You are on page 1of 12

~ .~/" ~:+~.

--

I,~. +

WEAR
ELSEVIER Wear 215 ~ IotAS) 267-278

On the finishing of Si3N4balls for bearing applications


Ming Jiang, R. Komanduri *
Me~'hunil'al uml Ae,rcrsp.~'e l~:n.g,meering. Okk~htumt .Vrule Univt,~sitv. Slillw~tter. OK 7407,~. USA
Received 8 January lt./~)7: accepted :g Sepleml'~r 1~)97

Abstract

The conventional meth~al of prcalucing of Si,N+, balls fiw bearing applications by grinding and lapping using diarmmd abrasive at low
speeds ( < it few hundred rpm) and higher hr,als ~several tens of N/ball ) is generally an expensive and time-consuming (q~cration ( ~verai
weeks ). It also leads to the formation of .~ralchcs. micr~:racks, and pits on the finished balls resulting from large radial and circumferemial
cracks and diskglgemem of grains. Since failure of ceramics initiates from such defects, the reliability of Si4Nz balls in service is of taime
concern. This paper deals with an ahemate technology lot finishing Si ~N.,balls for hybrid bearing applications using magnetic float polishing
( M FP ) process that overcomes some of the.~ limitations. A methtRlology fiw finishing of HtP'ed Si +N4 halls from the as-received comlition
by MFP is presented. It involves Ihe mechanical rermwal of material initially using harder a~asives with respect to the workmaterial (of
differenl materials of progressively lower harducsses and finer grain sizes) lol[owed by final cbena~mechanical polishing {CMPl using
preferably a softer abrasive filr obtaining superior :inish wilh minimal surface or subsurface defects, such as .~ratcbes. micro~racks. ~r pins
on the Si ,N~ balls. High material removal rates ( I ~ m / m i n ) with minimal subsurface damage is obtained with harder a~asives, such as B+C
ar SiC ( relalive Io Si ,N 4 ) due Io tile use of a flexible support system, small I~dishing loads ( = I N/ball ). and fine abrasives but high polishing
speeds ( compared Io conventional polishing ) by rapid accumulalion of minute aFramnts of material removed by microfracture. Final I~+lishing
of the Si ~N.~balls using a sol~er abrasive+ such as CeO: ( thai chemtb-mchanically rcacl with the SigN4 wt~kmaterial ) results in high quality
Si ~N+ balls of bearing quality with superior surface finish ( R, < 4 nm. R, < l).04 t~m ) and damage-free surf~e. It is fimnd thai CMP is very
effective fiw obtaining excellent surface finish ( R, ~ 4 nm and Rt = 41) nm ) on Si ,N+ ceramic material and CeO: in particular is one of most
suitable material l'~r this application. 1'~)8 Elsevier Science S.A.

Ko',ords: Magnetic Ihml [a~lishing: Finishing of ceramic balls: Silicon nitri&': Chemo-mcchanicat [adishing <CMPg: Cerium t~xide

I. I n t r o d u c t i o n as a polishing medium. In practice, it takes considerable time


( some 12-16 weeks) to finish a batch of ceramic balls. The
A critical factor affecting the performance and reliability long processing time and u.,~ of expensive diamond abrasive
o f ceramics for bearing applications is the quality o f the result in high processing costs, Also. the u ~ of diamond
resulting surface by polishing. It is well known that ceramics abrasive at high loads can result in deep pits. ~ratches, and
are extremely sensitive to surfi+ce detects resulting from microcracks, To minimize the surface damage. "gentle" pol-
grinding and polishing processes owing to their high hardness ishing conditions am required, namely, low level of con-
and inherent brittleness. Since fatigue failure of ceramics is trolled force and abrasives not much harder than the
driven by surface imperfections, it is paramount that the qual- workmalerial. Further, higher removal rates and shorter pol-
ity and finish of the ceramic bearing elements be superior ishing cycles can be obtained at high polishing speeds. This
with minimal defects so that reliability in performance of is accomplished in this investigation by a process k n o w n as
bearings in service can be achieved. This investigation magnetic float polishing. This process was originally devel-
focuses on the methodology fi)r obtaining superior finish on oped by Tani and Kawata { I I and s u h ~ q u e n t l y improved
Si+~N: balls using magnclic float polishing ( MFP} pr~ucess. significantly using the concept o f a float by Umehara and
Conventional polishing of ceramic balls generally uses low Kato 12 I. Umehara 13 I. Childs and Yoon [41. Childs et al.
tx~lishing speeds ( a few hundred rpm ) and d i a m o n d abn~sive 1.5-7 I. and Raghunandan et al. [ 8.9 I.
W h e n polishing advanced ceramics, such as SigN4 balls by
* ('orrespamdiug author. O k l a h t , n a Stale Univernity. Mechanical and
Aerospace l-ng.. 218 Engineering N~+rlh. Slillwatur. OK 74078. USA.
M F P with a harder abrasive, such as B.,C or SiC. high material
Tel.: + I-IIlO-I-4()5-744-5t.gXh fax: + I-(Hn-l-405-744-5720; e-mail: removal rates ( I # m / r a i n ) can he obtained with minimal
ranga(a'nnasler.ceal.okslale.edu subsurface damage due to the u ~ of a flexible support system.

O(M3.16JS/t)8/$ttj.|N) ~t5 It,~J8 Elsevier Science S.A. All rights reseD.cd


PIISnn43- ! 648! 97 ~OO236-6
268 M. Jitmg. R. Kommuluri/ Wear 215 (199812fi7-27,~

small polishing force ( I N/ball). fine abrasives hut high


V ....
polishing speed compared to conventional polishing. Higher
material pemoval results from the rapid accumulation of min-
ute amounts of material by microfracture of ShNa becau.~e of
high polishing speeds and light loads used instead of large
radial and circumferential cracks that results in the formation Jl
i t i Spindle

of pits in conven'.ional poli.~hing. As will be shown, the linal


polishing of the Si3N.~ balls with a softer abrasive, such as Shaft
CeO: results in superior surface finish (R,, < 4 nm, R, < (I.04 - i -,
/~m) with damage-l~e surface due to preferential removal .... G u i d e Ring
of material by chemo-mechanical polishing (CMP). it is
found that CMP is very effective in obtaining excellent sur- ...... Magnetic Raid
and Abrasives
face linish on Si.~N4 workmalerial and that CeO, is one of
R u b b e r Ring
most effective polishing media lot this application.
.... C e r a m i c Ball
-- Float

2. Magnetic float polishing and experimental ~t-up N S N S N S -- Aluminum Base

..... Magnet
The magnetic float polishing technique is based on the
Steel Y o k e
magneto-hydrodynamic behavior of a magnetic fluid that can
(a)
flout non-magnetic float and abrasives suspended in it by a
magnetic lield [ I0]. The forces applied by the abrasive to
the part are extremely small ( I N/hall) and highly control-
lable. Fig. I a and b are a schematic and a photograph of the
magnetic float polishing apparatus fi~r linishing advanced
ceramic balls. A bank of permanent magnets ( Nd-Fe-B ) are
arranged alternate N and S below an aluminum chamber
which is lilled with the required amount of magnetic fluid
and appropriate abrasive ( 5 - IOC;~by volume). The magnetic
fluid is a colloidal dispersion of extremely line ( 100 to 150
A) subdomain I~'rromagnetic panicles, usually magnetite
( Fe,O4 ). in a carrier fluid, such as water or kerosene. It is
made stable against particle agglomeration by coating the
panicles with an appropriate suffactant. In this investigation
a water ba.~ ti:rrofluid is u.~d.
When a magnetic tield is apptied, the magnetic particles in
the magnetic fluid 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 lield. The abrasive grains, the ceramic hall.,,,
and the acrylic float inside the chamber, all being non-mag-
netic materials, are levitated by the magnetic buoyant force,
The drive shaft is fed down to contact with the halls ( 3-point
contact ) and presse.s them d o g n to reach thu desired fiwce t~r
height. The balls are polished by the abrasive grains under
ihe action of the magnetic buoyancy levitational force when I:~g.. I~ ( a l The schematicplot of the magneticfloatladishing (MFP) appa-
the spindle rotates. Damage-freesurface on CeRlmic balls arc mlu-. used t'~rla~lishirlg Si ~N~bails. (b) Ph,~tographof the magnetic Ih~at
expected by the magnetic float polishing technique ~cause i~.~lishmg ! MFP) apparatusused I'-r I~lb,hing Si,N, b;llls.
the magnetic buoyant force ( [ N/hall) is applied via the
flexible float. The function of aco/lic Iloat here b, Io produce 3. Chem~mechanical polishing
more uni form and larger polishing pressure ( the larger brady-
ant Iorcc near the magtaetic po|cs can Ix- transmitted to the Chemo-mechanical action depends on the availability for
l'adi',hing area by this float I. An urethane rubber sheet is glued a ,~hort duration certain threshold pressure and temperature
to the inner guide ring to protect it thm~ wear. The material at the contact zone of the polishing prc~:ess to enable a chem-
of the drive shaft is austeniti stainless steel which is non- ical reaclion layer to be formed by the interaction o f the
magnetic. abrasive, workmaterial and the environmenl I I I I . This pr~-
M. ./tang. R. Kom,m, hlri l W e a r 215 f lg~,Sj 2h7-27~ Z69

ess is considered tribochemical polishing when there is no 4. Why cerium oxide (CEO..} is abt-,Lsive
superimposed mechanical action [12l. Thus, chemo--
mechanical action is very specific and proper choice of the It is well known that cerium oxide (CeO:) is an efficient
abrasive and the environment should be made for a given polishing agent for glass. It is also known that the fluid
workmaterial. Both thermodynamics and kinetics play an medium in which it is used is also very critical. For example.
important role on the rates of chemical reactions. Once the when polishing glass. CeO: is particularly effective in water
reaction products are formed, it is removed from the work- :md in alcohols only when hydroxyl groups are present, Thus,
material by subsequent mechanical action by the abrasive. cerium oxide slurry in water is invariably used for polishing.
Since material removal by this mechanism does not depend The oxide contains poiyvalent cerium atoms. Ce(IV) and
on the hardness but on the chemical potentials, it is possible Ce( IlII. which can provide chemical action with the work-
to remove material by abrasives substantially softer than the material, it appears thai when Ce( OH )L- i.e.. CeOz. 2HzO is
workmaterial. Theoretically any abrasive that can react with precipitated fresh, i.e.. in situ. in the polishing slurry form. a
the workmateri:d in a given environment and fi~rm a reaction soluble Ce( IV ) salt is probably involved in an equilibrium
reaction:
product can he used for CMP. However. some abrasives may
be harder than the workmaterial and some mechanical action S i O , + C e l O H ) =~CeO=,+Si(OHI
may occur in addition to the chemic-mechanical action. Mate-
The breaking and reforming of Si-O bonds is perhaps aided
rial removal by mechanical action may be satisfactory in
by the transfer of OH groupings to incipient fracture sites by
roughing, or even semi-finishing but in the final finishing
a transport mechanism using the relatively large and mobile
operation it is preferable to minimize the mechanical action
c~,~rdination sphere around oxophilic cerium atom [ 15 I.
that can affect the surface integrity. This is the reason why
Cerium is the most abundant element of the .,'are earths and
diamond abrasive was not considered in this investigation fi,r ranks around 25th in the listing of ahundance in the earth's
linishing Si~N.~ balls. Similarly. some abrasives react with a crust of all the naturally occurring elements. So. Ce is not
given workmaterial much more than others. For efficient particularly rare as compared to Ni or Cu. Cerium oxide has
removal of material, those with the highest reactions rates a high melting temperature ( 2750 K ~ but is a very soft mate-
would he preferable Ik~rchemo-mechanical action. Some idea rial ( Mohs hardness: 5 - 6 ) and hence cannot .~r-'ttch SigN4.
on this can be obtained by considering the chemical reactions Ce ions arc present in two stable valence states, namely, the
involved using thermiMynamic analysis, such as Gibbs free tetravalent C e " (Ceric) and the trivalent Ce * ' (Cerous).
energy of formation. The tetravalenl eerie ion is a strong oxidizing agent but can
A review of literature on chem(~mechanical polishing of be reduced by ferrous salts, hydrogen peroxide. When as.so-
various materials in general and that of Si,Na in particular elated with oxygen, it is completely stable as COO.. Ce_,O~ is
has been presented by Komanduri el al. [ [3] and may be unstable in air. water, and the like and readily converts to
referred to fi)r details. Also, the mechanism of chemic- CeO:. Ceria has the CaF_. structure with 8-ctx)rdinate cations
mechanical [x~lishing o f SigN4 with Cr_.O~ was reported by and 4-ct~rdinate anions. It can be visualized as a close-
Bhagavatula and Komanduri 114l in that it was shown con- packed cubic array of metal atoms with oxygen filling all the
elusively thai the role of Cr_,O~ was more than that of a mere tetrahedral holes. Ceria has been tested for acute effects and
catalyst (as rcporled in the literature by other researchers) found to have very low toxicity 1151. Hence. its u.~ is sate
and that Cr.O~ does play an active role in the chemical reac- from an environmental pt,int of view.
tion with Si,N., forming chromium silicate and chromium Since CeO_, can react cbemo-mcchanically with both SiLN4
nitride. This was shown by examination of the wear debris in base material as well as the glassy pha.~ that holds St,N4
particles together, this material is selected in the investigation
the SEM ( both the secondary electron images and the energy
fi)r final chemo-mechanicai polishing.
dispersion X-ray analysis) as well as X-diffraction of the
wear debris using a low-angle X-ray diffraction equipment.
In this investigation, this work is exte,zded to investigate the
5. Test procedure, polishing conditkms, and bah
action of CeOz abrasive for CMP of St,N4. It is fimnd that characterization
CoO: to be a superior abrasive than Cr:O, in linishing Si,N~
workmaterial, in that the finish obtained by Cr_.O, is ahout The uniaxially pressed-SigN., balls (CERBEC NBD-200
IO.7 nm R,, and O. 149 # m R, Fig. 4b). while that with CcOz from Norton Advanced Ceramics) in the as-received condi-
is about 3.8 nm R.1 and 0.029 p.m R, I Fig. 5f). as reported in tion had a nominal diameter of 13.4 ram. T h e ~ balls also
this investigation. Also. very few scratch marks, if :my. are contained nearly a 2(X)/.tm thick 5 mm wide band of mate-
round on Si,Nt surface when polished with CeO, even at a rial around the periphery at the parting plane resulting from
magnilication of 5(XX) to I O.tI(X)with the entire surface being the uniaxiatly pressing (HIP) process. These balls have to be
smooth without any pits. This is attributed to the use of sig- tinished It) a final size of 12.7 mm (0.5 in.). a sphericity of
nificantly sefler CeO., abrasive which will not scratch Si ,N.L 0.5/.tin, and best finish achievable. All the three factors were
in the final polishing. considered in the finishing of Si,N.L balls in this paper with
270 M. Jiang. R. gt~munduri / We.r 215 ( l~9SI 267-27,~

Table I a b r a s i v e used in this i n v e s t i g a t i o n . A s in m o s t f i n i s h i n g o p e r -


The chemical oml'a+sitionof NBD-2(VOsilicon nitride 1161 ations, there are three stages involved in magnetic float p o l
i sh i n g o f SigN4 balls, n a m e l y : ( I ) r o u g h i n g to r e m o v e as
AI C Ca Fe Mg O Si ,N~
much materM as possible without imparting serious damage
< n.5 _<U.88 ~ O,(M < o. 17 II.t~-1.0 2.3-3.3 97. I-t)4. I to the surface; (2) an intermediate stage of semifinishing
where size, sphericity, and surface roughness have to care-
lully monitored; and (3) final finishing where all three,
Table 2 namely, size. sphericity, and finish have to closely controlled
The n~hani-'dand the:finalpmFrlic~,of Si ,N~workpiccc I 16I
to meet the requirements.
Properly Value Table 4 lists the test conditions used for different stages of
polishing. Two coarser, harder abrasives. B+C ( 500 grit) and
FlexurM ,,trea~th ( M P a ) XIIO SiC (400 grit) ( i.e.. compared ~o St+N4 work material ) were
Welbull mtxtulus 03 used during the initial stages of polishing to reach the desired
Tonsil,: strength I MP'a,~ 4(1<) diameter at high removal rates and at the same time improve
Compressive ~.lrcnglh t GPa ) 3.l)
the sphericity f o r p r o p e r ball motion. After reaching the diam-
Hertz compressive stmnglh ( GPa ~ 28
H:udncss. Hv i I 0 kg ) ( ( ; P a ) 16.6 eter close to the desired diameter, an intermediate ( semifin-
FraClUl~ Iou~hll~.~s. K i , . M N m "' 4, I ishing) stage is utilized as a transition between the roughing
Den.qly ( g / c m ' I 3.16 and finishing stages, as the material removal rate is of prime
E!;zsfic m~Julus I (iPa) 320
concern in the first stage and surfi~ce finish in the final stage.
P(fi~.slm's ralio 0.26
"t'hcrmal expansion c~.'[licicnl a! _?O-I01X)"("t / " C ) 2/J x 10 " The two harder abrasives with a finer grit size were chosen
Thermal ~.'unduciivily at IIX)"C I Wire KI 29 for this intermediate stage, namely. SiC ( 1000 grit) and SiC
Thernml conducfivily at 5(XFC <W / m K | 21.3 ( 1200 grit). During this stage, the removal rates are much
Th~raml conductivily at I[I(RFC ( W h n K I t 5.5 lower and the finish much better than rough ing but the empha-
sis during this stage is the improvement of sphericity, in the
Table 3 final stage (prior to CMP), fine SiC abrasive (8000 grit ) is
"rhc propcaiesof variou,,ahra.~ive> used to approach the required diameter and sphericity and
remove ainu)st all the deep valleys f r o m the surface, This is
Al'~a.,.ive Density Kma~p ['lasli Mehing
followed by final polishing using a softer, theme-mechanical
g/on: ) hardnc~,.
+, mndulu~. I'~fint
I kg/mm: ) +(iPa I ("(") abrasive, namely, CoO, to produce the balls of required diam-
eter. sphericity, and linal ,,,urfac finish which is extremely
BA" 2.52 2l,IOil 450 245(I smooth and almost damage-free by preferentially removing
sit." 3.2 2500 420 241X) the peaks from the surlace.
('o: 7. I 3 625 165 25(~ The polishing shaft in MFP apparatus wasdriven by a high-
speed, high-prcci.,,ion at= b c a d n g sl)indic (1 S p i n d l e ) v, id~
emphasis on the tatter a.~l',cl, namely, i ~ s t finish achievable. a stepless speed regulation up to I 0 . 0 0 0 rpm. T h e magnetic
The large differences in the diameter between the as-received lield was measured by Gauss/Tesla meter. The pH value of
condition to the final size required is to remove all the reaction polishing enviromneul was measured by pH/temperalure
material that is fi)rmed on and near the surface during the meier. The polishing load was set up by measuring the normal
HIP'ing process. fi~rce with a Kistler's piezoelectric dynamometer connected
Tables I and 2 give the nominal chemical composition and to a charge amplilier and a display. To calculate the material
the mechanical properties, respectively of the St.,N+ ( NBD removal rmes, the weight reduction in the balls was measured
2(X)) balls, and Table 3 gives the various properties of the by measuring the weight before and after polishing at every
Table 4

SZ.I~ Abranl~ Abrasive Speed I_nild "l'illl~: RClllarks


I ~,~1/; I I rpn) J ~N/balll Imint
3"H~' Grit sil,e .~Ltc ~pm I

I Bil." ~(XI 17 It)",: 2(XlII 1.0 - Rmtghin~ ~high malcrial run~+.r,'alI


.',;i(" 40n "' ,~ I (],',,: 2,[Jlx~, I .o -
-~ .%i(" I|J[ J~l ~ IlJ + i 2INN) f .() .~() Scmi-lini.,,lfing t sph,:ri,:il,, and rL~u~:hm:s~,)
Sit I "iKI 3 It) +; 2(MN) I .O ,,~O
I:iaal Ihfi.,,hmg [silL'. ,,lflleri~.,ity, and Imi~,h }
('cO: _~ HP .; 21M)<) 1.2 1211

All ~hr~q~c~ u~:d. except ('(}:. ~crc oh, dined Ir~un Norma. ('{): i~ ()blamed froilt Aldrich ('hclmcal~.
M. Jian.. R. Komanduri / Wear 215 ~/tAt,~ 267-278 271

stage of the test using a precision balance. The surface finish micrograph. We al.,~ checked surface finish by ZYGO laser
of the polished balls was analyzed using a Form TalySurf interference microscope which is a non-contact measurement
120 L. Z Y G O laser interference micro.~ope, a Digital Nano- instrument. For the ZYGO laser interference microscope, the
scope ill atomic force microscope ( A F M ) , and an ABT 32 focus range is important and should include both peaks and
scanning electron micro~ope ( SEM ). The roundness of the valleys of the polished surface. Otherwise. the surface values
halls was measured using TalyRond 250. from ZYGO are unreliable. The stylus tip radius of AFM is
In this study, the finished balls h,e characterized for round- <0.08 p m and can easily be broken and not easy to be
heSS using a TalyRond 2 5 0 (cut-off: St) upr. Filler: 2CR ~. operated and u.~d very often, in this study. AFM is used fin.
and for surface features using a scanning electron micro~ope final high magnification evaluation of some random areas
(SEM). a Form TalySurf 120 L (cut-off: O+25 mm and 0.8 I =- 12). Based on the evaluation by all of TalySurf. SEM.
mm. evaluation length: 4 - 6 con.~cutive cut-off, Filter: ISO ZYGO and AFM characterization techniques, one can be
2CR). and an AFM. Although the latter three instrument more confident [hill the surface finish value shown are a reli-
measure or illustrate slightly different surface features, they able representation of the true surface quality.
are basically complimentary in nature. Their combined use
provides conlidencc on the data obtained. In this investiga-
tion. three randomly selected balls from each batch are traced 6. Results
three times at approximately three orthogonal planes using
the TalyRond and Form T;dySurf to provide the roundness Fig. 2a--c are SEM micrograph, a Form TalySurf profile,
and surface roughness, respectively. The TalyRond trace and an AFM image, respectively of a commercially finished
measures the maximum departure from a true circle of best Si ,Na ball surface ( considered as a master ball of ABMA
assumed magnitude and as such it denoted roundness. The Grade 3). From 1he SEM micrograph (Fig. 2a). it can be
sphericity of each ball. according to ABMA. is defined as the ~;cen that while some areas of the surface are extremely
maximum value o f the roundness measured on three orthog- smt~t)th, there are many fine .~ratcbes and .some pits. The
onal planes t)f the ball. Similarly. the surface finish of each AFM image of the polished surface ( Fig. 2c P more or less
ball is taken as the maximum value of three traces akmg three shows the same I~atures with an R., of = .5 nm and R, ..... of
orthogonal planes of the ball. =- 220 nm. Even though the TalySu.,t" profile of the sm(x)th
The surface roughness obtained by mechanical polishing region of the polished surface gives an R, of ~ 7 nm and Rt
generally has approximately a symmetrical prolilc. However, of = 7t) nm ( Fig. 2b) this may not reflect the actual suet'ace
when the peaks are smoothened preferentially leaving the roughness as can be seen from the SEM image (Fig. 2a).
valleys intact as in CMP of finishing of SigN4 giving a fairly From the SEM image, several deep pits can be seen the size
smooth ta'aring surf0ce, the surface roughness can be unsym- of such detects at the Pa)ttom would be ~maller than Ihe slylus
metrical. Many parameters have been propo~d ~o quantity tip radius ~2 # m ) of the TalySuif. As previously pointedout.
the various surface characteristics. It is necessary to ensure the values ohtained by TalySurf. AFM. and ZYGO would
that these values truly represent the surface features of inter- depend on Iheir abilily to analyze the dart from all the peaks
est. It is generally recognized that only R, is not enough to a~ well as the wdlcys. If tbe depth of field is not adequate fi~r
evaluate ihe surface finish and that I,a.'~thR~,and R, ( or R...... J a given magnification, the data would be in error on account
may be necessary. The R., value represents the average rough- of this. Consequently. care should be exerci.~d in the quan-
ness is a typical value of the measured surface but infi~rmation Ilia:lye evaluation of the satiate finish obtained at the.~ mag-
regarding lhe shape of the irregularities ( such as deep surface nifications although relative v M u e s and surface topography
dcl'ccts) is averaged out. The R, value is the vertical distance arc helpful in the analysis. Huh et al. 116l recently showed
he[wren the highest and lowest points of the roughness pro- the surface finish of a [a~lished surface of a commercial SigN4
file. It is not a typical value for the whole surface, but can ball {7/16 in diameter) using an AFM. an R., in the region
directly ~epresent the irregular surface defects, such as without defects as 1.8-2.8 nm and the regions including the
scratches and pits (deep valleys), which can have a signili- defects as I I - 18 nm ( with defect density medium and.~ratch
cant efl'~ct on the surface quality of advanced ceramic mate- marks sevcreL For smaller-sized balls ( I / 4 in diamelerl.
rials ( R, = Rp + R, = R,, ) fiw various applications. Ihcy rcl~rted an R., of 2.5 to 4 mn in the region without aqy
For o stylus instrument, such as TalySurf the stylus size detects and 35-40 nm including the regions with defects
and shape affect the accaracy of the profile, it would not be {with defect density large and scratch marks somc~. How-
possible to trace the complete profile of a deep valley espe- ever. the R ...... with defects can be many times this value (at
cially the I'a~ttom if the size of the valley is smaller than the least 10 limes). Fig. 3 is an AFM hnagc at higher magnifi-
tip radius, The stylus tip radius of TalySurf 250 used in this cation showing deep pits separated by sm(x)th regions of a
study is "- 2 pro1. However, SEM micmgraph can be helpful commercially linished ball.
to identify whether there arc surface delk,cts which can be Fig. 4a and b are an SEM micrograph and a Form TalySuff
relY,coted by stylus of TalySulf and whether the value from profile, respectively, ofa Si ~N~ball surface finished by Cr:O,.
TalySurf is a reliable for small-damage surface. TalySulf is From the SEM micrograph ,~Fig. 4a ~. it car: be seen that while
convenient to use for large area scanning with help by SEM much of the surfiic is very smtR)th there arc some line
,7, M. Jiaug. R. K~nmmdw'i / Wear 215 ( 1998~ 267-27bl

Im

(b) . .i i

-!- ~'- lOOltm


l O.2tim t Z ra~
It,s (1~)
217.711*~
II,I~! n~
* u~ ~ ~ r~sk~ss (lla) 5,0SI ml
I~: 7 nm tlaN k e l l e r (R~) 220,1~ tm
-i m -- $~r~a~e arelt
Rt: O.07 t i m ~'~rraoe ,tea 4irr

i:ig. 2. ~ a ) T h e S E M m i c r o g r a p b o f a c,',nmerial!)' linished Si~N.~ h~dl .~url'ace. ( b I T h e T a l y s u r f s u r f a c e mughne.,.s profile o f a c o m m e r c i a l l y finished Si ,N.t
ball .,url~ce. () T h e A F M i m a g e u l ' a c o m m e r c i a l l y linir.hcd Si ~N.~ ball .~urface.

k ,

(b) . . . . .
i zl~ LI /
0.2 p m
100,urn
i/iml
Ra: 1U.7 a m
n*mI
RI: 0.149 tim
J 2 rm~e "/5,151 l i
J Iiein -4.1711 mm I:ig. 4. ( a ) T h e SI:.M m i c r o g r u p h o f ;a Si iN4 hall ~urlac tl|li~hcd by C r : O ,
k (1~) I O . O g 7 iv* abrasive. ( b l The T a l y ~ u r f xurl'a~e r~ughne.,,~ prolil ~d" a Si ,N~ hall ~tlrI'~|c
IS4,an T m e s l m ~ s ( R a ) 5.172 *m
Max h e t g k t ( l i m a . ) ? 0 . 1 [ ~ 7 nt4 li,i,,hetl by 'r,O, ahraxi~ e.
.~ur face ar.l,a
I fatrf~ area d t r f

I-ig 3 TI~- AI:M illli~.~t." .3I h i g h e r lna~Tlilkc';,ltitm ,,ho'~-inl2 r..$tlt~tll alL'it mechanical abrasion leading Io abrasion and pilling. Also.
~'p'~t'alCtl b ) pilling o l a c , m m e r c i a l l ~ Imi~.hed Si ,N ~ hall ~.url~tc'e. soinc of the species generated during CMP with Cr.,O~ may
require carelid handling from an environmental point of view
~ratches and seven,l small pits. The surface finish values which is not the case with CeO_, [ 17 I. Hence, the use of CeO.,
obtained by the Fornl TalySurf are R.,= 10.7 nm and for polishing of Si ~N4 would be preferable both from CMP
R~=0.149 btm. A plau.,,ible explanation Ibr the observed and environnlenlal pot,its of view.
roughness with Cr:O ~abrasive, in spite of its CMP ability, is Table 5 gives the average surface finish and material
it..., higher hardness t harder than Si,N.~) and consequent removal rates obtained at progressive stages of polishing. The
M- Ji~mg. R. K.m~mduri / Wear 215 f I~x~Rt 267-27~ 273

Table 5
The average .~urface linish and material renmval rates during various stages of poli.',hing

Stage Abrasive Surface li.ish ( average ) MRR ( per ball ) M = c r i a ! r~inoval n~hani~m
R,, ( n m ) R, i # m ) ( mg/min ) ~/~m/min )

I B~C 5(X) 225 1.95 O.q~ 1.2 Microi'r.'wturc


SiC 4(x) 170 1.40 o.f~l. 0.8 Microfraulnre
2 SiC ilX)O 95 l).80 o,.~t 0.4 Submicrofraclur
S i C 12(X) 55 0.50 (i.2o 0.2 Submicrofr~cmrc
3 S i C 14000 15 o. 15 o.(~ - Suhmicrofrac~ur~
CeO., 4 {}.{)3 I}.() I - Tril'~chcmical

corresponding TalySurf surface finish profiles are shown in ZYGO are R, --- 3 . 9 nm and R,,,, = 0 . 0 2 1 / ~ m f o r the line scan
Fig. 5a-f. It can be seen Ihat the surface roughness as well as and by AFM are R, = 1.4 nm and R,,,,, =O.OI8/.tin for the
the material removal rates decca.~ as ~he hardness and grain area ~an. The SEM micrograph.~ (Fig. 7b) show es~ntially
size of the abrasive decrea~s due to a decrea~ in the size of smemth surface with practically no surface defects. Both AFM
brittle microfracture. For a harder abrasive with a fine grain and ZGYO also provide surface finish by averaging over an
size. the material removal is by submicro~opic fracture and area. The evaluation of surface topography by all the char-
therefi~re results in damage-free subsurface. Further CMP acterization techniques considered in this investigalion.
with a softer abrasive, such as CeO:, as will be shown, will namely, TalySurf. SEM. ZYGO. and AFM. gives cemfidencc
result in an extremely smooth surface. that the final surface is damage-free with a finish of R,,=4
Fig. 6a and b show the 3-D plot of the surface roughness nm and R, = 0.04 p.m ( cut-off: 0.25 ram. evaluation length:
using the ZYGO non-contacting l a i r interference micro- 4 - 6 con~cutive cut-off. Filter: ISO 2CR). if one ctmsiders
scope and the AFM profile, respectively of the final surface the AFM values, the surface finish R., would bc = 1.4 nm.
polished by sorer CeO_, abrasive. The surface finish values Fig. 7a is an SEM m i c r ~ m p h of a Si;Na ball surface
after the linal polishing obtained by Form TalySurf are obtained after mechanical polishing with a finer SiC abrasive
R=~---3.8 nm and R~=0.029 btm, while t h o ~ obtained by ( 8000 grit size) indicating that the material removal from the

(a)

. . . . . (b)

. . . . . I I

l~: 237 nm Rw 175 r i m


Rt; 2,U6 l~m Re I..~t gm
i::! ' '

(d)

...... = ,
t' Ii

i;" ~ : 98 =rim

Rt: {Y?P.~m
[~=: 5U n m
R~: 0.44 14m

(c) ". " - tO .. ~, -

..... ~. _ _ ..~ ~G.~,,rn ..... .___]o2~m


o , . . . . . ~ - ~ ~ . . "V'.'~.,,~"~r~'~ ~ r . . _ ~ , , ' - - ~ - - ' ~ .r - ,--.
100 prn "100~um
Ra: 16 ' a m R.: 3[] rim

Ri, 0 1.6 .urn RI: 0~'49 Hm


CeOz ='5gm) abrasive
Fig. 5. The T a l y S u r f surface r~,ughnc.~.~ pruli;s , f a Si ,N= hall after polishing by: [ a ) B=C (~(N) gril ~ abrasive. ( b ) SiC" i 400 grit ) ;,bra~ive. ~ ) SiC ( IO(~1
grit ) ahra.~i~e. ( d ) SiC" ( I ?(XI grit ) ;Ihra~ive. () SiC I X(XI() ~rit ) ahra~iv~:, and ( I ) ('cO: ( _~/.ml ) abram,ire. respectively.
274 llt. Ji.t~g. R. K.nt.ndt#ri / t~'r.r 215 1t{~q,'~) 2~7-.27,~

r,m

--48r ~ 0 L
B, ~65B

.em.mm

ILl

[
uJ

~i ~4 - r'-- "--"F-

at.rot
~o~tr

2;:? , r

(a)

Im

2 ramie i S . 5 9 3 me I
mean Z. | 0 2 me
Ilbes ( ~ q ) 3 . 2 3 9 me
l l q ~ r m a g k n e s s TRaJ 1 . 3 1 E mu
Ilal( ) 1 e i g h t O l i a N ) 18.433 rm
I
l
j
~rFIGe area !
SurFace area &aFt

I:i1:. ~ ~ u i " ~ Z Y ( ; O phlf i fIX} I:il~:uu ) ~d' Ih~: ~,urlac o f u .";i ,N, bull ;u'tcr ti.isl)intz hy ( ' 0 : 1 5 I " " ) abra~,iv. ~ b) The A I : M imup'-" . f ih .,,,rtacc t f f a
%i ,N, hall atIL'r ImtshinL~ h) (+el'): t .~ /.~m I ablusi~,v.
~ Jiung, R. Kimuuuhiri/ Weiir 215 t 1~8~ 267-278 "75

Fitz.7. ( a I The SEM inicrographoflhe surfacelif a St,N, ball after ~di.,,hlng Fig. 8. I a t The SEM mie.~l[mphor ll~ ~l;iill!,~,l ~n_,elshaft in i:lmii~i wilh
by SiC 1 8 t i l l ~.ril ) ahnisivc. I h ) The SEM miero[zrliph of lhc ~urlilce or a St;N4 halls in MYP ~howing Ihe pre:,cnce of B,C ~<;OO~.ri(I purlilc.,+ and
Si .N l hitll atier Fallishing By ('cO, ( 5 p.nit ahn:sivc. ahnl.,,ilm mark~,~m the ~,hal~malerial. (hi Th-- .-,?M micn~g~,~phof Ihc
',hlinle',',,,fuel~,lialihi co111a1wilh Si ,N, h:ill~,in MI-P~,llo~,.;ngIhprc~enc
~l' B+C I I.~lHIgril ppanich:.~and a~a~iml inark~~mii~c~lmltnialeri~'il.
workmaterial is predominantly by brittle fracture on a sub-
microscopic scale under the mechanical action o f the abra-
Fig, 9a a n d b s h o w | h e T a l y R o n d roundnes.,i tnlces o f ;in
sive. While some I~lishing scratches can he seen. the .~urface
as-received ball and the linishcd ball, rcspccti:'ely. It shows
is relatively free of pits that would normal!y form using dia-
that the roundness of the NBD-2(M) HIP-Si+~N~ balls was
mond abra.~ive+ Fig. 7b is a represenlative SEM mierograph
reduced from the as-received condition of = 2(X)/zm to a
of a Si ~N+ ball .~urhice after the surface has I',en finished by
tinal value of ~ 0.25 ,u.m. The sphericity of Ibe ball.~ ( i.e..
C M P with a softer abrasive, CeO_,, showing an extremely
maximum deviation of the roundness of :l ball laken in three
smt~)th surface with practically m) surfilce del~ccts, such as
orlhogonal planes) fi~rthis batch was fimnd lit be in the range
pits or scratch mark.~. Several areas of the Si +Na ball surface
of 0.30 ,urn.
were scanned and the micro~,raph shown in Fig. 7b was fimnd From the work presented here, it can bc seen that magnetic
to be a representative of the topography o f the surface. float IX)lishing technique can be used fiw linishing SieNa balls
Fig. 8a and b SEM micrographs of the polishing shaft after f r o m the as-received condition to the required diameter. ~phe-
polishing St,N+ balls with B~C 500 and B+C 1500 abrasives. riciuy ( < 0.3 /.tin). and surface linish ( < 4 toni wid-~out
They show that the abrasives are actually not embedded in scratches or pits on the s u r f ~ e . The actual polishing process
the shaft t,s considered by Childs et al. 171 but actually abrade from the as-received condition to the linal requirements can
the softer stainless steel polishi ng shaft. Thus. while the action be achieved in less than 20 h. This. however, dex~s not lake
of the abrasives is one of a two txuly abrasitm ( i.e., sliding into account the time taken filr the characlerization o f the
without rotation) as rightly [minted out by Child.,,, they are surfaces by various techniques mcnlioncd in this paper. The
not fixed but move relative to the polishing shall. The material methodology developed here incorporates tx~lishing condi-
remov;,I is due it) the r e l a t i v e speed between the abrasives tions and use of appn~priale abrasives ( i,)cluding grain sizes i
and the workmaterials. thai arc not se ve re enough at any siaLze to cause damage, such
276 M. Jiang. R. Komunduri / Wear 215 (19981267-2"/8

~...~" '~..~..j.
D

&

&

& 4

.! /:
z~

ql

4 ib

4..' &
&. . ,. .

I~ . .4

tOOI " '- II

~Ir q" II,v ~g

4
,ill & & ib *1 m~

Fig. eJ. 4.a t The TaiyRond roundnc.,,s pr,.~lilc o f an as-received Si ~N4 b~ll sho~ ing a 2(XI p.m 5 m m band at the parting line duc to the uniaxial pressing p n ~ e s s
t Roundne~.,,: 195 ptm ). ( b} The TalyRtmd rolmdnes~; pr.~lilc o i ' a Si ,N.~ bull after the tina| ~.lagc t}|"ptdi,lhing by C e O : 15 p.m } abrasive I Roundness: 0.25 p.m ).

as deep pits and cracks at and near the surface, so that the process is by mechanical microfracture because of higher
balls can be linished to the requirements without surface or hardness of the abrasive and the inherent hardness of the
near surface damage. Of cour.~, whether or not the surface workmaterial. Under these conditions material removal
can be linished absolutely smtmth also depends on the resid- occurs not by grain pullout, grain fracture, and large fracture
ual pcwosily of the uniaxially pressing-Si~N~ material. but by microfracture by cleavage. While chemo-mechanical
action may aiso occur, its contribution is considered to be
much smaller than the mechanical action, namely, microfrac-
7. Discussion lure by cleavage.
Childs et al j ? I have shown that in magnetic float polish-
7.1. Mech~miculp, di~hinj~ ing, material removal from the balls is accomplished by the
action o|" the abrasives embedded in the shaft due to sliding
The mechanism of material removal from the SigN, balls at the contact area between the drive shaft and the ball. It is
by liner grit, harder abrasives, such as B4C and SiC in MFP unlikely that when line abrasives are held between the SigN4
M. Jiang. R, Komanduri / Wear 215 Ill#H) 267-278 277

balls and the stainless steel shaft that the abrasives will get 8. C e a d m i o m
embedded in the shaft, as in the pre~nt case. If this is ,so, one
would never be able to remove material from .softer work- ( I ) Magnetic float polishing (MFP) combining mechan-
materials with loose abrasives, in an actual situation, the ical and CMP is an efficient and cost effective manufacturing
abrasive will abrade the soft stainless steel shaft much mote technology for woducing high quality SigN4balls for bearing
so than the Si3N~ workmaterial and it appears unlikely that applications due to high polishing speed,small and contwlled
the abrasives will be emhedded as Childs et al. considered polishing force, flexible support, and chemo-mecbanical
but would be moving relative to the polishing shaft forming action.
abrasion marks in the shaft. In fact wear on the stainless steel (2) There are three neces~ry stagesfor polishing, namely.
shah is as a result of it and may have to be ground periodically initial roughing stage where the emphasis is on high material
to improve sphericity, removal rate with minimal surface-subsurface damage, inter-
For larger*sized abrasives and higher loads, as in conven- mediate semi-finishing stage where material removal rate.
tional polishing with diamond abrasive, the finished surface sphericity, and surface roughness have to be closely moni-
is effectual by the formation of deep pits. grooves and cracks. tored, and final finishing stage with emphasis is on the
This will not be the ca~ with finer abrasives and lighter loads. required size, sphericity, and finish.
Higher material removal rates without subsurface damage is (3) High material removal rates ( I p,m/min) with mini-
feasible by magnetic float polishing hecau~ of high polishing mal subsurface damage ate possible using harder abrasives,
speeds and very flexible float system used [ 8 ], The low loads such as B~C or SiC dec to rapid accumulation of minute
used ( I N/bail), while c a u l s microcracking by cleavage, is amounts of material removed by mechanical microfracture at
small enough as to not cause larger cracks, or dislodge grains high polishing speeds and low loatL~ in the MFP process.
by grain pullout. Although material removal is by brittle fracture, it occurs on
a micro~ale due to low polishing force, flexible float system.
Pqd fine abrasives. The cracks generated ate localized and
Z2. C h e m t ~ - m e c h a n i c a l p o l i s h i n g suppres~d from propagating into microcracks. Conse-
quently, su~urface damage is minimized leading to the
higher strength of the workmateriai and reliability of the paris
The mechanism of material removal in the final stages of
in ~rvice.
polishing by softer cerium oxide is due to CMP. Thermal
(4) An advantage of the magnetic float polishing apparatus
analysis of flash temperature and flash duration [ 18 ] as well
u~d in this investigation is that it is capable of finishing a
as thermodynamic studies of the polishing process strongly
small batch ( 10-20 balls) to the finished requirements with-
suggest the possibility of CMP of Si~N~ by Cud., [ 19l. The
out the need for sorting them from a forge batch of balls or
details of the chemo-mechanical action of the Cud_, abrasive
u ~ different equipment a~ in conventional lapping. Such an
with the Si~Nj balls will he the subject of a subsequent pub-
apparatu~ would be beneficial especially when small batches
lication. Under the mechanical frictional action during pol-
are needed for specific low volume applications or for eval-
ishing, chemical reactions can be initiated between the Si.~N4
uation of materials in the development of new materials for
balls and the Cud_, abrasive in the pre~nce of water ( from
hearing applications.
the water based magnetic fluid) and the material is removed
(5) MFP can be acost effective process for finishing Si~N
by the chemical dissolution of material resulting in a reaction balls for hearing applications. The semifinishing and finishing
product that is subsequently removed by the mechanical stages can be accomplishing in about 4 h. The roughing stage
action of the abrasive. Since the hardness of C u d 2 abrasive depends on the amount of material to he removed from the
is = I/3 of Si3N4 workmaterial, it can hardly scratch or dam- as-received condition to the final requirements. In any c,~se,
age it and the material is removed by tribological interaction a batch of balls can be finished in ~- I6 to 20 h compared to
forming a reaction product. Thus. tribochemical action .several weeks by conventional polishing. Aim. diamond
instead of mechanical fracture is credited here for the abrasive is not required for the process. Faster polishing times
extremely smooth and damage-free surfaces accomplished and use of abrasives other than diamond would significantly
on the SigN4 balls, reduce the overall costs of manufacture of Si3N4 balls for
it has been suggested [7] that magnetic fluid grinding is bearing applications. Also, the implementation of this tech-
more likely to replace the roughing stage of finishing than the nology would not be very capital intensive as it can he used
final polishing stage. We have clearly demonstrated in this by modifying the existing equipment. Attempts are currently
investigatio, the'.! MFP can replace completely (both rough- underway to increase the batch size to finish about a hundred
ing and finishing) the conventional polishing starting from balls per batch with the next generation equipment.
the as-received balls and completely finish them to the final (6) The CMP depends on the polishing conditions u~d.
specifications in the same apparatus. The methodology for abrasive-workmaterial combination, and the environment.
finishing SigN4 balls by MFP is presented for the first time in Cud: is found to be the most effective abrasive ( superior to
this paper that involves actual finishing time an order or even Cr,~OO in the CMP of Si.~N 4 balls yielding a damage-
magnitude or more faster than conventional polishing. free surface with a finish of R~,<4 nm and Rt < 0.04 /,tin.
278 M. Jiang, R, K..umduri / Wear 21511998} 267-278

Also, it is possible that some of the ion species formed with Umehara, Magnetic fluid grinding mechanics. Wear 175 (1994) 189-
198.
CeO: abrasive can be much more safer than those with Cr.,O~ 161 T.H.C. Childs. S. Yoon. H.J. Mnhmood. The material removal mech-
from an environmental point of view, which will be discussed anism in magnetic fluid grinding of ceramic bali bearings, Proc. IME
in more detail in a sub~quent publication. (Lon ) 208 ( B I ) ( 1994 ) 47-59.
171 T.H.C. Chitds, S. Mahmood. H.J. Yoon. Magnetic fluid grinding of
ceramic hails. Tribal. Int. 28 {6) ( 19~;31 341-348.
181 M. Raghunandan. A. No~ri-Khajavi. N. Umehara. R. Komanduri.
Acknowledgements Magnetic float polishing of advanced ceramics. Trans ASME. J. of
Manuf. Sci. +Eng.. 119 (1~)6) 521-528,
This project is sponsored by grants from the National Sci- [91 N. Umehara. R. Komanduri. Magnetic fluid grinding of HIP-ShN4
rollers. Wear 192 (1996) 85-93.
ence Foundation on "Tribological Interactions in Polishing
I I0 ] R.E. Rosenswig. Ferrodynamics. Cambridge Univ, Press. New York.
of Advanced Ceramics and Glas~s," (CMS-9414610). 1985.
"Design. Construclion, and Optimization of Magnetic Field I 11 I N. Yasunaga, N. Tarumi. A. Obara. O. Imanaka. Mechanism and
Assisted Polishing'. ( DMI-9402895 ), and DoD's DEPSCoR application of the mechannehemica! txdishing method using .,;offer
Program on 'Finishing of Advanced Ceramics" ( DAAH04- Powder. in: B.H. Hnekcy. R.W. Rice (Eds.). Science of Ceramic
Machining and Surface Finishing--It, NBS Special Publications No.
9,5-1-0323). This project was initiated by an ARPA contract 5fl2, 1979, p. 171.
on "Ceramic Bearing Technology Program" (F33615-92- 1121 T.E. Fischer. Tribochemistry. Ann. Rev. Mater. Sci. 18 (1988) 303-
5933). Thanks are due to Drs, J, Lar~n Basse, B.M. Kramer. 323.
Ming Leu, and J. Lee of NSF and Dr. K.R. Mecklenburg of I 131 R. KomandurL N, Umehara, M. Raghunandan. On the possibility of
chemo-meclpmical action in magnetic float polishing uf silicon
WPAFB and Dr. W. Coblenz of ARPA for their interest and
nilride, Trans. ASME J. Tribal. ! 18 i IC~) 721-727.
support of this work. Thanks are also due to Dr. M. Raghun- fl41 SR. Bhagavatula. R. Komanduri. On chemo-mechanical polishing of
andan, a former graduate student at OSU. for his contributions silicon nitride with chromium oxide abrasive. Philos. Mag. A 74 (4)
to the research carried out at OSU. (1996) 1003-1017.
I1,~1 B,T. Kilboum, Cerium--A Guide to hs Role in ChemicalTechnology.
Molycarp, White Ptains. NY. 1992.
I 1~1 S.R. Ilah. T.E. Fischer. C. Burk. Ceramic bearing development--
References tribachemical flubbing of silicon nilride. VoL 4. Technical Report
No. WL-TR-~)6-40180 The Materials Direc:lorale, Wrighi Patterson
[ I] Y.TanLK. Kawata.Dcvcl,pmcntofhigh-efticientlinefinishingproc- AFB OH, March 1995,
ess using magnelic fluid. Ann. CIRP 33 ~'1~84) 217-220. 117] R.G. Reddy. R. Komanduri, On the nature of chemical specie.,~ in
121 N. Umehara. K. Kate. Principles,f magne,c fluid grinding of ceramic chemo-mech;mical polishing of SigN4 balh with Cr,O~ abrasive in
balh. Appl. Eleclromagn. Mater. I ( I ~ ) ) 37--43. magnetic Ileal polishing (MFP). manu~ript under preparation, Iq97,
131 N. Umehara. Magnetic fluid grinding--a new technique lbr finishing 1181 Z.B. ltou. R. Komanduri. Magnetic Field Assisled Polishing: Pun I,
advanced
ceramics. Ann. CIRP 43 ( ~ ) ( | 994 ) 185- | g8. Thermal Model. ASME Trans. J. of Tril'K)logy ( in press L 1997.
14j T.H.C. Childs. H.J. Y(~m. Magnetic fluid grinding cefl design. Ann, I lql M Jiang. R, Komanduri, On the Cherm~mcchanical Polishing
CIRP41 t I ) ~ 19921 343-346. (CMP ~ of Si~N.LBearing Balls with COO,. submitted t'or publication,
151 T.H.C. Childs. D.A. Jones, S MahmcNxl. K. KaRl. B. Zhang, N. i ~,~)7.

You might also like