High Temperatures ^ High Pressures, 2001, volume 33, pages 489 ^ 501
DOI:10.1068/htjr014
Optimised cell design for high-pressure synthesis of diamond
Chien-Min J Sung }
Kinik Company, Taipei, Taiwan; Taipei University of Technology, Taipei, Taiwan; fax: 886-2-8677-2171;
email: [email protected]
Received 1 February 2000
Abstract. More than 150 t of diamond grits (China makes and uses about half of the world
supply) are consumed each year in sawing constructional materials (eg concrete, granite, marble).
These diamond grits are typically synthesised under high pressure and high temperature with
graphite as the source material and a molten metal (eg iron, cobalt, nickel, or their alloys) as the
catalyst. Graphite and catalyst may be either stacked in the reaction cell as alternate layers or
they can be mixed as powders. The reaction cell may be small (< 50 cm3 ) with large gradients
of pressure and temperature, as in the case of a cubic press, or it can be large (up to 500 cm3 )
with relatively uniform distribution of pressure and temperature, as in the case of a large
(10000 t) belt press.
The current arts are such that if the reaction cell is small, and the charge contains layered
graphite and catalyst, then the diamond yield is about 1 carat cmÿ3 of the compressed cell, with
about 1/3 of the grits adequate in size and quality. If the reaction cell is large, the diamond
yield is about 2 carats cmÿ3 with about half as desirable grits. If powdered graphite and catalyst
are thoroughly mixed, and seeded crystals are dispersed uniformly in them, the diamond yield
in the latter case could be boosted to about 3 carats cmÿ3 with about 3/4 as high-grade products.
By mixing powdered graphite and catalyst and pressing the mixture in thin layers, catalystcoated diamond seeds may be planted uniformly in these disks. With the full utilisation of
the source materials, and the reaction volume, the production yield may attain more than
4 carats cmÿ3 with more than 3/4 as high-grade saw diamonds.
1 Introduction
Modern civilisation has been greatly aided by the widespread use of industrial diamonds
(figure 1). In particular, the building materials (eg concrete, granite, marble) may not be
so readily available without the prevailing applications of diamond saws, drills, and
300
Weight=t
200
synthetic
100
natural
0
1945 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000
Year
Figure 1. The world consumption of industrial diamonds has been increasing steadily, and it has
now approached 400 t (2 billion carats). Almost all industrial diamonds used are synthetic, and
among them the saw grits account for half of the quantity produced (mostly in China).
} Also known as James C Sung.
490
C-M J Sung
grinding wheels. With the employment of diamond, the unique stone tools, we can now
live comfortably in a revitalised stone age (Sung 1998).
The per capita consumption of stone in the world is about 8 kg of primary
materials (16 kg of quarried stone) and 8 m2 of finished products. For industrialised
countries, the per capita consumption is more than each person's body weight. It may be
interesting to note that each carat (0.2 g or 0.05 cm3 ) of diamond may cut about
105 cm3 of stone (eg marble) so the diamond/stone ratio is less than 1 ppm, that
represents the typical diamond content in a kimberlite pipe. Moreover, the diamond
segments used to cut the stone contain about 5% in volume of diamond grits. This ratio
(about 1 carat cmÿ3 ) is also similar to the high-grade saw diamond synthesised in a
high-pressure reaction cell. Furthermore, the amount of diamond actually worn out by
cutting the stone is less than 10% of that contained in diamond segments. This
proportion is comparable to the kerf loss for processing the stone (Sung 1999).
Saw grits are the largest application of industrial diamonds. Almost all of them are
synthetic grits, with an annual sale of over US $600 million. Saw grits are also the largest
(20 to 50 mesh, or 1 mm to 0.3 mm) in size, and hence the most difficult to synthesise.
The manufacturing processes for saw grits have been highly guarded industrial secrets.
Although there are over 20 countries (USA, Canada, S. Africa, Ireland, Sweden,
England, Germany, Russia, Ukraine, Poland, Czech Republic, Romania, Armenia,
Greece, China, Japan, S. Korea, N. Korea, India, Malaysia) and over 500 companies
(China has the majority) producing industrial diamonds worldwide, only three
companies in the world are capable of making high-grade saw diamonds. They are
General Electric (I was the responsible manager in the late 1970s and early 1980s) of
the USA, De Beers of South Africa, and Iljin (I guided the plant design and provided the
startup technology in the late 1980s) of Korea.
The most advanced technologies in growing saw grits today employ stringent
methods in controlling the uniformity of raw materials, and the steadiness of synthesis
processes. The state-of-the-art of diamond technology is such that the high-pressure
compressed reaction volume can be more than 200 cm3 (eg, for a press capacity of 5000 t
or more). Each cubic centimetre of the compressed reaction cell may grow up to
3 carats of diamond (17% of volume utilisation) with the graphite-to-diamond conversion
rate of up to 40%. However, even so, the produced diamonds still contain at least 1/3
that are not desirable in size or quality.
In this work I have explored the limit of growing saw diamonds by maximising the
number of good crystals that may be grown simultaneously in a reaction cell. The
approach is to make full use of the precious reaction volume by controlling the
nucleation site so the crystals can grow close to one another without interference.
Moreover, the optimum selection criteria of graphite and catalyst materials are
suggested, so the highest quality of diamond can be produced in the shortest possible
time.
2 High-pressure diamond synthesis
The above three industrial diamond leaders are synthesising saw grits under high
pressure with belt apparatus (Bovenkerk 1962, figure 2). Such apparatus can squeeze a
reaction cell to reach a pressure of over 5 GPa, that is about the pressure at the centre
of the Moon. An electric current is passed through the cell to heat its contents to over
1200 8C. Under such conditions, saw grade diamonds are synthesised in about 30 min.
The exact time is dependent on size and quality of the diamond desired (Sung 1997).
The reaction cell typically contains alternate layers of graphite and metal
(Bovenkerk 1960, 1961). Metal is used as the catalyst to accelerate the graphite-todiamond transition. The original list of catalysts included twelve transition elements
(Bovenkerk et al 1959). The most used catalysts today are alloys of iron, nickel, and
Optimised cell design for high-pressure synthesis of diamond
491
anvil
metal
die
gasket
Figure 2. Typical belt apparatus showing that the
reaction cell is enclosed in the high-pressure assembly. (Source: US Patent and Trademark Office;
redrawn from US Patent number 3 031 269.)
cobalt. For example, a popular alloy is Invar, Fe65Ni35 (Bovenkerk 1961; Kennedy et al
1976) or similar compositions, eg Fe70Ni30 (Strong 1960; Showa Denko Co. 1984).
Moreover, the catalyst may be deployed in a layered design with an iron disk sandwiched
between two nickel foils (Darrow 1967; Strong 1968). Such a design may cause the
nucleation of diamond to be delayed until the cell reaches a uniform temperature.
The ratio between graphite and metal may be about 3 : 1 by volume (Hall 1960). The
graphite could be as thin as 1.5 mm (Sumitomo Electric Co. 1985), and catalyst 0.13 mm
(Strong 1968).
Once the reaction cell reaches the pressure and temperature required for diamond
synthesis, the catalyst will melt, often at the eutectic temperature suppressed by the
dissolution of carbon from graphite. Within a few minutes of catalyst melting, diamond
nuclei will appear at the interface of graphite and catalyst. These nuclei will grow at a
rate of approximately 1 mm hÿ1. For growing high-strength saw grits of 30 to 50 mesh
(0.6 mm to 0.3 mm in size), the growth time is about half an hour (Sung 2000a).
3 Catalytic mechanism
Saw grits are synthesised in the stability of diamond. According to the common
understanding, graphite will dissolve into the molten metal and diamond will precipitate
out because of its lower solubility. However, this simple solution model cannot explain
why the graphite microstructure can drastically affect the type of diamond formed: for
example, nongraphitised carbon can only form diamond, if any, at a much higher
pressure. Moreover, amorphous glassy carbon must first crystallise to form graphite
before transforming into diamond (Hirano et al 1982). Furthermore, the higher the
degree of graphitisation of the carbon source, the faster the diamond can nucleate and
grow (Hirano et al 1984).
The solution mechanism is also incapable to explain why certain metals, eg copper
and zinc, cannot form diamond at all, although they may dissolve a minute amount of
carbon. The apparent dependences of diamond formation on graphite structure and metal
chemistry were attributed to the catalytic action. This catalytic mechanism was proposed by Gou (1972) based on the electronic interaction between the empty 3d orbitals of
transition metals and the unbonded (p bond) 2p electron of carbon (Sung and Tai 1997a).
According to Gou's model, when the catalyst metal melts, its atoms can assume a
configuration of pseudo-closest packing in short ranges, resembling the (111) face of
a face-centred cubic (fcc) lattice. If the metal atoms can cover roughly every other
carbon atom on the graphite basal plane, then their empty 3d orbitals may attract the
dangling 2p electrons of carbon. This attraction could pull every other carbon atom
toward the metal atoms. As the result, the graphite hexagon may pucker in such a way
C-M J Sung
2.46 A
492
B
A
R
C
R
y
y
rˆR
r
A
a
A
Figure 3. Puckering of graphite into diamond. The upper diagram shows that large catalyst atoms
can match with all other carbon atoms on the graphite basal plane and attract them to form a
`chair' configuration. The lower diagram depicts the domino effect of puckering to transform
a rhombohedral graphite (3R) into cubic diamond (middle), or a hexagonal graphite (1H) into
hexagonal diamond (lonsdaleite). (Sung and Tai (1997a, 1997b; reprinted with permission).
that the unmatched carbon atoms would move toward the opposite direction, ie toward
the next layer of the graphite basal plane. This puckering can deform a plane hexagon of
graphite to the chair type, the characteristic lattice configuration of cubic diamond
(Sung and Tai 1996).
If graphite has the right stacking sequence of hexagon layers, the carbon atoms
moving away may interact with matching carbon atoms in the next layer to form
diamond bonds. This puckering action may sweep through the graphite lattice like a
wave (figure 3). As the consequence of this domino effect, an entire grain of graphite
may transform into diamond (Sung and Tai 1997b).
Optimised cell design for high-pressure synthesis of diamond
493
The above model explains the catalytic mechanism of the graphite ^ diamond
transition. In order for this mechanism to operate, carbon must first form a graphite
structure, and the catalyst must possess empty d orbitals. Both requirements are
consistent with the empirical observations. For example, glassy carbon cannot form
diamond, nor can d-orbital full transition metals (eg copper, zinc). In order to facilitate
the formation of diamond, the graphite structure must be well developed, and the
catalyst must have empty d-orbitals.
However, the above model has two deficiencies. First, the common graphite is
predominantly 2H type, ie the stacking sequence is ABAB ... . As such, only half the
atoms can find matching ones in adjacent layers, hence the puckering of basal planes
cannot transmit to the next layer. In order to allow the puckering to sweep across the
structure, graphite must be either rhombohedral (3C) type with ABCABC ... sequence,
or hexagonal (1H) type with AAA ... sequence. 3C can pucker into cubic diamond; and
1H, hexagonal diamond (lonsdaleite) (Sung and Tai 1997b).
As diamond synthesised under high pressure is exclusively in the cubic structure,
it follows that ordinary graphite must first slide into the rhombohedral form
before puckering into diamond. Graphite basal planes are held together by the weak
van-der-Waals force. The activation energy for shuffling these planes is 0.076 eV, less
than 2% of that required to break a graphite bond (4.8 eV). This activation energy may
be easily overcome at a temperature above 900 8C. Moreover, the sliding of basal
planes can be conveniently accomplished when graphite recrystallises under the influence
of molten catalyst, a common phenomenon taking place before the formation of
diamond.
The more empty d orbitals a catalyst atom has, the stronger is its attraction toward
the carbon atom. As a result, when the deficiency of d orbitals is too high, there is a
tendency for the transition metal to form a carbide. Strong carbide formers (eg titanium,
vanadium) may not catalyse diamond formation.
In order for a catalyst to pucker the graphite basal plane, it must do so without
sticking to it permanently. Thus the golden rule for the diamond catalyst is ``touch and
go''. The best manifestation of this phenomenon is the dissolution of carbon in the
catalyst. When carbon atoms are dissolved as solute, they touch and go without forming
carbide. The ability for the catalyst to moderate its interaction with graphite is also
revealed in its solubility of carbon. Hence, the higher the solubility of a catalyst, the
stronger its interaction with carbon without forming a compound. In other words,
the solubility of carbon of a transition metal is the indicator of its catalytic power for
converting graphite into diamond. This power may be manifested in the thickness of a
graphite flake that it can pucker into diamond.
4 Ranking of catalysts
Based on the above model of atomic matching and ``touch and go'', the relative catalytic
power of transition metals can be calculated on the basis of their ability to nucleate,
grow, and form diamond. Moreover, if the catalyst has a lower melting point, it could
also reduce the pressure of the diamond synthesis. Lower synthesis pressures allow a
larger reaction volume for a higher output, or a longer carbide life (eg die) for a lower
cost. Hence, low melting point is another merit to be considered for selecting the
catalyst (Sung and Tai 1995/1996).
5 Diamond formation routes
According to the solution model, diamond nuclei ought to form inside the original
volume of molten catalyst. In an experiment performed in 1977, I observed that the first
diamond nuclei were actually precipitated out inside the original volume of graphite
where molten metal intruded. Moreover, macroscopic graphite flakes were recrystallised
494
C-M J Sung
in regions where diamond had not yet nucleated. However, as soon as a diamond
nucleus was formed, it served as the carbon sink and sucked up microscopic flakes
suspended in molten catalyst. As a result, microscopic flakes could not accumulate to
form macroscopic flakes.
Once a diamond nucleus reaches a critical size, it will continue to grow in the
graphite volume. However, there is always a thin film (< 0.1 mm) of molten catalyst that
separates diamond from graphite (figure 4). This thin film will wrap around the
diamond and it grows continually along with the diamond. Meanwhile, diamond is
feeding on microscopic flakes of recrystallised graphite that drift across the strait of
thin film.
It would appear that the sequence of events for diamond synthesis under high
pressure is as follows:
(i) molten metal invades into graphite in structurally weak regions;
(ii) graphite disintegrates to form microscopic flakes;
(iii)catalyst atoms penetrate microscopic flakes by intercalation, and then shuffle the
stacking sequence to become rhombohedral graphite;
(iv) rhomobohedral flakes are puckered under the influence of molten catalyst and they
stick together to form a diamond nucleus;
(v) the diamond nucleus grows by continually feeding on microscopic flakes;
(vi) as the diamond grows in size, molten catalyst is pulled in by the capillary effect, so
the thin metal envelope can expand continually around the growing diamond.
The above model suggests that diamond is not nucleated from supersaturated
solution of carbon atoms, but by puckering suspended graphite flakes. The growth of
diamond is also dominated by adding microscopic chunks of flakes instead of discrete
solutes. This conclusion is consistent with the observation that diamond facing graphite
grows nearly twice as fast as that facing catalyst (growing predominantly by adding
atoms). Moreover, when microscopic diamond is dissolved in molten catalyst, diamond
graphite
diamond
catalyst
Figure 4. Diamond is wrapped in a metal film that is impregnated with suspended graphite
flakes. These flakes are continually disintegrated from the surrounding graphite. The immersed
flakes are reshuffled in stacking sequence to form rhombohedral graphite. When these recrystallised flakes reach the diamond surface, they have been puckered and may attach to the growing
diamond in chunks measured in micrometres. The catalyst side of the diamond crystal grows
slowly by the attachment of dissolved carbon atoms. The unequal rate of growth is often
reflected in the asymmetrical appearance of the diamond crystal. Note that there is a temperature gradient across the catalyst envelope. Graphite, with electrical resistance ten thousand times
higher than that of molten metal, is hotter than the catalyst metal in the corner regions where the
graphite is incorporated to form rays of inclusions.
Optimised cell design for high-pressure synthesis of diamond
495
may not nucleate at all. This situation is demonstrated when gem diamond is grown
with a seeded crystal by the temperature gradient method. In this case, the growth rate
achieved by feeding dissolved diamond atoms is comparable to that of growing saw
diamond in the catalyst side. This growth rate is about half of that facing graphite.
54
e
edg
solid met
al
liquid m
etal
6 Layered cell versus powdered cell
The key to growing high-grade saw grits is to make sure that diamond nuclei form
uniformly and at the same time. If diamond nucleates randomly, the sparse region is
wasted and the dense region is overly packed with poor crystals. The result would be a
low yield with diamond ridden in defects (eg body inclusions and surface roughness).
Even if the nucleation is uniform, if nuclei continue to form, local pressure may be
reduced around the diamond already present. Moreover, the early formed diamond will
block the flow of electric current and cause more heat to be generated nearby. The
disturbed temperature field will make the subsequent diamond growth erratic. Thus, in
order to improve the diamond yield and quality, the nucleation of diamond must be
controlled.
Conventional arts of diamond synthesis depend on random nucleation at the
graphite ^ metal interface. Because of the inevitable formation of nuclei clusters, the
spacing of diamond must be sparse to allow diamond to grow without much interference.
If the reaction cell is small (< 50 cm3 ), as in the case of using a cubic press, less than 1
carat cmÿ3 of diamond may be formed. In this case, the graphite-to-diamond
conversion ratio is less than 20%. Even so, less than 2/3 of the saw grits may fall in the
desirable range of sizes. Worse still, because very few crystals are perfect, the proportion
of high-grade saw grits is less than 1/3.
If the reaction cell is large (> 50 cm3 ), with the proper design of high-pressure
assembly, the gradients of pressure and temperature inside the reaction cell may be
smaller (figure 5). Moreover, nucleation of diamond can be made relatively uniform, by
tightening up the specifications of materials, and by homogenising the surfaces of
Nucleation
Growth
cent
e
edg
re
52
centre
Pressure=GPa
small cell
large cell
Dissolution
etal
nd ^ m
diamo
l
e ^ meta
graphit
Graphitisation
50
1200
1300
1400
1500
Temperature=8C
Figure 5. Examples of P ^ T profiles that may exist in a large reaction cell compared to a small
one. The time arrows indicate the trend. Note that compression and heating accompany the
diamond growth.
496
C-M J Sung
graphite and catalyst. Nucleation may also be controlled by profiling the pressing force
and heating power. As a result, the diamond yield may increase to about 2 carats cmÿ3.
As the distribution of sizes and quality are tightened up dramatically, so the high-grade
portion in the driving sizes may attain 1/2. Even so, the reaction cell is still grossly
under-utilised because the diamond distribution in the cell is random.
The layered distribution of graphite and catalyst has several limitations for further
increase of diamond yield. Because the geometry is essentially two-dimensional, the
volume in the third dimension is not fully utilised. Moreover, graphite is structurally soft
but with high electrical resistance, so its pressure is low but its temperature is high. This
is in contrast with the catalyst, which is high in pressure but low in temperature. The
local pressure and temperature gradients may affect the diamond growth in such a way
that the crystals are growing faster in the graphite volume. Hence, diamond crystals are
usually asymmetrical.
Moreover, layered materials are intrinsically heterogeneous. In the case of graphite,
the firing temperature may vary from region to region. The microstructure of certain
regions may also change during the forming process (eg the skin effect caused by extrusion).
As a result, the graphite disks may have large differences of the degree of graphitisation,
the content of porosity, and the amount of ash from the centre to the periphery. In the
case of the catalyst, the compositional segregation is inevitable as the ingot solidified
slowly. The composition further varies during the mechanical rolling of the metal. Hence,
the concentration of key elements may vary by several percent from place to place.
The uniformity of materials can be greatly improved by using powders. In the case
of graphite, powders (typically < 50 mm) with high purity (impurities < 20 ppm) and
high degree of graphitisation (> 90%) can be produced. Moreover, powders of catalyst
(typically < 20 mm) are formed by gas spraying that can quench the molten alloy at a
rate of > 104 8C sÿ1, so the composition variation can be less than 10ÿ3.
With these apparent benefits, in recent decades, major diamond makers have turned
to using cells loaded with powder mixtures for further increases of diamond yield. As
the distribution of diamond in a powdered cell is essentially three-dimensional, the
growth of diamond is more symmetrical. Moreover, in the layered cell, the diamonds
tend to grow fast and continuously, but in the powdered cell, because the supply of
carbon nutrient is gradual, diamond grows more slowly and in an interrupted fashion.
As a result of this equilibrium growth, the diamond sizes are closer to one another.
Moreover, the amount and the size of inclusions in the diamond are greatly reduced.
Hence, the quality of the diamond (eg impact strength and thermal stability) is
significantly improved.
The use of powdered materials can drastically increase the contact surface areas,
and hence more interfacial sites may be available for diamond nucleation. Moreover,
small diamond particles (typically < 50 mm) may be added as the seed to bypass the
diamond nucleation, which is often difficult to control. If the materials can be prepared
correctly and the processes are optimised, each growing diamond crystal will be
completely wrapped in a thin (about 0.05 mm) envelope of catalyst metal. As a result, the
crystals are highly uniform in size and quality. Hence, the gross yield may be increased to
about 3 carats cmÿ3 of the cell volume. However, the major improvement is not in the
gross yield, but in the yield of driving size and desirable grade. As a result, up to a fivefold
increase of high-grade diamond may be obtained in the saleable product when compared
with the conventional technology based on the layered design of a reaction cell.
However, the powder method is not without problems. The densities of heavy catalyst
metal and light graphite powder differ by a factor of three, so they are intrinsically
difficult to mix well. Even if the mixing can be done properly, the subsequent transfer of
the mixture often causes them to segregate. Thus, intriguing methods have been
invented to homogenise the mixture. For example, the mixture may be consolidated by
Optimised cell design for high-pressure synthesis of diamond
497
Yield=carats
Good crystals=%
cold isostatic pressing. The densified body is crushed to powder. The latter is remixed,
and then consolidated by pressing again. Such a method can minimise the segregation
of graphite and diamond, but often with increased amount of contamination and
handling cost.
Powdered materials, with enormous amounts of surface area, are prone to oxidation
and hydration. As a result, their colour may change (eg with a blue-purple tint formed
on the surface when oxygen content is more than 100 ppm). In such a case, the
synthesis pressure will increase with the consequence that the diamonds formed become
finer. Moreover, powdered materials are susceptible to surface contamination that may
also cause erratic nucleation of diamond. In order to minimise the oxidation and
contamination, powders are typically purified in hydrogen atmosphere at high
temperature, and subsequently they should be stored under inert gases before use.
A more tricky problem with the powdered cell is that its operation window is much
narrower than that of the layered cell. Hence, its yield is much more sensitive to
material fluctuations and parameter drifts. For example, the sizes of graphite and
catalyst and their mass ratio may have a dramatic effect on the diamond yield and
quality, as depicted in figure 6. The strong dependence of diamond output on the powder
size and quality may require more stringent control of materials.
coarse
fine
Graphite size (mesh)
coarse
fine
Catalyst size
Graphite/catalyst
(mass ratio)
Figure 6. The possible effects of graphite size (left), catalyst size (middle), and graphite/catalyst
mass ratio (right) on diamond yield and quality in a reaction cell that contains a powdered charge.
Alternatively, if diamond seed crystals are used, they may be coated by the powder
of catalyst first, eg by electroless deposition or fluidised suspension (Sung 1995). The
presence of a metal envelope can assure the readiness for diamond growth. The coated
diamond seeds can then be mixed with graphite powder to reduce the chance of
segregation. This method may be quite effective, but the task of encapsulating diamond
seed with catalyst metal is not trivial.
7 Optimum cell design with positive nucleation control
Despite the fact that powdered cells may give high diamond yield and better crystal
quality, the layered cell has many intrinsic advantages, such as that it is simple to handle
and easy to control. Moreover, a layered structure is particularly suited for belt
apparatus of uniaxial design. In this case, the pressure is applied and the current is
transmitted perpendicular to the stacked layers. This uniaxial symmetry of cell design
can allow a more uniform field of pressure and temperature, two critical factors for
diamond growth. The problem with layered structure is nucleation control and interface
uniformity. As the result of random distribution and erratic growth, diamond yield and
quality may suffer.
498
C-M J Sung
The nucleation can be controlled by seeding smaller diamond particles (eg 50 mm in
size), for example, by fitting these diamond particles in holes drilled in the catalyst
(Showa Denko Co. 1984). A better way of controlled seeding is to guide diamond
particles by a template. The seeded crystals can then be pressed directly into either
catalyst or graphite (Sung 2000b, 2000c, 2000d). The diamond seed can be buried
halfway to allow for the protrusion into the graphite volume when the stack of graphite
and catalyst layers is assembled (figure 7). In such a configuration, the molten catalyst
can easily follow the diamond profile to form the metal envelope. This metal strait can
facilitate the transport of graphite flakes across so the diamond can grow.
graphite
diamond
catalyst
metal-coated diamond
metal-impregnated graphite
Figure 7. The planting of diamond seeds in thin layers of graphite or catalyst can dramatically
increase the diamond yield, throughput, and quality (left diagram). Layered graphite and catalyst
can be replaced by layered-catalyst-impregnated graphite or pressed mixture of them for further
increase of volume utilisation of the reaction cell (right diagram). In this schematic diagram, the
thin catalyst envelope that surrounds each diamond crystal is not shown
The seeded diamond may be overcoated with a thin layer of catalyst metal, eg by
electroplating, vacuum deposition (Chen and Sung 1991), or powder adhesion (Sung
1995). This catalyst envelope can draw other catalyst particles to grow with the diamond.
The catalyst sheath in the molten state is impregnated with suspended graphite flakes
that feed the growing diamond.
Assuming the saw grits to be grown are 40 to 50 mesh (0.4 mm to 0.3 mm), the
most popular size used in diamond saws, then the largest crystals (40 mesh) have a size
of 0.4 mm and a surface area of about 0.5 mm2. The thickness of a catalyst envelope is
less than 0.1 mm, hence, after the synthesis, the volume of the catalyst covering each
crystal is 0.05 mm3 , or less than 0.5 mg. Such an amount of catalyst can form a sphere of
about 0.4 mm in diameter, about the size of the diamond crystal itself.
Assuming the diamond-to-diamond distance is twice the largest diamond size,
then the growing area is four times that covered by the diamond, or 0.7 mm2. The
thickness of the catalyst can be 0.15 mm, more than twice that needed to envelop
the crystal.
The graphite thickness can be 0.3 mm, twice that of the catalyst. However, the mass
of the former is only about 40% of the latter.
In the above example of cell design, the diamond grown will span the total
thickness of graphite and catalyst. Each crystal will have a growing space of about
0.3 mm3. Hence, each cubic centimetre of the reaction cell will produce more than
3000 crystals, which is more than 4 carats for 40 mesh diamond (770 pieces per carat).
In this case, the graphite-to-diamond conversion rate is nearly 50%, a very high value
for growing good-quality diamond. However, because the distribution of diamond
crystals is regular, each diamond occupies less than 15% of the reaction volume, so there
is ample room for it to grow uninterruptedly. The free growth of diamond can allow the
development of euhedral crystal shapes with a minimum amount of inclusions.
Optimised cell design for high-pressure synthesis of diamond
499
In the above cell design it is assumed that the graphite layer and catalyst layer are
uniform in structure and composition. However, the reality is that these materials are
quite heterogeneous. Hence, a better design is to use powdered graphite and catalyst with
more uniform properties. Instead of mixing them and loading them into the cell, the
mixture is first pressed to form disks, and then these disks are planted with diamond
seeds that are coated with catalyst metal. The powder layers with uniformly distributed
diamond seeds are then loaded into the reaction cell for the diamond synthesis. Such a
design has the advantages of both the layered and the powdered cell. Moreover, the
positively planted diamond seeds guarantee that the diamond crystals grown are of the
same size and quality. Furthermore, because of the elimination of the catalyst layer,
the efficiency of cell volume utilisation is further improved. In such a cell, it is possible to
achieve a yield of 5 carats cmÿ3 with at least 4 carats of the desirable size and quality.
The diamond yield for this optimised cell design is compared with that for current
arts of diamond synthesis, as shown in table 1.
Table 1. Comparison of diamond yields.
Cell size
Small
Large
Large
Large
Large
Charge type
layer
layer
powder
layer
powder
Nucleation
uniformity
Seeding used
random
random
random
regular
regular
no
no
no
yes
yes
Yield=carats cmÿ3
gross
good
1
2
3
4
5
0.3
1
2
3
4
To achieve the optimised yield and quality, not only must the cell design be
streamlined, but the material selections are also critical. The catalyst chosen should
have a closest packing structure before melting, preferably in fcc configuration.
The interatomic separation should be close to 2.46 Ð, resembling the distance between
every other carbon atom. The catalyst should also have half-filled d-orbitals, similar to
iron, the most powerful catalyst for converting graphite to diamond.
Candidates of catalysts based on the above criteria include General Electric's
preferred Fe65Ni35 with a melting point of about 1300 8C, or the Chinese favour of
Ni70Mn25Co5 that melts near 1210 8C. Other possibilities are Ni70Mn30 (melting point
1200 8C), and Ni40Mn30Fe30 (melting point 1240 8C). It is preferable to keep the
melting point low so the diamond may be synthesised at lower pressures. One good
choice is to use Cu ^ Mn alloy. This alloy has an eutectic point at 871 8C for Cu66Mn34.
Although copper is not a catalyst because its d-orbitals are full, when it is combined
with manganese in an alloy, its electrons will be shared by the manganese. For example,
CuMn will have an electronic configuration resembling cobalt; and CuMn2 , like iron.
Moreover, the atomic size of copper is slightly smaller than iron, so it can balance the
oversized manganese, and make the alloy atoms match with the graphite atoms better.
Cu ^ Mn may also alloy with other more popular catalytic compositions, so as to
form Cu ^ Ni ^ Mn ^ Co, which may melt in the temperature range 900 ^ 1200 8C.
Examples are Ni70Mn20Cu10 (melting point 1190 8C), and Ni45Mn30Cu35 (melting
point 900 8C).
For the conventional synthesis, diamond tends to float on top of the catalyst, unless
a DC current is used to heat the charge with the anode pointing upward. In the seeded
growth, diamond is also located on the upper side of the catalyst. Hence, diamond will
grow upward toward the interior of graphite. As the largest crystal is 0.6 mm across in
our example, the graphite thickness may be kept at the same thickness.
500
C-M J Sung
The graphite used must be at least 85% ^ 90% graphitised. The degree of
graphitisation (G ) may be estimated from the separation [d…002† ] of basal planes based on
the following equation:
G=A ˆ 3:44 ÿ
d…002† =A
,
3:44 ÿ 3:35
where 3.44 Ð is the spacing of basal planes for amorphous carbon (L c ˆ 50 Ð), and
3.35 Ð is the spacing for pure graphite (L c ˆ 1000 Ð). The high degree of graphitisation
is achievable by sintering graphitisable carbon at 2500 8C for 12 h.
Based on the above cell design, each diamond crystal of 0.6 mm in diameter will
occupy a volume of 0.014 cm2 0:07 cm ˆ 0:001 cm3. The crystal of 0.6 mm is about
0.004 carat, so the reaction cell is capable of achieving nearly 4 carats cmÿ3.
Because the diamonds are all grown from seeded crystals, most of the diamond
crystals so formed are of the driving size and high quality. In essence, this productivity is
at least double that of the commercial diamond synthesis performed today.
8 Conclusions
The commercial production of saw diamond grits employs mainly two types of highpressure apparatus, small cubic presses as used mainly in China, and large belt
apparatus as adopted by world leaders. The former have small reaction cells with large
pressure and temperature gradients, and the latter possess large charge chambers with
more uniform pressure and temperature distributions.
If alternate layers of graphite and catalyst are used in a small reaction volume, the
diamond yield is typically 1 carat cmÿ3 with less than 1/3 of the grits falling in the
range of desirable sizes and quality. In the case of a large charge chamber, the diamond
yield can be increased to 2 carats cmÿ3 with about half the diamond as good grits.
On the other hand, if the layered graphite and catalyst are replaced by powdered charge,
and the latter is thoroughly mixed with uniformly distributed seeded crystals, the
diamond yield may be boosted to about 3 carats cmÿ3 with about 2/3 grits of high grade.
However, the layered structure is intrinsically more compatible than powdered charge in
large belt apparatus that employs uniaxial pressing force. If the former is optimised in
materials selection and geometrical design, the volume efficiency for diamond synthesis
would be higher than the latter. The powdered cell has the inevitable problems of
powder segregation between light graphite and heavy metal. This limitation makes it
harder to reach the highest diamond yield possible.
The efficiency of diamond synthesis with charge in a layered structure can be
increased by optimising the cell design. By choosing highly graphitised graphite as the
source material, and a low-melting-point solvent as the catalyst, the graphite-todiamond transition pressure and temperature can both be lowered. Moreover, by using
seeded diamond crystals planted at the interface of thin graphite and catalyst with an
optimised separation, these seeds can grow at the same rate without interfering with each
other. The result would be saw diamond grits essentially in the same size and with high
quality. Hence, the yield of high-grade saw diamond in the high-pressure reaction cell
could be as high as 4 carats cmÿ3 with virtually all grits as high-grade products of the
right size.
If powdered graphite and catalyst can be mixed and pressed to form layers, and
these layers are planted with diamond seeds that are overcoated with catalyst metal, such
a cell design has the merits of both layered and powdered cells. Moreover, the diamond
growth is positively controlled in location and time. As a result, the diamond yield may
attain 5 carats cmÿ3 with 4 carats in the desirable range of size and quality. This product
yield is more than twice that of the current state-of-the-art commercial processes.
Optimised cell design for high-pressure synthesis of diamond
501
The large production presses used today range from 5000 t to 10 000 t. The reaction
volumes range from 150 to 400 cm3. Layered cells of these sizes can yield diamond
of 300 to 800 carats. As discussed in this paper, this yield may be doubled by using
layers of mixed powder of graphite and catalyst, and planting into these layers a
diamond grid of seeded crystals.
References
Bovenkerk H P, 1960, US Patent number 2 941 252
Bovenkerk H P, 1961, US Patent number 2 992 900
Bovenkerk H P, 1962, US Patent number 3 031 269
Bovenkerk H P, 1969, US Patent number 3 423 177
Bovenkerk H P, Bundy F P, Hall H T, Strong H M, Wentorf R H, 1959 Nature 184 1094 ^ 1098
Chen S H, Sung C-M, 1991, US Patent number 5 062 865
Darrow, 1967, US patent number 3 310 501
Gou C C, 1972, unpublished report at Beijing 152 Factory and at the conference on ``The Formation
Mechanism of Synthetic Diamond Crystals'', Jengzhou, China
Hall H T, 1960, US Patent numbers 2 947 611, 2 947 608
Hirano S, Shimono K, 1982 J. Mater. Sci. 17 1856 ^ 1862
Kennedy C S, Kennedy G C, 1976 J. Geophys. Res. 81 2467 ^ 2470
Showa Denko Co., 1984, Japanese Patent number 59-164 609
Strong H M, 1960, US Patent number 2 947 609
Strong H M, 1968, US Patent number 3 407 445
Sumitomo Electric Co., 1985, Japanese Patent number 60-12 998
Sung C-M, 1978, in Proceedings of the Sixth AIRAPT International High Pressure Conference
Ed. K D Timmerhaus (New York: Plenum Press) pp 31 ^ 42
Sung C-M, 1995, in Fabrication and Characterization of Advanced Materials Eds S W Kim,
S J Park (Materials Research Society of Korea) pp 267 ^ 273
Sung C-M, 1997 High Temp. ^ High Press. 29 253 ^ 293
Sung C-M, 1998 Ind. Diamond Rev. 4/98 134 ^ 136
Sung C-M, 1999 Diamond Rev. Mater. 8 1540 ^ 1543
Sung C-M, 2000a J. Mater. Sci. 35 6041 ^ 6054
Sung C-M, 2000b, US Patent number 6 0396 41
Sung C-M, 2000c, US Patent number 6 159 286
Sung C-M, 2000d, Republic of China Patent number 115 958
Sung C-M, Sung M, 1996 Mater. Chem. Phys. 43 1 ^ 18
Sung C-M, Tai M F, 1995/1996 High Temp. ^ High Press. 27/28 523 ^ 546, 611 ^ 628
Sung C-M, Tai M F, 1997a Int. J. Refract. Met. Hard Mater. 15 237 ^ 256
Sung C-M, Tai M F, 1997b High Temp. ^ High Press. 29 631 ^ 648
Tsuzuki A, Hirano S, Naka S, 1984 J. Mater. Sci. 19 1153 ^ 1158
ß 2001 a Pion publication printed in Great Britain