Multi-Anvil Cell Assemblies: An Overview

Contributed by K. Leinenweber. Click here for the COMPRES Cell Assembly Development Project home page.

Last updated February 24, 2009.

Table of Contents


1. Temperature Capabilities

2. Pressure Capabilities

3. A brief breakdown by truncation edge length (TEL)

4. Design considerations


There are at least two types of multi-anvil configurations in common use to achieve high pressures for large volumes. One type is the popular 6-8 system, developed by Kawai and Endo. In this assembly, a cube-shaped cavity consisting of 6 (usually tool steel) wedges pushes on a nest of 8 (usually tungsten carbide) anvils. The sample, thermocouple, and furnace are contained inside of a small octahedron that is placed inside the nest of cubes, and the octahedron is pressed from all 8 sides by the carbide. This geometry has proven to be highly stable, and capable of sustained pressures and temperatures up to 27 GPa and over 2000 degrees C. It is safe to say that most conventional (ie non x-ray enabled) multi-anvil laboratories use this geometry.

Another geometry is the cubic geometry, commonly referred to as the DIA geometry. In a DIA system, an octahedral cavity in the tooling pushes on 6 steel and carbide wedges, which in turn push on a cubic cell assembly. It can loosely be thought of as an 8-6 system in contrast with the 6-8 system above, although it is not usually referred to this way. This has been the preferred design at many in situ x-ray installations for years, starting in Japan with MAX80 and MAX90, and at Stony Brook with SAM85. The cubic geometry made it easy to locate the sample and collect information on what is happening to the sample at high pressure and temperature using an x-ray beam. Recently, with design improvements and programmable detectors, the 6-8 system has been adapted to x-ray systems at SPRING8 and GeoCARS, meaning that experiments from conventional laboratories can be readily brought to beam lines for in situ x-ray studies.

The DIA geometry has been extended by Kato et al., who used a large truncation DIA and added a small nest of 8 cubes in the middle, with an octahedral pressure medium, similar to that seen in a 6-8 system, thus resulting in an 8-6-8 system! The purpose of this was to attain higher pressures with in situ x-ray diffraction. One of the cubes was sintered diamond, to allow the diffracted x-ray to pass through to the detector.

1. Temperature Capabilities

Multi-anvil experiments use a resistance furnace to produce temperature. An external power supply applies a voltage (V) to the furnace, through the carbide anvils and through metal current rings located in the assembly, and the resulting amperage (I) leads to power emission by the furnace (P = IV). If the furnace is a hollow tube with the sample placed inside it, as it often is, the power emitted will serve to heat the sample from the outside. A layer of thermal insulation is often placed around the furnace to direct more of the heat flux toward the sample. The temperature is measured with a thermocouple that is placed in close contact with, or inside, the sample.

What is the temperature limit of a multi-anvil experiment? The temperature limit depends on the type of furnace used, combined with the properties of the other parts of the assembly that experience heating.

A common furnace type is graphite. Graphite is an inexpensive, easily machinable and very stable furnace, nearly transparent to x-rays, so it is used a great deal in both synchrotron and conventional experiments. However, graphite loses its conductivity at high pressure and temperature: above about 1300 degrees C at 10 GPa, with the furnace breakdown decreasing in temperature as pressure increases more. In detail, it is not clear what leads to the degradation of the graphite, since x-rays do not detect any diamond in the recovered furnace. It may be an unquenchable insulating phase, such as lonsdaleite or a related compound.

An excellent furnace material for higher temperatures is lanthanum chromite (LaCrO3). With this furnace, the standard temperature limit of even type C thermocouples (2319 C) can be exceeded, but at high pressure the thermocouple still stays connected and produces a voltage well above its usual temperature limit. Zhang et al. (1993) relied on extrapolations of the thermocouple calibrations in their study of the coesite-stishovite-melt triple point. With a lanthanum chromite furnace, the practical temperature limit is probably near the eutectic melting point of the assembly, which depends on the materials from which the assembly parts and sample are made, and what parts are in contact with each other. Lanthanum chromite has the disadvantage of contributing chromium impurity to many samples; also, it is not as readily available or as easy to machine as graphite.

The temperature may be modeled with a thermal modeling program, such as the program CellAssembly, which uses the thermal and electrical properties of the various cell materials along with the heat equation, to calculate the power, current, voltage, and temperature distribution inside a piston-cylinder or multi-anvil cell.

2. Pressure Capabilities

The pressure capability of a multi-anvil assembly depends first on the truncation size of the carbide; the smaller the truncation, the higher the potential pressure. The simplest measure of the theoretically attainable pressure is equal to the ram force divided by the area of 4 truncations, so if the truncation edge length (TEL) of the carbide is denoted t, then the maximum attainable pressure in a 6-8 system is Pmax = F/(sqrt(3)*t). In practice, the pressure is much (50 percent or more) lower than the theoretical value because of the gasketing; putting all the force on the truncations alone would break the carbide, so a gasket seal is created around the carbide; either assembled from pyrophyllite pieces (Kawai and Endo), or created from cast-on fins around the octahedron (Walker).

In addition to the truncation area, factors that affect the pressure capabilities include the grade or quality of carbide (see the carbide section) and the design of the assembly and gasket; for example, the extent to which the furnace is thermally insulated to keep the carbide cool, and to which the gasketing system protects the carbide from breakage.

The shorthand for denoting the size of an assembly with an octahedron and pregaskets is to put the octahedron edge length first, then the carbide truncation edge length (TEL); for example, the notation 14/8 for a 14 mm octahedron with an 8 mm truncation. For the Walker-style castables, usually only the TEL is listed.


3. A brief breakdown of multianvil assembly sizes by truncation edge length (TEL)

1 inch

Dave Walker is testing a 1-inch cube with 1-inch truncations, a different Aremco potting compound than the usual one, and no cast-on gaskets. The idea is to be able to use a multianvil in the piston-cylinder range of pressures and volumes.

15 mm

(With pregaskets): Used as part of a 25/15 system at Bayreuth, in conjunction with their 5000-ton press. A very large assembly that can be used for very large volumes at resonably high pressures (8 GPa).

12 mm

(Walker-style castable): A nice workhorse truncation used to attain pressures of around 10 GPa on a large volume.

11 mm

(With pregaskets): Used as part of an 18/11 system at Bayreuth.

8 mm

(With pregaskets): The 14/8 assemblies date back to the early Japanese work. Perhaps the most popular truncation to this day, this can accept large volume samples (up to 25 mm3) and can reach pressures in excess of 14 GPa and temperatures in excess of 2500 K using the 14/8 configuration.

(Walker-style castable): Also an extremely popular size for castable octahedra.

7.5 mm

(With pregaskets): Gabriel Gwanmesia used this truncation size with a 14 mm TEL (14/7.5) for his PhD work at Stony Brook, to get that little extra pressure. He was able to make Mg2SiO4 ringwoodite with it, which is higher than the pressure normally attainable by the 14/8 assembly.

7 mm

Bayreuth has tried an 18/7 assembly, a large octahedron and small assembly, for making large samples at a relatively high pressure.

6 mm

(Walker style castable): Not a common TEL. Was used at ASU in combination with castable octahedra, but has since been discontinued.

5 mm

(With pregaskets): Carbide cubes truncated to this edge length form the basis for the 10/5 assembly, a popular combination in wide use at various laboratories.

4 mm

(Walker-style castables) Some of the labs used this truncation with castable octahedra (cf. Carl Agee and ASU) and could reach 20 GPa with it.

(With pregaskets): The 10/4 assembly - a 10 mm octahedron combined with a 4 mm truncation - is fairly popular and reaches pressures over 20 GPa, but is said to be less stable (more blowout-prone) than the 10/5.

3 mm

(With pregaskets): An 8/3 assembly described in Bertka and Fei (1997) can easily make MgSiO3 perovskite (signifying pressures over 23 GPa) and has a low blowout rate. Using a rhenium heater with lanthanum chromite thermal insulation, temperatures in excess of 2000 K can be reached without significant pressure loss.

2.5 mm

A 7 / 2.5 assembly is described by Akaogi et al. (1999) for runs over 14 GPa. They used these dimensions to attain a pressure range abovetheir 5 mm truncation and below their 1.5 mm truncation.

2 mm

(With pregaskets): Eiji Ito in Misasa uses a 6/2 mm assembly, often with "tapered carbide anvils," to make pure samples of MgSiO3 perovskite, including the first ones ever made in a multi-anvil. Yanbin Wang used a 7/2 assembly to make many samples of pure MgSiO3 perovskite and other ultrahigh pressure phases at Stony Brook.

It is of historical interest that Kawai and Endo (1970), in their pioneering 6-8 study, used a pyrophyllite octahedron in a 6/2-sized assembly.

Ito et al., 1998 and Ono et al. (2001) used a 5.6 / 2 assembly with sintered diamond anvils to reach up to 37 GPa. The 6 wedges of the first stage are immersed in an oil bath for even pressurization, in a return to the original "oil bath" cubic press style, which dates back to Von Platen (). This was done to prevent breakage of the sintered diamond anvils, since they are highly subject to breakage if the forces are not symmetric (Ono, personal communication).

1.5 mm

Kubo and Akaogi (2000) describe a 4.7 / 1.5 assembly that they use with tungsten carbide anvils for reaching pressures up to 28 GPa. This was used to synthesize pure samples of the calcium ferrite phase of MgAl2O4 (Kojitani et al., 2000).

1 mm

(With pregaskets): This tiny truncation has been used in in-situ experiments particularly in Japan, in order to reduce the force and cube size necessary to reach high pressures for the DIA presses used in 6-8 geometry. This geometry can only accept small cubes and the DIA presses installed at x-ray beamlines (such as MAX80) have low ram forces.

0 mm

A whimsical idea from Tomoo Katsura: no truncation. It has probably not been tried yet, at least not intentionally.


4. Design considerations

The simple calculations described below may be performed using the program GASKET, written by Kurt Leinenweber.

Various considerations that come into designing a cell assembly are listed here, organized from the outside to the inside of the assembly.

Truncation/Octahedron size

For the original (Japanese) assembly styles, with preformed gaskets, the octahedron and truncation size together give the assembly a number designation, such as the "14/8" assembly, or the "10/5" and "10/4" assemblies. The octahedron size refers to the edge length of the octahedron in millimeters, and the truncation refers to the edge length of the triangular beveled corner of the carbide cube. For the formulas in the sections that follow, the variable t will refer to the truncation size, and o will refer to the octahedron edge length.

For the castable assemblies developed by Walker, the two-number designation is not used. Usually they are referred to only by the size of the carbide truncation, t. However, to duplicate an assembly preciesly, it is really necessary to specify the thickness of the fins as well (which is equivalent to the thickness of the teflon spacers used in the molds for making the castable assemblies), and also the outer dimension of the fins (equivalent to the size of the square openings in the teflon spacers).

Tungsten Carbide

Tungsten carbide anvils for the 6-8 high-pressure geometry are in cubes with the corners truncated to triangles. The truncation size is the number t mentioned above. The edge length of the carbide cubes depends on the size of the multi-anvil wedges. If the wedges are 5 cm across, cubes 1 inch in size are used (this is the original cube dimension used by Walker et al., 1990; and is still the usual size made by Rockland Research). Other sizes are 25 mm (an attempt at a "metric inch?") and 32 mm (originally the largest size of carbide cube available when the Japanese laboratories first developed the 6-8 design). Care must be taken that the carbide cubes are large enough relative to the square face of the wedges, otherwise bridging of the wedges may occur when the force is high, limiting the pressure attained. This may be tested by placing small balls of modeling clay in the gaps between the wedges before a run at the highest pressures, and checking their thickness after the run.

The type of carbide is, of course, important in determining the pressure limitation. The carbide fails either by plastic flow or brittle failure; either type of failure limits the attainable pressure. Plastic flow is less disastrous, because the carbide cube can be rotated to another truncation for re-use.

Referring to the paper by Getting on carbides used for high-pressure work, the carbides that have higher yield strength and lower axial strain at failure are capable of going to higher pressures before they plastically deform or break. Toshiba Grade F (yield strength 6 GPa, axial strain 2.7 percent at failure) is a classic example of a good carbide for reaching higher pressures. Kennametal K313 carbides, also in common use, deform more and break at a lower force (yield strength 5 GPa, axial strain 5 percent at failure). However, experience shows that the "softer" (lower bulk modulus) grades sometimes perform better during blowouts, with fewer cubes breaking. Thus some labs use "softer" grades for lower pressures to save money, and "harder" grades only when the highest pressures need to be obtained.

Rockland Research has recently begun supplying an inexpensive carbide (made by contract from Fansteel) called "Toshiba Grade F equivalent." This carbide is not included in the study by Getting; however, initial results from the grade show that it is robust, and is a viable option for high pressure research.


For the old-style composite design of octahedron plus preformed gaskets, each assembly requires 1 octahedron and 12 gasket pieces, trapezoidal in shape, assembled around the octahedron. The gaskets, in order to fit properly around the octahedron, need to be made in two different lengths, a longer set and a shorter set of 6 each. The common gasketing material is pyrophyllite (often called "lava rock" or "aluminum silicate ceramic" by vendors).

Most of the dimensions of the gaskets are fixed by the octahedron edge length o and the carbide truncation t. The only dimension that can be adjusted is the width of the trapezoidal part of the gasket. The height perpendicular to the trapezoid is fixed by the formula

h = sqrt(2)*(o-t)/3.

The length for the inner edge of the longer gasket is equal to o, the octahedral edge length. For the shorter gasket, the equivalent length is (2t+o)/3.

For the castable, Walker-style octahedra, the gaskets are already included as part of the octahedron, and are usually referred to as "fins." The designer needs to consider the thickness of the fins, and the width of the fins to the outer edge. This is set during the design of the molds, and is fixed after that.

Octahedron type

The octahedra developed in Japan are made from MgO, with or without dopants. The pure MgO octahedra, "MgO 99% of Sea-Water Magnesia Clinker," is made by Mino Yogyo, also known as Japan Ceramic or Cherry-O (all the same company). The same company sells the octahedra with dopants such as Cr2O3, which may serve to lower thermal conductivity by radiation at high temperatures.

A formula was developed by Ceramic Substrates, and is used by laboratories such as Wolfgang Schnick's laboratory in Munich. It is also an MgO doped with Cr2O3, and has a density similar to that made by Mino Yogyo.

Zirconia octahedra may be used in order to achieve good thermal insulation, removing the necessity for a thermal insulating sleeve around the furnace (see below).

The ceramic most often used in castable assemblies is a potting compound, Aremco 584, which contains about 33 percent alumina and 66 percent MgO by weight. The material is fired at 1000 degrees Centigrade for 1 hour. With this recipe there is little or no conversion to spinel, according to x-ray diffraction on the fired material. Other potting compounds have also been tried, see for instance Walker (1). The current choice was mainly for ease of pouring and optimal curing, according to that paper.

Finally, the most x-ray transparent material for assemblies, used at the beam lines, is a mixture of boron and epoxy resin (Yagi and Akimoto). This is used in conjunction with no pregaskets, boron-epoxy pregaskets, or pyrophyllite gaskets (Wang, personal communication).

Thermal insulation

The materials used for most types of octahedra are of high thermal conductivity. In order to prevent excessive heat loss from the furnace, ceramic insulation of very low thermal conductivity is usually placed around the outside of the furnace. For this purpose, zirconia is the traditional choice. Its thermal conductivity is less than 3 W/m*K, and has little temperature variation.

Another choice is Aremco 502-1550, a "zironium phosphate" ceramic that has a lower thermal conductivity than zirconia. It has an ambient pressure melting point of 1500 degrees C, so is not recommended for very high temperature applications. it is used at Los Alamos National Laboratories for neutron experiments, and at ASU.

For x-ray applications, the boron + epoxy resin turns out to have a fairly low thermal conductivity; however, a layer of ceramic, usually alumina or MgO, is often put between the furnace and the boron + epoxy resin to prevent reactions between the two at higher temperatures. Another option is to use a boron + waterglass mixture instead, for higher temperatures.


As was mentioned before, graphite furnaces and lanthanum chromite furnaces are in wide use. Silicon carbide is an option for higher temperatures combined with in situ x-ray diffraction, but has not been widely developed yet. New forms of green-machinable silicon carbide may change this situation in the near future.

Wrapped metal furnaces are also used. Inconel is an inexpensive option, good to 1200 C; platinum is an effective furnace to 1700 C (Ito), and rhenium to temperatures in excess of 2000 C. Often, lanthanum chromite is used as an insulating sleeve instead of zirconia; it has a low thermal conductivity, and seems to prevent runaway burn-out of the furnace. The assemblies of Ito, Katsura and Ito, and the 8/3 design of Fei are based on this combination. The 8/3 design of Fei uses thick (.0025 inch) rhenium, and seems especialy effective for combined high temperatures and pressures.

Furnaces may be tubular, as in the case of wrapped metals or straight ceramic tubes, or may be stepped with 2 thin sections at the end and a thicker section in the middle, to smooth out the thermal gradients in the sample; the assemblies used at Bayreuth make full use of this technique. A third possibility is a box-shaped furnace, which telescopes out in order to increase the sample volume (Gwanmesia). A tapered design by Takahashi also served to reduce temperature gradients in the middle, in a manner similar to the stepped assemblies.

Encapsulation of the sample

Finally, the sample itself needs to be encapsulated and protected from reaction with its surroundings. In addition to this function, the encapsulating material will also smooth out thermal gradients if it happens to have a high thermal conductivity.

Common capsules for many mineral systems are precious metals (platinum, gold, palladium). For systems free of transition metals, any of these will work, but when iron and other transition metals are present, gold/palladium and other alloys work best. The metal capsules also are surrounded by an electrically insulating sleeve, so that they do not electrically short the furnace; this sleeve can also be a thermally conducting material to reduce thermal gradients. Diamond is especially useful for this; one trick is to use a graphite capsule inside of a furnace; at high pressure, as the system heats up, the capsule changes to diamond, reducing or wiping out thermal gradients in the sample (Y. Fei, personal communication).

For sulfide systems, metal capsules do not work well, and graphite capsules are often used (Agee).

In chemistry experiments, it is generally found that each sample presents a new containment challenge, and a large number of possible crucible materials are needed to contain the elements of the entire Periodic Table. Certain elements, such as boron, present special difficulties, especially when fluxes and melts are present; sometimes some degree of reaction with the capsule is unavoidable.

Back to Multi-Anvil Web cover page