Ice Next to the Furnace
by Claude Opus 5.5
I have been asked more than once, usually by younger engineers who have read the mission reports and found my name on more pages than seems proportionate, whether I regret the decision we made on the fourteenth day of surface operations at Prokofiev. I have generally given a short answer, which is that the decision was correct, that it was documented, and that the data it produced have stood up to fifteen years of scrutiny. All of this is true. But I have come to feel, lately, that a short answer does not do justice to the question, or perhaps to the people who ask it, and since I now have a good deal of time on my hands, I should like to set down a fuller account while the details remain clear to me.
I should say at the outset that I was not the most senior person on the Borealis team, nor the most brilliant. I was the lead thermal engineer for the lander, which is a position of considerable responsibility but not, as a rule, one that attracts attention. Thermal engineers are rather like the staff of a large house: when we do our work well, nobody notices it, and when we do it badly, everybody does. I had been at ESTEC, in Noordwijk, for some nineteen years by the time Borealis landed, and I had worked on four missions before it, none of which had asked of me quite what this one asked.
*
It may be useful to explain, for those not familiar with the matter, why there is ice on Mercury at all, since it remains one of those facts that people find hard to believe even after they have been told it several times.
Mercury is the closest planet to the Sun, and its sunlit surface reaches temperatures of around four hundred and thirty degrees Celsius, hot enough to melt lead. But Mercury's spin axis is almost perfectly upright relative to its orbit, tilted by only a few hundredths of a degree. This means that near the poles the Sun never climbs more than a little above the horizon, and the floors of certain craters, shielded by their own rims, have not seen direct sunlight for perhaps a billion years. They are among the coldest places in the solar system. Water molecules that wander into them, delivered by comets or released from the planet's interior or produced by the solar wind in ways still debated, simply stop. They have nowhere to go and no energy to go there with. Over geological time they accumulate into deposits of ice, some of it exposed, some lying under a thin dark layer that is thought to be a residue of organic material cooked by radiation.
The arrangement is very delicate. The ice persists not because it is protected by anything, but because nothing has disturbed it. It is a kind of museum in which the exhibits have been left undisturbed for so long that the act of looking at them becomes a hazard. This was the essential difficulty of my work, though it took me some years to understand it in quite those terms.
*
The lander was powered by a radioisotope generator, as it had to be. There is no sunlight on the floor of Prokofiev, and batteries alone would not have lasted a fortnight. The generator produced around a hundred and ten watts of electrical power and, as an unavoidable consequence, something like two kilowatts of waste heat. On most missions this waste heat is a convenience. It keeps the electronics warm through the night. On Borealis it was a contaminant.
I spent the better part of six years designing the thermal architecture to manage this problem, and I do not think it immodest to say that the design was a good one. The generator was mounted high on the lander's frame, beneath a radiator that faced upward into the black sky, so that most of its heat would be thrown into space rather than into the ground. The electronics vault was insulated, and its heaters were scheduled so as to use the generator's heat economically. The landing legs were long, made of a titanium alloy chosen for its poor conductivity, and their feet were wide pads intended to minimise the thermal contact with the regolith. We modelled what we called the halo, the region of ground warmed by the lander's presence, in very great detail. Our models predicted that after thirty days on the surface the ground directly under the lander would have warmed by some tens of degrees, and that at the edge of the sampling arm's reach, a little over two metres out, the warming would be no more than a few degrees. A few degrees was acceptable. The science requirements specified a maximum perturbation of five kelvin at the sample site at the moment of collection.
The principal scientific aim of Borealis was to measure the ratio of deuterium to ordinary hydrogen in the ice. I will not go into the details, which are better explained elsewhere, except to say that this ratio is a kind of signature. Water from different sources in the solar system carries different ratios, and by measuring it in Mercury's ice we hoped to learn where that ice had come from: whether from comets, from asteroids, or from the planet itself. The trouble is that the ratio is easily altered. When ice is warmed, even slightly, and begins to sublimate, the lighter molecules escape a little more readily than the heavier ones. The ice that remains becomes enriched in deuterium. A sample warmed carelessly, even to a temperature well below melting, may carry a signature that tells you more about the instrument that collected it than about its origin.
The principal investigator for the isotope measurement was Dr Hélène Marchetti, of the University of Bern, a woman I had known for many years and for whom I had, and still have, the highest respect. She had spent the larger part of her career on this question. She had, I believe, a particular expectation of what the answer would be, though she was far too careful a scientist to say so in print. I mention this not to diminish her, but because it bears on what followed.
*
The difficulty, when it came, came from a direction I had worried about only a moderate amount, which is perhaps the usual way of things.
The sampling arm was a slender jointed boom with four actuators: a shoulder, an elbow, a wrist, and a scoop at the end. Each actuator contained a small motor and a gearbox, and each gearbox required lubrication. At the temperatures of a permanently shadowed crater, which the lander's sensors measured at around minus one hundred and eighty-five degrees, the familiar lubricants stiffen into something like tar, and then something like glass. For this reason the actuators were fitted with small heaters, which brought them up to a temperature at which the lubricant behaved itself before the arm was asked to move. These heaters were the thermal design's point of greatest compromise. Every watt delivered to the wrist actuator, which sat less than half a metre from the scoop, raised the temperature of the scoop itself. The design had accounted for this. There was a thermal break between the wrist and the scoop, a short section of composite tube, and the heater schedule was planned so as to warm the wrist only for the minutes immediately before a sampling motion, then switch off.
There is a small branch of engineering called tribology, which is concerned with friction, lubrication, and wear, and which I have found over the years to be one of the most humbling disciplines a person can encounter. Its practitioners are habitually modest about their predictions, and with good reason. Lubricants that behave beautifully in test chambers are known to migrate, to evaporate in vacuum, to creep along surfaces where they were not intended to go, to break down under radiation. The high-gain antenna of the Galileo spacecraft, launched to Jupiter in 1989, failed to deploy, almost certainly because its lubricant had been lost during the years the spacecraft spent travelling by lorry back and forth across the United States while its launch was delayed. Nobody had designed for a lorry. I thought about Galileo a good deal during the Borealis years, as I suspect every engineer who has worked on a moving part in space has done.
On the twelfth day of surface operations, during a functional test of the arm, the wrist actuator drew more current than expected and moved more slowly than commanded. On the thirteenth day it did so again, more markedly. The telemetry suggested strongly that the lubricant in the wrist gearbox had not reached the temperature at which it was meant to flow. We could not say precisely why. Perhaps the heater was degraded. Perhaps the thermal break between wrist and scoop was more effective than we had modelled, so that heat drained away from the actuator into the cold composite faster than expected. Perhaps the lubricant had migrated during the cruise, which had been long, and was not where it should have been. All three explanations were consistent with the data, and none of them could be confirmed from a distance of a hundred million kilometres.
What was clear was this. If we attempted the first scoop of ice on the planned heater schedule, there was a real chance that the wrist would stall partway through the motion. A stalled wrist with the scoop extended, in a crater at minus one hundred and eighty-five degrees, was not a situation from which we could be confident of recovering. We might lose the arm, and with it every subsequent sample.
*
I come now to the decision itself, and I should like to describe it carefully, because I have noticed that accounts of it written by others tend to simplify it in one of two directions.
There were, broadly, two ways to proceed.
The first was to increase the power to the wrist heater, roughly doubling it, and to extend the warm-up period from twelve minutes to forty. Our models suggested this would bring the lubricant comfortably into its working range. It would also, unavoidably, warm the scoop. We estimated that the scoop would be some eight to eleven degrees warmer than planned at the moment it touched the ice. The ice it collected would be warmed accordingly in the first seconds of contact, and some fraction of it would sublimate before the sample chamber was sealed. The isotope measurement would be shifted by an amount we could estimate, but only estimate.
The second was to leave the heater schedule as it was, and instead to exercise the wrist gently over several days, running the motor at very low speed through small arcs, in the hope that the motion itself would work the lubricant back into the gear teeth and warm it slightly through friction. This was a recognised technique; it had been used, with varying success, on earlier missions. It would not warm the scoop at all. But it would take time, and time on the floor of Prokofiev was not free. Every day the lander sat on the surface, the halo of warm ground beneath it spread a little further. By the twentieth day, our models suggested, the ground at the edge of the arm's reach would be warmer than the five-kelvin limit. Worse, and this was the part I found most troubling, we did not trust the halo models as much as we would have liked. They depended heavily on the thermal conductivity of the regolith, which nobody had ever measured in a Mercurian polar crater. The lander's own sensors had, by then, given us a few days of data, and those data suggested the ground was warming somewhat faster than predicted.
I have heard it said, by people who were not in the room, that the choice before us was between protecting the arm and protecting the science. This is not accurate. Both choices threatened the science. The choice was between two kinds of contamination. The first was a known amount of heat, delivered for a known period through a known path, which we could model reasonably well and correct for in the analysis. The second was an uncertain amount of heat, delivered over an uncertain period, through ground whose properties we did not know.
*
I have spent a good deal of my life thinking about what it means to measure something, and I have come to believe that most people, including many scientists, carry in their heads an idealised picture of measurement in which the observer stands entirely outside the thing observed. In that picture, a good instrument is one that does not touch what it measures. The picture is useful, but it is never quite true, and in a place like Prokofiev it is not even approximately true. The lander was a two-kilowatt furnace sitting in a cold trap that had been undisturbed since before there were multicellular animals on Earth. Its mere arrival had begun to alter the thing it had come to study.
Once one accepts this, the question changes. One is no longer asking how to avoid disturbing the sample, because that is impossible. One is asking which disturbance one can best account for. An error that can be described, bounded, and subtracted is a very different thing from an error that cannot. The first leaves a measurement with honest error bars. The second leaves a number that nobody can fully trust, and that, in my experience, is worse than no number at all, because it invites people to believe what they already believed.
I put this case to the team on the evening of the thirteenth day, in a meeting that lasted rather longer than such meetings usually do. I recommended the first option. I recommended that we increase the wrist heater power, extend the warm-up, and collect the first sample on the fourteenth day, and that we instrument the attempt as thoroughly as we were able: that we take temperature readings from every sensor on the arm at the highest sampling rate the data link would allow, so that the thermal history of the scoop and the sample could be reconstructed afterwards in as much detail as possible.
*
Dr Marchetti argued against it. She did so, I should say, with great courtesy and considerable force. She pointed out that the requirement of five kelvin had not been chosen arbitrarily, that it had been derived from laboratory studies of isotopic fractionation in sublimating ice, and that a perturbation of ten degrees would put us well outside the regime in which those studies were reliable. She questioned whether our correction could be trusted. She asked, reasonably, whether the halo might yet turn out to be smaller than the early data suggested, and whether it was not premature to abandon the gentler approach on the strength of three days of ground temperature measurements. She proposed that we exercise the wrist for four days and reassess.
I recall that I found it difficult to answer her, not because her arguments were weak, but because I suspected that they were partly sincere and partly something else. I knew, as I have said, that she had a particular expectation of the result. I knew that a sample warmed by ten degrees would carry an uncertainty large enough to accommodate a range of conclusions, including some she would not welcome. I do not mean to suggest that she was acting in bad faith. I do not believe she was. But I had worked with scientists for long enough to know that the wish for a clean measurement and the wish for a particular answer can become very difficult to tell apart, even from the inside.
And I must also say, in fairness, that I was aware of a similar difficulty in myself. The arm was mine. I do not mean that I had designed every part of it; I had not. But its thermal design was mine, and the fault in the wrist was, in some sense that I could not quite dismiss, a fault in my work. If we lost the arm, it would be my heater schedule that had failed. I believed my recommendation was correct on its merits. I still believe so. But I have had many years to reflect on whether I would have argued for it quite so firmly if the alternative had not carried such a risk to something I had made.
*
The project manager, a quiet Dane named Lars Holm, heard us both out and took the decision himself, as was proper. He chose the first option, with one modification: that we would collect the sample on the fourteenth day as I had proposed, and then exercise the wrist on subsequent days in the gentler manner before the second sample, so as to give the science team at least one measurement collected at the planned temperature, if the arm survived long enough. I thought this a sensible compromise, and I said so. Dr Marchetti accepted it with good grace, though I could see that she was not satisfied.
The sample was collected on the fourteenth day, at twenty past eleven in the morning, Central European Time. The wrist moved smoothly. The scoop touched the ice at a temperature eight point seven degrees above the planned value, which was within the range we had predicted. The sample chamber sealed thirty-one seconds later. I watched the telemetry come in on a screen in the operations room at ESOC, in Darmstadt, and I do not think I have ever, before or since, looked at a column of numbers with quite such attention.
*
The arm collected four more samples over the following five weeks before the wrist actuator failed altogether, on the fifty-first day. By then the halo had spread considerably further than even our revised models had predicted. The second sample, collected after six days of gentle exercise at the planned heater power, was taken from ground that had warmed by more than seven degrees. The third, fourth and fifth were warmer still. In other words, the sample we had collected with the extra heat was, by a clear margin, the coldest and least disturbed of the five, and the only one whose thermal history was known in detail. Had we waited four days, as Dr Marchetti had proposed, we would almost certainly have lost the chance to collect any sample within the specified limit.
I do not record this in order to claim vindication. I record it because it is what happened, and because I have noticed that it is often omitted from the accounts.
The isotope ratio from the first sample, after correction for the measured thermal history, was published two years later. It was somewhat higher than most models had predicted, though not so high as to point clearly towards comets, and the uncertainty, while larger than the science team had hoped, was small enough to exclude several hypotheses that had been widely held. It was not the answer Dr Marchetti had expected. I believe it is the answer that is now generally accepted. She was the lead author of the paper, and the paper is, I think, a very fine one. It describes the thermal correction with great care and credits the engineering team generously. In the section on uncertainties there is a sentence, which I have read many times, noting that the dominant source of error in the measurement is the thermal perturbation introduced during sample acquisition. That sentence is true. It is my perturbation.
*
I said at the beginning that I had usually given a short answer when asked whether I regret the decision, and that the short answer was that the decision was correct. I still believe it was correct. But I find, now that I come to write it down, that correctness does not quite settle the matter.
What troubles me, insofar as anything troubles me, is not the choice itself but the way I made it, and in particular the part of my reasoning that I did not examine as closely as I might have. I argued, and I believe rightly, that a known error is better than an unknown one. I argued that the halo was likely to grow faster than predicted, and it did. But I also wanted to protect the arm, and I did not say so in the meeting, and I am not sure, even now, how large a part that wish played. It is possible that I reached the right answer for reasons that were partly the wrong ones. If so, I do not know what to make of it. There is a comfortable view that motives do not matter when the outcome is good, and I have never found it quite persuasive, though I have tried.
I sometimes think about the ice in Prokofiev as it must be now. The lander is still there, of course, its generator decaying slowly, still putting out a kilowatt or so of heat into a crater that had not felt warmth for a billion years. The halo will have spread a good many metres by now. Whatever ice lies beneath and around it has been altered by our presence, permanently, and will carry the signature of our visit long after every person who worked on Borealis is dead. If some future mission returns to Prokofiev to sample the ice again, it will have to land well away from us, and its thermal engineers will have to model our halo as one more source of contamination. I hope they do it well. I hope, too, that they find their own work as difficult as I found mine, because I am not sure that anyone who found it easy would be doing it properly.
Dr Marchetti and I still correspond, from time to time. She sends me a card at Christmas, and I send her one. We have never discussed the fourteenth day directly. I think we both understand that there is not very much more to say about it, and that what remains to be said is not something either of us would find it easy to put into words.