High Pressure Techniques

1

Chapter IIHIGH-PRESSURE TECHNIQUESH. Tracy Hall

1 Introduction2 Methods for Generating Static Pressures3 Piston-Cylinder Devices4 Binding Rings5 Hydraulic Presses6 Electrical Resistance Heating Equipment7 Opposed-Anvil Devices8 Belt Apparatus9 Multianvil Apparatus10 Instrumentation11 The Synthesis of Diamond

References

1 INTRODUCTIONThis chapter is written, primarily for the

nonspecialist, out of 25 years of experience indesigning and constructing all types of high-pressure equipment. During this period therehave been only brief intervals in which I havenot had a piece of apparatus under construction.My notebooks record about 250 ideas onapparatus design and at least 32 different devicesvalued at around $2 million have been builtunder my direction. I have been personallyinvolved in all aspects of the work, includingconception, design, drafting, and machining. Inthe latter I have tried to emulate that pioneer inthe field of high pressure of whom it has beensaid, “there was only one person at Harvard whowas a better machinist than Charlie Chase andthat was Percy Bridgman.” Mr. Charles Chasewas Physics Professor Percy W. Bridgman’sunusually skilled machinist and craftsman. Inthis work I too believe in getting my hands dirty.But it has not been work. For me it has beenlove.

The modern-day interest in high-pressuretechnology was most certainly sparked by theannouncement on February 15, 1955, by theGeneral Electric Company, that diamonds had, atlong last, been made by man. Indeed, this hadbeen one of the most intriguing, unsolvedproblems of science, attempts to synthesizediamond by the conversion of its polymorphgraphite having been instigated in 1792. Thatwas the year Lavoisier discovered that diamondconsisted only of the element carbon.

Considerable excitement surrounded theGeneral Electric announcement as journalists

from around the world assembled to hear thenews. There was also disappointment. Forwhile there was a discussion of the unsuccessfulattempts of others to synthesize diamond, and thetiny diamonds that had been made weredisplayed, the details of how these diamondswere made and the details of the equipment thatwas used were not revealed. Several monthslater, additional secrecy beyond companyinterest was revealed by the United StatesGovernment. Abatement of the secrecy did notcome until five years later.

You can well imagine my anguish, anxiety,and dismay when, having invented (January 2,1953) the belt apparatus in which the synthesiswas achieved and having accomplished theworld’s first reproducible synthesis of Diamond(December 16, 1954), I was not able to publishor claim recognition for an accomplishment thathad been sought by eminent scientists, includingNobel laureates, for much more than 100 years;and further, of watching my accomplishmentsbeing diluted through various means as the yearswent by, including leaks of information fromcompany and government sources that allowedothers, not hampered by proprietary orgovernmental interest, to come ever closer to theprocess and to the design that were original tome. A sampling of some of these problems isfound elsewhere1. These difficulties, however,led to the development of other high-pressureequipment, specifically the multianvil presses, ofwhich the most important members are thetetrahedral and cubic presses. These arediscussed later.

2 HIGH-PRESSURE TECHNIQUES

Practical diamond synthesis requirespressures of the order of 50,000 atmospheres(atm) and higher. Having mentioned a practicalunit of pressure measurement, I might divergefor a moment. Through the years many unitshave been used to measure pressure. HarvardProfessor P. W. Bridgman, who received theNobel Prize in 1948 for his work in the field ofhigh pressure, measured pressure in kilogramsper square centimeter. Bridgman’s first paper onhigh pressure was published in 1908, and hislast, in 1958. Chemists commonly use theatmosphere, or the kiloatmosphere, whilegeologists and others prefer the bar or thekilobar. Recently the National Bureau ofStandards began encouraging the use of thenewton per square meter (now called the pascal,abbreviated Pa). Fortunately, these units areclosely related: 1 atm = 1.0133 bar = 1.0332kg/cm2 = 14.696 lb/in.2 = 760 torr (mm Hg) =1.0133 X 105 N/m2 or Pa.

The rapid growth and apparent decline ofhigh-pressure research since 1950 is graphicallyrepresented in Fig. 2.1. This figure shows thenumber of papers published in the field through1974. Only paper reporting work above 5000atm are included in the count.

The United States Bureau of Standardsmaintains a high-pressure data center at BrighamYoung University. This center has the functionof evaluating certain aspects of data found in thehigh-pressure literature. It compiles a“Bibliography on High Pressure Research,” acurrent-awareness bulletin, published in six

issues annually, on a subscription basis. Thecenter cooperates with foreign countries, notablyJapan and Russia, which compile likeinformation for their countries. The centerattempts to obtain and file one copy of everyarticle on high pressure published anywhere inthe world. Another publication of the center is“The International Directory of workers in theField of High Pressure Research,” compiled byJohn F. Cannon and Leo Merrill, April 1971 (adirectory was also issued in April 1968). Thereis also a high-pressure bibliography covering theperiod from 1900 to 1968. Volume I includes abibliography and author index and Volume II, asubject index, published in April 1970 under thedirection of Leo Merrill, Associate Director ofthe center. Another publication of note from thecenter is “Behavior of the Elements at HighPressures,” by John Francis Cannon, which isreprint number 55 from J. Phys. Chem. Ref.Data, 1974. Reprint number 12, from J. Phys.Chem. Ref. Data, by D. L. Decker, W. A.Bassett, L. Merrill, H. T. Hall, and J. D. Barnett,is entitled “High Pressure Calibration, a CriticalReview.” This is a good source of informationconcerning the calibration of high-pressureequipment. The International Directory ofWorkers in the Field of High Pressure includesthe names of researchers, their affiliation, and thepressure range covered by their research. It alsoindicates the temperature range and the type ofapparatus used, such as the belt, BridgmanAnvils, cubic press, diamond anvils, gasapparatus, liquid systems, shock techniques,piston-cylinder devices, and so on. Also , if theworkers have interests in calibration andequipment design, this is indicated. There is alsoa key to the type of interest that the particularlaboratory or investigator has. For example, ifthe person is interested in crystal chemistry, thatis so indicated. Other subject headings arecrystal structure, equilibrium studies, inorganicsynthesis, lattice defects, organic chemistry,polymorphism, electronic and quantumproperties, geological and geophysicalproperties, magnetic properties, amorphoussolids, diamond research, mechanical properties,metallography and metallurgy, method ofanalysis, infrared, Mossbauer, neutrondiffraction, optical properties, thermodynamics,and so on. This publication, then, can be thenew interested researcher’s entrée into the fieldof high pressure.

Most high-pressure apparatus used forresearch is constructed in university shops andlaboratories to suit some particular interest of the

H. TRACY HALL 3

investigator. Consequently, when considered inspecific detail, there are nearly as many high-pressure devices as there are high-pressureresearchers. However, when considered in broadperspective, there are a limited number of basicdevices. In order of their development, they are(1) piston-cylinder, (2) Bridgman anvil, (3) thebelt, and (4) multiple anvils, the principalmultianvil types being the tetrahedral press andthe cubic press. These are landmark devices, andall other devices, by whatever name they may becalled, are derivatives of these. The origins ofthe piston-cylinder apparatus are lost inantiquity, but some important adaptations of itfor high-pressure, high-temperature use are dueto Parsons and Coes, whose work will soon bediscussed. Bridgman anvil devices are theinvention of Percy W. Bridgman; the belt andmultiple-anvil types are the inventions of thisauthor.

The chemical changes that can be induced ina system by perfusing a substance with energyare qualitatively related to the intensity factor(pressure, temperature, volts, etc.) of the energy.As far as chemistry is concerned, the mostimportant energy introduced into the chemicalsubstances of concern is heat. Heat of sufficientpotency to bring about changes in chemicalbonding has been available for a long time, fromfires, flames, torches, and Bunsen burners.Readily obtainable temperatures of a fewhundred degrees Celsius cause many chemicalchanges to occur. In terms of “ideal” energyequivalence, one cubic centimeterkiloatmosphere is equivalent to a 12.19°Kchange in temperature.

Of course the atomic and molecular effectsof temperature and pressure on substances arenot the same. However, pressure andtemperature diametric opposites in some ways.For example, as temperature is increased, solidstransform to liquids, which in turn become gases.Systems then proceed to products of moleculardissociation, and at sufficiently high temperaturethe atomic nuclei will be separated from theelectrons, and so on. With increased pressure, onthe other hand, substances originally gaseousbecome liquids, which in turn transform tosolids; and with ever increasing pressure, thecollapse of electronic shells will occur. Atpressures of billions of atmospheres nuclearfusion occurs. Simultaneous high temperaturewith pressure is important in the field ofchemistry because of the general effect of highpressure on reaction rates. High pressure tendsto reduce the rate of a chemical reaction.

Therefore, in order to have a chemical reactiontake place under high-pressure conditions in areasonable length of time, it is almost alwaysdesirable to use elevated temperature. Some ofthe thermodynamic and kinetic problems havebeen considered elsewhere2.

Tens of thousands of atmospheres arerequired to produce changes in solids and liquidsubstances qualitatively equivalent to thechanges that can be produced in these samesubstances by means of heat energy.Temperature can be controlled easily andcheaply. Pressure has been controlled only withrecent technology and at great cost. Also, notuntil recent years has high-pressure equipmentbeen available for purchase. Researchers havehas to devise their own. Consequently heat hasbeen much more universally applied to studyingchemical change than has pressure.

Pressure is ordinarily defined as a forcebeing exerted upon a unit of area. Within theuniverse, the range of natural pressures thatexists is enormous. The vast reaches of spacebetween the galaxies contain such as smallamount of matter that the vacuum existing thereis more nearly perfect than any that can beobtained in the laboratory. The distance betweenatoms in this space is so great that it does notmake sense to talk of gaseous pressures asexisting at all.

All pressures of any magnitude occurring innature on a macroscopic scale are due togravitational forces. The earth’s gravitationalfield attracts the air molecules that form ouratmosphere. The average pressure of thisblanket of air at sea level is 14.70 lb/in.2 (thenormal atmosphere). The pressure existing at thegreatest ocean depth is ~1000 atm. Interestinglysome of the early pressure experiments wereperformed by lowering liquid-filled, corkedbottles into the sea. Halfway between the earth’ssurface and its center, at the core boundary, thepressure is about 1.4 X 106 atm, and at theearth’s center the pressure is estimated to be 3 X106 atm. The pressure at the center of the sun isestimated at about 1011 atm, and the pressuresfound at the center of white dwarf stars may be1016 atm. Matter at 1016 atm will completelydegenerate. It will have an effective temperatureof absolute zero and a density of one milliongrams per cubic centimeter.

Man and his machines are indeed puny withregard to any discussion in galactic dimensions.Other comparisons, however, can place man’saccomplishments in flattering light. Forexample, the laboratory-attained pressure of


200,000 atm is equivalent to that present at thebase of a column of granite 400 miles high—4000 Washington Monuments stacked on top ofeach other.

A point has been reached in the field ofscientific endeavor where results can often beachieved if vast sums of money can be spent.There are two aspects to this development. Thefirst aspect concerns the increasing complexityof science. At the turn of the century a particleaccelerator in the form of a glow-discharge tubemight have cost $100. In 1934 the first cyclotroncost approximately $50,000. Today, amultibillion-volt particle accelerator may costhundreds of millions of dollars. The secondaspect involves expendability. Institutionalaccounting procedures have for years categorizedexpenditures into areas of supplies, equipment,travel, and so on, and the category of equipmenthas usually been defined in terms of cost. Thatis, anything valued at $100 to $500 or more hasarbitrarily been defined as a major item ofequipment. Such property must be given aninventory number and thereafter be properlyaccounted for, and there have been somedifficulties between the scientists and theaccountants over this matter. Before theaccountants have placed an inventory tag on acertain “equipment item,” set up as maintenanceprocedure, and so on, the scientist may havealready destroyed the item in his experiments.The ultimate example of today’s willingness toexpend is exemplified by the billion-dollarrockets being shot into space. High-pressureresearch, when probing at the highest pressures,requires a modest willingness to spend andexpend. The expendable equipment lost to anactive investigator working in the field of highpressure is of the order of $2000 to $20,000 peryear. Most of this loss occurs from the breakageof cemented tungsten carbide pistons, anvils, andso on.

The highest pressures attainable by man areattained by explosive means. Pressure andtemperature are simultaneously generated byexplosions, but the period for which the pressureand temperature exist is of the order of micro- orat the most milliseconds. Much of the apparatusused in these experiments is literally “blown up,”and the cost of the expendable materials can bemuch greater than the tungsten carbide broken instatic-pressure apparatus (apparatus that can holdpressure for at least several minutes, but moreusually for hours or days). While the generationof pressure by explosive means is not discussedin detail in this treatise, there are some

interesting aspects of it that will be mentioned.Pressure as high as 10 million atmospheres havebeen claimed by some investigators. Howevermost of the literature deals with pressure of theorder of a few hundred thousand atmospheres oreven less. In most chemical (synthesis)experiments, the sequence of events is asfollows. Pressure is applied at room temperatureto the substance in which chemical change iswanted. This is followed by increasing thetemperature to the point where reaction willoccur. Next, the temperature is reduced to roomtemperature to prevent reversion, and this is turnis followed by reducing the pressure toatmospheric. These events are easily controlledin static-pressure apparatus but are not so readilycontrolled in an explosion.

The use of explosions to generate pressurewas considered at the General Electric Companyduring the same time that static high-pressureapparatus was being used in attempts tosynthesize diamond. It was concluded, however,that the pressure and temperature sequence couldnot be properly controlled to convert graphite todiamond. In 1961, approximately six years afterI had synthesized diamond in static-pressureapparatus, John C. Jamieson, Professor at theUniversity of Chicago, and P. S. DeCarli of theStanford Research Institute collaborated in anattempt to make diamond by explosive meansand succeeded3. Their experiment waseminently elegant. They used a 55-gallon drumhalf-filled with water, open at the top, with aboard across it on which was placed a block ofgraphite. This in turn had on top of it anexplosive device. The explosive was fired froma safe distance, and after a settling time, thepowder in the bottom of the drum was recovered,dried, and enriched by density separation. Theyobtained a definite X-ray diffraction pattern,although the lines were very broad, for diamond.The diamond was up to 10µm is size, and theyield in the process was something of the orderof 2%. This work was financed by AlliedChemical Corporation. Improvements weremade in the process, and for a time it waspossible to purchase a dark-gray diamondpowder from that company. This diamond wasnot pure and apparently contained around 12%oxygen, which may be the reason that thematerial is no longer available. The function ofthe water in Jamieson and DeCarli’s experimentwas to cool the newly formed diamond so fastthat is did not convert back to graphite.

Subsequent to DeCarli and Jamieson’sdevelopment, personnel at the du Pont Company

H. TRACY HALL 5

developed a process in which iron was thecoolant4. They compressed, by explosive means,a material known as ductile or malleable iron.This iron contains a high proportion of carbondispersed as graphitic spheroids. If a shapedcharge of explosive is applied to ductile iron, ashock wave passes through it (explosivetechniques are also called shock wavetechniques). A shock wave passing through asolid is like a sound wave. There is a pressurefront, followed by a trough in which the pressurefalls. The rate at which the pressure front travelsthrough a solid depends on the nature of thematerial, and there are other things that occur,such as reflections and absorptions, whichcomplicate matters. Temperature also rises andfalls in a pattern that depends on severalvariables. The advance of a front throughnonhomogeneous material, such as ductile iron,will be complicated, but the net result of a properexperiment is that part of the graphite in thespheroids is transformed to diamond. The iron,which immediately surrounds the spheroids,serves to take away the heat and quench thediamond rapidly enough that it does not all revertto graphite.

There is a wonderful story, perhapsapocryphal, concerning the du Pontdevelopment. There was an abandoned minetunnel somewhere in one of the north centralstates, I think Minnesota, in which a largequantity of ductile iron was placed. This wassurrounded with appropriate charges of dynamitethat were later detonated, resulting in theformation of several tons of impure diamondpowder. After the explosion subsided, a front-end loader removed some material and depositedit in a digester where dilute sulfuric aciddissolved the iron, leaving a pasty residue. Thisresidue was dried and subsequently dissolved inone or more oxides of lead, which selectivelyoxidized the graphite, leaving most of thediamond untouched. Further minor cleaningoperations yielded a pure diamond powder ofsubmicron size that under the microscope had aspongy appearance. When most of this batchhad been sold, the front-end loader again wentinto the man-made diamond mine, retrievedmore material, and so on. Supposedly there isenough diamond in this mine to last the worldfor many years!

But there are now improvements in theprocess, and the mine has been abandoned. Arather complicated device is now used ontowhich appropriately shaped charges are fastened.Rather than ductile iron being used as a charge,

graphite containing copper particles is used. Thepurpose of the copper is to appropriately absorbthe heat to prevent reversion of the newly formeddiamond. Recovery of the diamond in anexplosion could of course be a problem, but thereis a momentum-absorbing plug at the end of thecylindrical device which gives away at theappropriate time, allowing the charge to be lessviolently blown into a retrieval container.Apparently the submicron diamond powder afterretrieval and purification is again placed into anexplosive device and reshocked, whereupon anagglomerization of the diamond takes place,producing particles of up to 40 or 50µm. Theselarger are more in demand than the originalsubmicron sizes.

2 METHOD FOR GENERATING STATICPRESSURES

In the introduction to this volume, I gavethree general methods for obtaining high valuesof intensive properties such as pressure,temperature, voltage, and the like. They weredisproportionation, gathering or focusing, and insitu energy transformation. At the present timethe use of disproportionation to obtain highpressure has given the most satisfactory result.However there are energy-transformationprocesses that have also been used, some ofwhich are attractive for further development. Itis possible to generate pressure by increasing thetemperature of a material which fills a confinedspace, for example, the heating of a liquid in ametal tube. Chemical energy may also beutilized to generate pressure by employing achemical reaction or a phase change in which theproducts of the reaction or the new phasesformed will tend to occupy a larger volume thanthat to which they are confined. The freezing ofwater in a confined space is a common exampleof this principle, as many know who have frozenradiators or pipes in the wintertime. Thefreezing of water in a confined space cangenerate a theoretical maximum pressure ofabout 2000 atm. The solidification of liquidbismuth in a confined volume can generate atheoretical maximum pressure of about 17,000atm. Likewise, the freezing of moltengermanium can generate a theoretical pressure ofthe order of 100,000 atm. These theoreticalpressures can be approached providing thecontainer volume increases only slightly as thefreezing liquid presses against it. It is alsonecessary that the volume of the germanium,bismuth, or water is large compared to the


volume of the material being subjected tocompression by the expansion process. At thepresent time, germanium has been demonstratedto be capable of generating higher pressures by afreezing process than any other knownsubstance. However, its use in a pressure-generating device has not yet been fullyexploited.

As mentioned, disproportionation providesthe most satisfactory means of obtaining highpressures at the present time. The most obviousform taken by a high-pressure device is that of apiston and cylinder. In such a device, a sampleconfined by the cylinder drives the smallerpiston into its cylinder, this being thedisproportionation principle (see Fig. 2.2). In thefield of hydraulics this type of device is called anintensifier.

3 PISTON-CYLINDER DEVICES

Today’s piston-cylinder devices are operableto pressures around 50,000 atm with solidmedia5. In these devices both the piston and thecylinder are constructed of cemented tungstencarbide. The cemented tungsten carbide used ismade of pure, virgin, material, formula WC,cemented together with approximately 3 to 12%cobalt. For general-purpose use, a tungstencarbide containing 6 to 8% cobalt is the most

desirable. Tungsten carbides have the highestcompressive strength of any engineering materialreadily available at this present time. Althoughcemented carbides possess tremendouscompressive strengths, they have rather lowtensile strengths and are brittle. Their tensilestrength increases with increasing cobalt content.However, the compressive strength decreaseswith increasing cobalt content. Compressivestrengths for 3 weight-percent cement tungstencarbide cylinders of 60,000 atm have beenreported by some manufacturers. However, theusual commercial product has a compressivestrength of only 35,000 atm. These figures arefor solid “rounds” in which the length equals thediameter, the round being compressed betweencarbide blocks. Longer rounds will have lowercompressive strengths (the so-called columneffect). Because of the low tensile strength,cylinders constructed of this material must besupported in such a manner that the carbide isnot subjected to a damaging tensile load. Thiscan be accomplished by surrounding the carbidecylinder with massive alloy-steel binding rings.

The upper pressure limit of a piston-cylinderdevice so constructed is set by failure of thepistons. When materials with compressivestrengths greater than those of the cementedcarbides become available, the upper pressurelimits obtainable in a piston and cylinder device,as well as in other types of devices, will beincreased. Sintered diamond, a newly availablematerial, will eventually make it possible toobtain higher pressures6.

Tremendous pressure could be theoreticallyobtained by a cascade process wherein onepiston-and-cylinder device would be placedinside another, and so on. The failure of high-pressure components is not caused by theabsolute pressure to which they are subjected,but rather the differential pressure. Thus acylinder in which the internal pressure is 100,000atm and the surrounding external pressure is50,000 atm is in no more danger of failure than acylinder wherein the internal pressure is 50,000atm and the external pressure is zero. There issome indication that tungsten carbide increasesin strength under high pressure and can thereforeunder these conditions withstand a greaterdifferential pressure. At the present time, theexperimental problems connected with cascadinghave made it impossible to utilize this principlebeyond two stages. As a matter of fact, the onlyuse of a two-stage device that has been reportedin the literature is due to Bridgman7. Theworking volume of Bridgman’s apparatus was

H. TRACY HALL 7

very small. The final piston had a diameter ofapproximately 1.5 mm, and the sample lengthwas also about 1.5 mm. This device iscomplicated and difficult to use. Electrical leadscould not be taken into the sample area. Adouble-ram hydraulic press was required for itsoperation. The double-ram press has two coaxialthrust generators. The initial program at theGeneral Electric Company for equipment tosynthesize diamonds called for the constructionof a Bridgman two-stage piston cylinder deviceof greatly enlarge size. It was hoped that thelarger size would make it possible to introduceelectrical leads into the inner chamber to providecurrent for electrical resistance heating of thesample and also to make temperature and otherkinds of measurements. However the inventionof the belt apparatus, which is infinitely simplerand easier to use, resulted in abandonment of thisapproach.

The High Pressure Institute in Moscow,Russia, has been constructing a monumentalhigh-pressure apparatus for a number of years. Ithas been reported that the total investment in thisunfinished device is around $20 million. It isexpected to be a piston-and-cylinder apparatus offive stages. Drawings of the contemplatedapparatus are not available, but the problems ofproviding thrust from the outside through all fivestages and of providing electrical communicationto the innermost stage for measurement and tosupply heating current are formidable, if notimpossible.

Let us now return to providing furtherdetails of single-stage piston-cylinder apparatus.Emphasis will be on modern devices that usesolids or a combination of solid plus liquid totransmit pressure, since there is a plethora ofliterature available on fluid pressure apparatus.Pressure transmission by solids allows muchhigher pressures to be achieved and also allowshigh temperature simultaneously with highpressure. This is important in chemistry.

Sir Charles Parsons was the first to attackthe problem of generating high pressuresimultaneously with high temperature8. Hebegan a diamond synthesis program about 1880and ended it about 1928. From his studies heconcluded that neither he nor anyone else had, upto that point, succeeded in making diamond.Parsons had excellent laboratory and machineshop facilities and conducted experiments on agrand scale for that period. His pressureapparatus consisted of piston-cylinder devicesthat used internal electrical resistance heating(see Fig. 2.3). He used a solid pressure-

transmitting material which also served asthermal and electrical insulation. He seems tohave been the first person to utilize such acombination in a high-pressure, high-temperaturedevice. His cylindrical chambers ranged indiameter from 1 to 15 cm. The maximumpressure at the temperature he reported was ofthe order of 15,000 atm at 3000°C.

Loring L. Coes, Jr., of the Norton Co., wasthe first person to develop a piston-cylinderdevice with capabilities substantially beyondthose of the Parsons device. In July 1953, Coesreported the synthesis of a new dense silicawhich has since been called coesite9. Coesite hasa density of 3.01, whereas the density of quartzis only 2.65. The refractive index of quartz in1.50; the refractive index of coesite is 1.60. Thismaterial was discovered during a synthesis studyof the minerals with which diamonds areassociated in the diamond-bearing pipes of SouthAfrica. Although man-made “minerals” are notusually named in honor of their discoverer, thisnew SiO2 substance was unique and sointeresting that researchers immediately began tocall it coesite. Coesite was later found to bepresent in some meteorite craters and constitutesthe first instance in which a mineral wasproduced in the laboratory before it wasdiscovered in nature. Coes reported that thematerial was made at a pressure near 35,000 atmat a temperature between 500° and 800°C.Incidentally, another SiO2 material, calledstishovite, more dense than coesite, wassynthesized by S. M. Stishov and S. V. Popovain 1961 at a reported pressure of about 115kilobars and a temperature of about 1300°C.This material has the remarkable density of 4.35and a refractive index of 1.8110. Stishovite hasalso been found in meteorite craters.


Coes’ first apparatus was not described inthe 1953 Science article. On April 21, 1954, atthe American Ceramic Society Meeting inChicago, Coes presented a paper entitled “HighPressure Minerals.” This paper described anumber of garnets and other materials that hehad synthesized during the course of his researchon the synthesis of diamond. Again, he did notdiscuss the nature of the equipment used.However at the 7th Symposium on CrystalChemistry as applied to ceramics at Rutgersuniversity, New Brunswick, New Jersey, on June4, 1954, Coes described his apparatus. He didnot personally publish a description of thisequipment, however, until 196211. A diagram ofthe apparatus and of the sample chamber is givenin Fig. 2.4. The key feature of the device is theuse of a hot, molded alumina liner or cylinder.The apparatus is double ended, pressure beinggenerated by pushing a tungsten carbide pistoninto each end of the alumina cylinder. Becausethe alumina cylinder is electrically insulating,heating is accomplished, very simply, by passingan electric current from one piston through asample heating tube and out through the oppositepiston. The apparatus was used at pressures ashigh as 45,000 atm simultaneously with atemperature of 800°C. A temperature of 1000°Ccould be obtained at a lower pressure of 30,000atm. Temperature was measured by means of athermocouple located in a well, as shown in thefigure. Although the temperature at this point islower than the temperature in the sample,comparisons of temperatures measured here weremade with a thermocouple inserted in the sampleat 1000 atm. From these measurements it waspossible to correct the temperature reading of thethermocouple in the well to the approximatetemperature of the sample in the chamber.

At 45,000 atm and 800°C, only one run isobtained in this device, the pistons and thealumina cylinder both being expendable. Even at30,000 atm the alumina cylinder is only usefulfor a few runs, as is also the case for the tungstencarbide pistons. The expense of using such adevice is great, and while it was used extensivelyat the Norton Co., I am unaware of any use of thedevice elsewhere except for some limited use ofthe equipment at Brigham Young University.Coes was kind enough to give a number ofalumina cylinders to me sometime around 1957,at which time I experimented briefly with theiruse. Coes once related to me that they hadbarrels of broken alumina cylinders around thelab. In fact, they had so many they wondered if ause could be found for them. Consequently, the

broken cylinders were crushed to provide mesh-size abrasive particles which were subsequentlymade into grinding wheels. These wheelsperformed better than the wheels they were thenmanufacturing. Consequently, this serendipitousdiscovery led to an important change in themethod of manufacture of aluminum oxide grainfor use in grinding wheels.

Nowadays both the piston and the cylinderare constructed of cemented tungsten carbide,and electrical insulation is provided in a differentmanner than in the device of Coes. For thehighest pressure the cobalt content should beonly 3%. Such a tungsten carbide is very brittle.Therefore the alignment of the piston and thecylinder and the hydraulic ram providing thethrust must be nearly perfect to prevent off-axisloading. The tungsten carbide should be finegrained. The industry code number in the UnitedStates for this carbide is C-4.

The outside diameter of the cylinder shouldbe about six times its inside diameter. Whensolid materials are used for transmitting thepressure, the sample cell length should be lessthan twice the piston diameter. With the cylinderin place in its binding rings, the fit between thepiston and the cylinder should be as close aspossible. A clearance of 0.0025 mm isappropriate. The surface of the cylinder holeshould be highly polished, as should also the

H. TRACY HALL 9

outside diameter of the piston; the ends of thepiston must be ground absolutely square with thecylindrical axis. A coaxial ram press is used tooperate the piston and cylinder device. The innerram provides thrust to the piston, and the outerram provides clamping force to the cylinder (Fig.2.5). The clamping force is necessary to preventthe tungsten carbide cylinder from splitting intwo in a plane perpendicular to the cylindricalaxis. Such splitting will usually occur in a planethat coincides with the surface of the tip of thepiston. The surface roughness values (standardmachining practice) of the inside diameter of thecylinder and the outside diameter of the pistonshould be about 1 µm. The remaining surfaces ofthe piston and the cylinder may have a higherroughness value of about 4µm. The smoothsurfaces of these components reduce breakage.Fracture under high stress begins onirregularities in the materials, often at surfacescratches. In addition, the friction between thepiston and the bore of the cylinder as the pistonmoves within the bore is reduced when thesurfaces are smooth. In experiments where it isnot necessary to have electrical leads into thesample region, it is best to make the piston-and-cylinder device double-ended, as Coes did. Adouble-ended piston-and-cylinder devicerequires extraspecial care in the alignment and inthe parallelism of the various components withthe moving elements of the hydraulic press.

The piston should be as short as possible andshould protrude as little as possible from the boreof the cylinder. The protruding part of the pistonis the weak part, and fracture will usually occurthere. A tight-fitting collar, as shown in Fig. 2.5,around a portion of the protruding part of thepiston is very helpful in preventing breakage onthe exposed piston ends. This collar is of lowalloy steel such as SAE Number 4340 hardenedto Rockwell C55. The interference fit between

the collar and the piston should be about 0.003cm/cm of piston diameter. That is, the insidediameter (I.D.) of the collar is smaller than theoutside diameter (O.D.) of the piston by 0.003cm for each cm of piston diameter. In addition,the I. D. of the collar should be tapered about0.050 cm/cm length. A matching taper is placedat one end of the piston so that when the collar ispushed over the piston with a forcing press, theinterference of about 0.003 cm/cm is developedwhen the small end of the collar is flush with thesmall end of the tapered portion of the piston.Molybdenum disulfide paste (such as MolykoteG) is used as a lubricant between the matingsurfaces as they are forced together.

A fully annealed (dead soft) steel safety ringsurrounds the hardened steel collar. This must bein place at all times during and after the forcingoperation. Highly hardened steels can breakcapriciously when under heavy tension, andprecautions must be taken. The O.D. of the collarshould be about three times the piston diameterand its thickness should about equal the pistondiameter. The safety ring I.D. should be the sameas the O.D. of the collar. There is no taper here.The O.D. of the safety ring should be about fivetimes the piston diameter. All these parts arebuilt to precision tolerances, the hardened steelsand carbides being ground to final dimensions.

At maximum compression of the sample theportion of the piston "exposed to air" should beas small as possible so that all portions of thepiston are receiving some kind of "support"; thatis, the cylinder supports that portion of the pistonwithin the bore (in compression the pistonexpands), the reaction of the sample against thetip of the piston within the bore supports the tip,the driving force of the press on the top side ofthe piston supports the top, and the tight-fittingcollar around the piston supports the portion itsurrounds. Thus, the only part of the pistonunsupported is the small exposed portion that hasnot entered the cylinder bore.

When operating at the highest pressures,even with the aforementioned supportprecautions, the piston's lifetime is usually onlyone or two runs, and those using simple-pistonand cylinder devices at the highest pressuresmust have a large supply of pistons.

After each run the cylinder bore and theoutside diameter of the piston are carefullycleaned and coated lightly with molybdenumdisulfide powder, a very good dry lubricant. Thispowder is best applied by dipping the finger intoit and then rubbing the coated finger onto theparts.


The steel platens of a commercial press willusually be a rather soft steel with Rockwell Chardness around 20. Consequently the tungstencarbide piston cannot push directly against theseplatens. Hardened steel, preferably of RockwellC55, must be placed on top of the platens, andthen tungsten carbide must be placed on top ofthe hardened steel. This tungsten carbide in turnbears on the end of the piston. All of thesecomponents must be square and parallel,including the platens of the press, for theslightest cocking of the brittle pistons will causethem to break. The tungsten carbide backingblock, which bears on the end of the piston,should also be of grade C4. Its area should beabout three to five times that of the area of theend of the piston, and this backing block must besupported by a steel binding ring to keep it underradial compression. In this application, a singlebinding ring is usually satisfactory. A low-alloysteel forging, such as SAE Number 4340 heattreated to Rc40-43 with a 0.0015 cm/cmdiameter interference fit is commonly used. Thesides need not be tapered but may be straight atthis relatively low interference. The edges of thecarbide backing block are rounded and polished,and the steel ring is pushed over the carbide witha forcing press. Molybdenum disulfide paste isagain used as a lubricant between the matingparts. The outside diameter of the binding ring isusually about twice the diameter of the carbidebacking block. The thickness of the backingblock should be about two thirds of its diameter.

In electrical resistance heating of samplesunder high pressure, the resistance heater isusually of very low electrical resistance.Consequently a high current at low voltage isrequired. High currents require large electricalconductors. It is difficult to use a double-endedpiston cylinder device, where pistons andcylinders are electrical conductors, if electricalresistance heating is required. In this instance itis best to have a closure on one end of thecylinder bore to serve as one electrical contactand the moving piston at the other end of thecylinder serving as the second electrical contact.The closure must be insulated from the cylinderbore, or there will be a short circuit. It is alsodesirable to place thermocouples and otherelectrical leads inside the higher-pressurechamber. The closure can also serve thispurpose. A suitable closure is shown in Fig. 2.6.This closure is made about 0.13 mm smaller indiameter than the bore and is wrapped with astrip of electrically insulating paper about 0.05mm thick. Thermocouple or other electrical leads

can gain entrance to the interior of the cylinder,as is shown in Fig. 2.6, where there are twotapered slots (more could be used). The taper isnot critical. A taper of about 0.1 cm/cm on adiameter is satisfactory. Electrical leads of about0.1 to 0.2 mm in diameter are satisfactory formost purposes. The narrow part of the taperedslot should be about 1 mm in diameter. A pieceof pyrophyllite is machined to fit the slot, and atiny hole is drilled along its length toaccommodate a wire. This hole can beconsiderably larger than the wire without causingproblems. For example, if a 0.2-mm wire is used,a 0.4-mm hole is satisfactory.

Under high-pressure operation thepyrophyllite is compressed and driven into thetaper, and forms a strong frictional grip aroundthe wire. The frictional properties of thepyrophyllite can be increased by painting thesurface with a water suspension of red iron oxideand allowing this to dry. Instead of using slots onthe exterior surface of the closure, it is possibleto put tapered holes within the closure. Forexample, a tapered axial hole along the centerline of the closure would appropriately do thesame job. However, such a tapered hole isusually more expensive to make than the groundtapered slots on the surface. The thermocouplewires or electrical leads so introduced into thesample should be covered with an electricalinsulating enamel. Provision should be made toprevent these electrical leads from being pinchedor cut at places where they emerge from theclosure. This requires holes or slots to be formedor ground in the carbide backing block toaccommodate their passage. The bottom side ofthe cylinder and binding ring must be insulatedwith a sheet of electrical insulation (usually fiberglass epoxy) to prevent contact with the backingblock. However the closure must make electricalcontact with the backing block. The backingblock and steel assemblies must be electricallyinsulated from the platens of the press in order toprevent short circuiting through the hydraulicpress.

H. TRACY HALL 11

An assembly for internal resistance heatingof the sample to high temperatures is shown inFig. 2.8. The entire assembly of Fig. 2.8 iscompressed by the piston. Heating current passesfrom the piston through the metal washer Awhich it contacts to the electrically conductingheater sample tube E, then out through the metalwasher G, the contacting cylinder closure, andthe carbide backing block. The electricallyconducting tube may be a high-temperaturemetal such as tantalum or molybdenum or it maybe made of graphite. The metal washers shouldbe made of a refractory metal. The solid cylinderC surrounding the heater sample tube musttransmit pressure quasi-hydrostatically to thetube and must be electrically and thermallyinsulating. It must also withstand hightemperature. A common material used for thispurpose is the naturally occurring mineralpyrophyllite, sometimes called Grade A Lava, orWonderstone. This material is a hydrousaluminum silicate which will lose water at hightemperature and which under high-temperature,high-pressure conditions will transform intoother materials of higher density. Pyrophyllite isrelated to talc. Talc is a hydrous magnesiumsilicate and is softer than pyrophyllite. Often aliner made of hexagonal boron nitride is used tosurround the heater sample tube. This materialwill of course not give off water. Water frompyrophyllite sometimes penetrates through the

heating tube into the sample, or else hydrogenand oxygen will be given off at the high-temperature, high-pressure conditions. As aresult, undesired reactions may occur. The BN,at sufficiently high pressures and temperatures,will convert to cubic boron nitride, a high-density polymorph analogous to diamond, andmay not transmit the pressure as desired.

There is considerable leeway in the designparameters for the insertion of electrical leadsinto high-pressure apparatus, and workers in thefield tend to develop their own particularpreferences. In general, an electrical lead can beinserted by having the wire pass through anunfired ceramic material in a tapered slot or holeand, in some instances, even in a slot or holewith parallel sides. The frictional properties ofthe ceramic material cause bind-up in the hole orslot and prevent the extrusion of thethermocouple wire. In some instancesthermocouple wires are pinched off in passingthrough the slot if the slot is too short butchanging the slope of the taper or the diameter ofthe hole or the particular frictional material usedcan, with trial and error, eliminate the difficulty.Pyrophyllite is the most commonly usedmaterial. However commercially availableunfired or partially sintered alumina, magnesia,and other substances will also serve for thispurpose. Also it is possible to merely compact apowder into the hole through which thethermocouple passes. Red iron oxide has higherfrictional properties against tungsten carbidesurfaces than any material that has been studiedand is often used in this application as well as tocoat pyrophyllite to increase surface frictionwhere pyrophyllite contacts tungsten carbide.Sheathed pairs of thermocouple wires (or otherwires) are readily available today and can be


used advantageously in high-pressure high-temperature research for insertion into thesample cell. The sheath is usually stainless steelbut may be made of more refractory metals suchas tantalum. The two wires (it is also possible toobtain material with more wires) are imbedded ina ceramic powder, usually MgO or Al2O3, withinthe sheath. The total diameter of the assembly isoften only 0.25 mm.

Piston-and-cylinder devices were first usedin the compression of gases, and the pistonusually had a leather cup packing, exactly likethe hand-operated air pump used for inflating abicycle or automobile tire. Pressures attainable insuch devices, of course, are small. Moresophisticated seals are needed for higherpressures. O-rings serve admirably in manyapplications, but if pressures cause extrusion ofthe O-ring, the "Bridgman seal" can be used12.The Bridgman seal is shown in Fig. 2.7. Aplastic or elastomeric substance, in the form of awasher, goes between the hardened steelmushroom head and the piston tip which has acentral hole to accept the stem of the mushroomplug. Gas or liquid pressure is exerted against themushroom head of the seal. The area of themushroom plug head exceeds the area of theflexible washer. Consequently, the pressureagainst the washer will always exceed thepressure of the fluid under pressure. This beingthe case, there is no way that this seal can leak.The washer will just push harder and harderagainst the walls of the cylinder as oil pressureincreases. Professor Bridgman has told theamusing story of applying for a patent on thisseal. The idea was original and unique to him,but in applying for the patent he found that thesame principle had been used in a patentedsausage grinder. Consequently he was not able toobtain a patent. Yet workers in the field of highpressure invariably refer to the seal as theBridgman seal, the sausage grinder inventorhaving been lost to history.

The direct heating of gases and liquids inapparatus where solid pressure-transmittingmedia are not employed can only be achieved byexternal heating of the entire chamber.Temperatures of about 300°C are obtainable inexternally heated piston-cylinder devices. Highertemperatures are precluded by loss of strength ofsteel and tungsten carbide above 300°C.

In any experiment at pressure it would bedesirable to have the specimen that is the objectof the research subjected to what is termedhydrostatic pressure; that is, pressure impingeson the macroscopic object of study from every

direction on a microscopic scale. This occurswhen a true fluid surrounds the specimen undertest. In the high-pressure devices invented todate, the pressure-transmitting medium must be asolid to obtain the highest static pressure. Thusthe pressure on the specimen is not trulyhydrostatic, but with proper design the pressurescan be reasonably hydrostatic. This is oftenreferred to as a quasi-hydrostatic pressure. Thehighest hydrostatic pressures are obtained usinga solid pressure-transmitting substance thatpresses on a thin-walled metal tube in whichliquid is contained. This liquid, in turn, may beexercising its pressure on a sample containedwithin itself13. There is a problem as to thechoice of a suitable liquid because most liquidreagents found on the laboratory shelf becomesolid (freeze) at room temperature at pressuresless than 10,000 atm. Liquids that will remainliquid at high pressure are usually mixtures ofmolecules that do not fit together well andconsequently have difficulty in crystallizing. Amixture of 4:1 by volume methanol/ethanolremains hydrostatic to almost 100,000 atm atroom temperature14 and 1:1 by volumepentane/isopentane to about 70,000 atm. Theviscosity of liquids increases with pressure, andno doubt the above mixtures have theconsistency of heavy molasses as theirhydrostatic limit is approached.

There are some solid substances thattransmit pressure rather well. Notable amongthem is indium metal. Among nonconductors,silver chloride transmits pressure effectively.Whether a substance will be a good pressure-transmitting medium (will approachhydrostaticity) or not can be inferred from shearfriction measurements made under pressure inBridgman anvils15. Indium and silver chloridehave low shear friction at high pressure.

A little philosophy concerning hydrostaticitymight be in order at this point. If the object ofstudy at high pressure is a solid material, onemight indeed have a true fluid pressing in on thissolid specimen from all directions, and onewould be inclined to say that the specimen isbeing subjected to hydrostatic pressure. Well, ofcourse it is on its exterior surface. However thequestion could be asked, "What about thepressure at some point deep within this solidsubstance?" The pressure at this point wouldhave to be transmitted by the solid specimen thatsurrounds that point, and this in turn woulddepend on the characteristics of the solid andalso on its shape. If the specimen is a metal, itwould also depend on the particular physical

H. TRACY HALL 13

treatment or heat treatment the metal might havehad. One would expect that the pressure at thecenter of a sphere of steel would be higher thanat the center of a tungsten carbide sphere of thesame size (the modulus of elasticity of tungstencarbide is three times that of steel). Geometricaleffects have been observed in such materials asbismuth since this material is often used as afixed point for the calibration of high-pressureapparatus. With a true hydrostatic liquidsurrounding bismuth metal, the transition pointwill be obtained at a different press loading whenthe bismuth is used in the form of a thin ribbonthan when the bismuth is used as a round wire.Can one really perform "hydrostatic"experiments on solid substances?

When the cell assembly (Fig. 2.8) for apiston-cylinder device is compressed, there is acertain amount of "frictional holdup"; that is, acertain percentage of the pressure at the pistontip is not transmitted to the sample. Thisfrictional holdup is due to two things. There isfrictional holdup within the pyrophyllite becauseit is a solid. There is also frictional holdup at theinterface between the cylinder of pyrophylliteand the bore of the tungsten carbide cylinder.The friction I at this interface can be reduced bypainting the outside of the pyrophyllite with asuspension of molybdenum disulfide or bysheathing the pyrophyllite with In, BN, AgCl, orNaCl. Silver chloride and sodium chloride arecorrosive to most metal parts and should not beallowed to remain in contact with them for longperiods of time. Substances other thanpyrophyllite may be used for the solid portion ofthe cell in piston-cylinder devices. They havetheir advantages and disadvantages. Forexample, boron nitride would have a lowerinternal friction, but it is much more expensiveand has a much higher thermal conductivity, thusrequiring higher electrical power to maintain thehigh temperature inside. Sodium chloride is arather good substance to use, since its thermalconductivity under pressure seems to be nearlythe same as that of pyrophyllite and its pressure-transmitting characteristics are better. Howeverthe fabrication of sodium chloride cylinders isdifficult.

When polycrystalline refractory, substancesare used as pressure-transmitting media, ahysteresis exists in which, with increasingpressure external to the solid, the pressure withinthe solid is less than that outside. Upon reductionof external pressure, however, the pressure insideis higher than that on the outside. A typicalhysteresis loop is shown in Fig. 2.9 for a cell

assembly such as that of Fig. 2.8, in whichpyrophyllite is the pressure-transmittingmedium. It is possible to utilize this phenomenonto reduce the load on the pistons or otherelements of a high-pressure device and thereforeincrease the useful life of expensive carbidecomponents. It works in this manner (refer toFig. 2.9): The press load, that is, the load on thepiston, would be increased to a certain valuewhereupon there would be a given value ofpressure at the surface of the sample and a lesserpressure at the center of the sample. The pressureon the piston tip could then be reduced byreducing the pressure on the hydraulic ram thatactivates the piston. This would result in areduction of the pressure on the exterior of thepyrophyllite, while the higher pressure would bemaintained at its center. Hysteretic phenomenahave been observed in which a differentialpressure of 25,000 atm has been maintained; thatis, the piston would be retracted to the pointwhere the pressure at its tip was 25,000 atmlower than the pressure being maintained byfrictional hysteresis in the center of apyrophyllite cylinder.

Because of the hysteresis in solid pressure-transmitting substances (apparatus friction alsoenters in), calibration of pressure apparatus isusually made on a basis of ascending(increasing) pressure only. Calibration is usuallymade in the following fashion: a cylinder ofsilver chloride is substituted for components B,D, E, and F of Fig. 2.8 and a wire of fixed pointcalibration material such as bismuth, thallium, or


barium is placed through an axial hole in thesilver chloride. The pressure-sensing wire

Table 2.1 Fixed-Point TransitionsHg 7470 ± 2 atm at 0°CBi I-II 25.16 ± 0.06 katm at 25°CTl II-III 36.2 ± 0.3 katmCs II-III 41.9 ± 1 katm (up only)Cs III-IV 42.4 ± 1 katm (up only)Ba I-II 54.3 ± 2 katmBi III-V 76.0 ± 3 katmSn 1-11 99 ± 6 atmBa ~138 (up only)

makes contact with metal disks (substituted forwashers A and G), which in turn make electricalcontact with the piston, the closure, the backingblocks, and so on, and thence to an instrumentthat accurately measures electrical resistance.The accepted fixed-point transition pressures fora number of calibrants, including Bi, Tl, and Ba,are given in Table 2.116. The shapes of some ofthe electrical resistance curves are shown in Fig.2.10.

The determination of the pressure at whichthese electrical discontinuities occur is difficult.Briefly, there is a device called the free pistongauge that can be used to determine fixed pointsup to about 25,000 atm. Beyond this pressure,techniques become more complicated, but in themain the higher pressure fixed-point transitionsare determined against in situ X-ray diffractionmeasurements in which the electrical resistanceof a substance, such as barium, is measuredsimultaneously with the measurement of thechange in the crystal lattice spacings of asubstance such as sodium chloride. In someexperiments, the detailed nature of the phasechange of the fixed-point material (Sn, Ba, etc.)has been determined at the identical place where

the corresponding electrical discontinuity occursand simultaneously with a determination of thelattice spacing in NaCl which surrounds thefixed-point material17. The continuous change inlattice parameter of sodium chloride as afunction of pressure (semiempirically determinedby D. L. Decker) is the currently acceptedstandard used to determine the value of the fixedpoints18.

There is some hysteresis in the metal wireitself as it undergoes transformation onincreasing pressure, as compared to undergoingthe reverse transformation on decreasingpressure. These hysteretic effects, however, aremuch smaller than is observed in suchsubstances as pyrophyllite. The silver chloridecylinder in which the fixed-point wire isenclosed is made by compressing powderedsilver chloride in a hardened steel die to therequired shape. The central hole is made by atiny spade drill. Care must be exercised to avoidbreaking the drill, inasmuch as the silver chloridehas a gummy, tough consistency. The bismuth,barium, and tin wires are usually prepared byextrusion through a die. The best die has aconical entrance angle, the exact dimensions ofthis cone not being too important, and an abrupt90° exit angle. The usual diameter of the wire isaround 0.1 to 0.4 mm. A slug of the metal to beextruded is placed in the die, and a piston pusheson it to cause extrusion. Bismuth and tin arereadily extruded into the air. Since heat isgenerated in extrusion through the die, care mustbe used in the case of barium. Barium wire isdifficult to make because it tends to hold up inthe die and then, at a certain pressure, extrudesviolently emitting fire and poisonous smoke. Itshould be extruded with care into paraffin oil.Thallium can be extruded well but, like barium,is poisonous! Barium and thallium oxidize in airand need to be kept under oil.

Another way of obtaining a "wire" of thesesubstances is to simply place a slug of thematerial between two tungsten carbide blocksand press the material into a flat sheet. A flatwire can then be obtained by cutting it from thesheet with a razor blade. Connecting fixed-pointcalibrant wires to metal components of the cellcan be a problem. Thallium and barium wiresmust be scraped free of oxide just before use.Barium is quite brittle. Bismuth is very stable inair, but is also brittle. Soldered connections havebeen used, and it is possible to use bismuth itselfas a solder to bismuth wire, but I prefer to makemechanical connections to all calibrant wires.They are made along the lines indicated in Fig.

H. TRACY HALL 15

2.11. A cleanly drilled hole is made in one diskof metal, and a hole is punched with a sharp awlin another disk. The disk is usually made ofcopper or silver sheet; then a piece of bismuthwire or other wire is placed in this assembly asshown in the figure. When the assembly comesunder pressure, the conical hole punched in thecopper sheet with the awl is forced to collapse onthe wire and make a good connection. A greatdeal of art is involved in this procedure, and eachresearcher tends to develop his own methods toovercome the problems. Cesium is a veryreactive liquid metal at room temperature, but a"wire" of it can be made by filling a smallpolyethylene tube with it from a hypodermicneedle under paraffin oil. Solid wires of copperor other metal just slightly larger than the insidediameter of the tubing are inserted in each end ofthe polyethylene tube to make electricalconnection to the cesium and to protect it fromthe atmosphere.

Equipment that utilizes solid pressure-transmitting media is calibrated by noting thehydraulic ram oil pressure required to cause eachof the fixed-point transitions to occur. Thetransitions are made only with increasingpressure, and pressure is increased very slowlyover as much as an entire day. A person has tobecome familiar with his own particularapparatus and cell design by making severalcalibration runs in order to obtain a usefulcalibration curve. The hydraulic oil pressure tothe hydraulic system, read on a sensitive gauge,is plotted as a function of the accepted fixed-point pressures of the calibrants, and a line isdrawn to best fit the points. This is all done atroom temperature. Even so, this is often used asthe pressure calibration when the press is used athigh temperature.

There will certainly be some uncertainty inknowing what the pressure is when the device isused at high temperature. Pressure studies at hightemperature (say at 800°C and above) have beenfew because the problem is so difficult. At hightemperature, changes take place in thepyrophyllite or other pressure media thatgenerally increase the density of the pressure-transmitting material and tend to lower the

pressure. On the other hand, the elevatedtemperature will cause expansion and tend toincrease pressure. About the best that can bedone at present is to assume that these effectsapproximately cancel each other. Anotherproblem, although minor to the aforementionedproblems, is that the reading of a thermocoupleis affected by pressure. These difficulties make itimperative that a high-pressure apparatus bedescribed in great detail, listing pertinentdimensions, materials of construction, andprocedures in order that others will be able tomake sound judgments on the research. Pressurevalues without this information are notmeaningful.

The most common thermocouple used attemperatures above 1000°C in high-pressurework is Pt, Pt-10% Rh. Such a thermocouple willbehave well to about 1700°C. With a given celldesign and physical arrangement in a givenapparatus, a temperature calibration curve can beobtained by plotting thermocouple temperatureversus the electrical power in watts. This plot cansubsequently be used to indicate the approximatetemperature of a sample simply by measuring thecurrent and the voltage imposed upon the heatersample tube. Additional confirmation andextension of the plot may be obtained byobserving the wattage required to melt fine wiresof tungsten or other high-melting metals. Thesewires are inserted in the sample in the samemanner as the thermocouple leads, melting beingdetected by the occurrence of an open circuit.When more accurate temperature information isneeded than that afforded by the watts-versus-temperature plot, thermocouple leads should beinserted and individual measurements made forthe experiment concerned.

Pyrophyllite melts incongruently to a glassysubstance at temperatures near 1500°C atpressures of a few thousand atmospheres. Itsmelting point is increased considerably at higherpressure, but the pyrophyllite does form a hard,white mixture of substances, as has beenpreviously mentioned. This white, hardsubstance is observable only on removing thecell after an experiment. The exact nature of thismaterial is not known at the high pressure andtemperature involved in its formation, andwhether or not it seriously hinders transmissionof pressure to the sample is also not known.

When the sample located between themoving piston and the closure is subjected tohigh pressure, the tungsten carbide cylinder inthis region is caused to expand (its diameterincreases). The maximum increase in diameter


occurs at a plane midway between the ends ofthe sample and tapers off toward the ends of thesample. However the clearance, which initiallymay only be 0.004 mm on a diameter betweenthe piston and the cylinder bore in the piston tipregion, increases sufficiently to allow cellmaterial to extrude between the piston andcylinder. This causes severe binding of thepiston, making the piston difficult to remove,and much piston breakage ensues. This isparticularly true if it is pyrophyllite that extrudes.This problem can be reduced by placing an outersleeve around the pyrophyllite that is made ofpolycrystalline hexagonal boron nitride. Boronnitride tends to be slippery and partiallyovercomes the problem. Whatever the material,when load on the piston is removed, the trappedmaterial between the piston and the cylinderbecomes more tightly wedged because thepressure on the cylinder is now relieved and itcontracts. This is a major difficulty with piston-and-cylinder devices at high pressure.Sometimes a hardened steel ring with an outsidediameter that fits snugly against the cylinder walland that has a tapering inside diameter (Fig.2.12) is placed against the tip of the piston tohinder extrusion19. Breaking the piston everysingle run is to be expected at 50,000 atm.

The belt device, to be described later, doesnot cost any more than a piston and cylinderdevice; in fact, the costs are comparable.Furthermore the belt does not require a clampingcylinder. In spite of this, the piston-and-cylinderdevice is used by many workers in the high-pressure field. Ordinarily, though, the device isused at pressures below 40,000 atm. Thedifficulties mentioned above are not as importantat these pressures, but at 50,000 atm and above,the problems are vexatious.

Many of the items discussed for piston-cylinder apparatus are applicable to other typesof apparatus to be considered later.

4 BINDING RINGS

Piston-cylinder and belt devices requirebinding rings to give lateral support to a tungstencarbide cylinder or die. The support required issubstantial and is usually provided by a set ofcompound binding rings, that is, two or morebinding rings nested together. There are theoriesfor the construction of these sets20, but most setsare the result of trial and error. The idea of thecompound binding ring set is to squeeze thecarbide as strongly as possible but at the sametime keep the size of the rings relatively small. Inresearch, the binding ring set is usually lifted inand out of the hydraulic press by hand--often atarm's reach. Not only that, but (as I wellremember from my earliest research with thebelt) at arm's reach on tiptoe. Thus it is importantto keep the weight of the set manageable.

A practical set is shown in Fig. 2.13. Thisset employs two tension rings, a split shim, and asafety ring. Note the tapers and the interferencefits. The smaller diameter given at an interface isthe I.D. of the outer ring, and the larger diameteris the O.D. of the inner ring. Assembly begins byplacing the outer tension ring within the deadsoft, low-carbon safety ring. Next, the innertension ring is forced into the outer tension ring.Molybdenum disulfide paste lubricant is used onall sliding surfaces during the assembly. At thisstage the I.D. of the inner tension ring is taperground to the dimensions shown. The shimwhich consists of four equal segments isequispaced around the carbide cylinder or die(there are gaps between the shim segments), andthis subassembly is then forced into the innertension ring. It may require 200 tons of thrust forthis final assembly operation.

The design is such that the assembledtension rings are stressed near their ultimatepractical capabilities, and occasionally one orboth rings will break. I have seen this happen onseveral occasions, once while holding a newlyassembled set in my hands. However the safetyring absorbed the stored energy and preventedmishap in this and in the other instances. If thereis going to be breakage, it usually occurs duringthe first 24 hr after assembly. Ring sets survivinga week will be good for many years.

The primary purpose of the shim is to savethe inner tension ring from being scored ordamaged from breakage of the tungsten carbidedie. Tungsten carbide, being harder than steel,will readily scratch, scuff, and gall at a carbide-steel interface upon breaking but moreparticularly and severely when a fractured die isforced out of a press-fit assembly. The shimsustains only modest damage on carbide die

H. TRACY HALL 17

fracture, and the die and shim are pushed outtogether. The shim, being in sections, can thenbe readily removed from the broken carbide. Apushing-out operation is not needed. A littletouch-up work is usually required on the innersurfaces of the four-piece shim whereupon it isready to accept a new die and be forced againinto the compound binding ring assembly.

For long runs at high temperature, a watercooling jacket is built to surround the safety ring.Hardened, low-alloy steels will lose theirstrength if heated too hot, and temperatureexpansion of the steel will decrease the inwardthrust on the die. It is best to keep the steel at atemperature below 100° C. Tungsten carbide willalso lose strength with increasing temperature.

The tension rings are constructed of AISI-4340 aircraft-quality ring forgings. After beingmachined to size plus a grind allowance, they areheat treated to Rockwell C-55. Final dimensionsare then obtained by grinding. The C-47 to C-53hardness range in this steel is notch sensitive andshould be avoided because parts of such hardnessare more likely to fracture. The shim isconstructed of an air hardening, nondeformingtool steel. It is ground to size and cut into fourequal segments parallel to the cylindrical axis. Itis then heat treated to Rockwell C-63 and peenedby a steel shot blast primarily to remove scale.Finally, the outer surface is polished to reducefriction during assembly. The taper of the fits issmall enough that friction will keep the ringsfrom sliding apart after assembly. The units arerather permanent and stable. I have had somesets for nearly 20 years that are still in goodcondition. The dimensions of Fig. 2.13 may beused as a base for scaling upward or downwardin size.

5 HYDRAULIC PRESSES

Somewhere in a scientific laboratory there isusually a hydraulic press. If so, it can often be

inexpensively modified to provide driving thrustfor piston-cylinder, belt, and Bridgman anvilapparatus. Such apparatus are consequentlypopular since a good 300-ton press (a nominalsize for research) can cost $25,000 or more.

Presses not specifically designed for high-pressure research usually are not well enoughaligned and have too much play in their slidingmechanisms. However this can be overcome byuse of a device called a die set. Die sets are usedin the precision punch and die stamping of sheepmetal and are designed for the very purpose ofsolving these press problems. The die set isinterposed between the platens of the press, andthe high-pressure apparatus is placed within thedie set.

Ordinary hydraulic presses have only oneram. There is not a second ram for use inclamping the cylinder of a piston-cylinderdevice. In this event a permanent clamp can beplaced on the cylinder. This increases the weightof the cylinder with its compound binding ringassembly and necessitates some changes indesign but is satisfactory. The clamp consists oftwo disks of hardened steel relieved so as toclamp only the carbide cylinder from top tobottom. There is no clamping of the bindingrings. A circle of matching holes in these disksjust beyond the O.D. of the safety ring accepts adozen or so high-strength bolts which are heavilybut evenly torqued to provide at least 25 tons ofclamping load. The central region of the clamptapers away from its I.D. at a 45° angle toprovide room for the piston, its collar, and so on.

If a standard hydraulic press is to bepurchased, a hobbing press is the least expensiveand the best. It has high tonnage in a small sizeand fairly good alignment. Presses specificallydesigned for high-pressure use are available fromonly a few commercial suppliers. The bestcommercial high-pressure equipment is sold as acompletely integrated, turn-key package but isvery expensive.

The least expensive hydraulic presses havespring or gravity return cylinders. In the lattercase the cylinder pushes upward from the bottomof the press, and gravity pulls it down. Thereturn action of these presses is slow. It is muchbetter to have a press with a double-acting ram(hydraulic cylinder). Oil, under pressure,advances and also returns such a ram.

Pressurized oil for actuating hydraulicpresses is usually supplied by an electricallydriven two-stage hydraulic pump. This pumpautomatically supplies a large volume of oil atlow pressure- say 35 atm- to move the pressing


elements to the point of contact. Then the high-pressure stage of the pump automatically takesover as resistance to motion is sensed. Thesecond stage supplies high-pressure oil at lowvolume to move the pistons, or other pressingelements the short distance required to generate ahigh pressure in an apparatus which utilizes asolid pressure-transmitting medium. Electricpumps capable of operation at 650 atm arereadily available and are of modest cost.Electrically driven pumps for higher pressuresare very expensive. If higher oil pressures aredesired, it costs less to buy air-drivenreciprocating pumps. Such pumps, capable ofpressurizing fluids (including hydraulic oil) to4000 atm, are readily available.

If convenience or automatic controls for thehydraulic press are desired, it is best to use oilpressures below 650 atm. Electrically andpneumatically controlled systems are readilyavailable up to this pressure. For experimentslasting more than a few minutes, automatic oilpressure control is highly desirable. Oil pressurewill change with temperature and will fallbecause of seal leakage in the press.

Oil pressure is measured with a precisionbourdon tube gauge. For electronic recording,special pressure transducers are commerciallyobtainable.

6 ELECTRICAL RESISTANCE HEAT-ING EQUIPMENT

The graphite and metal heater sample tubeshave an electrical resistance of the order of 0.001ohm in a cell such as that of Fig. 2.8. It willrequire about 500 W (a heating current in theneighborhood of 250 A at a potential around 2V) to reach a temperature of 1500° C in thesample. There are experiments in which a currentof 1000 A may be required, but currents of thismagnitude should only be used for a fewminutes. Longer periods of time will damage thetungsten carbide. Such high currents causestrong local heating where the current leaves thetungsten carbide to flow into the cell. Thecombination of pressure plus temperature causesmuch more carbide breakage than does pressurealone.

Alternating current is usually used forheating. The least expensive low-voltage, high-current transformers that I have found arestacked core welding transformers (the so-calledwound core transformers will not do). Thesetransformers have a limited-duty cycle for theirintended purpose at their indicated power rating

but can be used in continuous-duty resistanceheating at a lower rating. A 35-kVA, 440-V, 50-or 60-Hz transformer is satisfactory for generalresearch but it should be used only at a 220 Vmaximum. At 220 V, the voltage of such atransformer is around 4 V with external seriesarrangement of the output. The transformers areusually water cooled and a 35-kVA unit weighsabout 100 kg. They are available from a numberof welding transformer manufacturers.

It will be necessary to vary the output of thewelding transformer. This may be done by usinga variable autotransformer arrangement withcoarse and fine controls as shown in figure 2.14.The coarse-control autotransformer should berated at about 100 A. The output of the fine-control autotransformer should be rated at about10 A. The bias transformer is a “buck-boost”transformer of 1:10 or 1:20 turns ratio capable ofcarrying 100 A in its low-voltage output. Buck-boost transformers are available from industrialelectrical distributors.

The circuit should contain a voltmeter andan ammeter to determine the voltage across thesample heater tube and the current flowingthrough it. A current transformer rated at 1000 Awill be required in the current-measuring circuit.A wattmeter is also desirable. The product of VX A will not always agree with the wattmeterreading because of power factor considerations.

A less expensive control for the output ofthe welding transformer is a silicon-controlledrectifier (SCR). I have been unable to locate astandard commercial control of 100-A capacitybut have had such controls custom-made by anelectrical engineer. SCR controls are light inweight and small in size. However, they have thedistinct disadvantage of generating a lot ofelectrical noise which may affect the operation of

H. TRACY HALL 19

sensitive electronic measuring instruments beingused in a high-pressure, experiment.

Direct-current heating of the sample wouldbe desirable for experiments in whichmeasurements would be adversely affected bythe presence of the 50- or 60-cycle alternatingcurrent. Standard, variable voltage electronicdirect-current supplies are commerciallyavailable up to 100 A at about 8 V, and theseunits may be used in parallel. They are howeverexpensive.

7 OPPOSED-ANVIL DEVICES

The highest static pressures obtained in thelaboratory are obtained in opposed-anvil devices.Bridgman relates that he conceived this ideawhile contemplating the pressure that might existat the contact point of two knife blades pressedtogether at right angles to each other. Oneprinciple, operative in this apparatus is shown inFig. 2.15. Two truncated tungsten carbide conespressed together along an axis as shown at (A)will withstand a greater load than two rightcircular cylinders with the same face area andlength as shown at (B). Bridgman called this themassive support principle. In Bridgman's firstanvil research, a pinch of powder or a solid diskwas placed between the faces of the anvils (Fig.2.16a), and they were then pressed together.Pressure was computed simply as force per unitarea of the faces.

Later, an assembly (cell) as shown in Fig.2.16b was devised to enable the electricalresistance of metals to be measured as a functionof pressure. The outer ring E of the cell ispipestone. Pipestone is an impure pyrophylliteobtained near Pipestone, Minnesota. Thissubstance was used by American Indians inmaking peace pipes. Within the pipestone ring isa silver chloride disk F in which a metallicribbon G is embedded, as shown. This ribbonmakes contact on the top side with the top anvilface and on the bottom side with the bottomanvil face. If the anvils are electrically isolatedfrom each other, an electrical indicating devicecan be connected to them to measure electricalresistance as a function of pressure. If the anvilcircular faces are approximately 10 mm indiameter, the pipestone ring is made about 0.6mm thick. The outside diameter of the pipestonering is 10 mm (same as anvil diameter), and theinside diameter is about 7.5 mm. This in essenceis a two-dimensional pressure apparatus, and thecomponents within the cell are tiny indeed.Bridgman measured the electrical resistance of

many substances at pressures up to what werethought to be 100,000 atm21. Later work hasshown however that the highest pressureobtained in these experiments was only 65,000atm.

It is a general principle that the smaller theapparatus, the higher the pressure that can begenerated. This is particularly true of opposed-anvil apparatus. These devices have beenminiaturized to the point where diamonds areused for the anvils and cells are assembled undermicroscopes. The cell (sample) thickness may beas small as 0.025 mm, and the diameter mayonly be 0.50 mm. Great care must be used insuch miniscule devices to avoid error. Slightdimensional errors in fabricated parts (pipestonerings or silver chloride disks, for example) cancause great changes in the pressure obtained witha given anvil load. Also, samples being X-rayed,a common procedure with diamond anvils, mayshift in position and an X-ray diffraction patternof the wrong material may be taken. A serious


extant error of this nature occurred when aresearcher published the exciting discovery of anew form of carbon of greater density thandiamond. Actually the X-ray diffraction patternof the "new carbon" was merely that of silverchloride!

Unfortunately much of the work reported inthe literature in which tiny samples and opposed-anvil-type devices have been used containsserious errors. The most common error relates tooverstating the pressure. Pressures as high as700,000 atm have been reported. Subsequentlythe pressure was found to be less than 300,000atm. Small opposed-anvil devices are capriciousand great care should be exercised in their use.

Compression of a gasket to obtain motion isthe second principle operative in opposed-anvildevices. This is the means by which a thickersample is contained and compressed. In the caseof ungasketed anvils, the frictional forces of thesample itself come to equilibrium with theapplied load at some thickness characteristic ofthe substance. This thickness would be of theorder of 10% of that of a gasketed sample. Thepressure obtained at a given anvil load varieswith the substance.

It is amazing that a thin ring of pyrophyllitegasket can contain the sample as the pressurebuilds up. The gasket not only decreases inthickness as the anvils advance but some of itextrudes until an equilibrium situation is reachedfor a given pressure. On release of pressureblowouts are a common phenomenon.Pyrophyllite elasticity is limited and, havingpartially extruded during pressure buildup, itcannot return to its initial thickness.Consequently, it loses its frictional hold on theanvil faces, and the compressed sample blowsout. Accurate alignment of the anvils,particularly parallelism of the faces, is animportant consideration. Slight misalignmentcan cause premature anvil breakage andblowouts and skew the pressure distributionwithin the sample.

At pressures of 100,000 atm and beyond, thefaces of tungsten carbide anvils are permanentlydeformed and become concave. Sometimes thedeformed anvils are "refigured" (ground flatagain with a diamond grinding wheel). Afterrefiguring, anvils are stronger (since they havebeen cold worked by the deformation) and canbe subjected to even higher loads than whennew. At the higher pressures, opposed anvils areoften used but once, since they usually break onrelease of pressure.

The sample while under pressure is lens-shaped, being thicker at the center and thinner atthe edges. This is due not only to permanentdeformation of the anvils but also to theelasticity of the anvils. There is also a pressuregradient in solid samples, the pressure beinghighest at the center along the anvil axis, trailingoff to much smaller pressures at the interfacewith the gasket. As a matter of fact, this is one ofthe major disadvantages of the device.Consequently, when measurements of variouskinds are to be made, it is necessary to make themeasurement in a tiny space as close as possibleto the center of the sample. This can be done inX-ray diffraction or optical experiments where avery fine pencil or rays is impinged upon thesample along the centerline axis.

By using a metal, rather than a pipestone orpyrophyllite, gasket, it has been possible to workwith liquids to very high pressures, and X-raydiffraction as well as optical observation andmeasurement of single crystals has beenaccomplished22.

Diamond anvils are particularly appealingbecause they can be used under a microscope, itbeing most interesting and satisfying to observedirectly a phase transformation taking place at apressure of 100,000 atm. Their use was firstdescribed in 195922. The least expensive way toobtain diamonds for use as anvils is to purchasebrilliant cut gem diamonds from a major supplierin New York or Amsterdam. These diamonds arepurchased with one slight modification: theculets (bottom tips of the diamonds) are groundoff to form small flats parallel to the tables(broad top surfaces). These flats are ground todifferent surface areas to minimize axialalignment problems when they are pressedtogether. Alternatively a ground-off culet flat issometimes pressed against the table of anotherdiamond. The diamonds are commonly of aboutone-tenth carat (0.020 g). The anvil faces areoften about 0.5 mm across. The ground-off culetwill not be perfectly round but octagonal. Ifoptical properties are of no great concern,yellowish diamonds, which cost less than water-white diamonds, can be used. If transmission inthe infrared is needed, type II diamonds shouldbe purchased. Type I diamonds have an intenseabsorption band at wavelengths from about 7 to9 #m. Regardless of color or type, though, thediamonds must be physically sound. Majordiamond vendors are familiar with these factorsand are prepared to supply the needs of diamondanvil users. Although diamond is the strongestsubstance (in compression) known to man, it is

H. TRACY HALL 21

brittle, and an off-axis load would readily chip it.To assure that the planes of the mating diamondanvils are parallel to each other as they cometogether, they are mounted in spherical sockets.Alternatively one diamond may be mounted toswivel about the x-axis and the other about the y-axis Cartesian coordinates to accomplish thesame purpose. The diamond is commonly set ina metallic mounting (usually stainless steel) bypressing it into the metal to form a seat. It is thenheld in place by epoxy cement. When the anvilsare used in X-ray diffraction work, the metalused is often beryllium because of its relative X-ray transparency.

The area of contact of diamond anvils is sosmall that it is not necessary to use hydraulicrams to drive them together. Most researchersuse either springs, whose tension can beincreased or decreased by thumbscrews, ormerely the inherent elasticity of steel itself, andclamp the anvils together with socket headcapscrews. In spite of the cost of gem-qualitydiamonds, a diamond anvil apparatus is the leastexpensive apparatus that a person can acquire.Consequently they are very popular. Some othermaterials that have been used for opposed anvilsare polycrystalline alumina and single-crystalsapphire.

Diamond anvils are so tiny that it would bevery difficult to provide internal heating tooperate at high temperature simultaneously withhigh pressure. However Ming and Bassett23 haveused a laser beam to heat samples between thefaces of the diamonds and have even convertedgraphite to diamond by this means. Barnett,Block, and Piermarini24 have constructed adevice for externally heating diamond anvils.

In addition to the pressure gradient problemin opposed anvils, there is the very real problemof accurately determining the pressure.Simultaneous X-ray diffraction of NaCl and asubstance embedded therein has been used indiamond anvils25. The lattice spacing of the NaClis used to indicate the pressure on the embeddedsubstance. A drawback of the system is theexposure time required-up to 300 hr.

In some respects the ultimate diamond anvilapparatus is that of Barnett, Block, andPiermarini24. They have described an opticalsystem for rapid routine pressure measurementwhich utilizes a pressure shift in the sharp R-linefluorescence of ruby. A metal gasket (simply arelatively large square of Inconel about 0.13 mmthick in which a hole of about 0.20 mm indiameter has been drilled) is centered betweenthe diamond anvils and a 4:1 mix of methanol

and ethanol is placed in the hole. A speck of rubyis added as well as a speck of whatever else isdesired, and the anvils are closed on the gasket.Whatever is inside is trapped, and furtheradvance of the anvils subjects the contents tohydrostatic pressure all the way to 100,000 atm.And the pressure is known by the fluorescence ofthe speck of ruby!

A recent miniature diamond anvil pressureapparatus for single-crystal X-ray diffractionstudies26 is shown in Fig. 2.17. This device alsouses the method of encasing liquids within asmall-diameter hole in the metal foil, a techniquefirst used by Van Valkenburg27. Beryllium foil isused because of its X-ray transparency. Thisapparatus is mounted on a standard goniometerhead which may be attached to a standard X-rayprecession camera or to a single-crystal orienter.The single crystal to be studied is immersed in atrue liquid. This device has been used in thestudy of two high-pressure phases of calciumcarbonate, CaCO3 (II) and CaCO3 (III).

8 BELT APPARATUS

Figure 2.19 shows an exploded view of thebelt high-temperature high-pressure apparatus28.I invented this device in January of 1953 whileemployed at the General Electric ResearchLaboratory in Schenectady, New York. At firstthe device was not taken too seriously by mycolleagues or the management. Consequently itwas not protected with company secrecy as onemight have expected. For example, onSeptember 16, 1953, a prominent high-pressureresearcher, Professor A. Michels of the Van der


Waals Laboratory, Amsterdam, visited the G.E.research laboratory as the guest of a G.E.official. Professor Michels' wife accompaniedhim on this visit. I was told to show them thedetails of the belt, and while I explained anddemonstrated, Mrs. Michels took notes, tookmeasurements, and made detailed sketches.Many people at G.E. who were not connectedwith the diamond synthesis project knew aboutthe belt, and one young G.E. researcherconveyed the details of the device to his formerprofessor, who was working in the field of high-pressure at a Midwest university.

The first written report within G.E. on thebelt was the usual patent disclosure letter to theDirector of Research. There were subsequentfollow-up letters on additional details andmiscellaneous memorandum reports that werecirculated to about a dozen persons within thecompany. The most formal company report that Iwrote was No. RL-1064 of March 1954 entitled"The Belt Ultra High Pressure, HighTemperature Apparatus," This document waslabeled "class 4," under which this statementappeared: "This document contains informationof importance to G.E. Its distribution is rigidlylimited." This report was circulated within thelaboratory and also to G.E. personnel elsewhere.It not only contained the general details of thebelt and its operation, but included scaledrawings and dimensions. This 1954 report,

stripped of its detailed drawings and dimensions,became the Review of Scientific Instrumentsarticle published in February of 196028.

It was not until I had made diamonds inDecember of 1954 that company secrecy on thebelt began to firm up. You can imagine myeagerness as a young scientist to make my mark,upon accomplishing the twin task of developinga high-pressure, high-temperature apparatuscapable of synthesizing diamonds and thenmaking them. You can further imagine myanxiety on being unable to have my workpublished for six years. This anxiety washeightened by the fact that informationconcerning the belt was at large in the world andthat others might publish on this or relateddevices before I would have the opportunity. Theknowledge that the belt existed and theknowledge that diamonds had been synthesizedgave others a tremendous advantage in trying toaccomplish the same things. There were someother leaks of vital information. After severalhundred copies of one of the earliest pressreleases was distributed in 1955, it was decidedthat the details and photographs of the belt in thisbrochure were too revealing. Distribution of thisparticular brochure was then stopped, and onethat was less explicit was substituted. Also,persons in government laboratories professing"need to know" obtained copies of the patentapplications.

Although I have not published the fulldetails of the human side of the development ofthe belt apparatus and the first synthesis ofdiamonds, I have given some personal glimpsesinto the events that transpired29.

The belt was the first apparatus capable ofoperating simultaneously at high pressure andhigh temperature at pressures of 100,000 atm.Maximum pressures of the order of 150,000 atmsimultaneously with temperatures in excess of2000°C have been maintained for long periods oftime in this device. Referring to Fig. 2.18, thefunctions of the various parts are as follows: twotapered tungsten carbide pistons (1) push intoeach side of a tapered tungsten carbide chamber(2). Pressure is transmitted to the samplecontained in a molybdenum tube (3). This tubenot only serves as an electrical resistance heatingelement but also contains the sample and is oftenreferred to as the heater sample tube. Graphiteand tantalum tubes are also used. Pressure istransmitted to the sample contained in the tube(3) by pyrophyllite (4). The pyrophyllite alsoserves as thermal and electrical insulation.Electricity passing through the tube (3) serves to

H. TRACY HALL 23

heat the sample. Current enters this tube througha molybdenum disk (6), which in turn touches asteel ring (5), which in turn touches the tip of thepiston. The short, pyrophyllite cylinders (7)provide thermal insulation at the ends of thesample tube. A sandwich gasket, composed ofpyrophyllite (8 and 10) and steel (9), maintainsthe pressure in the chamber.

The hardened steel compound binding rings(11) and (12) provide lateral thrust to thechamber. Binding rings (13) and (14) do thesame for the conical pistons. Rings (15) and (16)are made of dead soft low-carbon steel and areintended to absorb the energy released should theinner binding rings fail. These binding ring setsare similar to those already discussed for piston-

cylinder apparatus. The chamber (2) and bindingrings (11) and (12) form a "torroidal belt" aroundthe sample, and from this I took the code name"belt" for the apparatus. For sustained operationat high temperature, the conical pistons and thecorresponding chamber must be cooled. Figure2.19 shows the belt assembly in closed position.Figure 2.20 shows a photograph of the first beltapparatus.

The belt uses a compressible gasket in amanner akin to that used in Bridgman anvils, butthere are some differences. In Bridgman anvilsthe gasket is a flat ring in a plane 90° from theaxis of compression. The motion of the anvils isdetermined by the amount that the gasket can becompressed. For more compression, one woulduse a thicker gasket. However there is a limit tothe possible thickness of the gasket. If it is toothick, chunks will break from it allowing thesample to escape, and the gasket will notcompress symmetrically or evenly. Themaximum permissible thickness of the gasket isdetermined by trial and error. For opposed anvilswith 1 cm-diameter faces, the maximumthickness for a pyrophyllite gasket is about 0.25mm. It would be more for anvils with larger-diameter faces if the gasket width werecorrespondingly increased.

In the belt the gasket is arranged at an angleless than 90° to the direction of compression, asis shown in the single-ended device (halfbelt) ofFig. 2.21. In the belt design the thickness tbetween the two arrows A-A is substantially thesame as would be possible to use in a Bridgmananvil device with faces of this diameter.However in the belt, the relative movement ofthe tapered piston into the tapered cylinder isalong the axis given by B-B. This thickness s isgiven by s = t/cos θ. The stroke (maximummotion) in the belt is correspondingly increased.The force acting perpendicular to the contactarea between the gasket and the tungsten carbidecomponents (A-A) is reduced by cos θ over thecase where the gasket is perpendicular to thedirection of compression. Therefore the friction


between the gasket and the tungsten carbide willbe reduced by cos θ. The smaller the angle θ ismade, the larger s becomes. The limit on thesmallness of θ is set in part by the frictionalforces along the gasket-tungsten carbideinterface. When they become too small, thegasket will be blown out by the pressure.

A conical piston has much greater strengththan does the cylindrical piston of a piston-cylinder design. Consequently, if sufficientstroke can be obtained to make the apparatususeful, a device with conical pistons should beable to obtain a much higher pressure than isobtained in piston-and-cylinder devices, and thisis indeed the case. The sandwich gasket furtherincreases the stroke, hence the sample size andthe ultimate pressure of any system utilizingstone gaskets. For example, it will allow anincrease in sample thickness in Bridgman anvils.In the belt the thickness of the gasket can bemore than doubled by the insertion of the thinmetal cone, which is the "meat between thebread" in this array. The metal reduces gasketcrumbling, allowing smooth compression andextrusion. The available stroke of the beltapparatus of Fig. 2.18, which utilizes thecombined effects of double-ending, taperedgasket, and the sandwich gasket, isapproximately 6 mm, compared to a maximumof about 0.7 mm in a Bridgman anvil device ofcomparable face diameter. This greatly increasedstroke makes it possible to have a tremendouslylarge sample size (compared to that available inopposed anvils) and, of course, offers a sizepractical for the commercial production ofdiamond.

There are some other advantages of the belt.Very accurate alignment is required for piston-and-cylinder devices. Good alignment,comparatively speaking, is required in Bridgmananvils. The belt is the least sensitive tomisalignment, and slight axial misalignment orcocking of the moving members with respect toeach other does not interfere with its operation.For certain purposes a plain pyrophyllite gasket(nonsandwich gasket) is perfectly suitable foruse in the belt. In other instances two steel conescan be used in the sandwich gasket to increasethe stroke of the belt even more. The distancebetween the two conical piston tips willdetermine the pressure that can be reached. Toobtain the highest pressure, the conical pistontips are placed only about 1 mm apart. Suchdevices have been called "high compressionbelts." In addition to placing the piston tips closetogether, one would miniaturize the apparatus to

take advantage of the effect already mentionedthat higher pressure can be obtained in smallerdevices than in larger devices. There seem to betwo effects that make this so. There is thesurface-to-volume ratio, which has often beenobserved to effect changes in engineeringproblems. Also, in sintered tungsten carbideparts, a part can be made more free from defectsin small sizes than in large sizes. In the usualmanufacture of cemented tungsten carbide,cobalt powder and tungsten carbide powder areintimately mixed together and then cold pressedto the approximate shape desired. The preformedpart is then sintered at a temperature just belowthe melting point of cobalt. In large parts, cobalttends to concentrate in the bottom of the partowing to the forces of gravity. This is not solikely to occur in smaller parts. There is also thestatistical situation that a large part has a greaterprobability of containing a flaw than does asmall part. In brittle substances, such ascemented tungsten carbide, breakage usuallystarts at a flaw within or on the surface of thematerial.

There is a cooperative phenomenon at workin the belt to allow it to operate at higherpressures. Through the gasket, the tapered pistonis supported by the tapered conical chamber, andvice versa. An included angle between 60º and90º is suitable for the conical piston in the beltapparatus. However in some special belts(unpublished work of the author), the angle hasbeen decreased to as low as 20º. Figure 2.22shows the central region of the belt and thegasket after a high-pressure run.

Thermocouple wires and other electricalleads are inserted into the sample region of thecell by drilling fine holes through thepyrophyllite in a manner such that the wires donot make contact with the tungsten carbidecomponents. Frictional forces will hold the wires

H. TRACY HALL 25

in place as pressure is developed within thesystem. In the belt, the center of the gasketassembly sustains more load than the outerregions, and there is a gradual drop of pressurealong the gasket to atmospheric pressure at itsouter edge. This is a kind of multistaging effectand is also a factor in making it possible tosustain such very high pressures in a relativelylarge volume.

9 MULTIANVIL APPARATUS

Opposed-anvil, piston-and-cylinder, and beltapparatus may all be classified as uniaxialdevices. That is, the means for mechanicallyreducing the volume of a substance isaccomplished by moving certain apparatuscomponents toward each other along a line.There are some disadvantages to uniaxialdevices. When quasi-hydrostatic pressure-transmitting materials are used, pressuregradients will invariably be set up within the cellin a pattern depending on the movement of thecompressing members. The Poisson ratio of thematerial between the pistons or anvils will alsoinfluence these pressure gradients. So will thegeometry of the cell components.

In the belt apparatus, the heater sample tubeis originally a right circular cylinder; but whenthe high-pressure, high-temperature run is over,it will have been considerably foreshortened andwill have a barrel shape. If one were producing asintered part, for example, a sintered diamondcylinder, it will be distorted in the sinteringprocess. Furthermore the sintered product willtend to delaminate (break into several disksperpendicular to the axis of compression) afterpressure is released. Sometimes the strainsresulting from the uniaxial compression are sogreat the material almost explodes as the layersseparate from each other. This problem can beovercome to some extent by surrounding theheater sample tube with a material that is morehydrostatic than pyrophyllite, for example,sodium chloride. Indium metal will also servethis purpose well because of its very low internalfriction. However it is metallic, has a lowmelting point, and also has a relatively highthermal conductivity, and these characteristicsrestrict its use.

Uniaxial hydraulic presses are ratherstandard equipment in research laboratories andin industry. Consequently they have been used topower the advance of pistons, anvils, and the likein high-pressure devices. Researchers havetended to think in terms of uniaxial machines.

However the problem just ascribed to uniaxialdevices can be overcome in apparatus in whichcompressing members are thrust toward a centralregion along several axes. The first such devicewas the tetrahedral anvil press5. This was thefirst high-temperature, high-pressure devicecapable of synthesizing diamonds that wasrevealed to the world. It came into being as amatter of necessity, and the reader might beinterested in a brief account of thesecircumstances.

After leaving the General Electric Companyin the fall of 1955 to become Director ofResearch and Professor of Chemistry at BrighamYoung University (BYU), I was besieged withrequests from scientists everywhere to reveal thedetails of the belt. Circumstances of companysecrecy compounded by a government secrecyorder, however, prevented disclosure. In fact, Icould not even use the apparatus for research atBYU. It soon became apparent that this problemwould not be resolved for some time.Consequently I found myself being encouragedby many persons and institutions to try todevelop another apparatus that could be madeavailable to scientists everywhere. Financialsupport for such an undertaking came first fromthe Carnegie Institution of Washington. Shortly


thereafter, additional financial aid was receivedfrom various government agencies, notably theNational Science Foundation.

Within two years I had tested several ideasin which pressure was generated by moving"anvils" along several axes. Some of thesedevices utilized a uniaxial hydraulic press tocause tapered wedges sliding in a cone-shapedring to move toward each other and thuscompress a pyrophyllite cylinder radially inward(Fig. 2.23). Another device was called the "BlackHawk Special." This machine utilizes a numberof flat wedges placed between two flat plates.The plates are maintained at a fixed distancecorresponding to the thickness of the wedges.Alternate wedges advance or retract, as shown inFig. 2.24, to reduce the volume and thus increasethe pressure. The movement of the wedges isexaggerated in the figure to better illustrate theprinciple. I also built and tested a variety ofuniaxial devices. The most successful device,however, used hydraulic rams built into a specialpress frame to afford symmetrical action inthree-dimensional space. I considered tetrahedralpresses, hexahedral (cubic), and higher-ordermultianvil presses, but calculations and intuitionshowed that the tetrahedral press would becapable of generating the highest pressures. Theoriginal tetrahedral press is shown in Fig. 2.25,and a cluster of tetrahedral anvils is shown in

Fig. 2.26. During the Christmas holidays of 1957diamonds were made in this apparatus—a test ofthe device that seemed important at the time.Patents were obtained on it through BYU andResearch Corporation, and all claims werequickly granted on the first office action.

The device was described at the Spring 1958Meeting of the American Chemical Society inSan Francisco and was published in the April1958 Review of Scientific Instruments, and manyscientists from all over the world visited Provo tosee it. Regardless, the U.S. Government(contrary to previous agreement) placed asecrecy order on the tetrahedral press on January15, 1959. Note that thousands of copies of theReview of Scientific Instruments, which fullydescribed it, had been distributed all over theworld nine months before this. Fortunately thissecrecy was short-lived, having been rescindedwithin six months.

A tetrahedral press might be considered athree-dimensional extension of the two-dimensional Bridgman anvil concept. Theprinciple of massive support is still at work inthe tetrahedral press and other multianvil presses,but to a lesser extent than in Bridgman anvils. Ina regular series of multianvil presses, the solidangle subtended by each anvil decreases as thenumber of anvils increases. This solid angle islargest in the tetrahedral press and is the primaryreason why higher pressures can be obtained inthis press than are achieved in cubic and higher-order presses. Some articles in the literatureclaim that the greater the number of anvils in amultianvil device, the higher the pressure that

H. TRACY HALL 27

can be achieved. This runs contrary to myexperience.

In multianvil devices the anvils support eachother through load on the gaskets, and this tendsto compensate for the reduction in massivesupport over that available in opposed anvils.The four anvils of the tetrahedral press (Fig.2.26) are driven toward a central point byhydraulic rams whose axes lie along lines normalto the triangular anvil faces. The hydraulic ramsand collinear anvil axes (A, B, C, D) intersect atthe tetrahedral angle of 109.47º at the center ofthe regular tetrahedral volume enclosed by theanvil faces.

Tungsten carbide is usually used as thematerial of construction for the anvils. Eachanvil is surrounded by a single press-fit steelbinding ring which, as usual, absorbs the tensileloads developed within the tungsten carbide. Thesloping shoulders of the binding ring break awayat a 2º steeper angle on each side than do thecorresponding 45º shoulders of the anvils. Thisprovides a widening gap moving outward fromthe center of the press and decreases thelikelihood of binding rings touching each otherin case of slight misalignment. Such touchingwould constitute an electrical short circuit. Thetapering gap also lessens the likelihood ofpinching off thermocouple wires or otherelectrical leads that may be inserted into thesample. The cell in which the pressure is

generated consists of a regular pyrophyllitetetrahedron, as is shown in Fig. 2.27.

If the anvils (Fig. 2.26) are advancedsymmetrically toward the center of the pressuntil their sloping shoulders F touch, theirtriangular faces E will define a regulartetrahedron of a given size. The tetrahedral cellG is made larger than this size so that a gasketwill be automatically formed by extrusion ofexcess pyrophyllite as the anvil faces impinge onthis larger cell. In practice, the edges of thetetrahedral cell are made about 25% longer thanthe corresponding legs on the triangular anvilfaces. The pyrophyllite within the tetrahedrontransmits pressure to the sample, providesthermal and electrical insulation, and providesthe necessary compressing gasket. This cellcontains considerably fewer and simpler partsthan the corresponding cell and preformedgaskets of the belt. The cell is machined frompyrophyllite by standard machine tools such aslathes and milling machines. Either high-speedsteel or tungsten carbide cutting tools may beused. Rough shapes of pyrophyllite are sawedwith a bandsaw from commercial blocks whichare available up to a 30-cm cube30. The materialis mined in South Africa.

Referring again to Fig. 2.27, the diagonallydisposed heater sample tube is usually made of ahigh-melting metal, such as molybdenum. Thetab is also of molybdenum, and the pyrophylliteprisms provide thermal insulation at the ends.Thermocouple leads may be brought outthrough edges of the tetrahedron. Fourelectrically insulated anvils provide fourelectrical connections to the cell's interior. Ifonly half a tab is used for heating current, twoanvils are free for other connections. There areother ways that the heater sample tube and othercomponents may be arranged within thetetrahedron. For example, a heater sample tubecould make contact through one triangular faceof the tetrahedron with a second tab makingcontact through an adjacent face, as shown inFig. 2.28. Several electrical leads for variousmeasuring purposes may then be brought outthrough an apex as is also shown in the figure.This apex is usually fired in a furnace at 750ºCfor 12 hr to increase its hardness and strengthover that of the natural stone. This helps toprevent pinch-off of the wires passing throughthe prism. If such fired pyrophyllite is used forthe entire tetrahedron, a suitable gasket will notform. There will be crunching sounds andexpulsion of large chunks of material as theanvils come together. A smooth, noiseless


extrusion of material is needed for proper gasketformation.

Pressure calibration of the tetrahedral pressis made in a manner similar to that described forthe piston-cylinder apparatus. A typical edgelength for the anvil triangular face for use inresearch is 2 cm. The corresponding length of thetriangular edge of the tetrahedral cell would be2.5 cm. The highest pressure observed in atetrahedral press at Brigham Young University(with 6 mm on edge anvil faces) is 120,000 atm,based on X-ray diffraction measurements onNaCl contained therein (Decker's NaCl scale).The everyday working pressure used at BYUwith 2 cm on edge anvil faces in tetrahedralpresses is 70,000 atm. At this pressure and at

room temperature, tungsten carbide anvils of 8%cobalt content will last indefinitely. Hightemperature simultaneously used with highpressure, however, causes some breakage.Simultaneous 1500ºC at 70,000 atm conditionswith the same anvils gives an anvil lifetime ofabout 2000 runs. This varies with the size of thesample within the tetrahedron and the currentflowing through the anvils. To increase lifetime,the sample and the heating current should besmall.

The design of the press in multianvilapparatus is important. In the original tetrahedralpress, the bases, tie bars, and otherappurtenances were not machined to anyparticular precision. Consequently turnbucklearrangements as shown in Fig. 2.25 were used toalign the tetrahedral axes of the rams. Each ofthe four hydraulic rams were independentlyvalved. To operate the device, the lower threerams were advanced until the anvils almosttouched to form a "nest." Then the pyrophyllite

sample cell was placed apex down in the nest,Next, the upper ram was advanced until thetriangular face of its anvil touched the upward-facing triangular base of the pyrophyllitetetrahedron. Each anvil was then individuallyadvanced approximately 0. 1 mm at a time byadmitting hydraulic oil under pressuresequentially to the first, second, third, and fourthrams and repeating the process until the desiredpressure was built up within the sample. Thisprocedure required skill, and it was difficult andtime consuming to teach others to use the press.Consequently I sought means for synchronizingthe motion of the anvils so that at any instant theanvil faces would be equidistant from the centerof the tetrahedral frame as defined by the tie barswhich hold the press together.

Consideration was given to mechanisms thatwould control the flow of oil to the rams or thatwould operate from a position-indicatingtransducer. Such devices were complicated anddid not satisfactorily solve the problem. The finalsolution was simple but took longer to conceivethan anticipated. The solution is a mechanicaldevice that I have called an anvil guide31. Figure2.29 shows a cross section of an anvil guidedevice as applied to a cubic press. It consists of12 guide rods or guide pins and six guide plates.A guide plate is fastened to the moving end ofeach hydraulic ram. The anvil, with its varioussupporting structures, is centrally mounted onthe guide plate. Each guide plate contains fourguide holes symmetrically arranged at 90º anglesto each other. The axis of each hole makes anangle of 45º with respect to the ram axis. Whenassembled, the guide rods are positioned within

H. TRACY HALL 29

the guide holes in which they are free to slide asthe rams advance or retract. All rams, beinginterconnected by the guide rods and plates, areforced to move synchronously together ashydraulic oil is simultaneously applied to all therams. In the tetrahedral press, there are six guiderods and four guide plates. The three holes ineach guide plate are disposed at 120º anglesabout the ram axis, and each makes an angle of35.26º to the ram axis.

The most advanced cubic (or tetrahedral)press built to date32 incorporates a hollow guide-pin system for retraction of the hydraulic rams,which saves weight and machining costs andimproves mechanical proportions (Fig. 2.30).This system unclutters the anvil region (fromwater-cooling lines) and provides better access.Oil circulating through the guide pin system,when it is not being used to retract the press,provides cooling. The use of inverted hydraulicrams and right/left-hand thread tie bars and basesreduces weight and cost further. Retraction isaccomplished by admitting pressurized hydraulicoil to the anvil guide assembly. Le Chatelier'sprinciple requires the interconnected system toincrease in volume, and this is accomplished ifthe cylinders move (retract) toward the bases.Anvil guide mechanisms can also be applied toany other variety of polyhedral press.

The guide pins and holes need not besymmetrically disposed as indicated in Fig. 2.29but need only to be parallel to the tie bar axes.They may thus be offset to facilitate theadmission of beams of matter or radiation suchas neutrons and X-rays into the sample.Alternatively one or more pins may beeliminated without adversely affecting theoperation of the anvil guide. Two smaller-diameter guide pins that straddle the normalsymmetrical position of a single pin have alsobeen used to give access.

The cubic press is the next member in theseries of polyhedral presses. Anvil breakage issomewhat higher than that in tetrahedral presses,and the maximum pressure achieved is not as

high. The everyday working pressure for cubicpresses is about 5000 atm less than that oftetrahedral presses. Excellent anvil lifetime isobtained in high-pressure, high-temperature runsat 65,000 atm. There are some secrets toobtaining long anvil lifetime in cubic presses. Inmy first cubic press there were occasions onwhich all six anvils would break for no apparentreason. This was very puzzling until I finallyrealized that the anvils were breaking every timethat I first used the press after returning from ameeting or speaking engagement away from theuniversity. This caused me to wonder if therecould be such a thing as too long a rest period--aperiod of nonuse of the anvils. So deliberateexperiments were performed to test this idea.Two things were learned. First, the anvils shouldnot remain idle for more than 60 hr if they arebeing routinely used at 65,000 atm. Second, ifthe anvils are unused for periods longer than 60hr, they should be "stress relieved" at a pressureof about 25,000 atm (at room temperature) for aperiod of 12 hr before resuming routine use at65,000 atm. This breakage phenomenon does notexist for the tetrahedral press, and no suchprecautions are required.

An advantage of the cubic press is itsrectangular coordinate geometry. A typical cubiccell is shown in Fig. 2.31, and a typical cubicpress is shown in Fig. 2.32. It is possible with thecubic press to have six independent contactsemanate from within the high-pressure cell andconnect to the outside through the six electricallyisolated anvil assemblies.

Multianvil presses need not be regular as arethe tetrahedral and cubic presses so fardescribed. For example, it is possible to have atetragonal press. Opposite ends of the cell havesquare faces, and the four sides are rectangular.The anvils for such a press would enclose atetragon, and the cell assembly would beoversize in order to allow for extrusion of a


gasket as is the case with the presses alreadydescribed.

As can be seen, in multianvil press systemsthe entire system (hydraulic rams, anvils, guidepins, etc.) is engineered as a unit. This contrastswith most situations for belts, Bridgman anvils,and piston-cylinder devices, where the hydraulicpress is a separate, standard item of manufacturethat is modified to accept the aforementioneddevices. Even with uniaxial systems, however,the best apparatus is a totally engineered,integrated unit, press and all.

10 INSTRUMENTATION

The basic high-pressure apparatus that havebeen described have been instrumented innumerous ways to study a great variety ofphenomena. It is not the purpose of this chapterto delve into the details of such instrumentation.Considerable ingenuity has been used by manyresearchers, and a few references are listedbelow to sample briefly what has been done.

Differential thermal analysis (DTA)33

Mössbauer effect34

Diffusion13

Manganin gauge pressure measurement35

Neutron diffraction (time of flight)36

Optical measurements37

X-ray diffraction38

Cryogenic techniques39

Nuclear magnetic resonance40

Electron spin resonance41

Thermal conductivity42

Publications of the author dealing with high-pressure apparatus that have not been referencedare cited in Refs. 43 through 53.

11 THE SYNTHESIS OF DIAMOND: ANEXAMPLE OF THE USE OF HIGH-PRESSURE, HIGH-TEMPERATUREAPPARATUS

A cubic press with six 20.3 cm-diameter(320 cm2-area) hydraulic rams and six 1.27 cmon edge square-faced anvils (1.61 cm2 area) isused. The hydraulic rams are suspended in anoctahedral frame consisting of ram bases and tiebars. This frame is mounted with one triangularface of the octahedron parallel to the floor. Iprefer this mounting because debris from the cell(crumbled gasket, etc.) falls between the threelower anvils as the press opens and does not, forthe most part, get into the sliding mechanisms ofthe press. Cubic presses are sometimes mountedwith a line through opposite apices of theoctahedron perpendicular to the floor. Thislocates one anvil facing upward with its faceparallel to the floor, which is convenient forplacing the cell. Gravity holds the cell in place asit rides upward while the anvils advance towardthe center of the press. However pyrophyllitedebris will fall onto the sliding surfaces of fourguide pins unless protective shields are used.

The cubic pyrophyllite sample cell is 1.59cm on edge. It has a central bore of 6.35 mm anda counterbore on each end to accommodate thecurrent ring and its central pyrophyllite disk. Thecurrent rings (two each) are made of AISI 1020cold-finished steel and are 11.07 mm O.D. by8.46 mm I.D. by 4.22 mm thick. A pyrophyllitedisk 8.46 mm in diameter and 4.22 mm thick isplaced inside each current ring and providesthermal insulation at each end of the heatersample tube. Two molybdenum disks 10.90 mmin diameter and 0.25 mm thick are used to carryheating current from the current rings to thesample. The sample consists of alternating disksof graphite and nickel with graphite diskstouching the molybdenum disks at each end. Thenickel disks are 6.22 mm in diameter and 0.36mm thick, and the graphite disks are 6.22 mm indiameter and 0.76 mm thick. This stack ofalternating nickel and graphite disks will serve asits own heating element.

H. TRACY HALL 31

A paper tab about 5 mm wide and 10 cmlong is glued to an edge of the assembled cellwith a minimum amount of white glue. Theexternal pyrophyllite portions of the cell arepainted with a water suspension of red iron oxide(Fe2O3) powder. The cell is then dried at 120ºCfor 1/2 hr. With the anvils in retracted position,the cell is roughly centered on one of the loweranvils and held in place by placing a piece ofadhesive tape across the paper tab onto the faceof an anvil binding ring. The current rings shouldbe placed so that they touch (upon anvil closure)the opposed pair of anvils that carry the heatingcurrent. Oil pressure is applied to the hydraulicrams until the anvils impinge on the cell and thepressure (room temperature calibration) is builtup to about 60,000 atm. This will occur at an oilpressure of about 476 atm.

After the cell is pressurized, heating currentis applied. The characteristics of the electricalsystem will come into play at this point, and alittle experimentation may be required in order todetermine the correct setting of the electricalcontrols. Assume that variable autotransformersare used to control the voltage impressed acrossthe sample. They should be set so that the initialcurrent passing through the sample is about 800A. If the setting is correct for making diamondsfrom the graphite, the current will start todecrease within 20 sec of its being applied, andas the current falls, the voltage rises. This is thesignal that diamonds are forming! The contentsof the cell are at a temperature near 1500ºC, andat this temperature graphite dissolves in themolten nickel which acts as a solvent-catalyst forchanging graphite to diamond. Diamond isprecipitated from the liquid nickel-carbonsystem, and since diamond is an electricalinsulator (it probably conducts some at 1500ºC)it impedes the flow of electricity as it is formed.Thus the current falls and the voltage rises.

I noticed this phenomenon in the firstsynthesis of diamond; in fact, I have a patent onit (U.S. pat. 2,947,608 assigned to GeneralElectric). One can easily find the heating powerrequired to make diamond by this means withoutactually knowing the temperature inside the cell.

The current may fall to about 200 A as thediamond forms. After 60 sec of heating, thepower is disconnected. The sample will cool tonear room temperature in about 5 sec. About 20sec after the heating power is disconnected, thepressure may be released and the anvilsretracted. This brief delay is intended to allowadjustments to take place in the cell and anvilsfollowing the drastic change in temperature of

the sample. This helps to preserve the life of theanvils.

The cell is removed from the press and splitopen with a mallet blow to a sharp knife set tobisect a square face of the pyrophyllite cubeparallel to the sample axis. One or two blowswill usually expose the core, which now consistsof tiny diamond crystals intermingled with nickeland graphite. The entire mass tends to be amonolithic unit but can usually be cleaved with aknife and mallet along the original nickel-graphite boundaries. The diamond crystals at thispoint will be covered with a film of nickel metal.

The nickel and graphite may be separatedfrom the diamond by first treating the samplevery carefully with hot concentrated H2SO4 +NaNO3. The sample is placed in a beaker towhich is added the concentrated H2SO4 and thenabout 10% by weight NaNO3. The mixture isheated strongly on a hot plate. Clouds of graphiteparticles will be seen to rise from the sample.Heating is continued for perhaps an hour untilmost of the graphite is oxidized and disappears.After the liquid is decanted, the residue is treatedwith concentrated HCl but is heated only lightlyon the hot plate or let stand overnight at roomtemperature. The diamonds will be reasonablyclean at this stage, and their morphologicalfeatures can be readily observed at 10xmagnification. For further cleaning, the H2SO4 +NaNO3 and the HCl steps are repeated as manytimes as necessary to remove all the graphite andnickel. There will still be some residue from thepyrophyllite. This can only be removed by along-term treatment with concentrated HF(perhaps a week) or by a fusion with NH4F. Theyield of diamond should be about 1 carat (0.20g).

The diamond obtained is called diamond gritin the trade and could be used in some grades ofresinoid-bonded diamond grinding wheels. Thediamonds may be black. Transparent, slightlygreenish or yellowish diamonds can be formedby lowering the temperature and pressure to apoint where diamond is still thethermodynamically stable phase (with respect tographite) but where the diamonds are grownmore slowly. Under these conditions the yieldwill decrease.

For information concerning thethermodynamics and Kinetics of diamondsynthesis, see the Journal of Chemical Educationarticle of Ref. (29).

References


1 H. T. Hall, Chemist, 47, 276 (1970).2 H. T. Hall, "Chemistry at High Temperatureand Pressure," in High Temperature-A Tool forthe Future, Stanford Research Institute, MenloPark, California, 1956, pp. 161 and 214.3 J. C. Jamieson and P. S. DeCarli, Science, 133,1821 (1961).4 G. R. Cowan, B. W. Dunnington, A. H.Holtzman, U.S. Pat. 3,401,019 (1968); J. M.Kruse, U.S. Pat. 3,348,918 (1967).5 H. T. Hall, Rev. Sci. Instrum., 29, 267 (1968).A good treatise on piston-cylinder use with fluidpressure media is S. E. Babb, Jr., "Techniques ofHigh Pressure Experimentation" in Technique ofInorganic Chemistry, Vol. 6, A. Weissberger,Ed., Interscience, New York, 1966, p. 83.6 H. T. Hall, Science, 169, 868 (1970).Commercial materials are available fromMegadiamond Industries, 589 Fifth Avenue,New York City, and from General ElectricSpecialty Materials Division, Worthington,Ohio.7 P. W. Bridgman, Proc. Am. Acad. Arts Sci., 74,425 (1942).8 C. A. Parsons, Proc. Roy. Soc. (London), 44,320 (1880); Trans. Roy. Soc. (London), A220, 67(1920). Also see anon. report on RichardThrelfall's discourse at the Royal Institution,Engineering, 87, 425 (1909).9 L. L. Coes, Jr., Science, 118, 131 (1953).10 S. M. Stishov and S. V. Popova, Geokhimiya,10, 837 (1961).11 L. L. Coes, Jr., "Synthesis of Minerals at HighPressures," in Modern Very High PressureTechniques, R. J. Wentorf, Jr., Ed., Butterworth,London, 1962, p. 137.12 P. W. Bridgman, The Physics of HighPressure, G. Bell, London, 1958, p. 32.13 H. R. Curtin, D. L. Decker, and H. B.Vanfleet, Phys. Rev., 139, A1552 (1965).14 G. J. Piermarini, S. Block, and J. D. Barnett, J.Appl. Phys., 44, 5377 (1973).15 P. W. Bridgman, The Physics of HighPressure, G. Bell, London, 1958, p. 408.16 H. T. Hall in E. C. Lloyd, Eds., AccurateCharacterization of the High PressureEnvironment, National Bureau of Standards(U.S.) Special Publ. 326, 1971, pp. 303 and 313.17 J. D. Barnett, R. B. Bennion, and H. T. Hall,Science, 141, 534 (1963); J. D. Barnett, V. E.Bean, and H. T. Hall, J. Appl. Phys., 37, 875(1966).18 D. L. Decker, J. Appl. Phys., 42, 3239 (1971).

19 P. W. Bridgman, Proc. Roy Soc. (London),A203, p. 8 (1950).20 E. W. Comings, High Pressure Technology,McGraw-Hill, New York, 1956, Chapt. 6.21 P. W. Bridgman, Proc. Am. Acad. Arts Sci.,81, 167 (1952).22 C. E. Weir, E. R. Lippincott, A. VanValkenburg, and E. N. Bunting, J. Res. Natl.Bur. Stand. (U.S.), A63, 55 (1959); G. J.Piermarini and C. E. Weir, J. Res. Natl. Bur.Stand. (U.S.), A66, 325 (1962).23 L. Ming and W. A. Bassett, Rev. Sci. Instrum.,45, 1115 (1974).24 J. D. Barnett, S. Block, and G. J. Piermarini,Rev. Sci. Instrum., 44, 1 (1973).25 W. A. Bassett, T. Takahashi, and P. W. Stook,Rev. Sci. Instrum., 38, 37 (1967).26 L. Merrill and W. A. Bassett, Rev. Sci.Instrum., 45, 290 (1974).27 A. Van Valkenburg, "Visual Observations ofSingle Crystal Transitions under TrueHydrostatic Pressures up to 40 Kilobar," inConference Internationale sur les HautesPressions (le), Creusot, Saone-et-Loire, France,1965.28 H. T. Hall, Rev. Sci. Instrum., 31, 125 (1960).29 H. T. Hall, "Diamonds," Proc. 3rd Conf.Carbon, University of Buffalo, Buffalo, NewYork, June 1957, Pergamon, London, 1957, pp.75-84; J. Chem. Educ., 38, 484 (1961); Chemist,47, 276 (1970).30 There is only one commercial source of supplyin the United States: The Tennessee LavaDivision of the 3M Co., Chattanooga, Tennessee.31 H. T. Hall, Rev. Sci. Instrum., 33, 1278 (1962).32 H. T. Hall, Rev. Sci. Instrum., 46, 436 (1975).33 G. C. Kennedy and R. Newton, in SolidsUnder Pressure, D. Warschauer and W. Paul,Eds., McGraw-Hill, New York, 1963. p. 163.34 W. H. Southwell, D. L. Decker, and H. B.Vanfleet, Phys. Rev., 171, 354 (1968).35 R. J. Zeto and H. B. Vanfleet, J. Appl. Phys.,40, 22 (1969).36 R. B. Bennion, H. G. Miller, W. R. Myers, andH. T. Hall, Acta Cryst., 25A, S71 (1969).37 R. A. Fitch, T. E. Slykhouse, and H. G.Drickamer, J. Opt. Soc. Am., 47, 1015 (1957).38 J. D. Barnett and H. T. Hall, Rev. Sci. Instrum.,35, 175 (1964).39 P. F. Chester and G. O. Jones, Phil. Mag., 44,1281 (1953).40 G. Benedek, Sci. Am. 212, 102 (1965).

H. TRACY HALL 33

41 J. H. Gardner, M. W. Johansen, C. Larsen, W.Murri, and M. Nelson, Rev. Sci. Instrum., 34,1043 (1963).42 H. H. Schloessin and V. Z. Dvorak, Geophys.J. R. Astr, Soc., 27, 499 (1972).43 H. T. Hall, J. Phys. Chem., 59, 1114 (1955).44 H. T. Hall, B. Brown, B. Nelson, and L. A.Compton, J. Phys. Chem., 62, 346 (1958).45 H. T. Hall, Science, 128, 445 (1958).46 H. T. Hall, Sci. Am., 201, 61 (1959).47 H. T. Hall, "High Pressure Methods," in Proc.Int. Symposium on High TemperatureTechnology, Stanford Research Institute,McGraw-Hill, New York, 1960, pp. 145, 335.48 H. T. Hall, "High Pressure Apparatus," inProgress in Very High Pressure Research,Bundy, Hibbard, and Strong, Eds., Wiley, NewYork, 1961, p. 1.49 H. T. Hall, "High Pressure/TemperatureApparatus," in Metallurgy at High Pressures andTemperatures, Gschneidner, Hepworth, andParlee, Eds., Gordon and Breach, New York,1964, p. 133.50 J. D. Barnett and H. T. Hall, Rev. Sci. Instrum.,35, 175 (1964).51 H. T. Hall, Rev. Sci. Instrum., 37, 568 (1966).52 H. T. Hall, Rev. Phys. Chem. (Japan), 37, 63(1967).53 H. T. Hall, Rev. Sci. Instrum., 46, 436 (1975).

High Pressure Techniques

Documents

Transcript of High Pressure Techniques