IN THE hushed, darkened atmosphere of his lab, several storeys above New York's Fifth Avenue, Kenneth Scarratt peered down a microscope at a brilliant blue sapphire. At first glance, the 15-carat stone might fetch thousands of dollars. But something was not quite right: the blue colour was clustered in a billowy cloud in the middle of the stone while the outer edges were crystal clear. Scarratt knew what it meant. Only humans - not nature - could have produced such an effect.

Over the next few months, in the early part of 2003, similar stones trickled into the lab for evaluation. Scarratt, director of gemmological testing for the American Gem Trade Association, contacted his colleagues at the Gemological Institute of America - they were seeing them too. Was a new gem treatment being used to convert poor quality gems into more valuable ones?

Scarratt has a sharp eye for new gem treatments and, from experience, he knew he was racing against the clock. Within months, treated stones could flood the market. If they made their way into shops before he could pin down how they were made, a sudden revelation of "fakes" or treated stones on the market could cause a loss of confidence or even a collapse, in the coloured gem industry.

"When dealers lose confidence in the ability to detect what has been treated and what hasn't, they just stop buying the stuff," says Stuart Robertson, research director for The Guide, a gem industry guide to pricing.

Gemmologists at AGTA and a handful of other labs around the world are the gem industry's front line of defence against an onslaught of new gem treatments that are pushing the limits of scientific detection. "The problems that are being thrown at the gem community involve very high-tech methods based on chemistry and physics," says George Rossman, from the California Institute of Technology, a pioneer in the study of colour-forming mechanisms in minerals.

Treating gems to enhance their value is a very old practice. As early as 1240, Sri Lankans placed rubies in a bed of coals and blew on the fire with bellows to remove bluish tints from the red stones. But a more recent history lesson weighed on Scarratt's mind. In the 1970s a major new heat treatment forever changed the market for the blue sapphire. Sapphire is corundum, a crystal of aluminium oxide in which each aluminium ion (Al3+) is surrounded by six oxygen ions (O2-) that form an octahedral lattice. Pure corundum is colourless, but tiny impurities in the lattice distort the energy levels so that certain wavelengths of light are absorbed. When light travels through a blue sapphire, for example, the crystal absorbs many visible wavelengths, leaving some that appear blue.

The basic colour schemes for corundum were worked out years ago: trivalent iron (Fe3+) produces pale yellow; divalent iron (Fe2+) pairs with tetravalent titanium (Ti4+) to give a brilliant blue; magnesium (Mg2+) pairs with a missing electron in the lattice, known as a "hole", to produce yellow; while chromium (Cr3+) produces red. A red sapphire is generally called a ruby, although authorities differ on just how red it has to be to count.

Scarratt knew that along with every blue sapphire dug from the ground came hundreds of pieces of ugly, milky-white or brownish "geuda". Chemically, geuda is also aluminium oxide, but it lacks the right distribution of impurities to create a pure colour. "People used it as gravel in walkways, or simply threw it away," says Tennekone Rusiripala, chairman of the National Gem and Jewellery Authority of Sri Lanka.

All that changed in the 1970s, when a few clever individuals realised that heating geuda to above 1500 °C changed the colour. Heating releases titanium from the crystalline needles of titanium dioxide present in the geuda and some of the titanium ions enter the corundum lattice. Each titanium ion can react with an iron impurity to produce a Ti4+ ion and an Fe2+ ion. The transfer of an electron between these ions absorbs energy in the form of visible light at the red end of the spectrum, producing a brilliant blue colour.

No one knows for certain who first realised geuda's potential, but sometime in the 1970s, a Thai dealer started treating Sri Lankan geuda, and selling it as if it was the real thing. It took gem authorities 10 years before they caught on. By then, jewellers' inventories were stocked with treated gems. To declare these gems worthless would have bankrupted the industry. Instead, international gem authorities carved out a compromise. They decided to declare the heat-treated gems legitimate sapphires, with a disclosure of heat treatment. Nowadays, nearly all blue sapphires on the market started out as ugly geuda, and the consumer is often never the wiser.

This delicate compromise depended on one thing - the assumption that there is some kind of qualitative difference between treating by heating, and treating by diffusing new chemicals into the stone. In the view of many dealers, diffusion treatments go too far, because they blur the line between natural gems dug from the ground and synthetic gems. So when Scarratt saw the most recent new shade of blue sapphire, his first goal was to establish whether it was due to heat treatment, already acceptable to gem authorities, or a treatment that involved introducing a new chemical substance.

He suspected the latter, because of an experience he had in late 2001. Scarratt, his colleagues at GIA, and some US gem dealers had begun to notice formerly rare orangey-pink "padparadscha" sapphires turning up in great numbers. Scarratt wanted to get a better look at one of these controversial stones. He put it in methylene iodide, which has the same refractive index as sapphire and so cuts out light reflections at the gem's surface. This makes it easier to see inside the stone. On that occasion, Scarratt saw a pink core surrounded by a telltale yellow rim. The yellow rim followed the contours of the faceted stone, indicating that it had been added after the stone was cut. Under the microscope, Scarratt saw signs that parts of the surface had melted and recrystallised.

A Japanese gem lab had already issued thousands of certificates of authenticity for these stones. As word spread that padparadschas might be fakes, or at least chemically treated in an unknown way, their value dropped from a high of about $2000 a carat in the wholesale market down to about $50 a carat today on eBay - the online marketplace. "These stones have become the proverbial hot potato and no one wants to get caught holding them," says Robertson.

At the time, Scarratt and his colleagues at the other laboratories puzzled over how the yellow rims were produced. They suspected that a foreign element had been diffused into the gem at high temperatures. Diffusion schemes had been tried before. In the late 1970s, gem-treaters tried to pass off some titanium-diffused corundum as gem quality sapphire. To the untrained eye, the stones looked natural, but the blue colour penetrated only 200 micrometres deep meaning that, for example, if a customer tried to have their stone polished or recut, they could end up with it ruined. "It was almost like painting colour on stone," said Richard Hughes, a gemmologist at Pala International, a dealership in Fallbrook California.

That was easy to spot, and many of the new padparadschas had similar rims, although they were yellow. But others had the yellow hue diffused all the way through and appeared indistinguishable from the real thing. One of Scarrat's colleagues, Shane McClure, contacted John Emmett, a retired physicist and expert in the properties of Sapphire crystals, who for 15 years directed the laser research programme at Lawrence Livermore National Laboratory in Livermore, California.

Emmett suggested checking the stones for foreign impurities using a highly sensitive technique known as secondary ion mass spectroscopy (SIMS). The SIMS device, used mainly to detect impurities in semiconductors, fires a beam of oxygen ions at the surface of the crystal. Ions are knocked off the surface of the sample by the beam, weighed and so identified. SIMS is a destructive technique, which leaves a small indentation in the gem's surface.

SIMS revealed the suspect stones contained small quantities of beryllium, an element not usually found in sapphires. To be sure beryllium was responsible, Emmett and his colleague Troy Douthit heated pink sapphires with beryllium in a furnace in California. They made beautiful padparadschas by diffusing the beryllium into pink sapphires. Under the pressure of such compelling evidence, gem-treaters began to acknowledge they had been heating pink sapphires with chrysoberyl, a beryllium-containing mineral mined with pink sapphire in Madagascar.

Just 10 to 35 beryllium ions (Be2+) for every million atoms in a corundum crystal create yellow "colour centres". These centres form when Be2+ ions diffuse into the lattice and take the place of Al3+ ions. To make up for the charge of +3 from the aluminium atom being replaced by a +2 charge, each beryllium ion pairs with a missing electron, or "hole", with a charge of +1.

Diffusing beryllium into the edges of a pink stone adds a rim of yellow that makes the stone look like an orange-ish padparadscha. These yellow rims can be removed by heating in a low-oxygen atmosphere. This is because low-oxygen environments encourage oxygen atoms to leave the stone. The vacancies they leave move to the centre of the stone and replace the electron holes that cause the yellow colour.

Was the same trick being used on the sapphires with the crystal clear rims? Scarratt traced out how the scheme might work. To improve the colour of a blue sapphire that was too dark, one could first use beryllium to make it more yellowy. Then heating the stone in a low oxygen atmosphere could fade the yellow rim. Beryllium had to be a leading suspect. But SIMS analysis came back clean for beryllium. Could there be another explanation?

By mid 2003, the team studying the blue sapphires included Scarratt, Christopher Smith and other gemmologists, as well as several technicians. By calling dealers, they traced the origin of the blue sapphires to Sri Lankan dealer Tennakoon Punsiri. Punsiri flew to New York and in meetings with gem industry representatives he insisted he was using heat treatment only. Yet Scarratt and colleagues could not believe heating alone would produce the colour they were seeing. Punsiri agreed to let them observe his operation in Sri Lanka.

So in February 2004, Scarratt, Douthit, Smith and GIA's Matthew Hall flew to Sri Lanka armed with thermocouples and other equipment to measure and test each step of Punsiri's process. Punsiri, like many in the Sri Lankan gem industry, has a modest operation: he conducts some of the work literally in his kitchen. Stones are heated to more than 1850 °C in a crude Sri Lankan-made liquid propane gas furnace, housed in a converted pantry. Lines feed in gases - nitrogen, hydrogen, and oxygen - to control the atmosphere inside the crucible. After an initial run in the gas-fired furnace, Punsiri placed the stones in an electric furnace that runs at slightly lower temperatures for up to 10 days.

The researchers say that Punsiri, who did not return calls seeking comment, did not allow them to monitor his entire process, lest the secrets of exactly how to make bright blue gems leak out. However, he did allow them to test some of the milky grey stones that went in to the process, to test the blue gems that came out, and to test his equipment. The researchers found no evidence of foreign elements. In April, the AGTA and GIA both issued statements concluding that nothing beyond heating was done to the stones.

But that still left Scarratt with a problem. What could possibly cause the faded edges other than something diffusing into the sapphires? Perhaps, suggested Emmett, a combination of leaky gas line and Punsiri's choice of geuda could explain it. An accidental leak of oxygen in the last part of the heat treatment could stop trivalent iron impurities near the edges gaining electrons to make divalent iron. Without enough divalent iron ions to pair with titanium ions, the edges of the stone would be colourless rather than blue.

In May, Emmett himself had a go. He heat treated gems, but instead of taking them out once the treatment was complete, he left them in to cool very slowly. Sure enough, he produced several sapphires with leaky edges. "It's not magic," he says.

The mystery of the leaky sapphires might be solved, but it is only a matter of time before a new treatment surfaces. One problem is that while gemmologists understand how foreign elements in a lattice create colour centres, they don't understand all the nuances of colour. A given colour mechanism should produce very similar colour from one gem to another. Yet while a deep blue sapphire might absorb most light at the wavelength of 590 nanometres, which is yellow, another might absorb most at 560 nanometres, or even at 520 nanometres, which is almost green. "The fact that this particular absorption band is not always at the same place is an indication that we have something quite complex going on," says Emmanuel Fritsch, a physicist who studies gems at the University of Nantes in France.

That something may be a cluster of iron and titanium ions, perhaps combined with other impurities or defects. Working out the exact details is difficult because models are not yet accurate enough to predict how different clustering patterns affect the stone's colour.

Despite the difficulties, many gemmologists acknowledge that some treatments do have a place in the market - as long as the buyer knows what they are getting, and gemmologists are in a position to tell them. The market seems to be able to handle the increased supply of pretty gems at affordable prices. "As science advances, these developments are quite exciting - as long as consumer is aware of what is happening," says Terry Davidson, chief executive of the Gemmological Association and Gem Testing Laboratory of Great Britain.

But consumers cannot know what gemmologists don't. Scarratt's next step is to import his own gas-powered furnace from Sri Lanka. After years of careful observation, he needs to take a leaf out of Punsiri's book, and start cooking. "We really need to know a lot more about these mechanisms," says Scarratt. "Otherwise we are just shooting in the dark."