The government acted yesterday to begin punishment proceedings against a heart researcher at Harvard University who faked experiments, according to Assistant Secretary for Health Edward N. Brandt Jr.

In a statement issued yesterday, Brandt also faulted Harvard for failure to curb the fraud sooner.

It took the university between six months and a year to dismiss the researcher after he admitted faking results.

Brandt said he would seek to bar John R. Darsee for 10 years from doing any federally funded research.

Brandt also took the unusual step of asking the Harvard teaching hospital where Darsee worked, the Brigham and Women's Hospital in Boston, to return $122,371 in federal funds for one of Darsee's projects.

National Institutes of Health officials said the sanction against the hospital and the attempt to get a 10-year "debarment" of Darsee were the most severe penalties yet sought under the agency's new science fraud regulations.

Darsee now has the right to a hearing. If he waives a hearing he may seek a negotiated settlement.

Darsee, 34, admitted to Harvard that he faked data in May, 1981, according to Daniel C. Tosteson, dean of Harvard Medical School. Tosteson also acknowledged in a letter to faculty members that a "prompt and thorough" effort was not made to investigate Darsee's work or dismiss him.

"I wish, in retrospect, those actions had been taken," Tosteson said.

Once considered a bright prospect in his field, Darsee was caught by three young colleagues who said they watched him fake data in experiments with dogs.

He altered records of blood tests from the dogs to cover up when his superior asked him for the raw data to back up one of his papers, according to NIH officials.

NIH officials said that since the original incident, which involved one project, at least nine other papers Darsee had published in scientific journals over several years were falsified.