For at least 15 years, the U.S. intelligence community has been trying to crack the puzzle of "yellow rain."

So when the answer seemed to present itself in the form of one broken leaf, a little piece of stem and a few stray flakes from a leaf, Secretary of State Alexander M. Haig Jr. announced in Berlin some "preliminary" news: the United States now has physical evidence from Southeast Asia of three potent chemical agents used in warfare in Democratic Kampuchea, formerly known as Cambodia. State Department spokesmen quickly added that these agents were probably made in the Soviet Union from a fungus, and were apparently used in Kampuchea, Laos and Afghanistan.

But Secretary Haig was making the announcement some five weeks before he originally planned, and too soon for some scientists and State Department officials to feel comfortable about the pronouncement.

There was an urge to make the news public from the beginning both because it might prevent further chemical attacks in Southeast Asia and because it was a political opportunity to call attention to Soviet actions, one source said.

But State Department officials say Haig was pressed into the announcement even earlier than planned because press accounts of the evidence had already begun to appear only two weeks after the first laboratory results came in.

There are also indications that the State Department feared that holding off any longer could have delayed the announcement until it coincided with this week's New York meeting between Haig and Soviet Foreign Minister Andrei Gromyko, thus making a sensitive meeting even more difficult.

So, without waiting for further lab work, the announcement was made. Predictably, the single sample quickly caused argument over the scientific validity of the State Department's conclusions, although the day after Haig's announcement officials said other samples have been collected for study.

"It is very hard to believe this," said Dr. Matthew Meselson, Harvard biologist and expert in chemical warfare. From a scientific point of view, he said it is "outrageous" to ask people to believe the report on the basis of a single sample.

Altogether, the State Department's evidence "is pretty shaky," said James R. Bamberg, a biologist at Colorado State University and the man who first identified and named T2. He said that it is possible that fungus in Southeast Asia grows T2 naturally, that the leaves were contaminated with spores when they were collected, and that on the trip to America the mold grew and produced T2. If the samples were in a container and refrigerated, he said, moisture and refrigeration are known to greatly boost the production of T2 in most fusaria, a type of fungus.

He said more samples are needed, as is--more importantly--some way of assuring that local fungi do not contaminate the samples.

On the other hand, Clifford Hesseltine, an Agriculture Department expert in the growth of the fusarium fungus, said it was very unlikely that the amounts of toxin found would be produced naturally on a living leaf, especially since there was no overgrowth of fungus on the leaf. If you have toxin, you've got to have mold and lots of it, Hesseltine said.

Journalist Sterling Seagrave, who has written a book on the controversy, said he learned from government officials that the samples were collected as carefully as possible under the circumstances; they were put into sealed plastic containers and were kept out of refrigeration. He said one of the first tests of the leaf and stem sample was for any fungi present. The tests showed no fungus, only the white powder which later proved to be the three fungus toxins.

Seagrave said it was far more likely that the toxin was put on the leaf than that it grew there.

The government is actively seeking others samples in Southeast Asia to corroborate the evidence of the leaf and stem.

One of the ironies of the report is that, after 15 years of seeking the unknown substance, the U.S. government's candidate for the mysterious agent is a chemical freely available in the United States. It can be bought through the mail; one company offers it and four other toxins in a research kit for $75.

The fungus it comes from is one of the most common on earth. But for 15 years, not knowing what to look for, government scientists could not find it in the evidence from the field.

The story of the leaf and stem, and how they triggered the dramatic Berlin announcement, started in 1979, when a team of doctors was sent to Southeast Asia by the Pentagon to try to discover the truth about the years of reports about a new, deadly chemical agent that had been used to gas and kill thousands of Hmong tribesmen in Laos.

Leader of the team was Col. Charles Lewis, chief of dermatology at the Brooke Army Medical Center in San Antonio, Tex. He and other medical specialists were sent to see the medical symptoms for themselves because they were unlike anything that could be attributed to known chemical weapons. Lewis identified a mix of three sets of symptoms. There were convulsions, like those produced by nerve gas; there was blistering of the skin and breathing passages, like mustard gas, and there was massive hemorrhaging, unlike any common weapon, and all three sets of symptoms were combined.

According to Seagrave, whose forthcoming book "Yellow Rain" describes the history of the case, Dr. James Vick of the Edgewood Arsenal labs suggested that the symptoms might not be from any known chemical agent. Natural biological poisons, such as snake venom and coral poison, could cause such a collection of symptoms. None, however, immediately seemed to fit all the symptoms or to be capable of causing them simply by inhalation or by absorption throught the skin, as in a gas attack.

Eventually the U.S. government began to press for collection of samples of the deadly substance, and a great variety of things that might be contaminated with the chemical agent--jackets, wood, leaves, dirt, blood and tissue--were gathered in Southeast Asia.

The recent samples added to reports of similar gas attacks and odd deaths going back as far as the civil war in Yemen in 1966, when Lt. Col. James Barrett, now a teacher in Annapolis, was military attache in Saudi Arabia. In an interview last week, he said that he was sent on a mission to collect samples of blood and tissue from victims of gassing being carried on by Soviet-supported Yemenis. Though the material was apparently never identified, its symptoms now match those reported from Afghanistan after the Soviets invaded late in 1979, and from Laos and Kampuchea in the past half-dozen years.

Finally last December, Seagrave and a government scientist simultaneously hit upon the possibility that the chemical agent might be one of a group called the mycotoxins, especially one called T2.

The State Department looked through the collected samples for one to test, and picked one that had come in during March of this year--a part of a leaf, an inch-long bit of stem, and three-quarter-inch leaf fragments. They were covered with "a white material resembling mold . . . in form, size, and agglomeration like a reagent-grade chemical," according to the description attached to the samples when they arrived.

This is one uncertainty in the State Department's evidence: the gas attacks are associated with a yellow powder, believed to be a silicon binder that can carry toxic chemicals. But how is it that the leaf sample has no yellow powder, only the bare toxin? The answer given by State Department officials is that possibly some delivery systems do not involve yellow powder, but were bombs or shells carrying just the toxin.

This summer a government researcher prepared the sample for tests by breaking the leaf in half. One half would be tested as is. To the other was added a small amount of T2 toxin, to be certain that if the substance were there the tests would detect it. A third leaf, gathered in the United States but looking like the one from Cambodia, was added to the set.

The three samples were sent to a laboratory where a large amount of agricultural material is regularly tested for T2 toxin, one source said. The samples were passed through a third party and the scientist conducting the test was not told where they came from.

What the lab found were three natural poisons--T2, nivalenol, and deoxynivalenol --all produced by fusaria fungus. The amount of at least two of the three toxins was about 20 times that found in natural outbreaks, according to the State Department. That would amount to about 1 to 160 parts per million, and probably more than 100 parts per million.

Eight parts per million can be fatal.

It was on Aug. 27 that the results of the tests were reported to an interagency chemical warfare group. At the same meeting a doctor sent to Southeast Asia to investigate the symptoms, but unaware of the test results, also suggested that the symptoms in refugees he had seen would best fit with T2 and other fungal toxins.

As it happens, T2 and the fusaria fungus have a long history in the Soviet Union. From the early 1930s through the middle 1940s, continuous outbreaks of fusaria fungus attacked grain. When people ate it, hundreds of thousands died. And the symptoms included blistering, hemorrhaging and convulsions.

Because of these disasters, Soviet scientists have studied the fusaria and T2 extensively. Of the papers in the open literature, according to Seagrave, a large percentage have to do with how the T2 might be effectively and economically produced in the laboratory.

Fusaria also exists in the wild in Southeast Asia, but the species there are not known to produce T2 toxin. Then again, as many scientists have pointed out, little or no work has been done to find out if any of the many strains that exist in Asia can produce T2 toxin.

The simultaneous reports in August stunned the interagency chemical warfare group. For years researchers in the defense establishment had been chasing the possible chemical agents which might produce the "yellow rain" symptoms. The possibility of natural toxins had been raised only a relatively few months before. But now, T2 seemed neatly to tie together all the years of puzzling reports.

The information was quickly delivered to the Pentagon, the CIA, the Joint Chiefs of Staff and Secretary Haig. Haig and others were excited by the discovery but, on the advice of scientists, decided at first to wait until more evidence could be gained before making any announcements. Less than two weeks later, Time magazine published the first sketchy report of the evidence.