Beryllium Diffusion Of

Over the past two years, the heat treatment of corundum involving lattice diffusion of beryllium (Be)at temperatures over 1800°C has become a major issue in the gem trade. Although initially onlyorange to orangy pink (“padparadscha”-like) sapphires were seen, it is now known that a full range ofcorundum colors, including yellow and blue as well as ruby, have been produced or altered by thistreatment. An extension of the current understanding of the causes of color in corundum is presentedto help explain the color modifications induced by Be diffusion. Empirical support is provided by Be-diffusion experiments conducted on corundum from various geographic sources. Examination ofhundreds of rough and faceted Be-diffused sapphires revealed that standard gemological testing willidentify many of these treated corundums, although in some instances costly chemical analysis bymass spectrometry is required. Potential new methods are being investigated to provide additionalidentification aids, as major laboratories develop special nomenclature for describing this treatment.

arly in 2002, it became apparent that corundumtreated by a new technique in Thailand hadbeen filtering into the marketplace unan-

nounced, particularly in Japan. It was subsequentlylearned that these stones had been traded for at leastsix months prior to this discovery, perhaps longer.The first announcement of this situation—an alertissued by the American Gem Trade Association(AGTA) on January 8, 2002—prompted substantialactivity in gemological laboratories worldwide.Quite rapidly it was demonstrated that this new pro-cess involved diffusion of the light element berylli-um (Be) into a wide variety of corundum types toalter their color.

The diffusion of beryllium into corundum cre-ates yellow, orange, or brown color components.The effectiveness of this process in turning pale-col-ored or nearly colorless corundum into vibrant yel-lows and oranges is dramatic. No less dramatic isthe alteration of pink sapphire to a “padparadscha”appearance or a vivid orange, as well as the conver-sion of bluish rubies to a fine red color. It also canreduce the amount of blue in dark blue sapphires,rendering them a more attractive color (figure 1).

Our initial interpretation—that this color alter-ation was caused solely by the diffusion of berylli-um into the stone in an oxidizing atmosphere—wasdenied by those involved with the treatment pro-cess, and was questioned by other gemologists.Their arguments hinged primarily on observationsthat apparently similar starting materials couldemerge from this process as a variety of colors, orcompletely unchanged. We believe that those obser-vations are correct, but their interpretation is not.To understand the unusual behavior of beryllium inthis material, we will have to examine far moreclosely the origin of color in corundum.

In addition to their changes in color, these stonesexhibit many other features—both internal and onthe surface—that indicate very high-temperature heattreatment and/or long periods of treatment. Takentogether, these features indicate that a new treatmentregime has been introduced into the jewelry trade.

84 BERYLLIUM DIFFUSION OF RUBY AND SAPPHIRE GEMS & GEMOLOGY SUMMER 2003

E

BERYLLIUM DIFFUSION OFRUBY AND SAPPHIRE

John L. Emmett, Kenneth Scarratt, Shane F. McClure, Thomas Moses, Troy R. Douthit, Richard Hughes, Steven Novak, James E. Shigley, Wuyi Wang, Owen Bordelon, and Robert E. Kane

See end of article for About the Authors and Acknowledgments.GEMS & GEMOLOGY, Vol. 39, No. 2, pp. 84–135.© 2003 Gemological Institute of America

We start this discourse with a short historicaldiscussion of corundum heat treatment and theconnection of this new treatment to previous pro-cesses. That is followed by a summary of recentprogress on beryllium diffusion. Next we delvedeeply into the causes of color in corundum,extending the current understanding, to elucidatehow a minute amount of this light element cancause such a variety of dramatic color alterations.To this end, we performed a series of Be diffusionexperiments in one of our laboratories. We alsostudied a large number of Be diffusion–treated sap-phires (some, both before and after treatment),untreated sapphires, and sapphires treated only byheat, using a variety of gemological and analyticalmethods. On the basis of the examinations andtesting conducted, we present criteria for determin-ing if stones have been Be diffused—some are quitesimple, and some require advanced analyticalinstrumentation.

Finally, we note that this new treatment processhas caused gemological laboratories to re-evaluatetheir thinking about corundum treatments in gener-al, and the manner in which the treated stonesshould be described on laboratory reports and otherdocuments in particular. Current descriptive lan-guage used on reports issued by the AGTAGemological Testing Center and the GIA GemLaboratory is presented.

Before we proceed, let us take a moment toexplain the nomenclature that we will use in this

article. The original diffusion process in whichtitanium was diffused from the outside of a pieceof corundum into the bulk of the stone, producinga blue layer under the surface, was called “surfacediffusion” by some gemologists (see, e.g., Hughes,1997, pp. 121–124). However, the term surface dif-fusion is used in many other disciplines to mean aprocess by which a material moves over a surface,rather than through the surface into the interior.As recommended by the International Union ofPure and Applied Chemistry (Kizilyalli et al.,1999), the scientifically correct term for the pro-cess by which a foreign material moves into andthrough a solid is lattice diffusion, previouslyreferred to in the scientific literature as bulk diffu-sion. For the purposes of this article, we will usethe term lattice diffusion or (in its shortened form)diffusion to describe this process.

BACKGROUND Corundum has been heat treated with moderatecolor improvement since antiquity. However,today’s modern heat-treatment techniques pro-duce dramatic results when compared with thesubtle changes of the past. The historic turningpoint was the discovery, apparently in the 1960s(see Crowningshield, 1966, 1970; Beesley, 1982),that the translucent milky white to yellow tobrown and bluish white sapphire from Sri Lanka,known as geuda, could be transformed to a fine

BERYLLIUM DIFFUSION OF RUBY AND SAPPHIRE GEMS & GEMOLOGY SUMMER 2003 85

Figure 1. The berylli-um diffusion processcan affect many colorsof corundum, includ-ing ruby and blue sap-phire. The Be-diffusedstones shown hererange from 0.40 to 5.05ct. Photo by Harold &Erica Van Pelt.

transparent blue by atmosphere-controlled high-temperature heating. This discovery was madepossible by the availability of simple furnacescapable of reaching temperatures in the ≥ 1500°Crange. The striking color change in the geudamaterial was caused by the dissolution of rutileinclusions in the stone, and by the inward diffu-sion of hydrogen from the reducing atmosphere.The importance of hydrogen diffusion was not rec-ognized until much later (Emmett and Douthit,1993). Eventually, tons of previously worthlessgeuda corundum were converted to marketabletransparent blue sapphire.

The benefits of this process with geuda sapphireled to successful experimentation with many typesand colors of sapphire (e.g., Crowningshield andNassau, 1981; Keller, 1982; Themelis, 1992; Emmettand Douthit, 1993) as well as off-color ruby. As aresult, the vast majority of rubies and sapphires trad-ed today have been heat treated.

Initially these gems entered the market withoutany form of disclosure. By the time buyers did learnthey were treated, the stocks of gem merchantswere full of stones with color created by a high-heatprocess. In the mid-1980s, international regulatorybodies such as CIBJO decided that, because the heattreatment of sapphire was a “traditional trade prac-tice,” such a treatment need not be declared in thecourse of trade. The concept that turning unattrac-tive corundum into a gem by heating it in an atmo-sphere-controlled furnace should be regarded as“traditional” was questionable from the outset, butthe situation had reached a stage where somethinghad to be done to allow trading to continue. Thus,the heat-treated geuda sapphires were placed in thesame category as the far milder historic corundumheat treatments.

Diffusion-treated corundum first appeared onthe world market in the late 1970s (Crowning-shield, 1979). This process marked a radical depar-ture from all earlier corundum heat treatment inthat it produced a thin outer layer of saturated bluecoloration in otherwise colorless or pale-coloredsapphire by diffusing titanium into the stone fromthe outside. As such, it represented the first success-ful attempt to add color to sapphire from an exter-nal source. The technology was developed by UnionCarbide Corp. and then sold to Golay Buchel.Eventually, treaters in Thailand also used this pro-cess (Kane et al., 1990; Hughes, 1991a,b). Nearly allsapphires produced by this method were blue, but afew orange stones also were seen (Scarratt, 1983). In

1993, GIA researchers reported on the experimentalproduction of red “diffusion-treated” corundum(McClure et al., 1993).

In October 2001, Australian gemologist TerryColdham informed one of the authors that a treaterin Chanthaburi, Thailand, had developed a newmethod to transform bluish red Songea (Tanzania)stones to a fine orange to red-orange (Coldham,2002; Hughes, 2002). Shortly thereafter, severalother sources confirmed this information. Thestones were to be marketed under a variety of newcolor names, such as “Sunset Sapphire” (“TreatedSongea sapphires . . . ,” 2002).

During a visit to Bangkok in November/December 2001, another author saw the new treat-ed orange sapphires in the gem market, as well as alarge volume of orangy pink treated stones similarin color to traditional “padparadscha” sapphires.He later obtained samples for study. When staffmembers at the AGTA Gemological TestingCenter (AGTA-GTC) examined these samples inNew York, they found that all showed evidence ofexposure to a high-temperature treatment (Scarratt,2002a). When these stones were immersed inmethylene iodide, many displayed unusual yellow-to-orange rims surrounding pink cores, which sug-gested that a yellow colorant was being diffusedinto pink sapphire.

On December 28, 2001, Ken Scarratt reportedhis observations to Richard Hughes, who thenexamined faceted sapphires just purchased inBangkok by Pala International. He found yellow-to-orange rims on most pieces (see, e.g., figure 2). Inearly January 2002, AGTA and Pala Internationalissued Internet warnings to their extensive mailinglists (Scarratt, 2002b).

Initially, the cause of these yellow-to-orangerims surrounding pink cores (with the overallcolor of the gems being orange to pinkish orange)was unknown. One of the authors (JLE) suggestedin January that the features resulted from latticediffusion of light elements, producing what arecalled “trapped-hole color centers” in the crystallattice of the corundum. While the color producedand the elements added were different, the processwas essentially identical to that used more than adecade earlier to produce a blue rim on sapphireby the diffusion of titanium. However, gemolo-gists at Bangkok’s Gem Research Swisslab (GRS)and the Gem and Jewelry Institute of Thailand(GIT) suggested that the color enhancement inthis orange to orangy pink corundum was due to a


change in the oxidation state of iron broughtabout by simple heat treatment (GIT, 2002;Proust, 2002; Weldon, 2002).

At the same time the orange color rims wereobserved, other unusual features were seen in thesestones. Significant overgrowths of synthetic corun-dum were noted on rough stones (Scarratt, 2002b).In addition, remnants of these synthetic over-growths were seen in stones re-cut and polished,raising the issue of what portion of the facetedstone was natural (McClure, 2002a). These fea-tures, plus unusual inclusion damage, suggestedthat an entirely new form of heat treatment hadbeen initiated in Thailand.

Toward the end of January 2002, the GIA GemLaboratory engaged one of the authors (SN) to ana-lyze the treated stones using secondary ion massspectrometry (SIMS, see Box A). During the AGTA-sponsored Gemstone Industry and LaboratoryConference held in Tucson in early February 2002,another of the authors presented the SIMS data,which showed that not only did these stones con-tain beryllium, but there also was a direct correla-tion between the depth of color penetration and thedepth of Be penetration (McClure et al., 2002). Atthe time, no standards for chemically analyzing theamount of beryllium in sapphire were available, andthus the absolute magnitude of the Be concentra-tion was incorrect in these data (Wang and Green,2002a); however, the relationship between berylli-um content and the color-altered zone was veryclear. Specifically, the concentration of Be decreasedfrom the edge toward the center of suspected treat-ed orangy pink sapphires, and the highest Be con-

centrations occurred in the yellow-to-orange sur-face-related color zones. These results, combinedwith the surface-conformal nature of the rims,proved that beryllium had been diffused into thestones from an external source for the purpose ofcreating the observed color.

Following the discovery that Be was the elementbeing diffused into the sapphires, two of the authorsbegan a series of experiments designed to replicatethe treatment being used in Thailand (see “Beryl-lium Diffusion into Corundum: Process andResults” section below). They guessed that the Thaitreaters would use chrysoberyl, since it was com-monly recovered with pink sapphire in Madagascar.Therefore, they heated Madagascar sapphires in thepresence of crushed chrysoberyl (BeAl2O4) to providethe beryllium, and found they could produce all thecolor modifications seen with the Thai treatment(Emmett and Douthit, 2002a). Simultaneously, sci-entists at D. Swarovski in Wattens, Austria, werecarrying out similar experiments (McClure, 2002b).By May 2002, both the Swarovski and Emmett/Douthit teams had independently reproduced theresults of the Thai process.

After the 2002 Tucson Show, many gemologistsbegan studying this material, which was available inabundance. On May 4, 2002, a meeting was held inCarlsbad, attended by several of the present authors,Dr. George Rossman of the California Institute ofTechnology, and representatives of the SSEF SwissGemmological Institute and the GemologicalInstitute of Thailand. At this meeting, the results ofGIA and AGTA studies and those of Emmett andDouthit were presented and debated. Subsequently,many contributions to the understanding of thesestones have appeared in the literature (Coldham,2002; Hänni, 2002; Hänni and Pettke, 2002; Pisutha-Arnond et al., 2002; Qi et al., 2002; Peretti andGünther, 2002; Fritsch et al., 2003).

On February 7, 2003, at the Tucson Show, AGTAheld a panel discussion on this new process; presen-tations were given by several of the present authorsand others (Shor, 2003). While Thai processors anddealers were still denying that the sapphires werebeing treated by Be diffusion (“Thailand . . . ,” 2002),the information presented by this panel was indis-putable. Don Kogen, president of Thaigem.com,attended this and other Tucson meetings on Be dif-fusion as the representative of the Chanthaburi Gemand Jewelry Association (CGA). He vowed to returnto his CGA colleagues with this information to helpresolve the situation.


Figure 2. Most of the Be-diffused stones seen early on,particularly in the pink-orange color range, showeddistinct yellow-to-orange surface-conformal layerswhen viewed in immersion. Photomicrograph byShane F. McClure; magnified 10×.


SIMS (secondary ion mass spectrometry) allows us tomeasure the trace-element concentrations in gemsdown to ppm (parts per million), a capability neverbefore needed by gemologists. SIMS is one of two com-mercially available techniques that can measure a widerange of elements down to ppm levels or below—theother is LA-ICP-MS (laser ablation–inductively coupledplasma–mass spectrometry)—without requiring specialpreparation or significant damage to the gem. It is theremarkable color alteration produced by as little as 10ppma of beryllium in ruby and sapphire that has forcedus to embrace such sophisticated analytical instrumen-tation. Embracing it has not been easy, as SIMS instru-ments cost $750,000–$2,000,000 each and LA-ICP-MSis more than $400,000, well beyond the financial capa-bility of most gemological laboratories. Thus, gem labstypically send samples to major analytical laboratories,recognizing that the cost of a single SIMS analysis canbe several hundred dollars.

The first commercial SIMS instrument was devel-oped to analyze the moon rocks brought back by theApollo astronauts in 1969, and the technique has beenused by mineralogists and petrologists ever since. TodaySIMS is also indispensable to the semiconductor indus-try for analysis of semiconductor wafers. SIMS operatesby focusing a beam of oxygen ions onto the sample in avacuum chamber. Ions of the sample material areknocked off the surface of the sample by the oxygen ionbeam and drawn into a mass spectrometer, which thensorts the ions according to their atomic weight andcharge. Once sorted, the ions are counted, and thecounts are converted into a quantitative analysis of thesample (for more information, see Hervig, 2002). FigureA-1 shows a SIMS instrument in operation.

As material (ions) must be removed from the stone,an important question is how much damage is done tothe gem during this kind of analysis. The oxygen ionbeam itself is very small, and is scanned over a square150 by 150 µm in area, with material removed to adepth of about 200 nm. This shallow depression on apolished gemstone is nearly impossible to see with theunaided eye, but can be observed with a gemologicalmicroscope if light is reflected off the surface at certain

angles (figure A-2). Although generally unnecessary, alight polishing can remove this very small depression.

As with all sophisticated analytical instruments,achieving accurate calibration is a primary issue forSIMS. Having a calibration standard made out of thesame type of material as the one being analyzed is by farthe best method. (Standards made from other materialscan, at best, provide relative concentrations.) One of theauthors (SN) has prepared synthetic sapphire standardsfor the analyses reported in this article, by ion implanta-tion. Widely used in the semiconductor industry, thistechnique can place a broad variety of elements into sap-phire with a concentration accuracy of a few percent orbetter (Leta and Morrison, 1980). Thus, we can achievean overall measurement accuracy of better than 25%—an impressive level when one considers that we are try-ing to measure only 5 to 50 atoms out of a million.

LA-ICP-MS has also been used successfully forthis purpose by other researchers. SIMS was chosenfor our research because several authors had experi-ence with this instrument in detecting light elementsin corundum.

BOX A: SIMS ANALYSIS

Figure A-1. Shown here is a SIMS instrument in use atEvans East. This PHI 6600, manufactured by PhysicalElectronics, employs a quadrupole mass spectrometerto detect and sort ions from the corundum sample.

Figure A-2. The ruby in these images wastested with SIMS and then examined at

25× magnification. In standard darkfieldillumination, no mark is visible (left).

With reflected light, however, the markleft behind by the SIMS instrument can beseen (right). The mark is only visible withcareful illumination. It would be extreme-

ly difficult to find with a hand loupe.Photomicrographs by Shane F. McClure.

Shortly thereafter, on February 20, 2003, theChanthaburi Gem and Jewelry Association issueda press release (CGA, 2003) in which they reportedthat their members had been introducing berylli-um into corundum by adding chrysoberyl to thecrucibles in which the stones were heated. Theyfurther stated that these Be-diffused stones wouldnow be sold separately from other heat-treatedcorundum.

The Be-diffusion process is, without question,the most broadly applicable artificial coloration ofnatural corundum ever achieved (see figures 1 and3). Analyses have indicated that the beryllium con-centration in the diffused region is only about 10to 35 parts per million atomic (ppma1) or 4.4 to15.5 ppmw (see table 1 for a comparison of ppmato ppmw for selected elements). That such minuteconcentrations of a foreign element could result insuch a wide variety of colors and color modifica-tions in corundum is truly unique. To understandhow this occurs, we need to examine the causes ofcolor in corundum much more closely than gemol-ogists have done in the past. Much of the informa-tion in the following section has appeared in thescientific literature over the past 40 years, or is theresult of studies by one of the authors (JLE) overthe past 15 years.

COLOR IN CORUNDUMThe range of color in natural corundum is veryextensive, with nearly all colors being represented(figure 4). Pure corundum is comprised only ofaluminum and oxygen, and is colorless. Moreover,its spectrum exhibits no absorption (i.e., the mate-rial is transparent) from 160 nm in the ultravioletto 5000 nm in the infrared. Gem corundum owesits many colors to impurities (typically trace ele-ments) that have replaced aluminum in the crys-tal lattice. These impurities can be the directcause of color, or they can chemically interactwith one another to cause color, or to modify thestrength (saturation) of a color. For the purpose of

this section, we will limit the discussion to col-orants of natural corundum and will ignore theadditional colorants used in the production of syn-thetic corundum.

The causes of color in corundum are manifoldand have been addressed in many publications. Agood general review is provided by Fritsch andRossman (1987, 1988); Häger (2001) recentlyreviewed the colors of corundum extensively in thecontext of heat treatment. The basic causes of color(also referred to as chromophores) are well known:Cr3+ produces pale pink through deep red as concen-trations increase, Fe3+ produces a pale yellow, Fe2+-Ti4+ pairs produce blue, and less well recognizedMg2+-trapped-hole pairs are responsible for yellow to


Figure 3. The Be-diffusion process is capable of causingdramatic changes in the color of corundum, such asaltering pink sapphire (left, before treatment) to pink-orange (right, after treatment). The samples weighapproximately 1 ct each; photo by Sriurai Scarratt.

TABLE 1. Comparison of ppma and ppmw forselected elements.a

Element ppma ppmw

Beryllium 1.000 0.442

Sodium 1.000 1.128

Magnesium 1.000 1.192

Silicon 1.000 1.378

Potassium 1.000 1.919

Calcium 1.000 1.996

Titanium 1.000 2.349

Chromium 1.000 2.550

Iron 1.000 2.739

Gallium 1.000 3.420

aCalculated as: ppma = (molecular weight of Al2O3) /5 • ppmw

(atomic weight of the element)

1 The more commonly used units for trace-element analyses areppmw (parts per million by weight), usually written as ppm. One ppmmeans that there is one microgram of impurity in one gram of crystal.In this article we choose to use the units ppma (parts per million atom-ic) to state trace-element concentrations. One ppma means that thereis one trace-element atom for each million atoms (for corundum, i.e.,400,000 Al atoms + 600,000 O atoms). These units are chosenbecause it is the concentration of trace elements that determines howthey chemically interact, not their relative weights.

orangy yellow (Kvapil et al., 1973; Schmetzer, 1981;Schmetzer et al., 1983; Boiko et al., 1987; Emmettand Douthit, 1993; Häger, 1993). A much widerrange of colors is produced by combinations of theseprimary chromophores. For example, green iscaused by Fe3+ yellow plus Fe2+-Ti4+ blue. Purple orviolet is a combination of Cr3+ red with Fe2+-Ti4+

blue, and orange can be a combination of the Mg2+-trapped-hole yellow plus the Cr3+ pink.

While these causes of color are understood invery general terms, a more detailed understand-ing is required to determine the impact of addingberyllium to the already complex set of naturaltrace elements in corundum. In addition, wemust understand the chemical interactions thatoccur among the various trace elements withinthe corundum crystal lattice. Although most ofus are more familiar with chemical interactionsin liquids than in solids, the latter nonetheless

occur during the growth of the corundum crystalin nature and during its heat treatment at hightemperatures.

First, we must note that there are two generalclasses of impurities in the corundum lattice: (1)those such as Cr3+ that have the same chemicalvalence (i.e., a positive charge of three) as Al3+ andare termed isovalent; and (2) those such as Mg2+ orTi4+ with a valence different from that of the Al3+

they replace—termed aliovalent. The most com-mon impurities in corundum are (by valence):

+1—Hydrogen (H)+2—Magnesium (Mg)+2—Iron (Fe)+3—Iron (Fe)+3—Chromium (Cr)+4—Titanium (Ti)+4—Silicon (Si)

Unfortunately, we know of no complete trace-element analyses (that is, down to the parts per mil-lion atomic—ppma—level) in the literature for anynatural corundum sample. Yet, as we shall see, aslittle as 5 to 10 ppma of an impurity can have a sub-stantial impact on the color of corundum.

Isovalent Ions. Cr3+ by itself in corundum producesa red coloration—from pale pink through deep red,depending on its concentration (McClure, 1962; fig-ure 5). The quantitative relationship between thestrength of the optical absorption and the concen-tration of chromium is well known, primarilybecause it has been carefully studied for ruby laserdesign (Dodd et al., 1964; Nelson and Sturge, 1965).

Fe3+ in corundum at high concentration ( 2,500ppma) produces a weak yellow coloration(McClure, 1962; Eigenmann et al., 1972; figure 6).The relationship between the strength of theabsorption and the Fe3+ concentration is not nearlyas well understood as that of Cr3+. The weak broadbands at 540, 700, and 1050 nm, and the narrowpeak at 388 nm, have been assigned to single Fe3+

ions, while the narrow peaks at 377 and 450 nmhave been tentatively assigned to Fe3+-Fe3+ pairs onthe nearest-neighbor lattice sites (Ferguson andFielding, 1971, 1972; Krebs and Maisch, 1971).However, this does not preclude the existence of ahigher-order cluster with additional ions or otherpoint defects. A quantitative relationship betweenthe strength of each of these iron-absorption fea-tures and iron concentration has not been com-pletely established.


Figure 4. Non-diffused corundum (here, approximately0.40 ct each), whether natural or heat treated, comes

in a wide variety of colors. Courtesy of Fine GemsInternational; photo by Harold and Erica Van Pelt.

Charge Compensation. The field known as defectchemistry provides a very useful framework forstudying charge-compensation mechanisms attrace-element concentrations and the chemicalinteractions among trace elements (Kröger, 1974;Kingery et al., 1976; Chiang et al., 1997; Smyth,2000). In this context a “defect” refers to any devi-ation from a perfect lattice. Thus impurities, ionvacancies, interstitial ions, and free electrons or(electron) holes are defects. To help explain theimpact of diffusion treatment, we will review herea few principles of defect chemistry as it relates tocorundum.

In discussing corundum coloration, it simplifiesmatters somewhat if we think about the charge ofan ion relative to the charge of a perfect lattice. Indefect chemistry, ions with excess positive charge(e.g., Ti4+ in Al2O3) are termed donors, because theymust “give up” an extra electron to enter thecorundum lattice. Similarly, ions with a relativecharge of −1 are termed acceptors, because theymust “accept” an electron from some other ion inthe lattice. Ti4+ (for example) has one extra positivecharge than the Al3+ ion it can replace in corun-dum, so we say that its relative charge is +1. Mg2+

has one less positive charge than the Al3+ it canreplace, so we say that its relative charge is −1. Ifboth magnesium (relative charge −1) and titanium(+1) are incorporated together, they will tend toattract each other and co-locate at adjacent or near-by sites in the atomic lattice, thereby balancingtheir charges. Acceptor or donor properties can alsobe ascribed to other point defects in the corundumlattice. Thus, oxygen vacancies (relative charge +2)and aluminum interstitials (+3) are donors, whileoxygen interstitials (−2) and aluminum vacancies (−3) are acceptors.

A crystal must be electrically neutral, whichmeans that the sum of all the relative positivecharges and electron holes equals the sum of therelative negative charges and free electrons.Consequently, when an ion with a valence of moreor less than +3 replaces Al, the excess or deficiencyof electrical charge must be compensated in someway so that the average positive charge still is +3.There are several ways charge compensation canoccur. During growth of corundum at low temper-ature, pairs of ions, one with a valence of +4 andone with a valence of +2, could be incorporatednear each other in a crystal. The total charge of thepair, +6, is the same as the two Al ions they wouldreplace. Thus the crystal remains electrically neu-

tral. Vacancies (atoms missing from lattice sites)and interstitial ions (atoms located between latticesites) can also charge compensate trace elements.


Figure 6. The 1.10 ct Australian sapphire on the left isunusual in that its color results only from a high con-centration of iron. The 1.40 ct heat-treated Sri Lankansapphire on the right is colored by Mg2+-trapped-holecenters. Iron causes pale yellow in sapphire primarilyby absorption of Fe3+-Fe3+ pairs as shown in this spec-trum. The narrow peak at 388 nm and the broadbands at 540, 700, and 1050 nm are attributed to Fe3+,while the narrow peaks at 377 and 450 nm areattributed to Fe3+-Fe3+ pairs. Photo by Maha Tannous.

Figure 5. These synthetic rubies and sapphires are col-ored only by chromium. As the concentration ofchromium increases (left to right), so does the satura-tion of the red. The spectrum shows chromiumabsorption in corundum. The samples weigh approxi-mately 2.5 ct each; photo by Sriurai Scarratt.

For example, an oxygen vacancy can charge com-pensate two divalent (+2) ions, and an oxygeninterstitial can charge compensate two tetravalent(+4) ions. Likewise, an aluminum (Al3+) vacancycan charge compensate three tetravalent ions, andan aluminum interstitial ion can charge compen-sate three divalent ions. In special cases, a freeelectron (relative charge –1) can charge compen-sate a tetravalent ion; or an electron hole (relativecharge +1, hereafter simply called a hole) can com-pensate a divalent ion. (A hole in corundum can beviewed as an oxygen ion carrying a charge of −1rather than –2.)

The way aliovalent ions are charge compensat-ed can be a large factor in determining the color ofcorundum. As noted above, when an aliovalent ionsuch as Ti4+ is incorporated into a corundum crystalduring growth, it can (in principle) be charge com-pensated by a divalent ion such as Mg2+, or by analuminum vacancy, or by an oxygen interstitial ion.However, it takes far less thermal energy to incor-porate a divalent ion than to produce an aluminumvacancy or an oxygen interstitial ion. Corundumgrowth in nature takes place at temperatures from250°C to 1400°C, depending on the growth environ-ment (G. Rossman, pers. comm., 2003). At thesetemperatures, it is far more likely that Ti will becharge compensated by a divalent ion (e.g., Mg2+ orFe2+), assuming one is available in the growth medi-um, than by a vacancy or an interstitial ion.However, when corundum is grown in the laborato-ry at temperatures near its melting point (~2045°C),or is heat treated at high temperatures, charge com-pensation by vacancies, interstitial ions, holes, orelectrons becomes more probable.

Aliovalent Ions. Fe2+ in corundum is an acceptor(Dutt and Kröger, 1975; Koripella and Kröger,1986). When not associated with other impurityions, it produces little or no coloration. Its majorabsorption bands are expected to be located in thenear-infrared portion of the spectrum but have notbeen observed. In corundum with a relatively highiron concentration, Fe2+ often is found paired withFe3+, leading to the creation of an intervalencecharge-transfer (IVCT) absorption (Burns, 1981;Nassau, 1983; Fritsch and Rossman, 1987) cen-tered at 875 nm (Krebs and Maisch, 1971; Smithand Strens, 1976). This band is very broad, withthe short-wavelength edge extending into the redregion of the spectrum, thereby producing a weakgray-blue coloration. Because this absorption fea-

ture is rarely seen in crystals that do not containtitanium, it is sometimes difficult to evaluatehow much of the blue coloration is contributed bythis band (as opposed to that contributed by theFe2+-Ti4+ pairs). Fe2+ also can be charge compensat-ed by tetravalent donor ions (e.g., Si4+ or Ti4+)forming donor-acceptor pairs; by hydrogen, whichis also a donor; or, in the absence of either of theseoptions, by an oxygen vacancy or an aluminuminterstitial ion.

Mg2+ in corundum is also an acceptor (Mohapatraand Kröger, 1977; Wang et al., 1983). It can be chargecompensated by tetravalent donor impurities (e.g.,Si4+ or Ti4+) or, in their absence, by an oxygenvacancy or a hole. Charge compensation by an oxy-gen vacancy can occur when the crystal grows inreducing conditions, or is heat treated in a reducingatmosphere (Emmett and Douthit, 1993). Whencorundum is grown or heat treated in oxidizingconditions, charge compensation of Mg2+ appearsto be by holes. The Mg2+-induced trapped hole (anO1- ion in the lattice) absorbs light very strongly inthe blue region of the spectrum (figure 7), whichleads to a yellow to orangy yellow color in naturalsapphire. In high-purity synthetic sapphire contain-ing only Mg2+, the color is a violet-brown. The dif-ference between these two colors is caused by thelocation of the trapped hole in the lattice. In thelatter case, the hole is trapped close to the Mg2+

ion. However, in natural sapphires, which contain


Figure 7. The absorption spectrum of the Mg2+ trappedhole, as seen here in synthetic corundum, producesthe strong yellow to orangy yellow color in some natu-ral sapphires that are heat treated without beryllium.

a variety of other impurities, the hole would prefer-entially associate with Cr3+ if available, or with Fe3+

in the absence of Cr3+, leading to an altered absorp-tion spectrum. As we shall see, in corundum Be2+

behaves much like Mg2+. The heat-treated yellowsapphires of Sri Lanka and the natural-color or heat-treated yellow portions of the sapphires from RockCreek, Montana, are both colored by the Mg2+-trapped-hole mechanism. As shown in figure 6, theweak yellow produced by Fe3+ and the orangy yellowof the Mg2+-induced trapped hole differ greatly inappearance. At this point we should note that theMg2+-related absorption center may be more com-plex than a simple Mg2+-induced trapped hole. Häger(1996) has shown that the addition of iron increasesthe absorption strength of the trapped hole in syn-thetic sapphire. The details of this interactionremain to be elucidated.

One of the unusual properties of the Mg2+-trapped-hole coloration in corundum is its sensitivi-ty to oxygen partial pressure (oxygen concentrationin the furnace atmosphere; Mohapatra and Kröger,1977; Wang et al., 1983; Emmett and Douthit,1993). If corundum containing Mg2+ is heat treatedin pure oxygen, the color saturation is maximized. Ifthat stone is then re-heated at an oxygen partialpressure of 10-3 atmospheres (atm), the color satura-tion is greatly reduced. This very strong relationshipbetween color saturation and oxygen partial pres-sure is not observed with other chromophores incorundum, such as Cr3+, Fe3+, and Fe2+-Ti4+.

Silicon, a common trace element and (as Si4+) adonor in corundum (Lee and Kröger, 1985), pro-duces no coloration but poses some unique issues.All of the other impurities discussed so far normal-ly occupy octahedral sites (i.e., a site surroundedby six oxygen atoms) in corundum and in manyother minerals. Si almost always occurs in a tetra-hedral site (surrounded by four oxygen atoms) inminerals, although there are no tetrahedral cationsites in the corundum lattice. So far, we do notknow whether Si occupies an octahedral site innatural corundum, or forms as micro- or nano-sized crystals of some simple aluminosilicate min-eral embedded in the corundum crystal. We doknow, however, that in Si-containing syntheticcorundum that has been heated to high tempera-ture and cooled rapidly (that is, compared to geo-logic cool-down times), at least a portion of the sil-icon is located on octahedral aluminum sites (Leeand Kröger, 1985). If Si actually occupies an alu-minum site, it will be “active”; that is, it will take

part in the charge-compensation process as adonor. If it is part of a nano-crystal of another min-eral phase, it will not. Since we are primarily con-cerned here with corundum that has been heattreated for tens to hundreds of hours at a very hightemperature during the Be-diffusion process, wewill make the assumption that low concentrationsof naturally occurring silicon (i.e., ≤100 ppma) arein solution on Al3+ sites. Proving this conjecturewill require additional studies. We may eventuallyfind that only a portion of the silicon is active.

Fe2+-Ti4+ acceptor-donor pairs, located on near-est-neighbor Al sites in corundum, produce thestrong blue color that we normally associate withsapphire (figure 8), resulting from strong absorp-tions at 580 and 700 nm. Figure 9 shows the Fe2+-Ti4+ absorption spectrum of a synthetic sapphiresample containing only Fe and Ti impurities atcolor-significant concentrations. This IVCT absorp-tion feature in corundum was first explained byTownsend (1968). More recent studies indicate thatthe actual defect cluster may contain other ionsand point defects in addition to single iron and tita-nium ions (Moon and Phillips, 1994).

The relative effectiveness of each of these traceelements, or trace-element pairs, in coloring corun-dum varies widely. Considering the color satura-tion in a 2 ct stone, in the course of earlier researchwe observed the following: 2,500 ppma Fe3+ willproduce a weak yellow color; 1,000 ppma Cr3+ willproduce a strong pink-red; 50 ppma of Fe2+-Ti4+

pairs will produce a very deep blue; and, finally, 15ppma of Mg2+-hole pairs will produce a strong


Figure 8. These magnificent blue sapphires (weighingapproximately 2–4 ct) owe their color primarily toFe2+-Ti4+ pairs. Courtesy of Gordon Bleck; photo byMaha Tannous.

orangy yellow (figure 10). Thus, to understand thecolor of corundum, we must be able to analyze forthese elements, and those that interact with them,down to the level of a few ppma, which is usuallynot possible, particularly for lighter elements, withenergy-dispersive X-ray fluorescence (EDXRF) spec-troscopy or electron microprobe analysis. To mea-

sure both light and heavy elements in corundum atsuch low levels, we have found it necessary to useSIMS or LA-ICP-MS (again, see Box A).

Interactions Among Elements. Natural corundumtypically contains most of the elements listedabove at some concentration levels, and most ofthese impurities will interact with each other toaffect the color. Because of the higher energyrequired to incorporate aliovalent ions into thecrystal lattice, the isovalent ions of Fe3+ and/or Cr3+

usually are incorporated at much higher concentra-tions (when they are available in the growth envi-ronment). Thus, for most corundum, the Fe con-centration usually exceeds, by at least one order ofmagnitude, the concentrations of Ti, Si, or Mg (aswell as other trace elements of less significance).

It is well known that, in the case of natural sap-phire, there is little or no correlation between theconcentrations of iron and titanium and the satura-tion of blue coloration. This is perhaps the mostobvious indication that the interaction amongimpurities is as important as the presence of theimpurities themselves in determining corundumcolor (Häger, 1992, 1993, 2001; Emmett andDouthit, 1993). With defect chemistry, we candescribe chemical reactions in solids in a way thatis directly analogous to what is done for chemicalreactions in liquids or gases (Pauling, 1956). Thus,once we know the relative positions of the donorand acceptor energy levels for corundum (see Box B),we can predict how the impurities will chemicallyinteract (Smyth, 2000). For many of the impuritiesin corundum with which we are concerned, the rel-evant energy levels have been determined by F. A.Kröger and his co-workers; an extensive compila-tion is contained in Kröger (1984) and its references.

From the defect chemical reactions and energy-level positions, we can formulate a series of rulesabout impurity interactions. These rules will notbe valid for all concentration levels or all tempera-tures. They become less valid as concentrations ofkey elements (such as Mg and Ti) exceed ~100ppma or at temperatures substantially above roomtemperature. A correct formulation requires solv-ing the equilibrium chemical defect reactions forall of the impurities simultaneously, which isbeyond the scope of this article. However, the fol-lowing rules are useful for a wide variety of situa-tions and, more importantly, will serve to illustratethe types of chemical reactions that occur amongimpurities in corundum:


Figure 10. The effectiveness of trace elements in color-ing corundum varies widely. These samples illustratethe effects of ~3,000 ppma Fe3+ (1.10 ct natural yellow

sapphire in center), ~1,000 ppma Cr3+ (193.58 ct syn-thetic ruby rod at top), ~50 ppma Fe2+-Ti4+ pairs (2.54 ct

blue synthetic sapphire wafer in center), ~20 ppmaMg2+ (40.04 ct purplish brown synthetic sapphire disc

in lower left), and ~25 ppma Cr3+ with ~25 ppma Mg2+

(large orange boule on right). Photo by Maha Tannous.

Figure 9. This is the absorption spectrum of Fe2+-Ti4+

pairs in synthetic sapphire containing only iron andtitanium.

• When Fe, Ti, and Mg are present, Ti will chargecompensate Mg before Fe.

• When Fe, Si, and Mg are present, Si will charge com-pensate Mg before Fe.

• When both Ti and Si are present with Fe and Mg, Tiwill charge compensate both Mg and Fe before Siwill charge compensate Mg and Fe.

• When the concentration of Mg exceeds the sum ofthe concentrations of Si and Ti, Mg will be chargecompensated by oxygen vacancies if the corundumformed or later was heat treated at low oxygen par-tial pressures. If it was formed or heat treated at highoxygen partial pressures, the Mg will be charge com-pensated by a trapped hole.

A simple example excluding silicon (Häger,2001) will illustrate how these rules can be applied.In all of the examples given here, we assume an Feconcentration substantially greater than the con-centration of Ti, Si, Mg, and Be, as this is typical ofnatural sapphire. Suppose we have an iron-contain-ing sapphire with 60 ppma Ti4+. If this stone grew orwas heat treated in a reducing environment, all 60ppma Ti4+ would form Fe2+-Ti4+ pairs, so the stonewould be deep blue. Another stone with exactly thesame iron and titanium content but also with 40ppma Mg2+ would be a lighter blue, since 40 ppmaof the Ti4+ charge compensates the Mg2+ leavingonly 20 ppma of Ti4+ to form Fe2+-Ti4+ pairs.

Now consider that the magnesium concentra-tion of our stone is 60 ppma. All 60 ppma of the Ti4+

must charge compensate the 60 ppma Mg2+, leavingnone to form Fe2+-Ti4+ pairs. Thus, if the stone hasonly a few hundred ppma of iron, it is nearly color-less; with more iron, it would be pale yellow orgreenish yellow.

To push our example one step further, supposethat the magnesium content is even higher, at 75ppma. Now there is 15 ppma excess Mg2+ that is notcharge compensated by Ti4+. If this stone formed orwas treated at low oxygen partial pressures, it willbe nearly colorless or pale yellow, depending on itsiron concentration (as for the 60 ppma Mg2+ case).This is because at low oxygen partial pressures, thecharge compensation is by oxygen vacancies.However, high oxygen partial pressure would resultin a strong orangy yellow color (figure 11), due tocharge compensation of the 15 ppma excess Mg2+ byholes forming the trapped-hole color center.

Using square brackets (such as [Mg]) to denoteconcentrations, we have summarized these interac-tions for iron-containing sapphire in table 2.

Aspects of this relationship between Fe, Ti, andMg have been discussed previously (Häger, 1992,2001; Emmett and Douthit, 1993). Häger (1993,2001) represented the charge compensation of tita-nium by magnesium as MgTiO3 (that is, Ti and Mgform a cluster occupying two nearest neighbor Alsites). While this is certainly possible, it is not at allnecessary. The charge compensation of these twoions only requires that they be in the same generalregion of the lattice.

The Mg, Fe, Ti relationship described (Häger,1992, 1993, 2001; Emmett and Douthit, 1993) isunfortunately an oversimplified model for naturalcorundum. The actual trace-element chemistry ofnatural corundum is often far more complex. For


Figure 11. This strong orangy yellow color zone in aheat-treated Montana sapphire is caused by the pres-ence of Mg2+ trapped-hole centers. Photomicrographby Shane F. McClure; immersion, magnified 10×.

TABLE 2. Relative trace-element concentrationsand color in iron-bearing corundum.

Compositiona Color

[Ti4+] >> [Mg2+] Very dark blue under reducing conditions[Ti4+] > [Mg2+] Blue under reducing conditions[Ti4+] ~~ [Mg2+] Nearly colorless or iron colored (pale yellow

to greenish yellow) under oxidizing orreducing conditions

[Mg2+] > [Ti4+] Nearly colorless or iron colored (pale yellow to greenish yellow) under reducing condi-tions; strong yellow to orangy yellow under oxidizing conditions

aKey to symbols:> “greater than” (approximately 5–25 ppma excess)>> “much greater than” (approximately 25–100 ppma excess)~~ nearly equal to


There are several different ways to picture interactions(chemical reactions) among ions in solids. One ofthese, known as the band model (Bube, 1992; Smyth,2000), has been exceedingly effective in describingsemiconductors. Surprisingly, it has also proved veryeffective in describing materials like corundum, whereall constituents other than aluminum and oxygen areat a concentration of 1% or far less. Figure B-1 shows aband-model energy-level diagram of corundum andsome of its trace elements. In this model, the energy ofan electron (in electron volts) is plotted on the verticalaxis. The extent of the horizontal axis is meant toimply only that the corundum energy levels are thesame everywhere in the crystal. The lowest energylevel on this plot is known as the valence band,because it contains all the valence electrons of corun-dum. The uppermost energy level is called the con-duction band, because if it contained a lot of electrons,the solid would conduct electricity. In corundum, theconduction band is empty at room temperature,which is why pure corundum is an excellent electricalinsulator. The region between the valence band andthe conduction band is called the band gap or forbid-den band, which has a width of approximately 9 elec-tron volts in corundum. In pure corundum, the ener-gy-level diagram consists of only the valence band andthe conduction band.

When trace elements (impurities) are added tocorundum, they can form discrete energy levels thatare often in the band gap. These energy levels areshown here as short lines to indicate they are local-ized around the impurity ion. To understand howthese ions interact with one another, we need toknow their positions relative to each other; theirabsolute positions in the band gap are of somewhatlesser importance. There is an agreed-upon conven-tion by which the ions are labeled: Donor ions, suchas titanium (see “Color in Corundum” section), arelabeled with their charge before donating their elec-tron, while acceptors are labeled with their chargeafter accepting the electron.

If we place magnesium into the corundum latticeon an aluminum site replacing an aluminum ion, itmust have an overall charge of +3 (like Al3+) for thecrystal to remain electrically neutral. But we knowthat the charge on Mg is +2. One method to accom-modate this discrepancy is to create a hole (a missingelectron) in the valence band near the Mg2+ ion. Thistype of charge compensation only occurs in highlyoxidizing environments. Since the relative charge ofa hole is +1, the hole plus the Mg2+ ion makes a totalcharge of +3 to satisfy the Al3+ site requirement. It isthis Mg2+-hole pair that absorbs light, giving rise to

the yellow and brown hues in Mg-containing corun-dum. Physically, the hole corresponds to one of theoxygen ions near the magnesium having a charge of–1 rather than the normal charge –2. A more struc-tural way of looking at this situation is to observethat an Al3+-O2− group has been replaced by an Mg2+-O1− group having the same total charge of +1.

If titanium, a donor, were introduced into other-wise high-purity corundum, its preferred valencewould be +3, and thus it would meet the chargerequirements of the Al3+ site. What happens if both Tiand Mg are in a corundum crystal in the sameamounts? The Ti3+ electron donor level is far abovethe Mg acceptor level, so an electron is transferredfrom Ti3+ (which becomes Ti4+) to the Mg2+-hole pair(which becomes just Mg2+). The electron and the holecombine, or annihilate each other, to refill thevalence band or, physically, the O1− near the Mg2+

becomes O2− as is normal. In this energy-level dia-gram, electrons move (fall) down to positions of lowerenergy. This interaction can also be pictured from thepoint of view of the hole rather than the electron. Inthat perspective, the hole is transferred from theMg2+-hole pair to the Ti3+, resulting in Mg2+ and Ti4+

exactly as before. Also in this energy-level diagram,holes fall up. It doesn’t matter which view is usedsince they are both the same, but usually we use theelectron point of view if there are excess donors andthe hole point of view if there are excess acceptors.

We now see a little more clearly what chargecompensation is, and how it takes place. By thismechanism of donor-acceptor electron (or hole) trans-fer, the two ions charge compensate each other as +2and +4, adding to +6, which is the charge of the twoAl3+ ions they replace. This electron transfer onlyoccurs because the Ti donor level is far above the Mgacceptor level and electrons always move down to thelowest available energy level. If the acceptor level wasabove the donor level, it would not occur. The Mg2+-Ti4+ pair does not absorb light in the visible region ofthe spectrum, so its formation does not produce colorin corundum. However, other pairs absorb lightstrongly and lead to strong coloration. Figure B-2Ashows the formation of a pair between the Ti3+ donorand the Fe2+-hole acceptor, where the Ti3+ becomesTi4+ and the Fe2+-hole pair becomes just Fe2+. Thisdonor-acceptor transfer creates the Fe2+-Ti4+ pairresponsible for the blue coloration in sapphire.

The Ti3+ donor level lies far above all acceptorlevels in corundum, and thus will always transferits electron (becoming Ti4+) to some acceptorimpurity if one is available. This is one of the rea-sons why the Ti3+ absorption spectrum is never

BOX B: UNDERSTANDING TRACE-ELEMENT INTERACTIONS IN CORUNDUM


observed in natural corundum.Now if there are two donors at different levels

and a single acceptor, the highest donor will preferen-tially charge compensate the acceptor. If the twodonor energy levels are close together, they will bothtake part in charge compensation of the acceptor.The same situation occurs in reverse if there are twoacceptors and a single donor. That is, the lowestacceptor preferentially charge compensates the donorunless they are both close together (see figure B-2).

Silicon in corundum is a little more complex (seediscussion in “Color in Corundum” section). In natu-ral unheated corundum, silicon probably exists asbroadly distributed nano- or micro-crystals of a silicatemineral. However, when the host corundum crystal isheat treated at a high temperature for a long time, allor part of the silicon will enter the corundum lattice insolution, replacing aluminum and producing thedonor states shown in figure B-1 (Lee and Kröger,1985). Si4+-Mg2+ pairs do not add color. Si4+-transitionmetal pairs in corundum have not been studied.

There are no published data on the position of theBe acceptor level. However, because the oxygen par-tial pressure dependence on the strength of the Be2+-hole pair absorption is very similar to that of theMg2+-hole pair, as is the absorption spectrum itself,the energy levels must be fairly close to each other.For the examples in this article, we have assumedthey are at the same level.

Now we can see the origin of the rules we havestated in the “Color in Corundum” section. If [Ti4+ +Si4+] > [Mg2+ + Be2+], there are more than enough

donor electrons to annihilate all trapped holes associ-ated with Be2+ and Mg2+, and thus the yellow col-oration cannot form. Likewise, if [Mg2+ + Be2+] >[Ti4++ Si4+] there are excess Mg2+ or Be2+ trapped-holepairs to cause the yellow coloration.

We can also see why the absorption spectrum ofBe in high-purity corundum is different from that inCr-containing corundum. In pure synthetic corun-dum, the Be-induced trapped hole is near the Be2+ ion,producing an orangy brown. However, if Cr3+ (a donor)is also present, the Be-induced hole is transferred tothe Cr3+ (holes fall up), producing a Cr3+ trapped-holepair and, thus, the characteristic orangy yellow col-oration. (The Cr3+ trapped-hole pair could combine toform Cr4+, but much more work is needed to deter-mine if it does.) The Fe3+ donor state is below that ofthe Cr3+ donor state, so if both ions are present the Be-induced hole again pairs with Cr3+, producing orangyyellow. In corundum without Cr3+ (or where Cr3+ isless than the excess holes available), holes would pairwith the Fe3+ to produce an Fe3+ trapped-hole pair (orperhaps Fe4+). The Fe3+ trapped-hole pair may beresponsible for the unusual “Post-it” yellow colorobserved in some Be-diffused sapphire. Additionalresearch involving Be diffusion into sapphire withonly iron as an impurity will be required to elucidatethis matter. By examining the relative energy-levelpositions of other donor and acceptor ions, however,we can deduce how they will interact.

Figure B-1. In this energy-level diagram of impuritiesin corundum, trivalent donor ions are shown on theleft, with divalent acceptor ions on the right. Acharge-compensating donor-acceptor pair can formwhen the donor level is above the acceptor level.There is no significance to the lateral positions of theenergy levels; the vertical positions are absolute,measured from the top of the valence band. Datafrom Kröger (1984) and references therein.

Figure B-2. In A, the Ti3+ donor exchanges an electronwith the Fe2+-hole acceptor, thus producing the Fe2+-Ti4+ pair responsible for blue in sapphire. In B, whenberyllium is diffused into A, the Be2+-hole acceptorpresents a more attractive site (lower energy) for theTi3+ donor electron than Fe2+, causing the electron tobe transferred to the Be2+-hole acceptor (or the Be-induced hole is transferred to the Fe2+). The transferresults in Fe2+ becoming Fe3+, eliminating the Fe2+-Ti4+

pair and the blue color. The Be2+-Ti4+ pair formed doesnot absorb light in the visible region of the spectrumand thus does not contribute color to the sapphire.

example, adding silicon, a common trace elementin corundum assuming that it is fully active, thecondition for the formation of the trapped-holecolor center changes from [Mg2+] > [Ti4+] to [Mg2+] >[Ti4+ + Si4+], so more magnesium is needed toobserve the yellow trapped-hole color. Zr4+ (zirco-nium) can be expected to react chemically in wayssimilar to Ti4+ and Si4+. Likewise, Ca2+ (calcium)and Ba2+ (barium) will react similarly to Mg2+. Theproblem becomes substantially more complexwhen we consider the possible roles of other tran-sition metal impurities such as vanadium (V),manganese (Mn), cobalt (Co), and nickel (Ni), eachof which can potentially exist in multiple valencestates. V and Mn could exist in corundum in 2+,3+, or 4+ valence states; thus, they could be eitherdonors or acceptors, or both. As donors, they couldsuppress the formation of trapped-hole color cen-ters by forming donor-acceptor pairs with excessdivalent ions. As acceptors, they would enhancethe formation of trapped-hole color centers byforming donor-acceptor pairs with excess tetrava-lent ions. Ni and Co could act as acceptors,enhancing the probability of trapped-hole colorcenter formation. Enhancing trapped-hole forma-tion increases the yellow color component, whilereducing trapped hole formation reduces the yel-low component.

Vanadium is commonly observed in Cr-con-taining corundum (Hänni and Pettke, 2002; Perettiand Günther, 2002), and Mn is often seen in trace-element analyses of corundum. There appears tobe no information in the literature on Ni and Co

concentrations in natural corundum. It is impor-tant to note that, in discussing these additionaltrace elements, we are not referring to concentra-tions high enough to contribute color on their ownpart. We are referring to concentrations in therange of 1 to 30 ppma, where they might play a sig-nificant role in the charge-compensation relation-ships among ions. Table 3 lists additional trace ele-ments that could potentially impact the color ofcorundum if they were in solid solution, whethernaturally occurring or from a diffusion process.Several appear in two or three columns, becausethey can potentially assume two or three valences.Thus, a quantitatively accurate analysis of thecolor of a ruby or sapphire would, in principle,require a complete trace-element analysis down toa level of a few ppma. Much additional trace-ele-ment data on a wide variety of natural corundum,as well as additional diffusion experiments, will beneeded before we fully understand which of theseions may be important. Inasmuch as most naturalcorundum exhibits color zoning (figure 12), andthus zoning of trace-element concentrations, suchan analysis would indeed be time consuming.Fortunately, we can understand the main featuresof how the addition of beryllium alters the color ofmost types of corundum by considering only theshort list of trace elements presented earlier: Fe,Cr, Ti, Si, Mg, and perhaps V.

Color Modifications Associated with BerylliumDiffusion. Now that we have an idea of how differ-ent elements interact in corundum, we can turn to


Figure 12. Variations in trace-element concentrationscreate the color zoning typically seen in many naturalsapphires, such as the orange and pink zones shownhere, which makes accurate chemical analysis verylaborious. Photomicrograph by Shane F. McClure;immersion, magnified 10×.

TABLE 3. Trace elements other than Fe, Mg, Ti, and Crthat could impact the color of corundum.

Valence

+1 +2 +3 +4

Hydrogen Beryllium Vanadium CarbonLithium Calcium Manganese SiliconSodium Vanadium Cobalt ManganeseCopper Manganese Nickel GermaniumSilver Cobalt Zirconium

Nickel TinCopper LeadZinc VanadiumCadmiumTinBariumLead

the specific case of beryllium and the color modifi-cations caused by the diffusion of beryllium intonatural corundum. As noted earlier, the presence ofBe has been established in the color-altered layer ofsapphires heat treated in Thailand by the “new”method (McClure et al., 2002; Wang and Green,2002a,b), at concentrations well above those mea-sured in unaltered regions. When beryllium is addedto corundum by diffusion, there is another layer ofcomplexity in color determination, because theberyllium concentration is not always spatially uni-form throughout the stone (see Appendix). The finalcolor may be some combination of the remainingcore of original color plus the Be-altered outer color.We will return to this point later, and for the pre-sent consider only Be-induced color modificationsresulting from a uniform beryllium concentration.

Drawing on our earlier studies (Emmett andDouthit, 1993) of Mg-induced trapped holes incorundum and subsequent research, our workinghypothesis was that Be2+ would position itself inthe corundum lattice by replacing an aluminumion, thus becoming an acceptor (again, see Box B).As such, it would act much like Mg2+, trappingholes and producing a strong yellow colorationunder exactly the same conditions. However, giventhe very small size of Be2+, we thought that its solu-bility in corundum would be lower than that ofMg. From experimental measurements on Be-dif-fused corundum (Hänni and Pettke, 2002; McClureet al., 2002; Peretti and Günther, 2002; Wang andGreen, 2002a,b; see also the “Beryllium Diffusioninto Corundum: Process and Results” sectionbelow), it appears that concentrations of diffusedberyllium range from about 10 to 30 ppma, andthat 10 to 15 ppma can produce a strong yellow col-oration. In many ways, then, the effect of diffusingberyllium into corundum is similar to that ofincreasing the magnesium concentration. Thus, wecan add the following to our list of rules:

• Ti4+ will preferentially charge compensate Be2+

before Fe2+.

• The condition for the formation of a yellow trapped-hole color center in an oxidizing atmospherebecomes [Mg2+ + Be2+] > [Ti4+ + Si4+].

With these simplified rules, we can predict mostof the color modifications that will result fromberyllium diffusion. First, let us examine the lightgreen and light blue sapphires typical of Songea(Tanzania), the Montana alluvial deposits, and thepale-colored or colorless sapphire that results fromheat treating some Sri Lankan geuda. In suchstones, we have the condition that [Ti4+ + Si4+] ≥[Mg2+] (that is, donor impurities are slightly greaterthan or about equal to acceptor impurities). Theaddition of beryllium as an acceptor changes theconcentration relationship to [Mg2+ + Be2+] > [Ti4+ +Si4+]—so there is an excess of acceptor impurities. Ina reducing atmosphere, the excess acceptor impuri-ties (Mg2+, Be2+) will be charge compensated by oxy-gen vacancies and thus produce no color. However,in an oxygen atmosphere, excess Mg2+ and Be2+ willtrap holes, producing a strong yellow to orangy yel-low coloration (figure 13). This is the reason thatberyllium diffusion of corundum must always becarried out in a highly oxidizing atmosphere. In thecases of Montana alluvial sapphire, Sri Lankan sap-phire, and Sri Lankan geuda, there are many stoneswith [Mg2+] > [Ti4+ + Si4+] that are nearly colorless orpale yellow as a result of slightly reducing condi-tions during formation. When these stones are heattreated in an oxidizing atmosphere, strong yellowcolors will be produced. When beryllium is diffusedinto these stones so that [Mg2+ + Be2+] >> [Ti4+ +Si4+], a very saturated orange will result.

Some of the first Be-diffused stones studied in theU.S. were pink/orange combinations, the result ofprocessing pink sapphires from Madagascar. PinkMadagascar sapphire contains Cr3+ far in excess of


Figure 13. Many of theoff-color sapphires from

Songea that do notimprove with standard

heat treatment (left) willturn a strong orangy yel-

low after Be diffusion(right). This stone

weighs 0.50 ct; photosby Elizabeth Schrader.

the added beryllium concentrations, so the Cr3+ con-centration is largely unaffected by heat treatment.Thus, we can apply our previous rules for Fe-con-taining corundum to corundum that also containsCr, taking into consideration the original Cr-pinkbody color. For example, when Be is diffused in anoxidizing atmosphere into a pink Madagascar sap-phire where [Ti4+ + Si4+] ≥ [Mg2+], the chemistrychanges to [Mg2+ + Be2+] > [Ti4+ + Si4+], which shiftsthe stone from donor- to acceptor-dominated; theresulting trapped holes produce a yellow coloration.When this yellow coloration permeates most or allof the original pink sapphire, an orange stone results(figure 14).

Some of the pink stones do not change colorwith Be diffusion. In that case, we have [Ti4+ + Si4+]>> [Mg2+], which becomes [Ti4+ + Si4+] > [Mg2+ + Be2+]from the Be diffusion; the material remains donor-dominated, does not trap holes, and does notbecome orange (figure 15). Earlier it was discussedthat other transition elements may also play a roleas donors or acceptors. As previously noted, thepink Madagascar sapphires contain 5 to 20 ppma ofvanadium. If V acts as a donor in these stones,which we would expect, it could also play a role insuppressing the formation of trapped-hole color cen-ters. Substantial additional analytical work andcareful optical spectroscopy will be required toresolve these issues.

We also need to understand the impact of Be dif-fusion into blue sapphire. As noted previously, darkblue natural sapphire results from the condition[Ti4+] >> [Mg2+]. (Note again that we are assumingan iron concentration far in excess of the Ti4+ con-

centration.) If Be is diffused into such stones, thecondition becomes [Ti4+] > [Mg2+ + Be2+], and thestone is still blue but much reduced in color satura-tion. Thus, it is possible to render unacceptablydark blue sapphire into marketable blue colors.

It is interesting to note that the earlier diffu-sion process—of titanium into otherwise colorlessor pale-colored sapphire (see, e.g., Kane et al.,1990)—is just the reverse of the process previouslydescribed. The sapphire typically used in the bluediffusion process is the very pale yellow, paleblue, pale pink, or completely colorless materialthat results from the unsuccessful heat treatmentof some Sri Lankan geuda. For this material, wecan presume the condition [Mg2+] ~~ [Ti4+ + Si4+],since it did not turn blue or yellow from the ini-tial heat-treatment process. Note that it also con-tains iron. When titanium is diffused into thestone, the condition in the diffusion layer shifts to


Figure 16. A dark blue color layer is formed near thesurface when Ti is diffused into sapphire. This 1.3-mm-thick section was sawn from a 1.45 ct stone, andshows a Ti diffusion–induced layer of blue color that is0.4 mm deep. Photo by Shane F. McClure; immersion.

Figure 14. When Be-induced yellow color is added tothe natural pink color of some Madagascar sapphires,a bright orange is produced, as can be seen in this 0.51ct pink sapphire that was cut in half with one halfthen being Be diffused. Photomicrograph by Shane F.McClure; magnified 10×.

Figure 15. These two pink preforms (the larger weighs0.77 ct) were reportedly subjected to the Be-diffusionprocess 10 times in Thailand, yet they did not changecolor. Photo by Maha Tannous.

[Ti4+ + Si4+] >>> [Mg2+]; that is, it becomes donordominated, and a very dark blue layer is formed(figure 16). Since titanium has a much greater sol-ubility in corundum than does beryllium, thefinal condition is [Ti4+ + Si4+] >>> [Mg2+], ratherthan [Ti4+ + Si4+] > [Mg2+]. Nevertheless, itbecomes clear that the mechanisms behind the Tiblue-diffusion process and the Be-diffusion processare essentially the same.

The foregoing discussion addressed the colormodifications in various types of corundumassuming that the added Be is essentially uniform-ly diffused throughout the stone. We now need toconsider the effect of non-uniform diffusion ofberyllium into a natural-color stone that hasessentially uniform natural coloration, such as theorangy pink (padparadscha-like) sapphires in whichthe final color appearance is produced by diffusinga layer of yellow into an originally pink stone.Depending on the duration of the diffusion pro-cess, the depth of penetration can vary from a fewtenths of a millimeter to the full thickness of thestone (see the Appendix and figure 17).

As a hypothetical example, suppose we startwith a 2 ct pink Madagascar sapphire, and exam-ine how the color could alter with diffusion time.It can be assumed that we are dealing with a stonewhere [Be2+ + Mg2+] > [Ti4+ + Si4+], and thus the Be-diffused areas will exhibit added yellow colorationfrom the trapped-hole color centers that form dur-ing the diffusion process. From our experiments(see the Appendix), we know that the diffusiondepth for this type of material is about 2 mm in 40hours at 1800°C. If the diffusion is conducted forone hour, the penetration depth will be roughly0.35 mm; the pink overall color of the stone willbe slightly modified in the direction of a padparad-scha color. In 10 hours, the penetration will beabout 1 mm and the stone will be the orangy pink

of a good padparadscha. Further increasing the dif-fusion time, and thus the depth to which the Bepenetrates, will successively yield pinkish orange,orange, and finally a very saturated deep orangecoloration as the beryllium penetrates the entirestone (figure 18).

Ruby is another important example of colormodification with diffusion depth. Many of therubies from Songea and Madagascar contain sub-stantial iron and titanium, and thus have a strongblue hue component as a result of Fe2+-Ti4+ pairs.As discussed previously, Be diffusion in an oxidiz-ing atmosphere shifts the equilibrium from [Ti4+ +Si4+] > [Mg2+] to [Mg2+ + Be2+] > [Ti4+ + Si4+], so theblue component becomes yellow. If these stonesare Be diffused all the way through, the entire bluecomponent is replaced with yellow, and the stonebecomes a very strong orange. However, if the Bediffusion only penetrates into the stone a relativelyshort distance, the yellow surface layer will visual-ly cancel out or offset the bluish core. Thus, the


Figure 18. The depth of the Be-diffusion layer directlyaffects face-up color. The stone on the far left has avery shallow diffusion layer. Moving right, each stonehas a progressively deeper layer, making each moreorange than the last. These stones weigh 0.80–1.89 ct;photo by Maha Tannous.

Figure 17. Depending on the duration of the Be-diffusion process, the induced color layer can penetrate to anydepth, including throughout the stone. This figure shows examples of shallow (left), moderate (center), and deep(right) penetration. Photomicrographs by Shane F. McClure; immersion, magnified 10×.

appearance will be a good ruby red. In this way,many attractive red rubies can be produced fromstones with a strong purple component (figure 19).

In principle, these stones can be detected easilyby observing the yellow rim in immersion.However, it is conceptually possible to create thesame visual effect without producing a yellow sur-face-conformal layer. After processing such rubieswith full Be penetration, so they turn orange, wecan reduce saturation of the yellow trapped-holecolor component by heating the stones at a loweroxygen partial pressure. Reducing the oxygen par-tial pressure will reduce the yellow coloration.Thus, by “tuning” the oxygen partial pressure in aheat-treatment process that takes place after the Be-diffusion process, an optimal ruby color may beachieved. This full-depth diffusion followed by anoxygen partial pressure post-process could also beused with our pink sapphire above to achieve pad-paradscha coloration without a visible diffusionlayer. Unfortunately, rubies and padparadscha-col-ored sapphires produced by this two-step methodwill be more difficult to detect.

It is also useful to consider the color modifica-tions resulting from partial Be diffusion into a bluesapphire where [Ti4+] > [Mg2+]. As we have alreadydiscussed, uniform Be diffusion into such stonesconverts them from blue to yellow. However, agreen overall color appearance results when theinward diffusion of beryllium converts only the out-ermost layer from blue to yellow. If this green stonewere to be heat treated subsequently in a reducingatmosphere, it would be blue with a colorless outerlayer (figure 20).

BERYLLIUM DIFFUSION INTO CORUNDUM: PROCESS AND RESULTSOver the course of the last year and a half, the authorshave conducted numerous experiments into both theBe-diffusion treatment of corundum and the charac-teristics of rubies and sapphires that have been treatedby this method. This first section describes thoseexperiments conducted primarily to explain thenature of beryllium diffusion and how it affects corun-dum. The following section reports on the examina-tion of known Be-diffused corundum samples, as wellas natural-color and heat-treated stones, to establishcriteria by which rubies and sapphires treated by thisnew diffusion process can be identified.

The extensive set of experiments in this first sec-tion were initiated to understand and then replicateThai results, to study the diffusion rates, and to deter-mine just how broadly applicable this new processwas to a wide variety of different types of corundum.By conducting these experiments in the laboratory,we could fully control all aspects of the process andnot have to guess the treatment conditions (whichhas been the case with stones processed in Thailand).

Materials and MethodsCorundum Samples. For the Be-diffusion experi-ments, we selected samples from large lots of


Figure 20. Partial Be diffusion of a blue sapphire fol-lowed by heating in a reducing atmosphere can createa colorless layer around the outside of an otherwiseblue stone, in effect lightening the overall color.Photomicrograph by Shane F. McClure; immersion,magnified 27×.

Figure 19. Be diffusion can produce good ruby colors(center, 5.05 ct) from strongly purplish or brownishred stones from some localities that typically do notrespond to standard heat treatment, such as thesmall stones from Songea in this image. Photo byMaha Tannous.

rough sapphire that were mined from several dif-ferent localities, including Australia, Madagascar,Tanzania, and Montana, as detailed in table 4. Ingeneral, we preferred to use small pieces (≤ 0.5 ct)to assure that we could diffuse beryllium com-pletely through the samples in 50 hours or less;however, we also conducted experiments on largerindividual pieces to measure diffusion depths.Many of the samples listed in table 4 are fromlarge, carefully collected mine-run parcels fromthe deposits indicated, and thus they contain awide range of stone sizes. From these lots weselected samples primarily on the basis of size,while endeavoring to maintain representativecolor and clarity.

For spectroscopic experiments on Be-diffusedThai-processed sapphire, we obtained a variety ofsamples (15 pieces) in Bangkok that ranged from0.25 ct to 2.1 ct. According to the treaters, theywere comprised of both Songea pale green and bluesapphire that had been processed to a vivid yellow,and Madagascar pink sapphire processed to pad-paradscha and saturated orange colors. As much aspossible, samples were chosen with the c-axis per-pendicular to the table facet to facilitate obtainingabsorption spectra that included the ordinary ray(o-ray) optical direction. On these 15 samples, the

culet was ground off parallel to the table and pol-ished to produce wafers from 0.5 to 1.8 mm thick.Samples with similar geometry were also preparedfrom rough material produced in our diffusionexperiments, or from rough pieces prior to diffu-sion treatment. Samples polished before diffusiontreatment required repolishing after diffusiontreatment. In total, about 30 samples were pre-pared and studied.

In addition to the natural sapphire samplesdescribed in table 4, four types of high-purity syn-thetic corundum were also used for the Be-diffu-sion experiments and spectroscopic testing (seetable 5). These samples were prepared as polishedwafers prior to Be diffusion, and repolished after-ward. In all cases the c-axis was perpendicular tothe polished face. We have not tested low-puritygem-grade Verneuil-grown sapphire.

Applying Beryllium Compounds. To apply berylli-um to the surfaces of the corundum samples fordiffusion, we used both molten fluxes (Emmett,1999) and dry powders as carriers for Be-containingcompounds. (A flux in this context is a high-tem-perature solvent for inorganic compounds.) While itis common knowledge that Thai heat treatersemploy fluxes in many processes, the actual formu-


TABLE 4. Rough corundum samples used for Be-diffusion treatment experiments.

Origin Color Lot size Avg. stone size Source

King’s Plain, New South Basaltic dark blue, transparent 10 kg 0.3–0.5 ct Great Northern MiningWales, Australia to opaque

Basaltic dark blue, transparent 0.3–0.5 ctto opaqueYellowish green 0.3–0.8 ct

Wutu Township, Shandong Basaltic dark blue 1 kg ~0.3 ct (large crystals Diversified MineralResourcesProvince, China sliced for wafers)Antsiranana, Madagascar Basaltic dark blue, dark green 250 g 0.3–0.6 ct Budsol Co., Hargem Ltd.Ilakaka, Madagascar “Pink” sapphirea 350 g 0.4–1.0 ct Wobito Brothers, Allerton

Cushman & Co., Paul WildDana Bar, Missouri River, Pale green, pale blue 1 kg 1.5 ct American Gem Corp.MontanaDry Cottonwood Creek, Very pale green, very pale 10 kg 0.7 ct American Gem Corp.Montana yellowish greenRock Creek, Montana Pale blue, pale yellowish green 10 kg 0.7 ct American Gem Corp.Yogo Gulch, Montana Medium blue, violet 200 g 0.2 ct American Gem Corp.Sri Lanka “Colorless” heat-treated geudab 100 g 0.3–1.0 ct Paul Wild

Pale blue, pale yellowish green Ruby, ranging from bluish red to brownish red

Umba, Tanzania A wide variety 600 g 3 ct Nafco Gems, Paul Wild

aThis parcel of “pink” sapphires showed a range of pink color saturation, and also included some violetish pink, pale yellow,pale blue, and colorless pieces.bThe sapphires in this parcel ranged from colorless to very pale blue, pink, or yellow, after heat treatment in an oxidizing or reducing atmosphere.

Subera Area C, Queensland,Australia

10 kg Great Northern Mining

Songea, Tanzania 600 g 0.3–1.2 ct American Gem Corp.

lations are closely guarded trade secrets. Ratherthan attempt to find out actual Thai formulas, wechose to formulate our own from calcium borate,calcium phosphate, sodium phosphate, or a sodiumfeldspar (a silicate mineral). To each main fluxcomponent we added small amounts of Al2O3 andZrO2 to raise the flux viscosity at high tempera-tures. To these fluxes we then added ~4% naturalchrysoberyl that had been finely powdered.Chrysoberyl, BeAl2O4, is 7.1% Be or 19.7% byweight BeO.

These flux mixtures were blended with water oralcohol containing a binder until they reached theconsistency of a heavy paint. For water mixtures,Varethane™ or methocel was used as a binder.(Methocel is a modified cellulose ether used as abinder in many industrial applications.) For alcoholmixtures, orange shellac was used as a binder. Littleconsistent difference was noted among the binders.The rough stones were dipped into the flux paintmixture and then allowed to dry. Polished corun-dum wafers were hand painted on both sides withthe mixtures.

Dry powders, that is, mixtures of a carrier anda beryllium compound, do not melt and flow likea flux. The dry powders used were very high puri-ty Al2O3 containing either 4% powderedchrysoberyl or 0.8% very high purity BeO. Wealso used a mixture containing 2.65% BeO. Twoapplication methods were employed. First, wecompletely buried the corundum samples in thepowder mixture, in an alumina crucible. This isthe original technique developed by UnionCarbide Corp. (Carr and Nisevich, 1975, 1976,1977). It works very well from the point of view of

diffusion uniformity, but sintering of the powderto a fairly solid mass made sample recovery labo-rious. In the second method, the powder wasmixed to the consistency of a paint, just as wasdone for the fluxes, and applied to rough stonesand polished wafers as previously described. Thislatter process makes sample recovery very easybut does not always result in completely uniformdiffusion.

Heat Treatment. Heat-treatment experimentswere conducted with an electrically heated furnacemanufactured by Thermal Technology of SantaRosa, California (figure 21). The modified type-1000A furnace uses a graphite heating element andgraphite insulation, and is capable of operation at2500°C (which far exceeds the 2045°C meltingpoint of corundum). For controlled atmosphereoperation, the furnace separates the heating ele-ment from the sapphires with a muffle tube madefrom Coors 99.8% Al2O3 ceramic. This ceramicmuffle tube becomes very soft at temperaturesover 1800°C, and it can collapse if it becomes toohot. This limits the temperature for the long runstypical of beryllium diffusion to 1800 to 1850°C.The sapphires are placed inside the muffle tubewhere any type of process gas can flow over them.A more detailed description of this equipment andhow it is used can be found in Emmett andDouthit (1993).

For the heat-treatment experiments describedbelow, the process gas used was pure oxygen. Thischoice was made because, as discussed earlier, thetrapped-hole coloration is sensitive to oxygen par-tial pressure, and is maximized in 100% oxygen.


TABLE 5. SIMS analyses of synthetic corundum showing the low levels of Fe, Ti, andCr (non-ruby) in ppma.a

Samplesb Na K Be Mg Ca Cr Fe Ga Ti

SS1, SG Czochralski-grown 0.48 0.07 1.46 0.16 0.02 0.8 0.04 3.1 1.27synthetic sapphireSS8, CSI Hemlite grade heat 0.39 0.14 13.4 0.83 0.04 0.25 0.03 0.94 1.98exchanger method–grownsynthetic sapphireSS3, SG Czochralski-grown 0.18 0.02 1.57 0.05 0.02 316 0.04 1.8 2.14light red synthetic rubySS18, SG Czochralski-grown 0.23 0.04 1.57 1.27 0.04 2580 0.13 7.5 2.00dark red synthetic ruby

aData taken before Be diffusion except sample SS8 which was measured after.bSG: Saint-Gobain Crystals and Detectors, 750 S. 32nd St., Washougal, WA 98671. CSI: Crystal Systems Inc., 27 Congress St., Salem, MA, 01970.

The coated sapphires were placed in shallowalumina crucibles made from Coors 99.8% Al2O3ceramic. When the stones were embedded in a drypowder, deeper Coors crucibles were used.Depending on the number of stones in a sample lot,up to 12 separate experiments could be run in thefurnace at the same time. For these experiments,temperatures of 1780°C or 1800°C were chosen.Run times from 25 to 100 hours at temperaturewere chosen, with 33 hours being typical.

Spectroscopy. The ultraviolet-visible and near-infrared (UV-Vis-NIR) spectra were recorded forthe range 200–1100 nm at room temperature usinga Hitachi U-2000 spectrophotometer. More than60 spectra were collected on natural and syntheticBe-diffused sapphire wafers. For those natural sap-phire samples where the color was not spatiallyuniform, reproducible spectra were assured by themethod described by Emmett and Douthit (1993).Spectra were recorded only for the o-ray (E⊥c).

There are three feasible methods for determin-ing trapped-hole spectra. In all cases we refer tothese methods as producing differential spectra.First, the spectra can be taken in the Be-diffusedlayer and in the center of the stone where berylli-um has not penetrated. If the spectrum from theunaltered portion of the sample is subtractedfrom that from the Be-diffused region, thetrapped-hole spectrum will result. This methodworks well with synthetic sapphire and ruby,where dopant concentrations are spatially uni-form, but it leads to inaccurate results on naturalcorundum samples, which are rarely uniform inchemistry.

The second method is to take one spectrumbefore Be diffusion and another after. If the beforespectrum is subtracted from the after spectrum, anaccurate trapped-hole spectrum results. This tech-nique works very well on uniform syntheticcorundum samples, and on natural samples if greatcare is taken to assure both measurements aretaken through exactly the same place in the sam-ple (Emmett and Douthit, 1993).

The third method, for use on a sample alreadydiffused, is to take two spectra on the sample equi-librated at two different oxygen partial pressures.As noted previously, the strength of the trapped-hole absorption spectrum is very sensitive to oxy-gen partial pressure. From a practical point ofview, the spectrum is taken on the Be-diffusedsample as received. Then the sample is heated to

1650°C for a few hours in an oxygen partial pres-sure of 10−3 atm, cooled, and the spectrum is takenagain. If the after spectrum is subtracted from theas-received spectrum, the spectrum of the trapped-hole center results. This works well on syntheticcorundum samples, and on natural corundumsamples if great care is taken to position the sam-ple as described above. Heating for a few hours at1650°C has little impact on a sample that has beenheated for a few tens to hundreds of hours at tem-peratures of 1800°C or above. For the workdescribed here, the second or third method wasused on all samples, as the first method proved tobe too inaccurate.

Chemistry. Secondary ion mass spectrometry withelement-in-sapphire standards (again, see SIMSBox A) was used at Evans East to determine Be andother trace-element concentrations in some of thenatural Be-diffused wafer samples used for absorp-tion spectroscopy, and in some of the syntheticsamples before and/or after Be-diffusion treatment.


Figure 21. Most of the Be-diffusion experimentsdescribed in this article were conducted using anelectrically heated furnace. Photo courtesy ofThermal Technology Inc.

Results and DiscussionIdentification of the Be-Induced Trapped Hole. InApril 2002, we initiated a set of experiments todetermine the origin of the color induced by Be dif-fusion. We believed that we were observingtrapped-hole-center absorption, similar to thatobserved much earlier in both natural and synthet-ic corundum (Mohapatra and Kröger, 1977; Wang etal., 1983; Emmett and Douthit, 1993; Häger, 1993).Because holes only form in corundum under highlyoxidizing conditions, the concentrations of Mg- orBe-hole pairs are very sensitive to the oxygen par-tial pressure at which they are equilibrated in thefurnace. An orange sapphire Be-diffused inThailand from pink Madagascar starting materialwas chosen for the first experiment. Figure 22shows the results. At an oxygen partial pressure of10−2 atm (1% oxygen, 1 atm total pressure), the Be-induced coloration has been reduced by about onehalf. This extreme sensitivity is not typical of thespectra of single ions in corundum.

From this experiment, we can also determinethe absorption spectrum of this Be-induced col-oration. By taking the differential spectrum of thefigure 22 spectra, we obtain the result in figure 23,which is the absorption spectrum of the Be-inducedcoloration in this sapphire. Note that it is quitesimilar to the Mg2+- trapped hole pair spectrum pre-

sented in figure 7. These two pieces of data, takentogether and compared with the well-studiedexample of the Mg2+-trapped hole pair, form thebasis of our identification of the Be-induced col-oration as arising from Be-induced trapped-hole for-mation. These spectral changes are reversible.When the sample was then equilibrated at an oxy-gen partial pressure of 1 atm, the original absorp-tion spectrum was restored. Subsequently, we test-ed a wide variety of sapphire samples from differentlocations over a wider range of oxygen partial pres-sures and temperatures with very similar results. Ineach case, the predominant 460 nm absorptionband appeared and showed the same oxygen partialpressure dependence (see “Absorption Spectra” sec-tion below). Additional confirmatory evidence ofthe induced trapped hole nature of these spectrawould be the change of the spectra with tempera-ture. It is known that such spectra broaden to thelow-energy side when the stones are heated severalhundred degrees above room temperature. Datathat tend to confirm this are shown in the“Gemological Testing and Identification” section,but additional studies would be useful.

Be-Diffusion Experiments. The next task was to dif-fuse beryllium into sapphire under controlled condi-tions to study the effects. Since the inception of thisproject, we have conducted 17 furnace runs devotedto beryllium-diffusion experiments, using differentsamples, carriers, and application methods, at tem-


Figure 22. These absorption spectra were taken of apink sapphire that was Be diffused, both as receivedand after a portion of the Be-induced coloration wasremoved by equilibrating the stone at a lower (10−2

atm) oxygen partial pressure. Figure 23. The absorption shown here represents thedifference (B–A) of the two spectra shown in figure 22.Because the chromium spectrum is unchanged by re-equilibration of the stone at a lower oxygen partialpressure, this differential spectrum shows the absorp-tion associated with the beryllium-induced coloration.

peratures from 1780°C to 1800°C and treatmenttimes of 25, 33, and (in two cases where the sampleswere embedded in powder) 100 hours. In each of the17 runs, from 1 to 14 separate experiments were

conducted for a total of 105 individual experiments.A subset of the 105 experiments is presented intable 6. In addition, another 21 furnace runs, for atotal of 53 individual experiments, were conducted


TABLE 6. Selected beryllium-diffused corundum experiments.

Experiment Samples No. of pieces Carrier Application method Temp. Time

1 Madagascar pink 7Sri Lanka near colorless heat-treated geuda 8

2 Madagascar pink 6Sri Lanka colorlessheat-treated geuda 9


4 Madagascar pink 6Sri Lanka colorless heat-treated geuda 4


6 Madagascar pink 3King’s Plain blue 3Dry Cottonwood Creek 3Subera 3Rock Creek 3Songea ruby 3Songea blue-green 3Sri Lanka colorlessheat-treated geuda 3

SS8 synthetic sapphire 17 SS1 synthetic sapphire 1

Madagascar pink 5Sri Lanka colorlessheat-treated geuda 6

Rock Creek 58 Madagascar pink 3

King’s Plain blue 3Dry Cottonwood Creek 3Subera 3Rock Creek 3Songea ruby 3Songea blue-green 3Sri Lanka colorless 3heat-treated geuda

SS8 synthetic sapphire 19 Songea blues, greens, 311 Al2O3 + 2.65% BeO Paint 1800 33

and ruby powder10 King’s Plain 363 Al2O3 + 2.65% BeO Paint 1800 33

powder11 Rock Creek 146 Al2O3 + 2.65% BeO Paint 1800 3312 Ilakaka, Madagascar 150 Al2O3 + 2.65% BeO Paint 1800 33

powder13 Shandong, China— 1 Al2O3 + 2.65% BeO Paint 1800 33

very dark blue wafer powder14 Synthetic ruby wafer 2 Al2O3 + 2.65% BeO Paint 1800 33

powder

Calcium phosphateflux + 4% chrysoberyl

Sodium phosphate flux+ 4% chrysoberyl

Paint 1800 25

Paint 1800 25

Paint 1800 25

Na-feldspar flux +4% chrysoberyl

Al2O3 + 4% chrysoberylpowder

Calcium borate flux+ 4% chrysoberyl

Paint 1800 25

Embedded in powder 1780 100

Al2O3 + 0.8% BeO Embedded in powder 1780 33powder

Al2O3 powder Embedded in powder 1780 33

Al2O3 + 2% MgO Embedded in powder 1780 100powder

(°C) (hrs)

to produce differential spectra, and to study oxygenpartial pressure dependence of the trapped-holeabsorption spectra. As first reported in Emmett andDouthit (2002a), experiments 1 through 4 repro-duced the full range of colors and diffusion-layerphenomenology that we had observed in the Thai-processed stones (see, e.g., figure 24). The stonesshowing color modification exhibited the typicalsurface-conformal color layer with a thickness ofroughly 1.4 mm. A few of the pink Madagascarstones did not change color. As discussed previous-ly, this lack of color modification can be attributedto the chemistry of the specific sapphire, where[Ti4+ + Si4+] > [Mg2+ + Be2+] or, more completely, [thetotal of all donors] > [Mg2+ + Be2+]. With the excep-tion of the silicate (sodium-feldspar) flux, there werefew or no discernable differences in the apparenteffectiveness of the fluxes. The corundum samplesin the silicate flux showed little or no color modifi-cation, and thus silicate fluxes were not pursuedfurther.

During this same time period, there was a largeamount of disinformation coming from Thailand tothe effect that neither beryllium nor chrysoberylwas involved in the new treatment process, andthat these dramatic color modifications were theresult of naturally occurring impurities. As a directconsequence, a more extensive set of experimentswas undertaken to significantly narrow the possi-bilities (Emmett and Douthit, 2002b).

The next experiment (No. 5) was designed toeliminate any possible effect from the flux itself.

To accomplish this, ground natural chrysoberylwas mixed with high-purity reagent-grade Al2O3powder. Another set of stones was packed in thispowder mixture in high-purity alumina cruciblesand heat-treated. The longer time for this experi-ment (100 hours vs. 25 hours for experiments 1–4)was chosen to assure that we could diffuse com-pletely through some of the smaller stones. Theresults of this experiment, in terms of induced col-oration, were just like the flux diffusion experi-ments, with the additional fact that the smallerstones were diffused, and thus colored, completelythrough. Again, a few of the pink Madagascar sap-phires did not show any color alteration.

Natural chrysoberyl is not a high-purity materi-al containing just BeO and Al2O3. It can containsignificant quantities of iron, chromium, and/or avariety of other elements. Since we had no low-level trace-element analyses on naturalchrysoberyl, it was at least conceptually feasiblethat trace-element impurities in the chrysoberylcould be responsible for the induced coloration incorundum. To address that possibility, high-purityreagent-grade BeO powder was procured, and mix-tures were made with the high-purity Al2O3 pow-der and used in experiment 6. A wide variety ofrepresentative sapphires were then heated for 33hours at 1780°C. The results of these experimentswere the same as the earlier experiments usingchrysoberyl (that is, all the orangy pink, orange,orangy yellow, and yellow colors were produced).Again, some of the pink Madagascar sapphiresremained unchanged in color.

We also wished to evaluate the possible role ofmagnesium. As discussed earlier, if [Mg2+] > [Ti4+ +Si4+], the trapped-hole yellow coloration will formwhen corundum is heat treated in an oxygen atmo-sphere. It is at least possible that Thai fluxes couldcontain some Mg compounds, which might diffuseinto the corundum causing the yellow trapped-holecoloration. To determine if magnesium diffusioncould be a factor in the Thai process, we added 2%MgO into the high-purity Al2O3 powder in experi-ment 7. Even after heating for 100 hours, there wasno induced coloration in the stones. This resultwas consistent with what we would expect, sincethe diffusion rate of magnesium into sapphire isvery low, and an MgAl2O4 (spinel) layer formsrapidly at the interface between the sapphire andthe powder.

As a final check, we conducted a “null” experi-ment (No. 8). We embedded a group of test sam-


Figure 24. Experiments 1–4 in this study were ableto reproduce every color and color layer that wehad observed in the Be-diffused samples processedin Thailand. Some of those colors are shown bythese corundum samples (0.4–2.1 ct). Photo byMaha Tannous.

ples from several different localities in high-purityAl2O3 powder with no other additives, andreplaced the entire inside ceramic surface of thefurnace to assure no beryllium contaminationfrom previous experiments. These samples wereagain heat treated for 33 hours at 1780°C. In thiscase, there were no substantive color modifica-tions at all.

To determine the diffusion rate of beryllium,we sectioned several of the larger stones from thevarious experiments described above, so we couldmeasure the thickness of the diffusion layer. Fromthese measurements, the chemical diffusion coeffi-cient was determined to be roughly 10−7 cm2/sec(see Appendix). While this is a rather crude mea-surement, we note that this diffusion coefficient isabout 1¼100 that of hydrogen in corundum (El-Aiatand Kröger, 1982), and about 100 times that of tita-nium (our measurements) in corundum, at thesame temperature. Also noted were minor varia-tions in diffusion depth for different stones pro-cessed for the same duration at the same tempera-ture. The variation of diffusion rate with impurityconcentration is well known for diffusion mea-surements in solids.

To further understand the role of Be2+ in corun-dum, we diffused Be into colorless, very high-puri-ty synthetic sapphire and high-purity syntheticruby, with Fe contents on the order of 0.1 ppma or

less and titanium contents of only about 2 ppma.Table 5 shows the low concentrations of Fe, Ti,and for the sapphire samples, Cr. These high-puri-ty materials were chosen in an attempt to ensurethat existing impurities in these materials couldnot cause any significant coloration. Beryllium dif-fusion into these types of materials resulted inconcentrations ranging from 13 to 35 ppma. Figure25 shows samples of both the Czochralski-grownand heat exchanger method (HEM)-grown synthet-ic sapphire after beryllium diffusion (for informa-tion on HEM, see Khattak and Schmid, 1984;Khattak et al., 1986). Both samples exhibit a strongorangy brown diffusion layer. Clearly, 10 to 30ppma beryllium is a very strong chromophore inboth natural and synthetic corundum, and there isno question that beryllium diffused into corun-dum in an oxygen atmosphere is the singlecausative agent of these color modifications.

After completing the experiments above, wewere curious as to just how universal this Be-induced color modification might be. Using a com-bination of aluminum oxide and 2.65% BeO pow-der mixed as a paint, we performed Be-diffusionexperiments on a wide variety of natural corundumsamples (again, see table 4). All of these experi-ments were run at 1800°C for 33 hours. The coloralteration produced is evident in these before-and-after pictures of mine-run sapphire and ruby fromSongea, Tanzania (experiment 9, figure 26); RockCreek, Montana (experiment 11, figure 27); King’sPlain, Australia (experiment 10, figure 28); andIlakaka, Madagascar (experiment 12, figure 29).Using the same procedure as for experiments 9through 12, we obtained similar results with sap-phire from Dry Cottonwood Creek and EldoradoBar, Montana; Subera, Australia; Shandong, China;and Umba, Tanzania. While this list is not exhaus-tive of sapphire deposits by any means, it doesimply broad applicability of the Be-diffusion pro-cess. Many additional photographs of the colorsproduced by Be diffusion into corundum are shownin Themelis (2003).

The reason that beryllium diffusion modifiesthe color of most corundum can be easily under-stood. Gem corundum usually is not strongly col-ored, which means that the net concentration ofeither Fe2+-Ti4+ pairs or Mg2+-hole pairs is not large.Consequently, reducing the effective concentra-tion of Fe2+-Ti4+ pairs or increasing the effectiveconcentration of Mg2+ by diffusing in 10 to 30ppma beryllium, has a strong impact on color.


Figure 25. These two high-purity synthetic sapphiresamples were Be diffused in the experimentsdescribed in the text. The cube was Czochralskigrown by Union Carbide Corp., while the cylinder (14mm in diameter) was grown by Crystal Systems Inc.using the heat exchanger method. The concentrationof beryllium in these samples is approximately 14ppma, which illustrates that beryllium is a very strongcolorant in sapphire. Photo by Maha Tannous.

Experiments on Blue Sapphire. From our analysisof the impact of beryllium diffusion on very darkblue sapphires where [Ti4+] >> [Mg2+], we find thatberyllium diffusion should produce the condition[Ti4+] > [Mg2+ + Be2+], substantially reducing theblue color saturation. To test this point (experi-ment 13), we cut a very thin (0.9 mm) wafer ofextremely dark blue sapphire from a Shandongcrystal (figure 30A). Figure 30B shows the same

wafer after Be diffusion in an oxygen atmosphere.The lighter blue areas have been rendered eithercolorless or pale yellow by the mechanisms previ-ously discussed, and the remaining blue areas aremuch less saturated. After subsequent normal heattreatment in a reducing atmosphere (figure 30C),the yellow areas have been rendered essentiallycolorless, while the previously colorless or paleblue areas in figure 30B are a more saturated blue.


Figure 26. These two groups of different colors of sapphire rough (top, 0.5–1.2 ct; and bottom, 0.8–2.5 ct), fromSongea, Tanzania, are shown before (left) and after (right) Be diffusion by the authors. Photos by J. Emmett.

Figure 27. Rough sapphire from Rock Creek, Montana, also responded well to Be diffusion, as seen in these imagesbefore (left) and after (right) Be-diffusion treatment. These samples weigh 0.4–1.4 ct. Photos by J. Emmett.

Overall the blue areas are now far less saturatedthan in the wafer before treatment. Thus, a simpleprocess for turning some types of very dark-bluesapphire into a much lighter blue is to diffuseberyllium into the crystal, followed by a standardreduction heat-treatment process. This process canbe conducted in 30 to 60 hours on pieces of roughsapphire sufficient to cut 2 ct faceted stones.Given the efficacy of this two-step processing, it isrecommended that high-value blue sapphires alsobe tested for Be-diffusion treatment.

Absorption Spectra. Figure 31 shows the differen-tial absorption spectrum of a Be-diffused high-puri-ty synthetic sapphire (sample SS8 in table 5) and asimilar spectrum of an orange Madagascar sapphirethat was Be diffused in Thailand from originallypink material (sample TD-1). While these two spec-tra both show strong absorption in the blue region,there are significant differences. The absorptionpeak of the SS8 synthetic sapphire is centered at420 nm, while that of the natural sapphire is at

460 nm. In addition, the SS8 material exhibits aweak broad band centered at about 700 nm. Togain some insight into these differences, we dif-fused beryllium into two synthetic rubies (SS3 andSS18) that contained different levels of chromiumbut otherwise were of very high purity. The differ-ential absorption spectra of both were essentiallythe same (see, e.g., figure 31).

Comparing the Be-diffused SS3 differentialspectrum with that of the Be-diffused natural sap-phire (TD-1), we note that both show peaks at 380nm and 460 nm, and neither shows the 700 nmpeak. The 460 nm peaks are essentially identical,while the 380 peaks exhibit different amplitudes.It thus appears that the spectral differencesbetween the SS8 synthetic sapphire and the sam-ple TD-1 natural sapphire are associated with thepresence of chromium. In high-purity Be-contain-ing synthetic sapphire, the trapped hole is associ-ated with the beryllium ion. However, in sapphirecontaining chromium in addition to beryllium,the Be-induced trapped hole will preferentially


Figure 28. Significant color alteration was also seen in this parcel of rough sapphire from King’s Plain, NSW,Australia (average 0.3–0.5 ct), shown here before (left) and after (right) Be-diffusion treatment. Photos by J. Emmett.

Figure 29. Rough sapphire from Ilakaka, Madagascar (average 0.4–1.0 ct), also showed dramatic color differences,as evident in this material before (left) and after (right) Be-diffusion treatment. Photos by J. Emmett.

associate with the Cr3+ ion (again, see Box B).Thus, the differential absorption spectrum in thiscase is that of a trapped hole near a Cr3+ ion, notthat of a trapped hole near an Mg2+ ion, a resultwe anticipated in the “Color in Corundum” sec-tion. There are two possibilities for the origin ofthis spectrum. First, we could be observing thespectrum of a Cr3+-trapped hole pair. Second, wecould be observing the spectrum of Cr4+ resultingfrom combining the hole and the Cr3+ ion.Determining which of these possibilities is cor-rect will require significant additional research.

In sapphire that does not contain chromium, butdoes contain Fe3+, beryllium-induced trapped holeswill preferentially associate with Fe3+ rather than

Be2+. The resulting absorption spectrum again canbe a Fe3+-trapped hole pair, or the spectrum of Fe4+.We would expect the absorption spectrum from thehole association with Cr3+ to differ somewhat fromthe spectrum of the association with Fe3+. However,all the natural sapphires for which we have bothSIMS data and careful differential spectroscopy con-tain chromium at ≥10 ppma. Thus it is understand-able why all exhibit the 460 nm absorption if it isindeed associated with a Cr3+-hole pair. To test thishypothesis, we need to diffuse Be into sapphire con-taining iron, but not containing chromium at morethan a few ppma. There are two possibilities forobtaining such material. First, it may be possible tofind such stones among the “colorless” heat-treatedSri Lanka geuda. Second, a high-purity iron-dopedsynthetic sapphire could be grown. We are workingwith a number of colleagues to pursue bothapproaches. Be diffusing such material should allowus to unambiguously observe the Fe3+-hole pairspectrum, and thus resolve remaining color issues.

GEMOLOGICAL TESTINGAND IDENTIFICATIONIt is essential that the role of the diffused Be incorundum be understood before any identificationtechniques, whether simple or complex, are pro-posed. The foregoing provides that foundation. Theexperiments described below were conducted tocharacterize Be diffusion–treated ruby and sapphire,and to provide possible identification criteria.

Materials and MethodsSamples. The number of samples (898) examinedfor the gemological identification portion of thisstudy is necessarily quite large, to examine asmany of the effects of the Be-diffusion process as


Figure 31. This figure shows the beryllium-induced absorption from sample TD-1 (a Thai Be-diffused pink Madagascar sapphire), sampleSS3 (a Be-diffused synthetic ruby), and sample SS8(a high-purity synthetic sapphire grown by theheat-exchanger method). The absorption spectrumof the Cr-containing samples differs significantlyfrom that of SS8, because the beryllium-inducedtrapped hole interacts with the chromium.

Figure 30. This thin wafer (0.9 × 16 mm) was cut from an extremely dark blue sapphire crystal from Shandong, China.It is illuminated from below so that the color is apparent. The wafer in its untreated state is on the far left (A). After Bediffusion in an oxygen atmosphere (B), most of the blue color has been removed or greatly lightened and a smallamount of yellow color has appeared. Subsequent standard reduction heat treatment removed the yellow color andslightly increased the saturation of the blue (C). The blue color has now effectively been lightened by this process.Photos by J. Emmett.

A B C

possible, and yet there are stones from some geographic areas, and in some color combinations,that remain to be studied.

From the 898 samples, 285 were known to beberyllium diffused, 178 were used for before-and-after experiments, and 435 were not diffused.Approximately half were faceted. The known-dif-fused samples were either acquired as such direct-ly from treaters or dealers in Thailand, wereproved to be Be diffused through chemical analysisor characteristic color zoning, or were treated byus. These samples were yellow, pink to orange tored, blue, purple, green, and colorless; they camefrom Madagascar (Ilakaka and unspecifieddeposits), Montana, Tanzania (Songea and Umba),and unknown localities.

The samples that were obtained for the before-and-after experiments were either colorless orshowed the following colors before treatment: lightyellow, yellowish orange, pink, brownish red, blue,purple, green, brown, and gray, with some multi-colored. These samples were sourced fromAustralia (Lava Plains), Laos, Montana, Madagascar(Ilakaka, northern, and unspecified deposits), SriLanka, Tanzania (Songea and Umba), Thailand(Kanchanaburi and unspecified deposits), Vietnam(Lam Dong, Quy Chau, and unspecified deposits),and unknown localities.

The samples that were not Be diffused werecolorless to yellow to orange, pink to purple tored, blue to green, brown, gray, and mixed palecolors; both natural-color and standard heat-treat-ed stones were represented. Specific informationon the sample population and the tests to whichthey were subjected can be found in the Be-diffu-sion samples table in the G&G Data Depository(http://www.gia.edu/gemsandgemology, click on“G&G Data Depository”).

Standard Gemological Testing. Routine gem test-ing was performed on a portion of the samplesexamined for this study, although not all sampleswere studied by all methods (again, see the Be-dif-fusion samples table in the G&G DataDepository). The tests included determination ofrefractive index, specific gravity by the hydrostaticmethod, and pleochroism, as well as observingabsorption spectra from desk-model spectroscopesand the reaction to ultraviolet radiation.

Since color distribution was anticipated to bethe most important identification criterion, weexamined virtually all samples with a gemological

microscope, using various types of lighting andimmersion techniques.

Our equipment included a horizontal binocularmicroscope configured with an immersion cell (fig-ure 32), and upright microscopes with both a stan-dard immersion cell and specially prepared,grooved, translucent white plastic trays. In allcases, the samples were immersed in methyleneiodide (di-iodomethane, R.I.=1.74). (Other immer-sion fluids, such as water or baby oil, may be used,but their lower refractive indices render them lesseffective.) Illumination of the sample with diffusedwhite light greatly facilitated the observation ofcolor gradients. Low magnification (10×) was usedto search for surface-conformal color zoning,whereas higher magnification (up to 63×) was usedto examine internal and surface features.

UV-Vis-NIR Spectroscopy. The UV-Vis-NIR spec-tra at room and liquid-nitrogen temperatures wererecorded for the range 190–900 nm using UnicamUV500 and Hitachi U-2000 instruments. Spectra fornatural, heated, and Be diffusion–treated stones—both faceted and in the form of polished wafers—were recorded in the direction of the o-ray (E⊥c).

Infrared Spectroscopy. The infrared spectra ofselected samples (including natural, heated, andBe-diffusion treated of various colors) were record-ed in the range of 6000–400 cm−1 using Fourier-transform infrared (FTIR) spectrometers (Nicolet560 and 760, Nexus 670, and Avitar) and variousaccessories (beam condenser, diffuse reflectance


Figure 32. This Be-diffused sapphire is shown in animmersion cell filled with methylene iodide; a diffus-er plate was placed between the light source and thecell. In conjunction with a horizontal microscope, thiswas one of the methods used to examine the stones inthis study. Photo by Ken Scarratt.


[DRIFTS], and attenuated total reflectance [ATR]).We employed an MCT-B detector and a KBr beamsplitter, using various resolutions (from 0.5 to 4.0cm−1), to observe the O-H region and other spec-tral areas of interest.

Raman Spectroscopy. Raman spectra were gath-ered using a laser Raman microspectrometer(Renishaw 1000 system) at both room and liquidnitrogen temperatures, with an argon-ion laser pro-viding excitation at 514.5 nm. We used this tech-nique to determine whether any Raman peakswere induced or changed as a result of Be diffusion,as well as to identify surface-reaching fluxes, syn-thetic overgrowths, and crystal inclusions.

Chemistry. Chemical analysis by energy-dispersiveX-ray fluorescence (EDXRF) was performed on an

EDAX DX 95 and a Spectrace QuanX. The fullrange of elements within the detection limits ofthese instruments was surveyed (from sodiumthrough the heavier elements). This technique wasprimarily used early in the study of these new dif-fused materials to see if there were any elementspeculiar to this treatment that were detectable bythese instruments (see, e.g., McClure et al., 2002).This did not prove to be the case, showing that thistesting technique is not useful for this purpose.

To search for light elements in the diffusedlayer, we also used SIMS analysis at Evans East(again, see Box A). Since February 2002, we havesubmitted 62 rubies and sapphires to SIMS analysiswith 153 separate analyses. Samples were selectedfrom stones known to have been treated inThailand by beryllium diffusion; stones treated tospecific parameters by two of the authors; stonesthat were known to be untreated; and stones thatwere submitted by clients for examination and lab-oratory reports.

Results and DiscussionVisual Appearance. We observed that many ofthe colors produced by Be-diffusion treatment arevery saturated, particularly those in the pink-orange, yellow, and orange ranges (again, see fig-ure 1). These colors have never been common innatural corundum, even among typical heat-treat-ed sapphires. The availability of large quantities ofsuch colors in one place, as several of the authorshave seen in Thailand, would immediately sug-gest that they have been treated by this new tech-nique. When a single stone is examined, however,color alone is not diagnostic, as natural and heat-treated sapphires and rubies exist in all the colorsseen in Be-diffused corundum (see Box C andchart figure C-31). Note, too, that not all sap-phires treated with Be diffusion produce such

Figure 33. Not all sapphires treated with berylliumhave an intense color, as this 14.38 ct Be-diffused sap-phire illustrates. It has a medium tone very similar tothat of some natural-color sapphires. It is interestingto note that the thin orange layer present in this stoneis visible even without immersion, which is not usual-ly the case. Photo by Maha Tannous.

Figure 34. A wafer cutfrom this 0.83 ct Viet-namese pink sapphire

(before treatment, left)was Be diffused for this

study (right). There was noobservable change in color

(the white material isresidual flux). Photos by

Elizabeth Schrader.


vibrant, saturated colors (figure 33). In fact, as dis-cussed in the “Color in Corundum” section, somestones do not change color at all, even thoughSIMS analysis shows that beryllium has penetrat-ed them (figure 34). In addition, as described earli-er, this new treatment can lessen the intensity ofcolor in some dark blue sapphires.

Optical and Physical Properties. The refractiveindex, birefringence, pleochroism, and specificgravity of our samples did not vary significantlyfrom those properties recorded for other types ofnatural or treated corundum. This is consistentwith the very low concentration of beryllium thathas been detected in the diffused stones (see table7). The Be diffusion–treated stones had a broadrange of reactions to UV radiation, from inert tostrong, and the fluorescence colors varied fromorange to reddish orange. This probably is due to

the wide variety of corundum being subjected tothis treatment. However, we did notice a some-what higher-than-expected occurrence of moder-ate-to-strong orange fluorescence to long-wave UVin the pink-orange stones. This reaction was by nomeans consistent, and in our experience it is notuncommon in non-diffused stones of the samecolor. Consequently, it cannot be considered a use-ful identification clue.

Most of the diffused stones we tested, particular-ly those in the yellow-to-orange range, showedweak-to-strong 450 nm iron-related absorption inthe desk-model spectroscope. All the samples witha pink or red component, as well as some orangestones, also showed Cr absorption in the red regionof the spectrum (692 and 694 nm). The Fe-relatedabsorption was stronger than expected for some ofthe colors tested. We believe this absorption is relat-ed to the starting material used for Be diffusion,

TABLE 7. Representative trace-element compositions of corundum by SIMS (ppma).a

Sample no. Color Notes Be Nab Mg Si Kb Ti Cr Fe Ga Geographicsource

Untreated

48718 Red 1.15 2.04 45.2 38.4 0.63 334 4300 6.55 62.9 Mong Hsu,Myanmar49081 Green 1.46 0.12 80.3 17.8 0.03 63.2 90.1 1461 16.1 Montana, U.S.49081 Orange 1.46 0.20 76.1 16.1 0.04 63.0 80.7 1358 13.0 Montana, U.S.49089 Yellow 1.48 3.96 35.8 17.7 0.17 29.1 4.61 2646 39.7 Unknown49089 Blue 1.52 0.16 9.20 18.2 0.03 120 2.57 1998 29.6 Unknown49102 Pink 1.42 0.10 28.5 40.3 0.07 742 1061 8.91 18.4 Vietnam

Before / after Be-diffusion treatment

49106 Yellowish Before 1.240.07 56.8 22.1 0.04 40.1 237 2496 20.7 Songea, Tanzaniabrown

49106T Orange After 17.4 0.54 43.9 15.4 0.41 28.6 222 1610 28.8 Songea, Tanzania45031 Pink Before 1.200.04 105 na 0.02 na 292 153 22.7 Unknown45002 Pinkish After 13.2 0.13 109 na 0.09 na 322 145 25.4 Unknown

orange45494 Pink Before 1.440.62 167 na 0.12 121 928 158 51.1 Unknown45493 Pinkish After 18.0 2.20 134 na 0.67 89.9 781 115 33.6 Unknown

orange

Be diffusion–treated samples

45033 Orangy Center 16.4 0.28 76.5 na 0.18 na 712 1056 19.0 Unknownred

45033 Orangy Edge 19.2 0.21 78.5 na 0.14 na 707 1083 19.5 Unknownred

45490 Yellow Center 3.14 0.03 11.8 na 0.01 15.8 3.93 1509 46.9 Unknown45490 Yellow Edge 24.5 0.04 5.20 na 0.01 20.0 3.91 1490 33.8 Unknown48415 Red 25.1 0.43 64.0 na 0.24 47.4 4118 2441 53.8 Unknown

ana = not analyzed. Accuracy of Be and Fe is ±10–20%; Ga, Mg, Si, Ti, Cr, Na, and K is about ±30%. A complete listing of theanalyses performed for this study can be found in the G&G Data Depository (www.gia.edu/GemsandGemology, click on “G&G Data Depository”).bHigh concentrations of Na and K in some analyses could be due to surface contamination.

rather than to the treatment itself. It must beemphasized, however, that this observation is in noway diagnostic of Be diffusion, as many natural yel-low or orange sapphires show similar absorption.

Color Distribution. Beryllium diffusion came to theattention of gemologists and the trade after yellow-to-orange three-dimensional surface-conformal lay-ers were seen on faceted orangy pink padparadscha-like sapphires (McClure et al., 2002; Scarratt2002a,b). We have never observed—either in thecourse of this study or in our many years of experi-ence—such a three-dimensional layer of color paral-

leling the surface of a natural-color polished gem-stone. To the best of our knowledge, it can only becaused by a diffusion treatment.

Therefore, careful observation of color distribu-tion became critical to this investigation. Exami-nation of the Be-diffused samples in our study withmagnification and immersion revealed that all of


Figure 35. A wide variation in the penetration depths ofsurface-conformal color layers can be seen in this groupof faceted Be-diffused sapphires (0.51–1.92 ct), immersedin methylene iodide. Photo by Maha Tannous.

Figure 36. The vast majority of predominantly red Be-diffused samples we examined showed a relativelythin, surface-conformal layer of orange color. The layerin this 5.05 ct stone is somewhat uneven, which sug-gests that it was recut after treatment. Photomicro-graph by Shane F. McClure; immersion, magnified 10×.

Figure 37. Many different combinations of Be-diffusedcolors are possible. Some, as in this 1.51 ct purple andorange briolette with an orange surface-conformallayer, are very unusual. Photomicrograph by Shane F.McClure; immersion, magnified 10×.

Figure 38. Natural color zoning in corundum does notconform to the faceted shape of the stone. Instead, itfollows the structure of the corundum crystal, creatingangular zones and parallel bands, as can be seen in thisgroup of non-diffused rubies and sapphires (0.14–1.29ct). Photo by Maha Tannous; in immersion.


the fashioned orangy pink and pinkish orange sam-ples showed color zones that related directly to thefaceted shape of the stone (see, e.g., figure 35). Insome, this zone penetrated as much as 80% of thestone (again, see figure 17 right and chart figure C-1); those stones with the deepest penetrationappeared most orange. The vast majority of the pre-dominantly red faceted samples also showedorange surface-conformal color zones (figure 36).The few blue Be-diffused sapphires we have exam-ined all showed a thin, colorless surface-conformallayer overlaying typical blue hexagonal zoning.

However, most of the Be-diffused yellow andorange samples were colored throughout, with lessthan 10% showing a layer of surface-conformalcolor. Sapphires of other colors, such as green andpurple, typically revealed orange or yellow surface-conformal layers overlying purple, green, or blue

central colors (figure 37).The color distribution in an untreated sap-

phire, or even in a sapphire that has been heatedby standard methods, generally will follow thetrigonal crystal structure of corundum. Stones cutfrom such crystals may show angular or straightcolor banding that conforms to the corundum’scrystal morphology and related trace-element dis-tribution, rather than to the shape of the facetedstone (figure 38). In addition, some natural-colorstones are faceted to use isolated zones of color ina way that results in an unusual color distribution(e.g., a dark blue zone located in the culet of anotherwise colorless stone or, in the case of a pad-paradscha sapphire, an orange zone retained alongone edge to produce the optimum color).Although we have never observed a natural colorzone that conformed to the faceted surface of thepolished stone, a Be-diffused piece of pink or redrough also might be cut in such a way that theorange color component does not conform to thefacets of the finished stone, but rather merelyforms a concentration of orange, for example,along one or several edges (figure 39). The gemolo-gist must develop a keen sense for color distribu-tions that are not completely surface conformal,which will indicate that further testing (such asSIMS) is needed. Therefore, any near-surface con-centration of yellow or orange color should beviewed with suspicion.

In several of our samples, the surface-conformalcolor layer was present in conjunction with natu-ral angular or banded color zones; these even over-lapped one another in some cases. In suchinstances, the two types of zoning were easily dis-tinguished and the relationship of color to facetedshape proved that a color-causing element hadbeen diffused into the stone from an external

Figure 40. Occasionally Be-diffused corundum showsboth surface-conformal color layers and natural colorzones that correspond to the crystal structure, as seenin this 1.29 ct briolette. Photomicrograph by Shane F.McClure; immersion, magnified 10×.

Figure 39. A Be-diffused stone will not always have a complete surface-conformal layer of color, which could causeconfusion with natural color zoning. We had the diffused piece of rough on the left cut into the 1.74 ct stone in thecenter. As can be seen with immersion on the right, the orange diffusion layer was evident only as concentrationsjust under the surface along the edges of the table and one side of the pavilion, and not all the way around the stone.Left and center photos by Maha Tannous; photomicrograph on the right by Shane F. McClure, magnified 10×.


source (figure 40 and chart figure C-4).

Magnification. Inasmuch as standard gemologicaltests, other than the observation of a surface-confor-mal color layer, have not provided proof of Be diffu-sion, we believed that other evidence obtained frommicroscopic examination could be critical to identi-fication. We cannot stress strongly enough that thefindings represented here indicate changes in inclu-sions that have occurred due to exposure to extremeheat—heat higher than is typical for what we referto as standard heat treatment (i.e., above 1780°C).While actual temperature data from Thai treaters isunavailable, we have observed alterations in inclu-sions that we rarely see with standard heat treat-ment. Such evidence cannot unequivocally proveberyllium diffusion. However, these types of fea-tures, considered together with the other evidenceoutlined in this article, can provide a strong indica-

tion that a stone has been Be-diffusion treated.

Included Crystals. Almost all of the Madagascarsamples contained small crystals of zircon, whichis very common in corundum from that locality(see, e.g., Schmetzer, 1999). In all the Be-diffusedsamples that contained zircon inclusions, wefound that those inclusions were always signifi-cantly altered. What were once transparent sub-hedral crystals (rounded grains, as shown in fig-ure 41) were transformed into opaque white irreg-ularly shaped masses, often with a crackledappearance and a gas bubble in the center (figure42 and chart figure C-7). We have seldom seenthe destruction of zircon inclusions in sapphiresthat have been subjected to typical heat treat-ments, regardless of their locality of origin. Thefact that all such inclusions in Be-diffused stonesare significantly altered is important. It meansthat the temperatures being used for Be diffusionare significantly higher than those used for stan-dard heat treatment. Therefore, if zircon crystalsappear to be unaltered, it is unlikely that the sap-phire host has been subjected to the high temper-atures typically used for Be diffusion (Buttermanand Foster, 1967).

Figure 43 shows the Raman spectra of (A) a zir-con inclusion inside a natural pink sapphire and (B)a zircon in the same sapphire after Be-diffusiontreatment. Spectrum A has strong Raman scatter-ing at 1018, 983, 441, 364, 224, and 200 cm−1. Inspectrum B, all these peaks are gone, which indi-cates that a phase transformation has occurred. It isvery likely that the peaks at 192 and 181 cm−1 arefrom ZrO2 (baddeleyite). None of the other peaksmatched any minerals in our database that couldcrystallize from the partial melt (such as quartz,tridymite, cristobalite, or mullite). These results

Figure 41. Undamaged zircon inclusions in Madagascarsapphires typically appear as small transparent sub-hedral (rounded) crystal grains. Photomicrograph byShane F. McClure; magnified 40×.

Figure 42. At high tempera-tures, zircon inclusions

become significantlyaltered. Typically, they are

transformed into opaqueirregular masses, often witha crackled appearance anda gas bubble in the middle.

Photomicrographs by KenScarratt; magnified 40×.

indicate that the partial melt was cooled rapidly,and the zircon inclusion eventually transformedinto ZrO2 and an Al2O3-SiO2-rich glass, followingthe Be-diffusion treatment. These findings are inagreement with those recently published byRankin and Edwards (2003), who provide an in-depth discussion of what happens to zircon crystalsat such high temperatures (see also Butterman andFoster, 1967).

Note that the melt created by these highlyaltered inclusions often flows into the surround-ing strain haloes. As it cools within the halo, itforms the phases mentioned above and createscrossing fern-like structures that are reminiscentof the structure of devitrified glass (see chart fig-ure C-10). These recrystallized haloes were seenin many of the Be-diffused stones, but they wereparticularly common in the orange, yellow, andpink-orange samples. A Raman spectrum of thesubstance within the halo exactly matched spec-trum B in figure 43, indicating that it, too, wascaused by the destruction of zircon.

Other types of crystals were seen in many of thesamples tested, but in all cases they were damagedor significantly altered by the treatment, and inmost cases they were visually unidentifiable,shapeless white masses (see chart figure C-28). Wealso saw numerous instances of large gas bubblesthat were trapped in internal cavities filled withwhat appeared to be a transparent glass (figure 44).

Internal Diffusion. We saw evidence of internal lat-

tice diffusion (i.e., the diffusion of a color-causingagent from an inclusion into the surrounding stone;Koivula, 1987) in many of our study samples, par-ticularly in the orange, yellow, and red colors. Thisphenomenon generally occurs in association withrutile (TiO2), and is extremely common in blue sap-phires that have undergone standard heat treat-ment. At high temperature, rutile inclusions breakdown and release Ti, which diffuses into the struc-ture of the corundum, reacts with existing Fe, andproduces a localized blue coloration (see chart fig-ures C-17 and C-18).

In many of the untreated Songea rubies andsapphires we examined, the rutile inclusionswere in the form of euhedral crystals. In untreat-ed stones from Madagascar, they appeared most-ly as needles or clouds of fine, dust-like parti-cles, as are commonly seen in corundum sam-ples from many other localities. However, all ofthe Be-diffused stones from Songea in this studythat contained rutile crystals showed a typicalinternal diffusion pattern consisting of a hexago-nal or irregularly shaped crystal surrounded by aspherical blue halo (figure 45). The blue colorcontrasted sharply with the yellow, orange, ororangy red bodycolor of the stones. Some of theareas of spotty blue coloration created by rutilesilk in Be-diffused stones from Madagascar stillcontained a visible pinpoint inclusion (chart fig-ure C-19).

The important observation is that although inter-nal diffusion is quite common in heat-treated bluesapphire, it is not at all common in yellow, orange,and red corundum. While it is true that yellow sapphire from some localities (e.g., Montana) is


Figure 43. Shown here are the Raman spectra of (A) azircon inclusion inside a natural pink sapphire, and(B) a similar inclusion in the same sapphire after Be-diffusion treatment. Due to partial melting and reac-tion with the surrounding corundum, the zircon in theBe-diffused sample was transformed to ZrO2 (badde-leyite) and an Al2O3-SiO2-rich glass.

Figure 44. These former mineral inclusions werereduced to a glass-like melt containing gas bubblesafter the host sapphire was Be-diffusion treated.Photomicrograph by Ken Scarratt; magnified 40×.


The most critical issue for gemologists and the gem tradeis how to separate these treated stones from natural-colorcorundum and corundum that has been heated by “stan-dard” methods, without the addition of external ele-ments. Although our research has shown that all the Be-diffused stones can be identified by mass spectroscopy,especially SIMS analysis, this technique is expensive andtime consuming. However, we have found many other,more readily obtainable clues that provide either proof orstrong indication of Be-diffusion treatment.

This box and the accompanying chart summarizethe clues a gemologist can use to help determine if asapphire or ruby has—or has not—been subjected to Bediffusion. While we believe these indicators to be com-mon to all colors of Be-diffused corundum, to date wehave not examined a sufficient number of Be-diffusedblue sapphires to draw general conclusions about them.Observations requiring magnification are best seen witha gemological microscope. All numbers in parenthesesrefer to images on the chart.

DIAGNOSTIC EVIDENCEColor Zoning – The only diagnostic evidence of Be-diffu-sion treatment immediately available to the gemologistis the presence of a color zone that conforms to the exter-nal faceted shape of the stone. This color zone may beseen in all types of Be diffusion–treated corundum andcan penetrate to any depth, including throughout thegem (figure C-1). The Be-diffused conformal zone is yel-low, orange, or occasionally colorless (C-2). It surroundscentral cores that are either colorless or some other color,most often pink or red (C-3). It also may be seen to over-lay natural zoning (C-4). Observation of this featurerequires the use of immersion, with methylene iodide(what gemologists call 3.32 liquid) the best medium.

HIGHLY INDICATIVE EVIDENCEThe features in this section do not prove that a stone hasbeen Be-diffusion treated. However, they do prove that itwas subjected to very high temperatures (i.e., higher thanthose typically used for the standard heat treatment ofcorundum) and are a strong indication that Be-diffusiontreatment may have occurred. Therefore, observation ofone or more of the following features in a ruby or sap-phire means that the color origin is highly suspect.

Highly Altered Zircon Inclusions – Zircon crystals incorundum usually survive standard heat treatment,retaining their initial appearance as small transparentangular or rounded grains (C-5). The extreme alteration ofsuch crystals to white, formless masses (C-6)—often con-taining a gas bubble (C-7)—is an excellent indicator thatthe stone has been exposed to very high temperatures.

Internal Recrystallization – Several forms of internal re-crystallization occur as a result of the extreme tempera-

tures used in Be-diffusion treatment. One of theserelates to the zircon inclusions mentioned above.Discoid fractures around zircon crystals are commonlycreated during normal heat treatment as the crystalexpands, even when the inclusion itself is not damaged(C-8); the circular fractures around zircons and othercrystals are often glassy and reflective (C-9). Such frac-tures also often occur around zircon crystals during Be-diffusion treatment. However, as the zircon is destroyed,the resulting fluid melts and flows into the planar frac-tures of the discoid. Then, as the stone cools, the meltpartially recrystallizes, leaving a characteristic appear-ance that we had not seen in sapphires prior to this newtreatment: a plane of transparent material interwovenwith fern-like patterns reminiscent of the structure seenin devitrified glass (C-10), occasionally displaying irides-cent colors (C-11). The discoids also may contain areashealed by the regrowth of corundum (C-12), and neartheir center is always a white mass (usually with a gasbubble) that used to be a zircon crystal (C-13).

The second type of recrystallization involves themineral boehmite. Recrystallized corundum in thechannels left behind by destroyed boehmite inclusionsoften has a characteristic “roiled” appearance (C-14),particularly in transmitted light. These now-filled chan-nels often contain voids that may be spherical (C-15) orhave a bone-like shape (C-16).

Internal Diffusion – Internal lattice diffusion aroundrutile crystals and “silk” is very common in heat-treatedblue sapphires. This typically manifests itself as roughlyspherical blue haloes around the crystals (C-17) and spottyblue coloration where the rutile “silk” used to be (C-18).In our experience, however, this phenomenon is extreme-ly rare in rubies and yellow-to-orange sapphires (C-19)since these stones are heated in an oxidizing atmospherethat usually removes blue color. Its presence in these col-ors can be considered a direct result of exposure to hightemperatures. However, since yellow sapphires fromsome localities (such as Montana) are routinely heated—but not diffused—at temperatures high enough to causesome spotty blue coloration, these characteristics indicatethe possibility—but not proof—of Be diffusion.

Synthetic Overgrowth – Synthetic corundum can grow inany high-temperature treatment environment. However,the appearance of synthetic overgrowth on faceted Be-dif-fused rubies and sapphires is different from any we haveseen before. Specifically, remnants of synthetic over-growth occasionally have been seen in surface-reachingcavities as tiny platelets, but not covering entire facets.Although this synthetic material is often removed duringfinishing, portions remain on the surfaces of some stones,particularly those that are yellow to orangy red, providingan important clue to the use of high heat.

The synthetic crystals that grow during Be-diffusiontreatment are often randomly oriented, tend to be platy

BOX C: A PRACTICAL GUIDE TO RECOGNIZING BERYLLIUM-DIFFUSED RUBIES AND SAPPHIRES


and hexagonal in shape (C-20), and are much larger andfar more numerous than any typically found on stonesheat treated by standard methods. When they growtogether, they form a solid layer that often looks like anaggregation of irregular blocks (C-21), sometimes withtiny voids between them and cloud-like inclusionsalong the crystal interfaces (C-22). This tends to create aroiled appearance in the surface layer (C-23). While thisusually can be seen with darkfield illumination, it iseven more readily visible with transmitted light (C-24).Cross-polarized light also helps reveal such features, dueto the randomness of the crystal growth. Since the ori-entation of many of the synthetic crystals is differentfrom that of the natural host, in crossed polarizers theymay appear light while the host material is dark (C-25).

We have recently noted increased synthetic growthon the surfaces of a number of Mong Hsu rubies, some-times lining cavities and sometimes having the sameappearance as that found on Be-diffused stones (C-26).The cavities exhibited linings of tiny hexagonal plateletsand were often filled with glass residues. This indicatesthat some of these stones are now being treated at highertemperatures than were typically used before. However,we have not seen this type of growth in yellow andorange heat-treated sapphire that has not been subjectedto Be-diffusion treatment.

Other Inclusions – Other unidentifiable severely dam-aged crystals that do not appear to have been zirconshave been noted in some Be-diffused corundum (C-27).Although the original form of these white, shapelessmasses is unrecognizable, the sheer amount of damage(C-28) is clear evidence of exposure to extreme tempera-tures. This is also true of internal cavities that are nowfilled with a transparent glass-like material containingspherical gas bubbles (C-29).

EVIDENCE THAT IS NOT INDICATIVE Some features of Be-diffused corundum overlap toomuch with corundum that has not been treated by thisprocess to help in identification, although they may sug-gest that a stone is suspect.

Color – The colors produced by Be diffusion are oftenvery saturated and “unnatural” (C-30). Yet all of the col-ors produced by Be diffusion can occur naturally or as aresult of standard heat treatment (C-31). Thus, for anyindividual stone, color by itself does not indicate Be dif-fusion, but it can be a reason for further testing.

Be diffusion produces colors in rubies that are oftena slightly orangy red, or occasionally even pure red (fig-ure C-32). These colors mimic many iron-rich rubiesthat are found in Africa and elsewhere.

Healed Fractures – Artificial healing of fractures clearlyoccurs during the Be-diffusion process (C-33). However,it also commonly occurs during the heating of rubies,particularly from Mong Hsu. Therefore, it cannot beused to identify the presence of Be diffusion.

Dissolution – Dissolution of the surface is very commonon Be-diffused stones. Surfaces that have not been repol-ished often have a melted appearance (C-34). However,this is a common feature on most corundums that havebeen heated at high temperatures, particularly whenfluxes are used, so it is not distinctive of Be diffusion.

INDICATIONS OF NO EXPOSURE TO VERY HIGH TEMPERATURESThe presence of these inclusions proves that the stonehas not been exposed to high heat and therefore couldnot be Be diffused.

CO2 Inclusions – Internal “voids” that contain water anda bubble of carbon dioxide (CO2) are quite common insapphires from some localities, particularly Sri Lanka (C-35). Because CO2 expands when heated, these inclusionscannot survive the very high temperatures necessary forBe diffusion. Therefore, the presence of undamagedinclusions would prove that a stone has not been Be-dif-fusion treated.

Internal voids filled with other liquids will also notsurvive high-temperature heat treatment. The presenceof undamaged liquid-filled voids of any kind proves thatno Be treatment has occurred (C-36).

Included Crystals – As discussed above, the presence ofundamaged zircon crystals in a sapphire is a good indica-tion that Be diffusion treatment has not taken place.Likewise, in our experience, it is unlikely that any crys-talline inclusion could survive the temperaturesrequired for Be diffusion without being significantlyaltered. Therefore, the presence of transparent, angular,or rounded solid grains of any mineral would be anexcellent indication that Be diffusion has not takenplace (C-37 and 38).

Rutile Needles – Needle-like inclusions of rutile, oftenreferred to as “silk,” are common in corundum frommany localities. These needles usually survive thelower-temperature heat treatments that are performedon some sapphires. However, they typically do not sur-vive the higher-temperature treatments to which mostblue sapphires and rubies are subjected, including thevery high temperatures necessary for Be-diffusion treat-ment. The presence of unaltered rutile needles meansthat a stone has not been exposed to this (or any other)high-temperature treatment (C-39).

SEND IT TO A LABORATORYWhat if you have a sapphire that is colored throughoutand does not show any of the indications listed above?These are the most difficult situations. We recom-mend that you send it to a qualified gemological labo-ratory and have its chemistry analyzed by the tech-niques mentioned in this article. We will continue toinvestigate alternative methods to identify this treat-ment process.

routinely heated to temperatures high enough(without adding chemicals) to cause such spottyblue coloration, this phenomenon is seen onlyrarely in heat-treated stones. Its presence does notunequivocally prove beryllium diffusion, but itdoes provide solid evidence of exposure to veryhigh temperatures.

Flux-Assisted Synthetic Growth. During flux-assisted heat treatment (with or without diffusion),material from the surface of the corundum (as wellas the alumina crucible and the boehmite inclu-sions) will dissolve to some degree into the moltenflux (which may itself contain alumina). On cool-ing, the flux becomes supersaturated in aluminaand crystallizes out onto the nearest convenientsurface, often on the stones themselves (figure 46and chart figure C-20).

We noted synthetic corundum overgrowth(identified by Raman spectroscopy and micro-probe) on the surfaces of some of the yellow,orange, and orangy red Be diffusion–treated sam-ples in this study. The overgrowth was particularlyobvious on rough or preformed samples and oftenremained after the stones were recut followingprocessing. The extent of this overgrowth rangedfrom single small areas (usually around the girdle)to numerous large areas covering 10–20% of thestone’s surface. The typical appearance of thisovergrowth is depicted in figure 47 (McClure,2002a; Scarratt, 2002a; Moses et al., 2002). The“blocky” texture may be observed even at low-

power (10×) magnification with transmitted light,but it may be missed if only reflected light is usedto examine the specimen.

Pre-existing fractures in corundum—as well asthose created during Be diffusion—and boehmitechannels are often sealed, or “healed,” by the sameprocess. The molten flux enters the fracture orchannel, and the surfaces of the fracture or chan-nel dissolve into the flux, where the dissolvedmaterial is held in solution until the temperatureis reduced. On cooling, the alumina then crystal-lizes out of the flux onto the walls of the fractureor channel, sealing it while at the same time trap-ping residues such as gas bubbles, flux remains,and glass. This “flux-assisted healing” of fractureswas well documented in rubies even before theintroduction of the new Be-diffusion treatment


Figure 46. Large areas of platy synthetic corundumcrystals were present on the surface of some Be-dif-fused samples examined for this study. Photomicro-graph by Shane F. McClure; magnified 20×.

Figure 45. All the Be-diffused sapphires in this studythat contained rutile crystals exhibited blue internaldiffusion haloes around those crystals, which is veryunusual in yellow and orange sapphires. Photomicro-graph by Ken Scarratt; magnified 20×.

Figure 47. Blocky synthetic overgrowth produced dur-ing the Be-diffusion process was still present on thesurface of some treated stones, even after repolishing.Photomicrograph by Ken Scarratt; transmitted light,magnified 30×.

(Hänni, 2001).In the Be-diffused stones, the artificially healed

fractures (figure 48) often resembled the “finger-prints” or “feathers” seen in flux- or hydrothermal-ly grown synthetic rubies and sapphires. A “roiled”appearance could be seen within the healedboehmite channels, which were aligned with edgesof intersecting parting planes. These tubes oftencontained high-relief internal cavities that variedfrom spherical to a “dog bone” shape (see chart fig-ures C-15 and C-16).

These observations are not surprising, as theconditions and equipment used to heat treat or lat-tice diffuse elements into corundum closely resem-ble those used for corundum synthesis, except thatthe production of flux-grown synthetic corunduminvolves lower temperatures (around 1300°C) andthe use of platinum crucibles rather than aluminaones (Nassau, 1980).

UV-Vis-NIR Spectroscopy. As discussed earlier,UV-Vis-NIR spectroscopy was used to confirm thetrapped-hole nature of color in these Be-diffusedstones, but we have not found it useful as an identification method. We have examined the UVand visible-range absorption spectra of dozens ofsamples, both at room temperature and at liquid-nitrogen temperature, but have failed to find a Be-diffusion identification criterion. The difficulty isthat there appears to be little difference in thespectra of trapped holes caused by naturally occur-ring magnesium and those induced by berylliumdiffusion. This method is therefore of no use in theidentification of Be-diffusion treatment.

Infrared Spectroscopy. Much natural corundumcontains detectable quantities of hydrogen, whichcan be present in several forms (Rossman, 1988).Hydrogen can bond with oxygen in the lattice toform O-H, resulting in sharp absorption bands at3367, 3310, 3233, and 3185 cm−1 (Beran, 1991).Hydrogen can also exist in corundum as H2O, lead-ing to a broad absorption band at ~3400 cm−1. Itcan also be present as water within fluid inclu-sions, and, finally, it can exist as a component ofan included solid phase, such as diaspore (Smith,1995), which is one of the hydroxides of alu-minum. Since corundum is transparent from theUV region to 2000 cm−1 (5000 nm) in the infrared,even trace amounts of hydrogen are detectableusing infrared spectroscopy. Some of the variationsobserved in the IR spectra of natural sapphires are

shown in figure 49.During the course of this study, we collected IR

spectra on 50 Be-diffused sapphires. In none of thesespectra did we observe hydrogen-related features.This result was expected, since hydrogen rapidlydiffuses out of corundum at high temperatures (e.g.,1800°C) in a hydrogen-free atmosphere (i.e., the oxy-gen atmosphere used for Be diffusion). An exampleof this effect is illustrated in figure 50. The bluespectrum shows the O-H absorption of a naturalpink sapphire from Vietnam. After Be diffusion at


Figure 48. Healed fractures such as these were seen inmany of the Be-diffused corundum samples weexamined. They often appeared similar to thosefound in flux- or hydrothermally grown syntheticrubies and sapphires. Photomicrograph by Shane F.McClure; magnified 25×.

Figure 49. The O-H infrared absorptions in sapphiremay vary in appearance, as illustrated by the selectionof five spectra shown here. The top four spectra arederived from heat-treated blue sapphires, and the bot-tom spectrum is from an unheated yellow sapphire.

1800°C for 33 hours in an oxygen atmosphere (redspectrum), the O-H absorption features have entire-ly disappeared. Figure 51 shows the H2O absorptionband (Rossman, 1988) in a natural light yellow sap-phire of unknown origin (blue spectrum). After thesame Be diffusion parameters noted above, the H2Oabsorption features are absent (red spectrum).

As shown earlier, some types of dark bluebasaltic sapphire can be greatly lightened by a two-step process of Be diffusion followed by normalheat treatment in a reducing environment.Because the reduction heat treatment is carriedout in an atmosphere that contains hydrogen, thehydrogen will diffuse back into the crystal andappear in the absorption spectrum as O-H, withthe absorption features enumerated above. Thus,an attractive blue Be-diffused sapphire couldexhibit the O-H absorption features of a natural-color sapphire or one that has been heat treatedwithout diffusion. Consequently, the presence ofO-H absorption features in a blue stone is not evi-dence that it has not been Be diffused. This is onlyan issue in blue sapphire, since heat treatment of ayellow or orange stone in a reducing atmospherewill remove the Be-induced color.

Our data show that a ruby or a yellow, orange,or orange-pink sapphire that has undergone Be dif-fusion contains no structurally bonded hydrogenand little if any molecular water. However, thistells us only that the stones have been heated inan oxidizing atmosphere, the same atmosphere

required for Be-diffusion treatment. It does notnecessarily mean that Be diffusion is involved.Still, it can be a helpful piece of data when com-bined with other factors.

Raman Spectroscopy. No additional peaks werediscovered and no changes in the normal corun-dum spectrum were observed in specimens ana-lyzed before and after Be diffusion treatment.

Raman analysis also proved that the over-growth seen on some of the samples was indeed(synthetic) corundum. An unexpected result wasthe identification of a tiny grain of material on thesurface of one stone as synthetic cubic zirconia,the feed source of which may have been a zirconinclusion or a flux additive.

Chemistry. EDXRF spectra on stones that were notre-polished following Be diffusion consistentlyrevealed the presence of both Ca and Zr, elementsthat are not typically found in corundum. The Zrmay be related to exposed zircon inclusions or tozircons that were accidentally included in the heat-treatment process. The Ca may indicate that calci-um borate is a component of the flux mixture (see“Beryllium Diffusion into Corundum: Process andResults” section).

SIMS analysis readily identified the presence of


Figure 50. The infrared spectrum of a natural-colorpink sapphire from Vietnam containing hydrogen asO-H (blue) shows strong absorption bands at 3367,3310, 3233, and 3185 cm−1. After Be diffusion at1800°C for 33 hours in an oxidizing atmosphere, thesehydrogen-related bands were eliminated (red).

Figure 51. This infrared spectrum of a natural-coloryellow sapphire from an unknown source (blue) showsa broad absorption band centered at about 3300 to3400 cm−1 that is indicative of molecular water (H2O),possibly contained in the stone as an inclusion. AfterBe diffusion at 1800°C for 33 hours in an oxidizingatmosphere, this H2O-related absorption wasremoved (red).

beryllium at the surface of all the samples of Be-diffused rubies and sapphires analyzed. Table 7shows representative analyses2 of a variety of ourcorundum samples (with full analyses given in thedata depository SIMS table). When there is no Bein the stone, the SIMS analysis indicates about 1.5ppma because of interference from aluminumions. This sets the lower limit of what can bemeasured, and thus a measurement of 5 ppma ormore gives high confidence that a stone containsberyllium. The Be concentration for this processranges from approximately 10 to 35 ppma.Consequently, our data confirmed that very littleBe is needed to affect the color of corundum. Ifmeasurements are much below 10 ppma, thesample should be re-measured on another facet.

Overall, these analyses demonstrate a clear corre-lation between color modification and elevated Becontent.

Comparison to Ti Diffusion TreatmentThere are definite differences in diagnostic fea-tures between sapphires that have been diffusiontreated with titanium (see, e.g., Kane et al., 1990)and those that have been diffused with beryllium.In the earlier diffusion-treated stones, the color-causing elements (mainly Ti to produce blue) hadlimited penetration into the surface of the treatedstone (again, see figure 16). However, Be diffusesmuch faster than Ti (see Appendix), so the depthof the color layer produced can vary greatly, evenextending throughout the stone.

Ti-diffused corundum will typically showpatchy surface coloration with immersion (figure52), since the very shallow color layer is often cutthrough during repolishing. Immersion will alsoreveal color concentrations at facet junctions andhigh relief (figure 53), as well as “bleeding” of colorinto the stone from the surfaces of cavities, frac-tures, and fingerprints (figure 54). None of thesefeatures has been seen in stones that have been dif-fused with Be. In contrast, the layer of color con-forming to the outside shape of the stone is usuallynot visible in faceted Ti-diffused sapphires, evenwith immersion. However, this surface-conformallayer is easily visible—when present—in stones


2The very first analyses were conducted and published (McClure etal., 2002) without the benefit of a Be-in-sapphire analytical standardwith which to calibrate the SIMS equipment. Those analyses indicateda Be content roughly 10 times higher than we now know to be cor-rect. An accurate Be-in-sapphire standard was then prepared by oneof the authors (SN) and analyzed at Evans East, which allowed the pre-vious data to be recalculated (see, e.g., Wang and Green, 2002a).Although the precise concentrations changed for all the samplesreported, the ratio of beryllium content between known treated stonesand known natural-color stones remained the same. Thus, the originalcorrelation of elevated Be content to color modification in treatedstones was correct, though the absolute Be concentration was not.Since April 2002, all analyses have been run with the Be-in-sapphirestandard. All other elements in the table are also calibrated by ele-ment-in-sapphire standards.

Figure 52. Titanium-diffused sapphires often showpatchy surface coloration in immersion, caused by theshallow penetration depth of Ti. This is not seen in Be-diffused corundum. Photo by Shane F. McClure.

Figure 53. The facet junctions of this 1.77 ct Ti-dif-fused sapphire appear darker blue when seen inimmersion. This phenomenon is not seen in Be-dif-fused corundum. Photo by Shane F. McClure.

that have been Be diffused (figure 55).

WHERE DO WE GO FROM HERE?We have been investigating other nontraditionalmethods of Be detection with the goal of develop-ing a low-cost screening methodology for thesetreated stones. We have no solution yet, but out-line below a number of possibilities that are being

studied. We hope that this discussion will stimu-late gemologists around the world to focus theircreative talents on this problem.

Electrical Conductivity. Magnesium at 15–40ppma in corundum gives rise to measurable electri-cal conductivity as the temperature is raised mod-erately above room temperature (Ramírez et al.,2001; Tardío et al., 2001a,b,c). A similar effectshould be produced by beryllium. The trappedholes induced by Mg2+ (and potentially by Be2+),which are responsible for color formation in corun-dum, also give rise to electrical conductivity whenthe stones are heated. One of the authors (OB) per-formed initial experiments on a Be-diffused cube ofCzochralski-grown sapphire using electrically con-ductive epoxy for electrodes. Be-diffused syntheticsapphire was chosen to eliminate possible interfer-ence with other impurities. Measurable conductivi-ty was achieved at about 160°C over a range of 500to 3,000 volts, similar to what was observed byTardío et al. (2001a,c).

Measurements were then made on two naturalfaceted sapphires—a 1.6 ct Be-diffused yellow, anda 5.46 ct heat-treated but not Be-diffused yellow.The Be-diffused sample showed conductivity oftwo to three orders of magnitude higher than thenon-Be-diffused sapphire, providing some indica-tion that this method has potential.

There are many issues that will have to beaddressed before the value of this technique can bedetermined. It will be necessary to prove that theconductivity measured is a bulk conductivity andnot surface conductivity that would only be depen-dent on stone finish and the size of the stone. Apractical method of applying electrical contactswill have to be developed. A significant amount ofdata will have to be acquired on the effects ofgeometry and beryllium concentration to provethat Be can be identified independently from natu-rally occurring Mg, and to delineate regions ofvalidity and regions of uncertainty. In principle,though, it is a technique that could be automated.

Low-Temperature Heating. During discussions ofthe Be-diffusion problem with Thomas R.Anthony of General Electric Co. (pers. comm.,2002), he suggested that the Be-induced color mod-ification may not be stable at temperatures in therange of 400°–600°C. It is known that the absorp-tion due to the Fe2+-Ti4+ pair is substantiallyreduced at temperatures of 500°C and above, and


Figure 54. “Bleeding” of color into surface-reachingfractures and cavities is often visible in sapphires thathave been diffused with Ti. This was not seen in anyof the Be-diffused samples examined in this study.Photomicrograph by John Koivula; magnified 15×.

Figure 55. The actual surface-conformal color zonethat is produced by Ti diffusion is usually not visi-ble, even in immersion, as seen with the stone onthe right in this image. This is distinctly differentfrom the Be-diffused stones, in which the surface-related color layer (if present) is often easily seenwith immersion (left). Photomicrograph by ShaneF. McClure; magnified 10×.

the Be-induced trapped-hole coloration couldexhibit similar, useful color alterations with tem-perature. These “thermochromic” effects are acommon property of minerals. With increasingtemperature, absorption bands become broader andshift to a lower energy (i.e., higher wavelength).

The Be trapped-hole center causes selectiveabsorption from 600 nm to the high-energy side, pro-ducing additional yellow coloration. However, colorappearance varies depending on the color centerinvolved, so this yellow color differs in tone fromthat typically caused by Fe in natural corundum.Because of the different natures of these two absorb-ing centers (Burns, 1993), they could react differentlyto heat. Therefore, heating Be-diffused sapphires andnatural sapphires with similar (or even identical) col-ors could reveal a temperature range within whichthe colors of treated and untreated corundum tem-porarily become very different, possibly yielding auseful method of identification.

To test this idea, a natural colorless corundumwas Be diffused at 1800°C for 33 hours at a highoxygen partial pressure. It turned intense yellowdue to formation of Be trapped-hole centers. It wasthen heated to 600°C with three other natural yel-low sapphires colored by Fe. The Be-diffused yellowsapphire began to exhibit a strong brown colorationstarting at about 400°C (figure 56). Figure 57 showsthe changes in its absorption spectrum over therange from room temperature to 600°C. The shiftin coloration from yellow to brown is a result ofthe increasing absorption in the red region of thespectrum. The original color was restored aftercooling. The three naturally Fe-colored sapphiresshowed no visual changes in color when heated.

More than 30 pinkish orange Be-diffused sap-


Figure 56. These photos show a Be-diffused sapphire onthe left (0.95 ct) and three non-diffused (Fe-colored)sapphires on the right (0.28–0.33 ct). Note that at about400°C, the Be-diffused sapphire took on a distinctbrown coloration, whereas the natural sapphires didnot show any change. Photos by Wuyi Wang.

Figure 57. The absorption spectrum of the Be-dif-fused yellow sapphire in figure 56 changed as thetemperature increased. The increased absorption inthe red at high temperature is responsible for thebrown coloration.

Be-diffused Natural

25ÞC

100ÞC

200ÞC

300ÞC

400ÞC

500ÞC

600ÞC

phires also were tested for visual color change onheating, and all exhibited a strong brown col-oration at ~500°C. Hughes (1987b, 1988) notedthat heat-treated yellow sapphires from Sri Lanka(Mg2+ trapped-hole coloration) shift to a darker,browner color when heated to moderate tempera-tures. Thus it remains to be determined if thisheating process can distinguish the yellow Be-induced coloration from the yellow Mg-inducedcoloration of heat-treated yellow sapphire.

It is important to note that this procedure alsorequires several caveats, and may not apply acrossthe full range of colors and types of corundumbeing treated by the new diffusion process.However, if applicable, it may provide a means torapidly identify large groups of treated sapphires—both loose and mounted in jewelry, since the tem-perature range being used would not affect the met-als used in jewelry.

Luminescence to Ultraviolet Radiation. Recentwork outlined in Fritsch et al. (2003) has noted dif-ferences in luminescence to ultraviolet radiationthat may provide helpful indications for identifyingBe-diffusion treatment, particularly using a special-ized microscope for viewing this property that wasdesigned by Franck Notari of GemTechLab inGeneva. Junko Shida, of the GemmologicalAssociation of All Japan, recently presented usefulfluorescence observations to members of theLaboratory Manual Harmonization Committeethat she and her staff had made when examiningBe-diffused corundum using laser tomography.They found that, with this technique, the Be-dif-fused areas fluoresced a distinctive orange (J. Shida,pers. comm., 2003).

DURABILITY AND STABILITYOF THE TREATMENTOne of the first questions most jewelers andgemologists ask about a gemstone enhancementis: Is the process permanent and stable when sub-jected to typical jewelry cleaning and repair proce-dures? To address this question, we tested threefaceted Be diffusion–treated orange to pinkishorange sapphires (all of which had been treated inThailand) to determine the durability and stabilityof this process under a wide range of conditions towhich sapphires might be subjected during jewelrymanufacturing or routine cleaning and repair. Thesamples used were: (1) a 1.12 ct medium dark

pinkish orange oval mixed-cut, (2) a 0.43 ct darkorange oval mixed-cut, and (3) a 0.83 ct intensereddish orange princess cut. All three sapphireswere mounted in 14k gold rings (the first in a four-prong white head, the second in a six-prong yellowhead, and the third in a four-prong yellow head)before the durability testing began. Details on thetests used and the results observed are provided inthe durability table of the G&G Data Depository(http://www.gia.edu/GemsandGemology).

Given the low concentrations of Be required tochange the color of corundum, we did not expectthese diffused stones to behave any differentlyfrom natural-color or heat-treated samples.Indeed, Be-diffusion treatment did not appear tohave any impact on the durability or stability ofthe three samples tested when exposed to routinecleaning and jewelry repair procedures. Neitherultrasonic nor steam cleaning produced any dam-age in the treated stones, which were examinedwith magnification both before and after thecleaning procedure.

In addition, these three Be-diffused stonesshowed little or no surface etching as a result ofretipping. However, it is well known that the useof borax-containing chemicals (both fire coat andflux) contributes to moderate to severe surfaceetching of corundum, so other researchers mayfind different results using a different flux or high-er temperatures. The best way to avoid this poten-tial problem is to unmount the stone for anyrepair procedure that would expose it to heat.

If etching due to exposure to borax-containingcompounds is severe enough, the stone mighthave to be repolished. Of particular importancefor diffusion-treated stones with shallow colorlayers is the fact that repolishing or recuttingcould remove some or all of the diffused layer,resulting in a significant change in the appearanceof the stone (figure 58). This precaution is espe-cially pertinent to Be diffusion–treated stones inthe orangy pink color range, which may have ashallow color penetration. Therefore, we recom-mend great caution in repolishing or recutting Bediffusion–treated stones.

LABORATORY NOMENCLATURE ANDTRADE DISCLOSURE RECOMMENDATIONSAs with other gem treatments, a clear declarationof the lattice diffusion process is necessary,whether Be or Ti is involved, to insure trade stabil-


ity and to maintain consumer confidence in gems.In the case of the two gemological laboratories rep-resented by authors of this article, GIA and AGTA,laboratory reports indicate the exposure of a gemto a treatment process such as lattice diffusion,and note if synthetic corundum is present on thesurface. This is consistent with the nomenclatureadopted by the Laboratory Manual HarmonizationCommittee (LMHC), of which our two labs aremembers. For lattice diffusion, the LMHC conclu-sion states that the stone shows: “. . . indicationsof heating, color induced by lattice diffusion froman external source.” For the flux-assisted healingof fractures, a quantitative descriptor (such asminor, moderate or significant) is applied, as iscommonly used for heat-treated rubies.

It is also important that synthetic overgrowthson corundum are properly characterized anddeclared on laboratory reports. The reasons for thisclosely parallel those for clearly describing the fill-ing of cavities and fractures with glass on reportsand, as a consequence, within the industry (Kane,1984; Scarratt and Harding, 1984; Scarratt et al.,1986; Hughes, 1987a; Scarratt, 1988, 2000; “Glassfilled rubies . . . ,” 1994; “Glass filling . . . ,” 1994;Robinson, 1995). Specifically, synthetic over-growths created during the Be-diffusion process: (1) are created artificially and, therefore, are unnat-ural additions to the stone; and (2) unnaturallyincrease the weight of a finished stone by bothadding the weight of the synthetic material andallowing a larger stone to be cut than would other-wise be possible.

Gem dealers and retailers often need moredescriptive terminology to communicate with theconsumer. In some instances, one could say that

beryllium was introduced into the stone to createcolors that are otherwise very rare in nature. Asimple and accurate description might be “artifi-cially colored with beryllium.” Additional disclo-sure information is needed for those stones whererepolishing may affect the color or where syntheticmaterial may be present on the surface.

SUMMARY AND CONCLUSIONSWe have shown that the “new process” fromThailand is the diffusion of beryllium into ruby andsapphire at very high temperatures in an oxygenatmosphere. Unlike the much older titanium diffu-sion process, which produced a thin blue layerunder the surface, Be diffusion can penetratedeeply, sometimes coloring the entire volume ofthe stone. We have shown that beryllium diffusedinto corundum, at high temperature, is the solecause of the color modifications observed. A widevariety of colors are produced by this process,depending on how the Be trapped-hole pair inter-acts with the internal chemistry of the stone. Thusfar, the colors produced or modified by this processare predominantly yellow, yellowish orange,orangy pink to pinkish orange, orange, and orange-red to red (ruby). Many are very attractive, and haveexcellent commercial potential (figure 59). We havealso shown that some types of dark blue basalticsapphire can be lightened very significantly by Bediffusion. Thus, the majority of sapphire colors inthe marketplace could potentially be products ofthe Be-diffusion process.

However, color modifications are not the onlyimportant consequences of this process. We havenoted significant amounts of synthetic corundum


Figure 58. To illustrate the consequences of recutting incompletely Be-diffused stones, we took one of these twoclosely matched treated sapphires and had it recut as if it had been chipped and needed significant repair. The0.71 ct stone on the right side of the pair was recut to 0.40 ct, and clearly has lost most of its orange component.Photos by Maha Tannous.

overgrowth on many of these stones, much ofwhich may remain after re-cutting. Some facetedstones have shown 10–20% of the surface cov-ered with this synthetic material, raising ques-tions concerning just how “natural” such a stoneis. In addition, fractures and boehmite channelsare often sealed or “healed” by this process

Identification of Be diffusion–processed stonesranges from very simple observations to quitecomplex chemical analyses. Standard gemologi-cal properties are not changed by Be diffusion,nor does the durability of the stone appear to beaffected. If the diffusion layer penetrates only aportion of the radius of the stone, it may be seeneasily by immersion in a high R.I. liquid withdiffuse background illumination. Observation ofa three-dimensional surface-conformal colorlayer is considered proof that the stone has beendiffused. If there is only a portion of a color layer,however, because the stone was recut or original-ly diffused in the rough, care must be taken notto confuse Be-diffusion treatment with naturalcolor zoning.

In some sapphires, diffusion fully penetratesthe stone—and identification becomes substan-tially more difficult. However, the inclusion scenemay provide important indicators of the high heatrequired for Be-diffusion treatment. Careful exam-ination with magnification and diffuse reflectedlight may reveal synthetic growth on surfaces andin healed fractures and channels. Highly altered

zircon inclusions are also indicative of the veryhigh temperatures used in this process, as is thereplacement of boehmite with synthetic material.Internal diffusion, in the form of blue halosaround rutile crystals in yellow and red corun-dum, is another important indicator. Althoughthese features do not prove Be diffusion, they doindicate that the stone should be subjected toadditional testing.

SIMS and LA-ICP-MS can both unequivocallyidentify the presence of beryllium in color-alter-ing concentrations (5–30 ppma). While these areexpensive tests, they are necessary for potentiallyhigh-value stones that do not exhibit a surface-conformal color layer.

As a part of the study of this process we havedecided to replace the term surface diffusion,which has been used by gemologists and somelaboratories for years, with lattice diffusion to beconsistent with the usage in other scientific disci-plines. In their official reports, both the GIA labo-ratory and AGTA-GTC describe Ti- and Be-dif-fused stones with the wording “. . . indications ofheating, color induced by lattice diffusion from anexternal source.” Thus, both the earlier titaniumblue diffusion and the new beryllium diffusion aredescribed the same way, as they are exactly thesame process, just different diffusing elements.

The implications of Be diffusion to the entiregem community (from miner to consumer) arequite far reaching. It appears that virtually any


Figure 59. As shown bythis attractive jewelry,Be-diffused sapphiresclearly have significantcommercial potential, aslong as they are properlydisclosed. Photo byMyriam NaftuleWhitney; courtesy ofNafco Gems.

color of ruby and sapphire can be reproduced bythe Be-diffusion process. Transparent corundumactually is a common commodity. Only corun-dum in attractive colors and color saturation lev-els is rare. There are large deposits of sapphirethat can produce large stones in unmarketablecolors. Conventional heat treatment canimprove only a very small percentage of suchmaterial. However, it appears that much, if notmost, of this material can be Be diffused to pro-duce attractive colors (figure 60). We haveobserved that Be diffusion beneficially affects50–95% of mine run material, not just a smallportion. Thus, if we cannot develop a simplelow-cost detection method, gem sapphire willbecome as common as blue topaz, supply willexceed demand, and prices will fall radically,damaging the entire gem trade. Today, there is nosimple, low-cost alternative to SIMS or LA-ICP-MS, but we are working on the problem and strong-ly encourage others to do so as well.

Technology will always advance, and attemptingto slow or stop it is as rewarding as trying to sweepback the sea with a broom. However, as a commu-nity we need to come together to understand eachnew process as it is developed, and bring it to thegem market with full disclosure, allowing aninformed marketplace to determine value. Only inthis way can our miners, treaters, dealers, and jewel-

ers prosper. Only in this way will the consumershow us the trust and support we have enjoyed inthe past.


Figure 60. Many attractive sapphire colors that werepreviously very rare will now be made abundant withBe diffusion, opening up uses in jewelry designs thatwere previously not available. The total weight of theBe-diffused sapphires in this pin is 3.50 ct. Photo byMyriam Naftule Whitney; courtesy of Nafco Gems.

ABOUT THE AUTHORS

Dr. Emmett ([email protected]) is a principal of CrystalChemistry, Brush Prairie, Washington; Mr. Scarratt is labo-ratory director, AGTA Gemological Testing Center, NewYork; Mr. McClure is director of Identification Services, GIAGem Laboratory, Carlsbad, California; Mr. Moses is vicepresident for Identification and Research, GIA GemLaboratory, New York; Mr. Douthit is a principal of CrystalChemistry, Los Altos, California; Mr. Hughes is the authorof Ruby & Sapphire and webmaster at Palagems.com,based in Fallbrook, California; Dr. Novak is manager ofDynamic SIMS Services, Evans East, East Windsor, NewJersey; Dr. Shigley is director of GIA Research in Carlsbad;Dr. Wang is a research scientist, GIA Gem Laboratory,New York; Mr. Bordelon is a gem merchant and design-er/manufacturer of gemological instruments in New Orleans,Louisiana; and Mr. Kane is president, Fine GemsInternational, Helena, Montana.

ACKNOWLEDGMENTS: The authors thank gem dealersTerry Coldham, Mark Smith, Joe Belmont, Tom Cushman,Rudi Wobito, Hans-Georg Wild, Markus Wild, Bill Larson,Eric Braunwart, Santpal Sinchawla, Jeff Bilgore, and Roland

Naftule for graciously supplying stones and other materialsfor these experiments. They thank Junko Shida for detailson the Gemological Association of All Japan’s experiencewith laser tomography and supply of, and information con-cerning, Be-diffused blue sapphires. A number of stimulat-ing conversations with Professor George Rossman of theDivision of Geological and Planetary Sciences at theCalifornia Institute of Technology, Pasadena, California, arealso greatly appreciated, as are conversations with MilanKokta and Jennifer Stone-Sundberg of Sant-GobainCrystals and Detectors, Washougal, Washington. YianniMelas provided invaluable assistance in the early stages ofthis research. Dr. Thomas R. Anthony of the GeneralElectric Company, Schenectady, New York, is also thankedfor his assistance. Garth Billings and Dave Witter arethanked for their support of the experiments. Don andCorey Johnson of Eaton-Turner Jewelry, Helena, Montana,kindly provided assistance with the durability testing.Thomas Moses, Shane McClure, and Kenneth Scarrattacknowledge with thanks the aid of their laboratory staffmembers. John Emmett, Troy Douthit, and Owen Bordelonexpress their appreciation to the AGTA for partial financialsupport of their work.


The motion of atoms through solids is fundamentalto our technological society. The heat treatment andcase hardening of steels, the doping of semiconduc-tors for integrated circuits, and the production of pad-paradscha-like colored sapphires from pink ones allrely on the transport of atoms through solids. Themechanism by which atoms or ions move throughsolids is known as diffusion.

How actually does an atom move through asolid that is, after all, solid? It turns out that eventhe most perfect crystal has places in its latticewhere atoms are missing. These holes are termed“vacancies.” In the case of corundum, which ismade up of aluminum and oxygen atoms, even themost perfect crystal sample will have missing alu-minum and oxygen atoms, and the higher the tem-perature the more missing atoms there will be. Aforeign ion can move through solid corundum byjumping from one vacancy to the next. Since thevacancies are randomly distributed, the diffusingion will “random walk” through the crystal, some-times moving forward or back, up or down, or rightor left. Averaged over a large number of jumps, theion always moves away from regions of high con-centration of diffusing ions to regions of low con-centration. Jumping requires energy, which is sup-plied by heat, so an ion makes more jumps per sec-ond at high temperatures than at low temperatures.Thus, diffusion rates increase dramatically withtemperature due to the increase in the number ofboth vacancies and jumps per second.

All types of heat treatment of corundum involvesome type of diffusion process. It is the movement ofpreexisting ions that makes heat treatment work. Forexample, the dissolution of rutile needles into thehost corundum crystal is simply titanium diffusinginto corundum (and aluminum diffusing into rutile)at high temperature. In this case, the diffusioninvolves ions that are already in the host corundum(Koivula, 1987). In this article, however, we are pri-marily concerned with the diffusion of a foreign ele-ment (Be) into corundum from an outside source.Thus, diffusion is the physical process, and diffusiontreatment—the term we use to describe the enhance-ment method in this article—refers to artificially col-oring a gem.

How fast does diffusion occur? In corundum itdepends on the size (ionic radius) of the diffusing ionrelative to the size of the vacancy, the ionic charge ofthe diffusing ions, the temperature during treatment,and a variety of other less important factors. Becauseof the “random walk” nature of diffusion, the depth

of penetration does not increase linearly with time.Rather, it increases with the square root of time.This means that if the penetration depth of a givenion at a given temperature is 1.0 mm in one hour, itwill be 3.16 mm in 10 hours, and 10 mm in 100hours. Thus at a particular temperature we can saythat the diffusion depth x (in cm) is related to time by

x = Dt

where D is the diffusion coefficient in units ofcm2/sec, and t is the time in seconds over which thediffusion is conducted. The diffusion coefficient D isa measure of the ease with which an ion movesthrough the corundum lattice. At 1750°C, for exam-ple, the diffusion coefficients for the following ionsare approximately: H+ ~10−5, Be2+ ~10−7, Ti4+ ~10−9,and Cr3+ ~10−13 cm2/sec. Diffusion coefficientsincrease exponentially with temperature.

Figure Ap-1 shows the concentration profiles atthree different times for beryllium (assuming D =10−7 cm2/sec at 1750°C). The shape of these curvesis typical of the diffusion process. If we wish to fullysaturate a stone with a foreign ion, we need toknow how long it will require. Figure Ap-2 is a plotof the average concentration in a spherical sampleas a function of time t, the diffusion coefficient D,and the diameter d, of the stone. Continuing ourexample, if we assume D to be 10−7 cm2/sec, then itwill require 97 hours to completely saturate a 5-mm-diameter spherical sample of corundum, but75% of complete saturation can be achieved inabout 16 hours, which is one-sixth of the time forcomplete saturation.

What determines the magnitude of the diffu-sion coefficients, which, as we have noted, canvary over many orders of magnitude from elementto element? At a given temperature, aliovalentions (see “Color in Corundum” section) exhibitdiffusion coefficients in relatively pure corundumthat are between 1,000 and 10,000 times those ofisovalent ions. As discussed above, a diffusing ionrequires vacancies into which to jump in order tomove. In pure corundum, or corundum containingonly isovalent impurities, vacancies occur onlyfrom heating the lattice. These thermally inducedvacancies exist only in extremely low concentra-tions, even at very high temperatures. Thus thediffusion of isovalent ions is very slow. In con-trast, aliovalent ions make their own vacanciesand do not have to rely on thermally inducedvacancies. This is due to the fact that aliovalentions must be charge compensated to dissolve into

APPENDIX: WHAT IS DIFFUSION?


the corundum lattice (see “Color in Corundum”section). In diffusion situations, this charge com-pensation is usually accomplished by the forma-tion of vacancies. For example, to place three Ti4+

ions in the corundum lattice requires the forma-tion of one aluminum vacancy to charge compen-sate them. At any measurable Ti4+ concentration,therefore, the concentrations of these vacanciesexceeds by orders of magnitude the concentrationof thermally induced vacancies. Thus the aliova-lent ion can diffuse orders of magnitude fasterthan isovalent ions.

Ion size, or the effective “radius” of an ion, has astrong effect on the diffusion coefficient. The effec-tive radius (Shannon, 1976) of the Al3+ site in corun-dum is 53 picometers (1 pm = 10−12 meters). Be2+ onthis site, at 45 pm, is much smaller. Mg2+ on thissite, at 72 pm, is much larger. Ando (1987) deter-mined that the diffusion coefficient of Mg2+ in sap-phire—at low concentrations and at 1800°C—isabout 2.5 × 10−10 cm2/sec. Our estimate of the Be2+

diffusion coefficient (see “Beryllium Diffusion intoCorundum: Process and Results” section) at thesame temperature, based on the depth of the colorboundary, is roughly 10−7 cm2/sec—400 times greaterthan that of Mg2+. Recently there has been concernin the gemological community about the possibilitythat boron or lithium might be diffused into sapphire(see, e.g., Shor, 2003). Diffusion coefficients for B3+

and Li+ in corundum have not been measured, butboron is both isovalent and very small (27 pm), so itsdiffusion coefficient is hard to estimate but is expect-ed to be smaller than that of titanium or magnesium

because it is isovalent. Thus Be3+ would diffuse muchmore slowly than Be2+. Lithium is both large (76 pm;very close to the size of Mg2+) and aliovalent andcould be expected to diffuse somewhat like magne-sium or titanium.

In most diffusion situations, the concentrationof the diffusing ion is much higher in the flux orother carrier outside the crystal, than just inside thecrystal. The concentration just inside the crystal isnot determined by the concentration outside, butby the solubility of the foreign ion in corundum.The solubility of foreign ions in corundum alsoincreases with temperature. Solubilities vary wide-ly; for example, at 1750°C, titanium solubility is aseveral hundred ppma while that of beryllium is afew tens of ppma.

A good overall reference on diffusion is Borg andDienes (1988).

Figure Ap-1. The depth of beryllium diffusion into sap-phire is shown here calculated for three different pro-cessing times—1 hour, 10 hours, 100 hours. Diffusiondepth does not increase linearly with time, but withthe square root of time.

Figure Ap-2. This diagram shows the time required tosaturate, by diffusion, a sphere of material of diameterd in cm. D is the diffusion coefficient in units ofcm2/sec, and t is the diffusion time in seconds. Notethat to achieve 100% saturation requires approximate-ly six times as long as to achieve 75% saturation. Anexample of how to use this graph is as follows: Supposethat we wish to determine the degree of Be saturation(and thus color saturation) that will be achieved in a 5-mm-diameter sapphire that we will Be diffuse for 25hours at 1750°C. Thus d = 5 mm (or 0.5 cm), t = 25hours (or 90,000 seconds), and D = 10−7 cm2/sec (or0.0000001 cm2/sec). Putting these numbers into therelation on the horizontal axis, 2 Dt/d, we get 0.375.A value of 0.375 on the horizontal axis corresponds toabout 85% saturation. Note that the curve is very gen-eral in that it applies to any ion at any temperature dif-fusing into any spherical host, as long as the diffusioncoefficient is known at the temperature of interest.

REFERENCESAndo K. (1987) Magnesium-impurity diffusion in Al2O3 single crystals.

Nonstoichiometric compounds. In C.R.A. Catlow and W.C.Mackrodt, Eds., Advances in Ceramics, Vol. 23, American CeramicSociety, Westerville, OH, pp. 149–158.

Beesley C.R. (1982) The alchemy of blue sapphire. Jewelers’ Circular-Keystone, Vol. 153, No. 8, pp. 102–103.

Beran A. (1991) Trace hydrogen in Verneuil-grown corundum and itscolour varieties: An IR spectroscopic study. European Journal ofMineralogy, Vol. 3, No. 6, pp. 971–975.

Boiko B.B., Shkadarevich A.P., Zhdanov É.A., Kalosha I., Koptev V.G.,Demidovich A.A. (1987) Lasing due to color centers in an Al2O3:Mgcrystal. Soviet Journal of Quantum Electronics, Vol. 17, No. 5, pp.581–582.

Borg R.J., Dienes G.J. (1988) An Introduction to Solid State Diffusion.Academic Press, San Diego, CA, 360 pp.

Bube R.H. (1992) Electrons in Solids, 3rd ed. Academic Press, San Diego,CA, 329 pp.

Burns R.G. (1981) Intervalence transitions in mixed-valence minerals ofiron and titanium. Annual Review of Earth and Planetary Sciences,Vol. 9, pp. 345–383.

Burns R.G. (1993) Mineralogical Applications of Crystal Field Theory.Cambridge University Press, Cambridge, UK.

Butterman W.C., Foster W.R. (1967) Zircon stability and the ZrO2-SiO2phase diagram. American Mineralogist, Vol. 52, No. 5/6, pp.880–885.

Carr R.R., Nisevich S.D. (1975) Altering the appearance of sapphirecrystals. U.S. patent 3,897,529, issued July 29.

Carr R.R., Nisevich S.D. (1976) Altering the appearance of corundumcrystals. U.S. patent 3,950,596, issued April 13.

Carr R.R., Nisevich S.D. (1977) Altering the appearance of corundumcrystals. U.S. patent 4,039,726, issued August 2.

CGA (2003) Chanthaburi gem traders and heaters agree to discloseberyllium treatment, CGA Press Release, February 20.

Chiang Y.-M., Birnie D.P. III, Kingery W.D. (1997) Physical Ceramics.John Wiley & Sons, New York, 522 pp.

Coldham T. (2002) Orange sapphires or just lemons? AustralianGemmologist, Vol. 22, No. 7, pp. 288–293.

Crowningshield R. (1966) Developments and highlights at GIA’s Lab inNew York: Unusual items encountered. Gems & Gemology, Vol.12, No. 3, p. 73.

Crowningshield G.R. (1970) Developments and highlights at GIA’s Labin New York: Unusual fluorescence [sapphire with unusual fluores-cence]. Gems & Gemology, Vol. 13, No. 4, pp. 120–122.

Crowningshield R. (1979) Developments and highlights at GIA’s Lab inNew York: Treated blue sapphires. Gems & Gemology, Vol. 16, No.5, p. 147.

Crowningshield R., Nassau K. (1981) The heat and diffusion treatmentof natural and synthetic sapphires. Journal of Gemmology, Vol. 17,No. 8, pp. 528–541.

Dodd D.M., Wood D.L., Barns R.L. (1964) Spectrophotometric determi-nation of chromium concentration in ruby. Journal of AppliedPhysics, Vol. 35, No. 4, pp. 1183–1186.

Dutt B.V., Kröger F.A. (1975) High-temperature defect structure of iron-doped α-Alumina. Journal of the American Ceramic Society, Vol.58, No. 11/12, pp. 474–476.

Eigenmann K., Kurtz K., Günthard H.H. (1972) The optical spectrum ofα-Al2O3:Fe3+. Chemical Physics Letters, Vol. 13, No. 1, pp. 54–57.

El-Aiat M.M., Kröger F.A. (1982) Hydrogen donors in α-Al2O3. Journal ofApplied Physics, Vol. 53, No. 5, pp. 3658–3667.

Emmett J.L., Douthit T.R. (1993) Heat treating the sapphires of RockCreek, Montana. Gems & Gemology, Vol. 29, No. 4, pp. 250–272.

Emmett J.L. (1999) Fluxes and the heat treatment of ruby and sapphire.Gems & Gemology, Vol. 35, No. 3, pp. 90–92.

Emmett J.L., Douthit T.R. (2002a) Beryllium diffusion coloration ofsapphire: A summary of ongoing experiments. American GemTrade Association, http://agta.org/consumer/gtclab/treated-sapps04.htm.

Emmett J.L., Douthit T.R. (2002b) Understanding the new treatedpink-orange sapphires. Pala International, http://palagems.com/treated_sapphire_emmett.htm.

Ferguson J., Fielding P.E. (1971) The origins of the colours of yellow,green and blue sapphires. Chemical Physics Letters, Vol. 10, No. 3,pp. 262–265.

Ferguson J., Fielding P.E. (1972) The origins of the colours of natural yel-

low, blue, and green sapphires. Australian Journal of Chemistry, Vol.25, pp. 1371–85.

Fritsch E., Chalain J.-P., Hänni H., Devouard B., Chazot G., GiulianiG., Schwarz D., Rollion-Bard C., Garnier V., Barda S.,Ohnenstetter D., Notari F., Maitrallet P. (2003) Le nouveau traite-ment produisant des couleurs orange à jaune dans les saphirs(French with English abstract). Revue de Gemmologie, No. 147,pp. 11–23.

Fritsch E., Rossman G.R. (1987) An update on color in gems. Part 1:Introduction and colors caused by dispersed metal ions. Gems &Gemology, Vol. 23, No. 3, pp. 126–139.

Fritsch E., Rossman G.R. (1988) An update on color in gems. Part 2:Colors involving multiple atoms and color centers. Gems &Gemology, Vol. 24, No. 1, pp. 3–15.

GIT (2002) GIT information on heat-enhanced orange-pink (padparad-scha-like) sapphires from Madagascar. Gem and Jewelry Institute ofThailand, http://www.git.or.th (document no longer availableonline; hard copy on file with authors).

Glass filled rubies increasing (1994) Jewellery News Asia, No. 119, pp.66–70.

Glass filling in ruby from Burma (1994) Indian Gemmologist, Vol. 4,No. 1, pp. 1, 24.

Häger T. (1992) Farbgebende und “farbhemmende” Spurenelemente inblauen Saphiren. Berichte der Deutschen MineralogischenGesellschaft, Beihefte zum European Journal of Mineralogy, Vol. 4,p. 109.

Häger T. (1993) Stabilisierung der Farbzentren von gelben natürlichenSaphiren. Berichte der Deutschen Mineralogischen Gesellschaft,Beihefte zum European Journal of Mineralogy, Vol. 5, p. 188.

Häger T. (1996) Farbrelevante Wechselwirkungen von Spurenelementenin Korund. Dissertation, Universität Mainz, Mainz, Germany.

Häger T. (2001) High temperature treatment of natural corundum.International Workshop on Material Characterization by SolidState Spectroscopy: The Minerals of Vietnam, April 4–10, Instituteof Materials Science, Hanoi.

Hänni H.A. (2001) Beobachtungen an hitzesehandeltem Rubin mit künst-licher Rissheilung. Zeitschrift der Deutschen GemmologischenGesellschaft, Vol. 50, No. 3, pp. 123–136.

Hänni H.A. (2002) Orange treated sapphire—towards finding a name fora new product. Journal of the Gemmological Association of HongKong, Vol. 23, pp. 23–28.

Hänni H.A., Pettke T. (2002) Eine neue Diffusionsbehandlung liefertorangefarbene und gelbe Saphire. Zeitschrift der DeutschenGemmologischen Gesellschaft, Vol. 51, No. 4, pp. 137–152.

Hervig R.L. (2002) Beryllium analysis by secondary ion mass spec-trometry. In E.S. Grew, Ed., Beryllium: Mineralogy, Petrology, andGeochemistry, Reviews in Mineralogy and Geochemistry, Vol.50, Mineralogical Society of America, Washington, DC, pp.319–332.

Hughes R.W. (1987) Glass infilling of cracks in ruby. ICA Lab Alert,Vol. 4, p. 1.

Hughes R.W. (1988) Identifying yellow sapphires—two important tech-niques. Journal of Gemmology, Vol. 21, No. 1, pp. 23–25.

Hughes R.W. (1991a) Thailand taking the heat. JewelSiam, Vol. 2, No. 2,pp. 42–48.

Hughes R.W. (1991b) There’s a rumble in the jungle—The sapphire face-lift face-off. Gemological Digest, Vol. 3, No. 2, pp. 17–31.

Hughes R.W. (1997) Ruby & Sapphire. RWH Publishing, Boulder, CO,511 pp.

Hughes R.W. (2002) The skin game: Orange sapphire treatment raisescontroversy. The Guide, Vol. 21, No. 2, pp. 3–7.

Kane R.E. (1984) Natural rubies with glass-filled cavities. Gems &Gemology, Vol. 20, No. 4, pp. 187–199.

Kane R.E., Kammerling R.C., Koivula J.I., Shigley J.E., Fritsch E. (1990)The identification of blue diffusion-treated sapphires. Gems &Gemology, Vol. 26, No. 2, pp. 115–133.

Keller P.C. (1982) The Chanthaburi-Trat gem field, Thailand. Gems &Gemology, Vol. 18, No. 4, pp. 186–196.

Khattak C.P., Schmid F. (1984) Growth of large-diameter crystals byheat exchanger method for optical and laser applications. In S.Musikant, Ed., Proceedings of the SPIE Conference on Advances inOptical Materials, 260 pp.

Khattak C.P., Scoville A.N., Schmid F. (1986) Recent developments insapphire growth by heat exchanger method (HEM). In R.W.Schwartz, Ed., Proceedings of the SPIE Conference on Infrared andOptical Transmitting Materials, 181 pp.


Kingery W.D., Bowen H.K., Uhlmann D.R. (1976) Introduction toCeramics. John Wiley & Sons, New York, 1032 pp.

Kizilyalli M., Corish J., Metselaar R. (1999) Definitions of terms for dif-fusion in the solid state. Pure and Applied Chemistry, Vol. 71, No.7, pp. 1307–1325.

Koivula J.I. (1987) Internal diffusion. Journal of Gemmology, Vol. 20,No. 7/8, pp. 474–477.

Koripella C.R., Kröger F.A. (1986) Electrical conductivity of Al2O3: Fe +Y. Journal of the American Ceramic Society, Vol. 69, No. 12, pp.888–896.

Krebs J.J., Maisch W.G. (1971) Exchange effects in the optical-absorptionspectrum of Fe3+ in Al2O3. Physical Review B, Vol. 4, No. 3, pp.757–769.

Kröger F.A. (1974) The Chemistry of Imperfect Crystals. North-HollandPublishing Co., Amsterdam, The Netherlands, 988 pp.

Kröger F.A. (1984) Defect related properties of doped alumina. SolidState Ionics, Vol. 12, pp. 189–199.

Kvapil J., Perner B., Sulovsky J., Kvapil J. (1973) Colour centre formationin corundum doped with divalent ions. Kristall und Technik, Vol. 8,No. 1/3, pp. 247–251.

Lee C.H., Kröger F.A. (1985) Electrical conductivity of polycrystallineAl2O3 doped with silicon. Journal of the American Ceramic Society,Vol. 68, No. 2, pp. 92–99.

Leta D.P., Morrison G.H. (1980) Ion implantation for in-situ quantita-tive ion microprobe analysis. Analytical Chemistry, Vol. 52, pp.277–280.

McClure D.S. (1962) Optical spectra of transition-metal ions in corun-dum. Journal of Chemical Physics, Vol. 36, No. 10, pp. 2757–2779.

McClure S.F. (2002a) Gem Trade Lab Notes: Bulk-diffusion treated sap-phire with synthetic overgrowth. Gems & Gemology, Vol. 38, No. 3,pp. 255–256.

McClure S.F. (2002b) Swarovski scientists identify possible source ofberyllium in new treated sapphires. GIA Insider, May 3, 2002,http://www.gia.edu/newsroom/issue/2798/1046/insider_newsletter_details.cfm#3.

McClure S.F., Kammerling R.C., Fritsch E. (1993) Update on diffusion-treated corundum: Red and other colors. Gems & Gemology, Vol.29, No. 1, pp. 16–28.

McClure S.F., Moses T., Wang W., Hall M., Koivula J.I. (2002) GemNews International: A new corundum treatment from Thailand.Gems & Gemology, Vol. 38, No. 1, pp. 86–90.

Mohapatra S.K., Kröger F.A. (1977) Defect structure of α-Al2O3 dopedwith magnesium. Journal of the American Ceramic Society, Vol. 60,No. 3/4, pp. 141–148.

Moon A.R., Phillips M.R. (1994) Defect clustering and color in Fe, Ti: α-Al2O3. Journal of the American Ceramic Society, Vol. 77, No. 2, pp.356–357.

Moses T.M., Hall M., Wang W. (2002) Gem Trade Lab Notes: More bulkdiffusion—rubies and orange sapphires. Gems & Gemology, Vol. 38,No. 4, pp. 342–344.

Nassau K. (1980) Gems Made by Man. Chilton Book Co., Radnor, PA. Nassau K. (1983) The Physics and Chemistry of Color. John Wiley &

Sons, New York, 454 pp.Nelson D.F., Sturge M.D. (1965) Relation between absorption and emis-

sion in the region of the R lines of ruby. Physical Review, Vol. 137,No. 4A, pp. 1117–1130.

Pauling L. (1956) General Chemistry. W. H. Freeman and Co., SanFrancisco, CA, 710 pp.

Peretti A., Günther D. (2002) The color enhancement of fancy sapphireswith a new heat-treatment technique (Part A). Contributions toGemology, Vol. 1, pp. 1–48.

Pisutha-Arnond V., Häger T., Wathanakul P., Atichat W. (2002) LabNews: A brief summary on a cause of colour in pink-orange, orangeand yellow sapphires produced by the new heating technique. Gem and Jewelry Institute of Thailand Gem Testing Lab,http://www.git.or.th/eng/eng_lab/eng_lab_news/eng_news_on_22_november_2002.htm.

Proust M.A. (2002) New sapphire treatment still a mystery. ColoredStone, Vol. 15, No. 2, pp. 12–13.

Qi L., Yuan X., Lin S., Wang Z. (2002) Colouring mechanism of high-temperature heat-treated orange sapphire in Madagascar [inChinese]. Journal of Gems & Gemology, Vol. 4, No. 4, pp. 1–5.

Ramírez R., Tardío M., González R., Kokta M.R., Chen Y. (2001)Electroluminescence in magnesium-doped Al2O3 crystals. RadiationEffects & Defects in Solids, Vol. 154, pp. 295–299.

Rankin A., Edwards W. (2003) Some effects of extreme heat treatment on

zircon inclusions in corundum. Journal of Gemmology, Vol. 28, No.5, pp. 257–264.

Robinson N.L. (1995) Thais get burned by glass fillings. Colored Stone,Vol. 8, No. 4, pp. 1, 6.

Rossman G.R. (1988) Vibrational spectroscopy of hydrous components.In Hawthorne F.C., Ed., Spectroscopic Methods in Mineralogy,Reviews in Mineralogy, Vol. 18, Mineralogical Society of America,Washington, DC, pp. 193–206.

Scarratt K. (1983) Notes from the Laboratory: Diffusion treated corun-dum. Journal of Gemmology, Vol. 18, No. 6, pp. 523–530.

Scarratt K. (1988) Notes from the laboratory: Glass filled feathers andcavities in ruby. Journal of Gemmology, Vol. 21, No. 3, pp. 131–139.

Scarratt K. (2000) Gem News: Heat-treated rubies: Glass or not glass?Gems & Gemology, Vol. 36, No. 1, pp. 75–76.

Scarratt K. (2002a) Is it pink or is it padparadscha? Rapaport DiamondReport, Vol. 25, No. 5, pp. 103, 105, 109.

Scarratt K. (2002b) Orange-pink sapphire alert. American Gem TradeAssociation, http://www.agta.org/consumer/gtclab/orangesap-phirealert.htm.

Scarratt K., Harding R.R. (1984) Glass infilling of cavities in naturalruby. Journal of Gemmology, Vol. 19, No. 4, pp. 293–297.

Scarratt K., Harding R.R., Din V.K. (1986) Glass fillings in sapphire.Journal of Gemmology, Vol. 20, No. 4, pp. 203–207.

Schmetzer K. (1981) The colour of natural corundum. Neues Jahrbuchfür Mineralogie, Monatshefte, No. 2, pp. 59–68.

Schmetzer K. (1999) Ruby and variously colored sapphires from Ilakaka.Australian Gemmologist, Vol. 20, No. 7, pp. 82–84.

Schmetzer K., Bosshart G., Hänni H.A. (1983) Naturally-coloured andtreated yellow and orange-brown sapphires. Journal of Gemmology,Vol. 18, No. 7, pp. 607–622.

Shannon R.D. (1976) Revised effective ionic radii and systematic studiesof interatomic distances in halides and chalcogenides. ActaCrystallographica, Vol. A32, pp. 751–767.

Shor R. (2003) Gem News International: AGTA corundum panel. Gems& Gemology, Vol. 39, No. 1, pp. 58–59.

Smith C.P. (1995) A contribution to understanding the infrared spectraof rubies from Mong Hsu, Myanmar. Journal of Gemmology, Vol.24, No. 5, pp. 321–335.

Smith G., Strens R.G.J. (1976) Intervalence-transfer absorption in somesilicate, oxide and phosphate minerals. In R.G.J. Strens, Ed.,Physical Chemistry of Rocks & Minerals, John Wiley & Sons, NewYork, pp. 584–612.

Smyth D.M. (2000) The Defect Chemistry of Metal Oxides. OxfordUniversity Press, New York.

Tardío M., Ramírez R., González R., Chen Y., Kokta M.R. (2001a)Electrical conductivity in magnesium-doped Al2O3 crystals at mod-erate temperatures. Radiation Effects & Defects in Solids, Vol. 155,pp. 409–413.

Tardío M., Ramírez R., González R., Chen Y., Kokta M.R. (2001b)Enhancement of electrical conductivity in α-Al2O3 crystals dopedwith magnesium. Journal of Applied Physics, Vol. 90, No. 8, pp.3942–3951.

Tardío M., Ramírez R., González R., Chen Y., Kokta M.R. (2001c) Hightemperature semiconducting characteristics of magnesium-doped α-Al2O3 single crystals. Applied Physics Letters, Vol. 79, No. 2, pp.206–208.

Thailand stands by thermal enhancement claim (2002) Jewellery NewsAsia, No. 220, pp. 80–90.

Themelis T. (1992) The Heat Treatment of Ruby and Sapphire. GemlabInc., Clearwater, FL, 236 pp.

Themelis T. (2003) Beryllium-Treated Rubies & Sapphires. TedThemelis, Bangkok, Thailand, 48 pp.

Townsend M.G. (1968) Visible charge transfer band in blue sapphire.Solid State Communications, Vol. 6, pp. 81–83.

Treated Songea sapphires hit the market (2002) Jewellery News Asia,No. 209, p. 62.

Wang H.A., Lee C.H., Kröger F.A., Cox R.T. (1983) Point defects in α-Al2O3:Mg studied by electrical conductivity, optical absorption, andESR. Physical Review B, Vol. 27, No. 6, pp. 3821–3841.

Wang W., Green B. (2002a) Gem News International: Update on Be-dif-fused corundum. Gems & Gemology, Vol. 38, No. 4, pp. 363–364.

Wang W., Green B. (2002b) An update on treated natural corundum.GIA Insider, December 20, 2002, http://www.gia.edu/newsroom/issue/2798/1482/insider_newsletter_details.cfm#2.

Weldon R. (2002) Corundum conundrum. Professional Jeweler, Vol. 5,No. 3, pp. 36–37.


Documents

Beryllium Diffusion Of