The second document was much longer and was merely confidential; indeed, it was in the form of a draft letter to Science, awaiting final approval before publication. Its snappy title was 'Space Vehicle Discovery: Anomalous Orbital Behavior'.

Then followed a dozen pages of mathematics and astronomical tables. Floyd skimmed through these, picking out the words from the music, and trying to detect any note of apology or even embarrassment. When he had finished, he was compelled to give a smile of wry admiration. No one could possibly guess that the tracking stations and ephemeris calculators had been caught by surprise, and that a frantic cover-up was in progress. Heads would doubtless roll, and he knew that Victor Millson would enjoy rolling them – if his was not one of the first to go. Though to do him justice, Victor had complained when Congress had cut funds for the tracking network. Maybe that would get him off the hook.

'Thank you, Colonel,' said Floyd when he had finished skimming the papers. 'Quite like old times, having classified documents. That's one thing I don't miss.'

The colonel placed the envelope carefully back in his briefcase, and activated the locks.

'Dr Millson would like you to return his call as soon as possible.'

'I know. But I don't have a secure circuit, I've some important visitors coming shortly, and I'm damned if I'm driving down to your office in Hilo just to say I've read two documents. Tell him that I've studied them carefully and await any further communication with interest.'

For a moment it looked as if the colonel was going to argue. Then he thought better of it, made a stiff farewell, and departed morosely into the night.

'Now, what was all that about?' asked Caroline. 'We're not expecting any visitors tonight, important or otherwise.'

'I hate being pushed around, particularly by Victor Millson.'

'Bet he calls you back as soon as the colonel reports.'

'Then we must switch off video and make some party noises. But to be perfectly truthful, at this stage I really don't have anything to say.'

'About what, if I'm allowed to ask.'

'Sorry, dear. It seems that Discovery is playing tricks on us. We thought the ship was in a stable orbit, but it may be about to crash.'

'Into Jupiter?'

'Oh no – that's quite impossible. Bowman left it parked at the inner Lagrange point, on the line between Jupiter and Io. It should have stayed there, more or less, though the perturbations of the outer moons would have made it wander back and forth.

'But what's happening now is something very odd, and we don't know the full explanation. Discovery's drifting more and more rapidly toward Io – though sometimes it accelerates, and sometimes even moves backward. If it keeps this up, it will impact within two or three years.'

'I thought this couldn't happen in astronomy. Isn't celestial mechanics supposed to be an exact science? So we poor backward biologists were always being told.'

'It is an exact science, when everything is taken into account. But some very strange things go on around Io. Apart from its volcanoes, there are tremendous electrical discharges – and Jupiter's magnetic field is spinning round every ten hours. So gravitation isn't the only force acting on Discovery; we should have thought of this sooner – much sooner.'

'Well, it's not your problem anymore. You should be thankful for that.'

'Your problem' – the very expression that Dimitri had used. And Dimitri – cunning old fox! – had known him much longer than Caroline.

It might not be his problem, but it was still his responsibility. Though many others had been involved, in the final analysis he had approved the plans for the Jupiter Mission, and supervised their execution.

Even at the time, he had had qualms; his views as a scientist had conflicted with his duties as a bureaucrat. He could have spoken out, and opposed the old administration's shortsighted policies – though to what extent those had actually contributed to the disaster was still uncertain.

Perhaps it was best if he closed this chapter of his life, and focused all his thoughts and energies upon his new career. But in his heart he knew that was impossible; even if Dimitri had not revived old guilts, they would have surfaced of their own accord.

Four men had died, and one had disappeared, out there among the moons of Jupiter. There was blood on his hands, and he did not know how to wash them clean.

Dr Sivasubramanian Chandrasegarampillai, Professor of Computer Science at the University of Illinois, Urbana, also had an abiding sense of guilt, but one very different from Heywood Floyd's. Those of his students and colleagues who often wondered if the little scientist was quite human would not have been surprised to learn that he never thought of the dead astronauts. Dr Chandra grieved only for his lost child, HAL 9000.

Even after all these years, and his endless reviews of the data radioed back from Discovery, he was not sure what had gone wrong. He could only formulate theories; the facts he needed were frozen in Hal's circuits, out there between Jupiter and Io.

The sequence of events had been clearly established, up to the moment of the tragedy; thereafter, Commander Bowman had filled in a few more details on the brief occasions when he had re-established contact. But knowing what happened did not explain why.

The first hint of trouble had been late in the mission, when Hal had reported the imminent failure of the unit that kept Discovery's main antenna aligned to Earth. If the half-billion-kilometre-long radio beam wandered off target, the ship would be blind, deaf, and dumb.

Bowman himself had gone out to retrieve the suspect unit, but when it was tested it appeared, to everyone's surprise, to be in perfectly good order. The automatic checking circuits could find nothing wrong with it. Nor could Hal's twin, SAL 9000, back on Earth, when the information was transmitted to Urbana.

But Hal had insisted on the accuracy of his diagnosis, making pointed remarks about 'human error'. He had suggested that the control unit be put back in the antenna until it finally failed, so that the fault could be precisely located. No one could think of any objection, for the unit could be replaced in minutes, even if it did break down.

Bowman and Poole, however, had not been happy; they both felt that something was wrong, though neither could pinpoint it. For months they had accepted Hal as the third member of their tiny world, and knew his every mood. Then the atmosphere aboard the ship had subtly altered; there was a sense of strain in the air.

Feeling rather like traitors – as a distraught Bowman had later reported to Mission Control – the human two-thirds of the crew had discussed what should be done if their colleague was indeed malfunctioning. In the worst possible case, Hal would have to be relieved of all his higher responsibilities. This would involve disconnection – the computer equivalent of death.

Despite their doubts, they had carried out the agreed programme. Poole had flown out of Discovery in one of the little space pods that served as transporters and mobile workshops during extravehicular activities. Since the somewhat tricky job of replacing the antenna unit could not be performed by the pod's own manipulators, Poole had started to do it himself.

What happened then had been missed by the external cameras, which was a suspicious detail in itself. Bowman's first warning of disaster was a cry from Poole – then, silence. A moment later he saw Poole, tumbling over and over, spinning away into space. His own pod had rammed him, and was itself blasting away out of control.

As Bowman admitted later, he had then made several serious mistakes – all but one excusable. In the hope of rescuing Poole, if he was still alive, Bowman launched himself in another space pod – leaving Hal in full control of the ship.