The Justice Department said Halliburton had recommended to BP, the British oil company, before the drilling that the well include 21 metal centralizing collars to stabilize the cementing. BP chose to use six instead. During an internal probe after the accident, Halliburton ordered workers to destroy computer simulations that showed little difference between using six and 21 collars, the government said, after which the company continued to say that BP was neglectful to not follow its advice.

The development was not entirely unexpected after the first phase of the civil trial in New Orleans. Lawyers representing businesses and others that suffered from the spill had long accused the company of conducting undocumented cement tests and hiding the results. BP had accused Halliburton of destroying evidence of its cement testing.

But during the trial this year Thomas Roth, a senior company executive who was in charge of cementing operations when the spill occurred, acknowledged that because of the well design and other factors, “the cement placement was going to be a job that would have a low probability of success.”

Timothy Quirk, a Halliburton laboratory manager, testified that he conducted stability tests on cement samples from a similar blend that had been used in the well after the accident. Following instructions from a colleague, he said he did not prepare a laboratory work sheet. “It was unusual,” he said. He also acknowledged that he had thrown out his notes.

Later tests showed that the cement was not stable.

The failure of the cement foam seal set off a complex and ultimately deadly cascade of oil and gas up the well casing that exploded into flames to engulf the Deepwater Horizon rig. The blowout preventer, which is supposed to contain a well bore breach, also failed.