.G., Sulev, G.S., 1967, Finishing
Machining of part by Ferromagnetic Powder in
Magnetic Field. Naukaitechnika, (in Russian).
(97) Konovalov, E.G., Sakulevich, F.J., 1974, Principles
of Electro-Ferromagnetic Machining,
Naukaitechnika, (in Russian).
(98) Kunz, R.R., Clark, H.R., Nitshin, P.M., Rothschild,
M., Ahern, B.S., 1996, High Resolution Studies of
Crystalline Damage Induced by Lapping and Single-
Point Diamond Machining of Si(lOO), J. Mater. Res.,
1 1 : 1228-1237.
(99) Laczik, Z., Booker, G.R., Mowbray, A., 1996a,
Assessment of Residual Subsurface Polishing
Damage in InP Wafers by Photolulminescence,
Journal of Crystal Growth, 158: 37-42.
(100) Laczik, Z., Booker, G.R., Mowbray, A., 1996b,
Residual Polishing Damage and Surface Quality of
Commercial 1nP Wafers: A Scanning PL Study,
Materials Science and Engineering, 842: 217-224.
(101) Lange, F.F., 1974, Strong, High Temperature
Ceramics, Ann. Rev. of Mater. Sci., 4: 671-696.
(1 02) Lawn, B., Wilshaw, R., 1975, Indentation Fracture:
Principles and Applications, J of Mat. Sci. 10: 1049-
1081.
(103) Lawn, B.R., Evans, A.G., 1977, A Model for Crack
Initiation in Elastic/Plastic Indentation Fields, J of
Mat. Sci. 12: 2195-2199.
(104) Lawn, B.R., Evans, A.G., Marshall, D.B., 1980,
Elastic-Plastic lndentaion Damage in Ceramics: The
MediadRadial Crack System, J. Am. Cer. SOC., 63:
574-581.
(105) Leatherman, G.L., Nathan Katz, R., 1989,
Structural Ceramics: Processing and Properties, in
Superalloys, Super Composites, and Super
Ceramics, Academic Press Inc. , New York, 671-
696.
(106) Leistner, A.J., 1993, Teflon Laps: Some New
Developments and Results, Applied Optics, 32:
3416-3424.
(107) Liu, C.W., Dai. B.T., Tseng. W.T., and Yeh, C.F.,
1996, Modeling of the Wear Mechanism During
Chemical Mechanical Polishing, J. of Electrochem.
SOC.143: 716-721.
(108) Lonardo, P.M., Trumpold, H., De Chiffre, L., 1996,
Progress in 3D Surface Microtopography
Characterization, Annals of CIRP, 45/2: 589-598.
(109) Lucca, D.A., Rhorer, R.L., Komanduri, R., 1991,
Energy Dissipation in the Ultraprecision Machining
of Copper, Annals of CIRP, 40/1: 69-72.
(110) Lucca, D.A., Rhorer, R.L., Maggiore, C.J., Seo,
Y.W., 1995, Assessment of Subsurface Damage in
Ultraprecisoin Machined CdS by Ion Channeling,
Annals of CIRP, 44/1: 513-516.
(111) Lucca, D.A., Maggiore, C.J., 1997, Subsurface
Lattice Disorder in Polished Il-VI Semiconductors,
Annals of CIRP, 46/1: 485-488.
(112) Maehata, H., Kamada, H., 1987, Electrolytic-
Abrasive Mirror Finishing, Precision Engineering, 9:
31-43.

(113) Makedonski, B.G., Kotshemidov, A.D., 1974.
Schleifen im Magnetfeld, Fertigungstechnik und
Betrieb, 24: 230 - 235.
(114) Malkin, S., 1989, Grinding Technology, Ellis
Harwood Ltd, Chichester, West Sussex, U. K.
(115) Malkin, S., Hwang, T.W., 1996, Grinding
Mechanisms for Ceramics, Annals of CIRP, 492:
569-580.
(116) Marsh, D.M., 1964, Plastic Flow in Glass, Proc.
Royal Society, 279A: 420-435.
(117) Marshall, D.B., Lawn, B.R., Evans, A.G., 1982,
Elastic Plastic Indentation Damage in Ceramics:The
Lateral Crack System, J of Am. Cer. SOC.,65: 561-
566.
(118) Marshall, D.B. and Lawn, B.R., 1986, Indentationof
Brittle Materials, Microindentation Techniques in
Materials Science and Engineering, ASTM STP 889,
P.J. BIau and B.R. Lawn, Eds., 26-46.
(119) Matsuo, T., Touge, M., Yamada, H., 1997, High-
Precision Surface Grinding of Ceramics with
Superfine Grain Diamond Cup Wheels, Annals of
CIRP, 46/1: 249-252.
(120) McKeown, P.A., 1987, The Role of Precision
Engineering in Manufacturing of the Future, Annals
of CIRP, 3612: 495-501.
(121) McKeown, P.A., Carlisle, K., Shore, P., Read,
R.F.J., 1990, Ultraprecision, Hlgh Stiffness, CNC
Grinding Machines for Ductile Mode Grinding of
Brittle Materials, Infrared Technology and
Applications, SPIE: 301-313.
(122) Meinel, A.B., 1982, Cost Relationship for Non-
conventional Telescope Structural Configurations,
J. Opt. SOC.Am., 72: 14-20.
123) Miyashita, M., 1988, Review on Ultraprecision
Grinding Technology, Ultraprecision in
Manufacturing Engineering, Springer-Verlag, Berlin,
Germany, 41-57.
124) Mori, Y., Ikawa, N., Okuda, T., Yamagata, K., 1976,
Numerically Controlled Elastic Emission Machining,
Technology Reports of the Osaka University, 26:
283-294.
(125) Mori, Y., Ikawa, N., Sugiyama, K., 1978, Elastic
-
Emission Machining Stress Field and Fracture
Mechanism, Technology Reports of the Osaka
University, 28: 525-534.
(126) Mori, Y., Ikawa, N., Okuda, T., Sugiyama, K.,
Yamauchi, K., 1983, Numerically Controlled Elastic
Emission Machining - Motion Analysis of Fluid and
Distribution of Film Thickness, Japan Soc.of
Mech.Engg., 49.
(127) Mori, Y., Ikawa, N., Sugiyama, K., Okuda, T.,
Yamauchi, K., 1985a, Elastic Emission Machining -
2nd Report - Stress Field and Feasibility of
Introduction and Activation of Lattice Defects,
Japan Society of Precision Engineers, 51: 1187-
1194.
(128) Mori, Y., Okuda, T., Sugiyama, K., Yamuchi, K.,
1985b, Numerically Controlled Elastic Emission
Machining - Consideration of Machining Process by
Motion Analysis of Powder Particles in Fluid, Japan
Society of Precision Engineering, 51: 1033-1039.
(129) Mori, Y., Yamauchi, K., Endo, K., 1987a, Elastic
Emission Machining as Evaluated by Atomistic
Techniques, Proc. of the 6th Int. Conf. on Prod.
Eng., Osaka, Japan, 58-63.
(130) Mori, Y., Yamauchi, K., 1987b, Elastic Emission
Machining, Precision Engineering 9: 123-128.
Keynote Papers
(131) Mori, Y., Yamauchi, K., Endo, K., 1988, Mechanism
of Atomic Removal in Elastic Emission Machining,
Precision Engineering, 10: 24-28.
(132) Nakagawa, T., Suzuki, K., Uematsu, T., 1986,
Highly Efficient Grinding of Ceramics and Hard
Metals on Grinding Center, Annals of CIRP, 3511:
205-210.
(133) Nakagawa, T., Suzuki, K., Uematsu, P., Kimura, M.,
1988, Develpment of a New Turning Center for
Grinding Ceramic Materials, Annals of CIRP,
37/1:319-322.
(134) Nakao, M., Hatamura, Y., 1996, Development of an
Intelligent Face Grinding Machine to Fabricate
Ultraflat Surfaces on Thin, Brittle Substrates,
Annals of CIRP, 4511: 397-401.
(135) Nakasuji, T., Kodera, S., Hara, S., Matsunaga, H.,
Ikawa, N., Shimada, S., 1990, Diamond Turning of
Brittle Materials for Optical Components, Annals of
CIRP, 39/1: 89-92.
(136) Namba, Y., Tsuwa, H., 1977, Ultra-Fine Finishing of
Sapphire Single Crystal, Annals of CIRP, 26/1: 325-
329.
(137) Namba, Y., Tsuwa, H., 1978, Mechanism and Some
Applications of Ultra-Fine Finishing, Annals of CIRP,
2711: 511-516.
(138) Namba, Y., Tsuwa, H., 1980a, Float Polishing of Mn-
Zn Ferrites and Characteristics of Finished
Surfaces, Paper presented at the 4th International
Conference on Production Engineering, Tokyo,
1017-1022.
(139) Namba, Y., Tsuwa, H., 1980b, Surface Properties of
Polished Stainless Steel, Annals of CIRP, 2911:
409-412.
(140) Namba, Y., Tsuwa, H., 1980c, Ultrafine Finishing of
Ceramics and Metals by Float Polishing, Laser
Induced Damage in Optical Materials, NBS Special
Publication No. 620, Boulder, CO, 171.
(141) Namba, Y., 1982, Float Polishing and Defects of
Fused Silica, in Laser Induced Damage in Optical
Materials, Gaithersburg, MD, U.S. Dept. of
Commerce, National Bureau of Standards, 130.
(142) Namba, Y., Yonezawa, T., 1982, Polishing Method,
U. S. Patent No. 4358295.
(143) Namba, Y., Tsuwa, H., Wada, R., 1987, Ultra-
Precision Float Polishing Machine, Annals of CIRP,
3611: 21 1-214.
(144) Namba, Y., Wada, R., Unno, K., Tsuboi, A., 1989,
Ultra-Precision Surface Grinder Having a Glass-
Ceramic Spindle of Zero-Thermal Expansion, Annals
Of CIRP, 38/1: 331-334.
145) Namba, Y., Yamada, Y., Tsuboi, A., Unno, K.,
Nakao, H., 1992, Surface Structure of Mn-Zn Ferrite
Single Crystals Ground by an Ultraprecision
Surface Grinder with Various Diamond Wheels,
Annals of CIRP, 41/1: 347-351.
146) Namba, Y., Abe, M., 1993, Ultraprecision Grinding
of Optical Glasses to Produce Super-smooth
Surfaces, Annals of CIRP, 42/1: 417-420.
(147) Nanz, G., Camilletti, L.E., 1995, Modeling of
Chemical-Mechanical Polishing: A Review, IEEE
Trans. on Semiconductor Manufacturing, 8: 382-
389.
(148) Nelson, J., 1981, Non-axisymmetric Mirrors
Produced by Stressed Mirror Polishing, Precision
Engineering, 3: 7-10.
593
149) Newton, Sir Isaac, 1952, Opticks, Dover
Publications, Inc., New York, based on the 4th
Edition, London published in 1730
150) Nogawa, H., 1988, Ceramic Processing - State of
the Art of R & D in Japan, ASM Int., Metals Park, OH
151) Nouruzi-Khorasani, A., Lunn, M.A., Jones, I.P.,
Dobson, P.S., 1990, Surface Damage of CdTe by
Mechanical Polishing Investigated by Cross-
Sectional TEM, J. Cryst. Growth, 102: 1069-1073.
(152) Ohmori, H., Nakagawa, T., 1990, Mirror Surface
Grinding of Silicon Wafers with Electrolytic In-
process Dressing, Annals of CIRP, 39/1: 329-332.
(153) Ohmori, H., Takahashi, I., 1994, Efficient Grinding
Technique Utilizing Electrolytic In-Process
Dressing for Precision Machining of Hard Materials,
Advancement of Intelligent Production, 315-320.
(154) Ohmori, H., Nakagawa, T., 1995, Analysis of Mirror
Surface Generation of Hard and Brittle Materials by
ELlD Grinding with Superfine Grain Metallic Bond
Wheels, Annals of CIRP, 44/1: 287-290.
(155) Onchi, Y., Matsumori, N., Ikawa, N., Shimada, S.,
1995, Porous Fine CBN Stones for High Removal
Rate Superfinishing, Annals of CIRP, 44/1: 291-
294.
(156) Palmquist, S . , 1962, Archiv. fur das
Eisenhuettenwesen,33: 629-633.
(157) Parks, R.E., Lam, P., Kuhn, W., 1985, The Large
Optical Generator: A Progress Report , Proc of the
SPIE, 542.
(158) Parks, R.E., Evans, C.J., 1994, Rapid Post-
Polishing of Diamond-Turned Optics, Precision
Engineering, 16: 223-227.
(159) Parks, R.E., Evans, C.J., Roderick, D.J., Dagata,
J., 1997, Applications of the Rapidly Renewable
Lap, Proc. SPIE, 31-34 ( in press).
(160) Preston, F.W., 1927, The Theory and Design of
Plate Glass Polishing Machines, J. SOC. Glass
Tech., 11: 214-256.
(161) Puthanangady, T. K., Malkin, S., 1995,
Experimental Investigation of the Superfinishing
Process, Wear, 185: 173-182.
(162) Puttick, K.E., Shahid, M.A., Hosseini, M.M., 1979,
Size Effects in Abrasion of Brittle Materials, J.
Phys. D: App. Phys., 12: 195-202.
(163) Puttick, K.E., Hosseini, M.M.,1980. Fracture by a
Pointed Indenter on Near (1 11) Silicon, J. Phys. D:
App. Phys., 13: 875-880.
(164) Puttick, K.E., Rudman, M.R., Smith, K.J., Frank,
A., Lindsay, K., 1989, Single Point Diamond
Machining of Glass, Proc. Roy. SOC.A, 426: 19-30.
(165) Puttick, K.E., Whitmore, L.C., Chao, C.L., Gee,
A.E., 1994, Transmission Electron Microscopy of
NanomachinedSilicon Crystals, Phil. Mag A, 69: 91-
103.
(166) Raghunandan, M., Umehara, N., Nbori-khajavi, N.,
Komanduri, R. 1997, Magnetic Float Polishing of
Advanced Ceramics, Trans ASME, J. of Engg. for
Ind. (in print).
(167) Raghunandan, M., Komanduri, R., 1997, Finishing
of Silicon Nitride Balls for High-speed Bearing
Applications,, Trans ASME, J. of Engg. for Ind.
(accepted for publication).
(168) Rayleigh, Lord ,1901, Polish, from the Scientific
Papers by Lord Rayleigh (John William Strutt),
Dover Publications Inc., New York.
594
(169) Rosensweig, R.E., 1985, Ferrohydrodynamics,
Cambridge University Press, Cambridge, U.K.
(170) Runnels, S.R., Renteln, P., 1993, Modeling The
Effect of Polish Pad Deformation on Wafer Surface
Stress Distribution During Chemical Mechanical
Polishing, Dielectric Science and Technology.
(171) Runnels, S.R., Eyman, L.M., 1994, Tribology
Analysis of Chemical Mechanical Polishing, J.
Electrochem. SOC,141: 1698-1701.
(172) Runnels, S.R.,1994, Feature Scale Fluid Based
Erosion Modeling for Chemical Mechanical
Polishing, J. Electrochem. SOC., 141: 1900-1904.
(173) Runnels, S.R., Olavson, T., 1995, Optimizing Wafer
Polishing Through Phenomenological Modeling, J.
Electrochem. SOC.,142: 2032-2036.
(174) Saito, T.T., 1976, Diamond Turning of Optics, Proc.
of the SPIE,
(175) Saito, T., 1978a, Precision Machining of Optics,
Proc. of the SPIE, 159.
(176) Saito, T.T., 1978b, Diamond Turning of Optics: the
Past, the Present, and the Exciting Future, Optical
Engg., 17: 570-571.
(177) Sakata, O., Nikulin, A., Hashizume, H., 1993, X-
Ray Evaluation of Microroughness of
Mechanochemically Polished Silicon Surfaces, Jpn.
J. Appl. Phys., 32: L616-L619. .
(178) Sakulevich, F.J., Kozuro, L.M., 1977, Magneto-
Abrasive Machining of Fine Parts, Vyssaja Skola,
Minsk (in Russian).
(179) Sakulevich, F.J., Kozuro, L.M., 1978, Magneto-
Abrasive Machining, Naukaitechnika, Minsk (in
Russian).
(180) Samuels, L.E., 1971, Metallographic Polishing by
Mechanical Methods. (2nd ed.), American Elsevier,
New York.
(181) Semba, T., Tani, Y., Sato, H., 1996, Development of
Melamine-Bonded Diamond Wheels with High
Porosity for Smooth and Mirror Finishing of Die
Materials, Annals of CIRP, 45/1: 315-318.
(182) Shaw, M.C., 1995, Precision Finishing, Annals of
CIRP, 44/1: 343-348.
(183) Shaw, M.C., 1996, Principles of Abrasive
Processing, Oxford University Press, Oxford, U.K.
(184) Shibata, T., Ono, A., Kurihara, K., Makino. E.,
Ideda, M., 1994, Cross-Section Transmission
Electron Microscope Observations of Diamond
Turned Single-Crystal Si Surfaces, Appl. Phys.
Lett., 65: 2553-2555.
(185) Shibata, T., Fujii, S., Makino, E., Ideda, M.,1996,
Ductile-Regime Turning Mechanism of Single-
Crystal Silicon, Precision Engineering, 18: 129-137.
(186) Shinmura, T., Takazawa, K., Hatano, E.,
Matsunaga, M., 1990, Study on Magnetic Abrasive
Finishing, Annals of CIRP, 3911: 325-328.
(187) Shinno, H., Hashizume, H., Ito, Y., Sato, C., 1992,
Structural Configuration and Performances of
Machining Environment-Controlled Ultra-Precision
Diamond Turning Machine "Capsule", Annals of
CIRP, 41/1: 425-428.
(188) Singhal, S.C., 1975, Effect of Water Vapor on the
Oxidation of Hot-Pressed Silicon Ntiride and Silicon
Carbide, J. Am. Ceram. SOC.,58: 17.
(189) Soares, S.F., Baselt, D.R., Black, J.P., Jungling,
K.C., Stowell, W.K., 1994, Float-Polishing Process
and Analysis of Float-Polished Quartz, Applied
Optics, 33: 89-95.

(190) Stowers, I.F., Komanduri, R., Baird, E.D., 1988,
Review of Precision Surface Generating Processes
and Their Potential Applications to the Fabrication
of Large Optical Components, SPIE's 32 Annual
International Technical Symposium on Optical and
Optoelectrical Applied Science and Engineering,
San Diego, CA.
191) Strickler, R.O.S., Booker, G.R., 1963, Surface
Damage on Abraded Silicon Specimens, Phil. Mag.,
Ser. B, 1: 859-876.
192) Suga, T., Suzuki, S., Miyazawa, K., 1989
Mechanochemical Polishing of Sintered Silicon
Nitride, JSPE, 55: 2247-2253 (in Japanese).
193) Sugita, T., Ueda, K., Kanemura, Y. 1984, Material
Removal Mechanism of Silicon Nitride During
Rubbing in Water, Wear, 9 7 1-8.
(194) Suh, N.P., 1986, Tribophysics, Prentice-Hall,
Englewood Cliffs, NJ.
(195) Suzuki, K., Tochinai, H., Uematsu, T., Mishiro, S.,
1993, A New Grinding Method for Ceramics Using a
Biaxially Vibrated Nonrotational Ultrasonic Tool,
Annals of CIRP, 42/1: 417-420.
(196) Suzuki, K., Ogura, M., Uematsu, T., Imanaka, O.,
1994, Precise Polishing of 3-0 Surface with an
Elastic Conductive Tool by Utilizing Electrophoresis
Phenomenon, Proc. of the 7th Int Conf on
Production/Precision Engineering, Chiba, Japan,
527-532.
(197) Suzuki, H., Murakami, S., 1995, An Ultraprecision
Grinding Machine for Non-Axisymmetric Aspheric
Mirrors, Nanotechnology, 6: 152-157.
(198) Syn, C.K., Taylor, J.S., Donaldson, R.R., Shimada,
S., 1988, Ductile Brittle Transition of Cutting
Behavior in Diamond Turning of Single Crystal
Silicon, Proc. Jap. SOC. Prec. Eng., Kawasaki,
Japan.
(199) Tabor, D., 1970, The Hardness of Soids, Proc. of
the Institute of Physics, F. Physics in Technology,
1: 145-179.
(200) Tabor, D., 1986, Indentation Hardness and Its
Measurement: Some Cautionary Comments, in
Micro-indentation Techniques in Materials Science
and Engineering, ASTM STP 889 Eds. P.J. Blau and
B.R. Lawn, 129-159.
(201) Takazawa, K., Shinmura, T., Hatano, E., 1983,
Development of Magnetic Abrasive Finishing and Its
Equipment, MR 83-678. Proc. of the SME'S 12th
Deburring and Surface Conditioning Conference,
Orlando.
(202) Takazawa, K., Shinmura, T., Hatano, E., 1985,
Advanced Development of Magnetic Abrasive
Finishing and Its Equipment, Proc. of the SME'S
Deburring and Surface Conditioning Conference '85,
Orlando, 30-46.
(203) Takeuchi, Y, Sawada, K., Sata, T., 1996,
Ultraprecision 3-D Micromachiningof Glass, Annals
Of CIRP, 45/11401-404.
(204) Tani, Y., Kawata, K., 1984, Development of High-
Efficient Fine Finishing Process Using Magnetic
Fluid, Annals of CIRP, 33/1: 217-220.
(205) Tani, Y., Kawata, K., 1986, Development of a High-
Efficiency Finishing Process Using 'Liquid-Bonded
Wheel', Annals of CIRP, 35/1: 215-218.
(206) Taniguchi, N., 1974, On the Basic Concept of Nano-
Technology, Proc of the ICPE, Tokyo,Japan, 2: 18-
23.
Keynote Papers
(207) Taniguchi, N., 1983, Current Status in, and Future
Trends of, Ultraprecision Machining and Ultrafine
Materials Processing, Annals of CIRP, 32/2: 573-
582.
(208) Taniguchi, N., 1989, Energy-Beam Processing of
Materials, Clarendon Press, Oxford, U. K.
(209) Taniguchi, N., 1993, The State of the Art of
Nanotechnology for Processing of Ultraprecision
and Ultrafine Products, 1993 ASPE Distinguished
Lecture, Precision Engineering, 16: 5-24.
(210) Tomita, Y., Eda, H., 1996, A Study of the Ultra
Precision Grinding Process on a Magnetic Disk
Substrate-Development of New Bonding Materials
for Fixed Abrasives of Grinding Stone, Wear, 195:
74-80.
(211) Tornizawa, H., Fischer, T.E., 1986, Friction and
Wear of Silicon Nitride and Silicon Carbide in Water:
Hydrodynamic Lubrication at Low Sliding Speeds
Obtained by Tribochemical Wear,Trans ASLE 30:
41-46.
(212) Tonshoff, H.K., Brinksmeier, E., 1980,
Determination of the Mechanical and Thermal
Influences on Machined Surfaces by
Microhardness and Residual Stress Analysis,
Annals of CIRP, 29/2: 519-530.
(213) Tonshoff, H.K., Brinksmeier, E., Hetz, F., 1987,
Detection of Microcracks, Annals of CIRP, 36/2:
545-552.
(214) Tonshoff, H.K., v. Schmieden, W., Inasaki, I.,
Kbnig, W., Spur, G., 1990, Abrasive Machining of
Silicon, Annals of CIRP 39/2: 621-635.
(215) Tonshoff, H.K., Hartmann, M., Przywara, R., Klein,
M., 1994, Defects i n Silicon-Wafers:
Characterisation and Prevention, Advancement of
Intelligent Production, 627-632.
(216) Touge, M., Matsuo, T., 1996, Removal Rate and
Surface Rougness in High-PrecisionLapping of Mn-
Zn Ferrite, Annals of CIRP, 45/1: 307-310.
(217) Tseng, W.T., and Wang, Y.L., 1997, Re-
examination of Pressure and Speed Dependencies
of Removal Rate During Chemical-Mechanical
Polishing Processes, J. of the Electrochemical
Society, 144: L1517.
(218) Tsuwa, H., Aketa, Y., 1974, Elastic Emission
Machining as an Ultra Fine Finishing Method, Proc.
of the Int. Conf. on Production Engineering
Research, Tokyo, Japan, Part 11, 33-38.
(219) Tsuwa, H., Ikawa, N., Mori, Y., Sugiyama, K., 1979,
Numerically Controlled Elastic Emission Machining,
Annals of CIRP, 28/1: 193-197.
(220) Ueda, K., Amano, A., Ogawa, K., Takamatsu, H.,
Sakuta, S., Murai, S., 1991, Machining High-
Precision Mirrors Using Newly Developed CNC
Machine, Annals of CIRP, 40/1: 555-558.
(221) Uematsu, T., Suzuki, K., Wu, M.H., Suzuki, K.,
Imanaka, O., 1993, Efficient Mechanochemical
Polishing for Silicon Nitride Ceramics, Machining of
Advanced Materials, NlST Special Publication No.
847 (edited by S. Jahanmir), 409-413.
(222) .Umehara, N., 1988, Study on Magnetic Fluid
Grinding, Ph.D. Thesis, Tohoku University, Sendai,
Japan, (in Japanese).
(223) Umehara, N., Kato, K., 1990, Principles of Magnetic
Fluid Grinding of Ceramic Balls, Applied
Electromagnetics in Materials, 1: 37-43.
595
(224) Umehara, N., 1994, Magnetic Fluid Grinding-A New
Technique for Finishing Advanced Ceramics,
Annals of CIRP, 43/1: 185-188.
(225) Umehara, N., Kobayashi, T., Kato, K., 1995,
Internal Polishing of Tube with Magnetic Fluid
Grinding-Part 1, Fundamental Polishing Properties
with Taper-Type Tools, Journal of Magnetism and
Magnetic Materials, 149: 185-187.
(226) Urnehara, N., Komanduri, R., 1996, Magnetic Fluid
Grinding of HIP-Si3N4 Rollers, Wear, 192: 85-93.
(227) Van Vlack, L.H., 1982, Materials for Engineering -
Concepts and Applications, Addison Wesley, MA.
(228) Venkatesh, V.C., Tijnshoff, H.K., Nakagawa, T.,
Marinescu, I.D., 1995, Observations on Polishing
and Ultraprecision Machining of Semiconductor
Substrate Materials, Annals of CIRP, 44/2: 61 1-
61 8.
(229) Venkatesh, V.C., Fang, F., Chee, W.K., 1997, On
Mirror Surfaces Obtained With and Without
Polishing, Annals of CIRP, 46/1: 505-508.
(230) Vitali, G., Pizzuto, C., Zollo, G., Lucca, D.A., De
Luca, L., 1996, Low-Power Pulsed-Laser Annealing
of the Damaged Surface Layer of
Chemomechanically Polished CdS, Jpn. J. Appl.
Phys., 35: L1558-LI561.
(231) Vora, H., Orent, T.W., Stokes, R.J., 1982,
Mechano-Chemical Polishing of Silicon Nitride., J.
Amer. Ceram. SOC.,65: C14O-Cl41.
(232) Vora, H., Stokes, R.J., 1983, Study of Mechano-
Chemical Machining of Ceramics and the Effect on
Thin Film Behavior, Report to the Office of Naval
Research (Final Technical Rept. No. NO0014-804-
0437-2).
(233) Vorburger, T.V., 1987, Measurements of
Roughness of Very Smooth Surfaces, Annals of
CIRP, 36/2: 503-509.
(234) Wang, Z., Zhou, Z., Pang, T., Wang, H., Pang, N.,
Shunzeng, Z., 1994, A New Liquid Used in
Ultraprecision Polishing of InP Wafer, Advancement
of Intelligent Production, 539-543.
(235) Warnock, J., 1991, A Two Dimensional Process
Model for Chemomechanical Polish Planarization, J
of Electrochem. SOC.,138: 2398-2402.
(236) Warnock, J., 1994, J. of Electrochem. SOC.,
141 :1900.
(237) Watanabe, J., Suzuki, J., 1981, High Precision
Polishing of Semiconductor Materials Using
Hydrodynamic Principle, Annals of CIRP, 30/1: 91-
95.
(238) Whitehouse, D.J., 1994, Handbook of Surface
Metrology, Inst. of Physics, Bristol, UK.
(239) Wills-Moren, W.J., Modjarrad, H., Read, R.F.J.,
1982, Some Aspects of the Design and
Development of a Large High Precision CNC
Diamond Turning Machine, Annals of CIRP, 31/1:
409-414.
(240) Wills-Moren, W.J., 1988, Experiences in the
Precision Machining of Brittle Materials,
Ultraprecision in Manufacturing Engineering,
Springer-Verlag, Berlin, Germany, 3-21.
(241) Wills-Moren, W.J., Wison, T., 1989, The Design and
Manufacture of a Large CNC Grinding Machine for
Off-Axis Mirror Segments, Annals of CIRP, 38/1:
529-532.
596
(242) Wyatt, O.H., Dew-Hughes, D., 1974, Metals,
Ceramics, and Polymers, Cambridge University
Press, Cambridge, U.K.
(243) Xie, Y., Bhushan, B., 1996, Effects of Particle
Size, Polishing Pad and Contact Pressure in Free
Abrasive Polishing, Wear, 200: 281 -295.
(244) Yamaguchi, K., Horaguchi, I., 1995, Development
of Directionally Aligned SIC Whisker Wheel,
Precision Engineering, 17: 5-9.
(245) Yamoaka, H., Uruga, T., Arakawa, E., Matsuoka,
M., Ogasaka, Y., Yamashita, K., Ohtomo, N., 1994,
Development and Surface Evaluation of Large Sic
X-Ray Mirrors for High-Brilliance Synchrotron
Radiation, Jpn. J. Appl. Phys., 33: 671 8-6726.
(246) Yanagi, K., Tsukada, Y., Yamada, R., 1992,
Surface Topography Assessment of High-Precision
Coated Abrasive Tape and Estimation of Its
Machining Performance, Annals of CIRP, 41/1: 377-
380.
(247) Yasseen, A.A., Mourlas, N.J., Mehregany, M.,
1997, Chemical-Mechanical Polishing for
Polysilicon Surface Micromachining, J.
Electrochem. SOC.,144: 237-242.
(248) Yasunaga, N., Obara, A., Imanaka, O., 1974,
Mechanochemical.Polishing of Single Crystals with
Soft Powers, Paper presented at the Int. Conf. on
Production Engg., Tokyo.
(249) Yasunaga, N., Obara, A, Tarumi, N., 1977, Study of
Mechanochemical Effect on Wear and Its
Application to Surface Finishing, Res. Electrotech.
Lab, 776: 50-134 (in Japanese).
(250) Yasunaga, N., Obara, A., Imanaka, O., 1978, Study
of Mechanochemical Effect on Wear and Its
Application to Surface Finishing, J. of JSPE, 44: 77-
83 (in Japanese).
(251) Yasunaga, N., Tarumi, N., Obara, A., Imanaka, O.,
1979, Mechanism and Application of the Mechano-
Chemical Polishing Method Using Soft Powder, in B.
J. Hockey & R. W. Rice (eds.), Science of Ceramic
Machining and Surface Finishing II, National Bureau
of Standards, Gaithersburg, MD, 171-183.
(252) Yoshikawa, H., 1967, Brittle-Ductile Behavior of
Crystal Surface in Finishing, J. of JSPE, 35: 662-
667 (in Japanese).
(253) Yoshioka, J., Koizumi, K., Shimizu, M., Yoshikawa,
H., Miyashita, M., Kania, A., 1982, Surface
Grinding with a Newly Developed Ultra Precision
Grinding Machine, SME Tech. Paper MR82-939.
(254) Yoshioka, J., Hashimoto, F., Miyashita, M., Daitoh,
M., 1984, High-Precision Centerless Grinding of
Glass as Preceding Operation to Polishing -
Dressing Conditions and Grinding Accuracy, SME
Tech. Paper MR84-542.
(255) Yoshioka, J., Hashimoto, F., Miyashita, M., Kanai,
A., 1985, Ultraprecision Grinding Technology for
Brittle Materials: Application to Surface and
Centerless Grinding Processes, ASME PED-16.
(256) Yu, T.K., Yu, C.C., Orlowski, M., 1993, A Statistical
Polishing Pad Model for Chemical Mechanical
Polishing, Paper presented at the IEDM93.
(257) Zhong, Z., 1992, New Grinding Method for Aspheric
Mirrors with Large Curvature Radii, Annals of CIRP,
41/1: 335-338.
(258) Zhong, Z., Venkatesh, V. C., 1994, Generation of
Parabolic and Toroidal Surfaces on Silicon and
Silicon-Based Compounds Using Diamond Cup
Grinding Wheels, Annals of CIRP, 43/1: 323-326.