Note : Les descriptions sont présentées dans la langue officielle dans laquelle elles ont été soumises.
60S~0,~ 3 ~
7~_~LE
I:)IAMOND POS~ESS~JG ENHAN~D lHERMAL ~NDUCIlV17~
The present ~nven~on relates tO the preparadon of polycrystalline diarnond and more
par~culaAy ~ polycrys~alline diamond possessing cnhanced Ihermal conductivity.
10High thennal conducdvity diamond, such a~ high quality sclected ~ype 1I natural
diamond~ is charac~eriz~d by a vety high degree of purir" and is ~ o have a thcnr~l
conduc~ity at 25-C (298-K) on thc order of aboul 21 wats/~n-K. Such high thcrrnal
conduc~ mond is u~seful, fot e~tamplc, as a heat sink ma,~ , such as in the backing
of semi-conduc~ Despi~ i~ high CQStS, type II natu~al d~nd has been employ~d as a
1$ heat sink material because it has the highest them~al conductivity of diamonds.
Conven~onally~ duccd high pressure/high tempeTanlre (HP/~I~) synthetic high quali~y,
low nitrogen type Il diamonds can be produced with a simila~y high therrnal conduc~ivity.
For thc most pa~t, diamonds prepared by low-p~essu~e chemical vapor deposition (CVD)
- p~cesscs are not single c~ys~al di~nond and have materially lower ~he~nal conductivities,
20 typically on thc aTder of 12 wuts/~n-K ~t about 300K (herein~ftcr sornetimes referred to
, ~ Since di~wnd is usually an eleca-ical insulasor, f.e. ele~ic~ly non-conducting,
heat is conduc~ed by phonons. Anything that shortens the phonorl mean free path ~i.e.
latdec ~,ribra~orl modes) de~cs thennal conducdYity. In 98% ~ nanlral diamonds (type
25 Ia), m~ogen impuri~e~ ~catter phonons. Thi~ reduces the mean f~c phonon pa~h and,
thu3, shc ~mal conducti~ r, to nc~r 8 waats~/cm K In polys:~ystalline diamond ~ypical of
that made by CVD p~ocesses, thae a~E muny dcfects, ~uch as, for example, nvins, grain
botmda~ies, vacancie~, ~nd dislocations, ~hat reduce the phonon mean ~ree path. The
~'! t~ conducd~ r of CVD diamond is remarlkabl in one scnse in ~at it is ~bout 60% of
30 dlc th~mal conductivi~ of highly perfect diamond.
With ~e~pect to polycrystalline diamond (in fih3l, compact, or other form~, ~hennal
cor~ucdvity is known tv bc af~ected by, ~r cxample, unp~des, isot~pic e~ s, and
bound&~y sca~in& tO name jus~ a few fac~s. In fact, grain bound3ly scat~enng has bcen
. ~ bclievcd to b~ do_n~nt in the lower ~hernul conduc~-~ity of polycrystalline diamond
"
.
, ~ .
.'''~ , '
~... .
~o~
compared to single crystal diamond. Enhancement of th~ thermal conducti~vity of
polycrystaUine diamond, then, is a necd that yet exists in the art.
r~ad ~çrnent o~he Tnvention
S Broadly, the presenl invention is directed tO polycrystalline diamond of improved
therrnal conductivity. The novel polycrystalline diamond consists essentially of at least
99.5 wt-% isotopically-pure carbon-12 or carbon-13. The inventive polycryslalline
diamond is forrncd &~m at least 99.5 wt-% isotopically-pure carbon-12 or carbon-13.
Single-crystal isotopically-pure carbon-12 and carbon-13 diamond are known to
possess improved thermal conductivity. Polycrystalline diamond, however, possesses
thennal conducdvity patterns deleteriously irnpacted by, for example, impurities, isotopic
effects, and grain boundary scattering. In fact, grain boundaly scaKering would lead ~he
skilled artisan to believe that the thennal conducdvity of polycrystalline diamond would bs
subs~antially unaffected by the isotopic naturc of the diamond itself. Unexpectedly,
however, isotopic effects were discovered to predominate in impacting the therrnal
conduc~vity of polycrystalline diamond consisting essentially of isotopically-pure carbon-
12 and carbon-13. This is true whether the isotopically-pure polycrystalline diamond is
grown directly or wheth individual iso~opically-pure carbon-12 or carbon 13 diamond
crystals are subjected to sintering for ~orming a polycrystalline structure, e.g. Iayer or
compact, thercof.
Detailed lXsOE~Qf the Invçn~Qn
Heat conductivity in diamond is qui~e complicated, especially considering the
parallel and se~ies paths that it çan take. It should be understood that theorics on heat
2S conducdvi~ in di~mond, then, are inconsistent in the li~erature and, likely, are incomplete.
Thus, much of ~he the~y expounded he~in should b¢ intelpret~ accordingly. Rcgsrdless
of ~ the~ic~ a~poul!ded herein, synthesis of polycrystalline carbon-12 and carbon-13 has
been achicve~ and the unexpectcd thermal conduc~ confimled.
The dcscription that follows is directed par~cularly to 12C diamond, but it holds
equally tr~e for 13C diamond as well. Since diamond is an insulator, hea~ is conducted by
phonon~ ~qua~on I, ~low, sets fonh the th~nal conduc~Yity of polycsystalline diamond
in ~nn~ of ~hG ~pecific heat (C), phonon veloci~ , and fnean f~ec path of phonons O-
(~) K = (113)CV~, or K
It has bcen shown prcviously that both the specific heat and the phonon vel~i~ (the sound
vclocity) are thc samc in high quality diamosld arld diamond made by chemical vapor
deposidon (CVD) tcchniques. Consequently, all th~ Yariatisn in thermal conductivity
.
-2~
60S~7~
between different grades of diamond occurs because of diffetences in the mean free path of
phonons in dif~rens grades of diamond. The mean free path of a phonon is given by the
following ~quation:
S (II) 11~ phonon-phonon + IJ~ grain-boundaries + 1/~ disloca~ions +
1/~ vacancies ~ impurities + 1/~ isotopes + ......
wherc, scattering caused by phonon-phonon interactions, grain boundaries, dislocations,
vacancies, impurities, and isotopes arc included explicitly, while other possible scattering
centcFs (e.g., small voids) are repsescnted by ".. ".
Estimates of so~e of thc phonon mcan frce path length ~om themlal diffusivity
data of natu~al isotope abundance high quality diamorld and isotopically pure high quali~
` ~ diamund can be made. The avc~age phonon veioeity at room ~empcrature in diamond is
equal to the sound velocity of 1.38 x 10~ cm/sec. The specific heat of diamond at room
temperatu1e is reported to be 6.195 joules/g. For isotopically-purc high quality diamond,
the phonon mean ~ee pa~h is limited principally by a phonon-phonon scattering. From
equa~on I, we find that A phonon-phonon is 0.17 microns.
For natural isotopc abundance high quality diamond, thc phonon mean free path isdetermined by both phonon-phonon and phonon-isotope scattering, and is equal to 0.09
microns. From shis value an~ equation 11, W5 can denve the mean free path of isotope
scattering ~ isotopes to be 0.19 microns.
For polyc~ystalline CVD diamond, addilional phonon scattering centers come into
play and the themlal conducd~i~ is decneased to approxima~ely 12 wa~ts/cm-'K. which
gives a phonon mean fiec path of 0.05 micr~ns. Sevcral observa~ons about the magnitude
2S of this phonon mean free path shoold be made. Firs~, elimination of scattering centers,
which are much more widcly spaced than 0.05 microns, will not affcct the therrnal
conduc~vi~q acco~ding to equa~on n. Thus, elin~ination of grain boundanes in CVD or
othcr polycrys~alline diamond material having a gTain size o~ 10 mic~ns, for example, will
only increase ~c ~herma~ eonduc~ r by Q596.
Secondly, although ~limination of grain boundaric~ by using epitaxial growth on
diamond o~ hete~cpita~dal growth on a ~oreign substrate will not affect the therrnal
conducti~fity, such growth may lead to a lower concen~a~on of dislocations by starting
with perfect seed cFystals and ~he~eby increase the thcmlal conduc~ivity. From etch pitch
studics it has been estimated that the dislocation density in typical CVD diamond material
excecds 108 disloca~ions/cm2. Thc phonon mean free path ftom scattering off of
disloca~ons should be less than I micron. Elin~inadon of all dislocations should increase
the thennal conduc~vity by grealer than 5%.
"
-3-
,, , ' , '
,~ 60æ~
:` ~
A reduction in the numbers of grain boundaries can be achieved through the control
of nucleation during the initial stages of diamond growth. This can be accomplished by a
vanety of means. Heteroepitaxy would allow single crystal films, if successful. Even if
polycrystalline material was forrned, it would have fewer grain boundaries than standard
5 CVD diamond grown on Si, Mo, etc. Suitable substrates for heteroepitaxy would be Ni,
Cu/Ni alloys, CBN (cubic boron nitride), ana CBN films grown epitaxially on Si. Another
approach is to seed the substrate with diamonds. Using CVD diamond to grow
homoepitaxially, it should be possible to control the orientation grain boundaries of the
film. Reducing the grain boundaries and the dislocation density would eliminate phonon
10 scattering and increase the thermal conducdvity of the resulting film.
Probably the largest scatter of phonons and CVD diamond are vacancies and
vacancy clusters. Because CVD diamond is deposited at a temperature of about 900'C,
which is less than 1/4 the melting tempcrature of diamond, there is not much solid-state
diffusion during deposition. This lack of defect mobility causes a large amount of atomic
,~ 15 defects, such as vacancies, to be frozen during growth. Current CVD technology,
however, militatcs against improving this condition.
One scatteling center that is easily removable from CVD diamond are carbon-13
isotopes when mak;ng isotopically-pure carbon-12 (and carbon-12 isotopes when making
isotopically-pur~ carbon-13). Knowing the mean free palh of isotope scattering, equation
20 II can be used to esdmate the change in thermal conductivity that can be expec~ed by
eliminating uowanted isotopes fiom conven~ional CVD material with a therrnal conductivity
~' of 12 watts/cm--K. Deletion of ~ isotopes equals 0.19 microns in equation II and
- substitution of ~e enhanced A in equation I shows that the thermal conductivity of CVD
diarnond should increass &om 12 to 15 watts/cm--K when it is made of isotopically pure
25 carbon-12. Th~ thermal conductivity for isotopically-pure car~on-13 similarly should
increase to a~und lS watts/cm--K.
Lascr flash diffusivity IR detection sys~em dasa was generated fiom about 0.5 mmthick dislcs of CVD diamond which was greater than 99.5 wt-% isotopically pure carbon-
12. Onc side of thc disk was blackened and a laser polse irnpac~ed the~eon. Diffusivity or
30 the time rabe of temperature decay, was detectcd by an inf~ared detector on the reve~e side
of the samplo. The measurement was made at room tempera~ure, l~iZ., 2S-C. A natural
abundance isotopc sample also was tested. The natural isotope sarnple testcd at 8 watts/cm-
C while the iso~spically pure samplc tested at 12 watts/cm--C. Thus, the thermal
conductivity of polyclystalline isotopically pure carbon-13 material unexpectedly has a
35 much higher th~mal conductivity than the polycTystalline diamond made from natural
- abundance isotopes. The value of the the~mal conductivi~ can only be improved by paying
atten~on to dislocations, vacancies, vacancy c3usters, and like factors that tend to depress
the th~nal conductivity of the polycrystalline diamond pieces. Controlling grain.~ .
, .
, ' '
~Q7~
60 D 0531
boundanes also is important as obvious loss of therrnal conductivit,Y is expenenced, though
not ncarly to thc degree with isotopically pure polycrystallinc diamond than with natural
isotopc abundant polycrystalline diarnond.
As noted above, the isotopically-pure polycrystalline diamond can be grown by
SCVD ~echniques, or can be grown by high pressure/high temperature (HP/HT) techniques
including growing the polycrystallinc diam~nd directly, or growing the polycrystalline
diarnond and then sintering the diamond to form an appropriate piece. Though HP/HT
techniques are well known in the art, reference to the following patents provides details on
such processing conditions: U.S. Pat. No. 3,141,746; 3,381,~28; 3,609,818; 3,745,623;
103,831,428; and 3,850,591, Ihe disclosures of which are expressly incorporated herein by
referencc.
With rcspect to conventional CVD processes useful in the present invention,
hydr~carbon/hydrogen gaseous mixtures are fed into a CVD reactor as an initial step.
Hydrocarbon sources can includç the methane series gases, e.g. methane, ethane, propane;
15unsanlrated hydroearbons, e.g. ethylene, acctylene, cyclohexene, and benzene; and the
likc. Methane, however, is preferred. Use of either carbon-12 or carbon-13 for these
hydrocarbon sour~es is made in accordance wilh the precepls of thc present invendon. The
mol~r ratio of hydTocarbon to hydrogen broadly ~anges &om about 1:10 to about 1:1,000
vith about 1:100 being prefe~d. l~is gaseous mixture opdonally may be diluted with an
20inert gas, e.g. argon. The gaseous mixture is at least pardally deeomposed thermally by
one of sevcral tecbniques known in thc ar~. One of thesc techniques involves the use of a
hot filament which normally is fo~mcd of tungsten, molybdenum, tantalum, or alloys
- thereof. U.S. Pat. No. 4,707,384 illustrates this process.
The gaseou3 mixturc partial decomposition also can be conducted with the
25assistancc of d.c. discharge or radio frequency electromagnetie radiadon to generate a
plasma, such æ proposed in U.S. Pats. Nos. ~,749,587, 4,767,608, ~nd 4,830,702; and
U.S. Pat. No. 4,434,188 with respect to use of n~icrowaves. Thc substrate may bebombarded with cl¢c~rons dunng the CVD dccomposition p~cess in a~cordance with U.S.
Pat. NQ 4,740,263.
3ûRegardless of the particular method used in generadrlg thc pamally decomposed
gaseou~ mihcture, the substratc is maintained at an clcvated CVD diamond-folrning
ten~c~at~ which typieally ranges from abou~ 500 to 1100~ C and ps~ferably in the rangc
of about 850 to 950J C where diamond growth is at its highes; rate in order to minimize
grain size. Pr~ssu~es in thc range of firom aboult 0.01 to 1000 Tog, advantageously about
35100-800 Tolr, are taught in thc art, with reduced pressL~ being prefe~ed. DetaiJs on CVD
,! processcs additionally can be reviewed by referencc to Angus, e~ al., "Low-Pressure,
MetastaUe G~ h of Diamond and 'Diamondlike' Phases", 5cfencc, ~ol. 241, pages 913-
921 (August 19, 1988); and 13achmann, e~ al., "Dia~nond Thin Films", Chemical and
-5-
6 OSD~
Engineering New3, pages 24-39 (May 15, 1989). The disclosurcs of all citations hercin
a~e expressly inco~orated herein by referencc.
`:
"
.
','
~,
; '
~'.
,^-` ' .
i..
d
,. .
.`
''
'j'~'
' ,' ' - .
.r
... .
'
Y~ ' ' , :
, ., . ' , ,