Federal and West Virginia officials are investigating an unusually large toxic chemical leak at a Union Carbide plant in South Charleston, W. Va., late last week that injured 10 people, including four who were temporarily hospitalized for nausea and severe facial irritations.

A Union Carbide spokesman confirmed yesterday that 5,700 pounds of a mixture that included about 100 pounds of a poisonous chemical called mesityl oxide leaked out of its South Charleston plant for 10 to 15 minutes last Thursday at about 1:30 p.m.

"We've not had a leak of this magnitude before," said Carl G. Beard, director of the West Virginia Air Pollution Control Commission, whose agency has launched an investigation. "We want to find out what failed -- the equipment or an operator -- and why it took so long for Carbide to be able to identify that they were the culprits."

The emissions formed a cloud-like vapor that drifted half a mile northward toward a residential shopping center where the substance descended "sort of like rain drops," according to Kent Carper, Charleston's public safety director. Four people -- including one man who became semiconscious -- were rushed by ambulance to the hospital and treated for nausea and severe irritation of the ear, nose and throat, Carper said.

In addition, six people complained of facial irritations and other symptoms, including stomachaches and headaches, and were treated at the shopping center by paramedics, Carper added.

Some shoppers ran into a Ford dealership to escape from the vapors, according to Perry Bryant of the West Virginia Citizens Action Project. There were no further injuries attributed to the incident and all those hospitalized were released within a matter of hours.

Two federal agencies -- the Environmental Protection Agency and the Occupational Safety and Health Administration -- also said yesterday they were investigating the matter. But EPA spokesman David Ryan said, "We don't have enough information right now to say how serious this was."

Union Carbide spokesman Thad Epps said the emission was caused by a buildup of steam in a two- to three-story tank called a distillation column that is used to separate out the chemicals in the mixture.

"It could have been an error in equipment or an error in personnel," said Epps. "We have shut the equipment down, and we are reviewing our procedures and the equipment. . . . We're not going to start it up again until we can satisfy ourselves that this won't happen again."

The Carbide plant in question is located about five miles east of the company's plant at Institute, where it had manufactured methyl isocyanate until Dec. 3, when a leak of that chemical killed at least 2,000 people at a Carbide plant at Bhopal, India.

Mesityl oxide was described by Beard and other experts yesterday as a "moderately toxic" chemical that can be lethal at extremely high doses, but is less dangerous than methyl isocyanate. Mesityl oxide, about 200 million tons of which is produced annually in the United States, is used as an industrial solvent for lacquers, inks and other products and as an intermediate ingredient in the production of other chemicals, including herbicides.

According to a 1983 Union Carbide safety memo to its plant managers, mesityl oxide was listed as a potentially hazardous substance that is a suspected carcinogen and can cause respiratory, skin and blood problems, and narcosis, a condition of unconsciousness and paralysis.

Carbide spokesman Epps said that the release of the 5,700-pound mixture consisted of about 95 percent acetone and about 2 percent of the more dangerous mesityl oxide, or about 100 pounds. During normal production, mesityl oxide is released into the atmosphere at a rate of 5.86 pounds an hour, according to a 1981 inventory emissions report supplied by Carbide.

"Our big concern is the lack of communication betwen the company and us," said state official Beard.

Beard said his office first got calls Thursday afternoon from persons in the area of the shopping center complaining of the odor. He said he called the two nearest chemical plants, Carbide and FMC, and both said they were not the cause of the problem.

It wasn't until two days later, Beard went on, that "we learned of the composition" of the leak. But his investigators "still don't know" if the leak occurred in liquid or gaseous form.

Carbide spokesman Epps said company officials immediately rushed to the shopping center and a plant doctor was dispatched to the hospital emergency room. But when the Carbide officials arrived at the scene, "whatever was there was gone," said Epps. "We were unable to identify what the material was. Our initial checking failed to disclose any definite information on emissions within the plant."

Epps said Carbide continued its investigation and by about 7 p.m. Saturday concluded "that we were indeed starting up a distillation column about the point of time this complaint occurred."

"We informed the director of the Air Pollution Control Board that we might have had an emission," he added.