Read our COVID-19 research and news.

How NASA Tried to Save Its Prime Planet Spotter

Kepler deputy project manager Charles Sobeck.


NASA’s Kepler spacecraft has been a workhorse for planet hunters since its launch in 2009. Sifting through Kepler data, astronomers have discovered 130 new extrasolar planets and identified hundreds of planetary candidates, showing that the universe is teeming with planets.

On Thursday, NASA announced that the Kepler spacecraft would no longer be able to continue doing science as before. That’s because engineers have been unable to revive the functionality of two of the spacecraft’s four reaction wheels, which are necessary to keep it precisely pointed at target stars. Science caught up with Charles Sobeck, deputy project manager for Kepler at NASA’s Ames Research Center in Moffett Field, California, to learn more about what happened to the spacecraft and what lies in store for it. The interview has been edited for clarity and brevity.

When did you first realize there was a problem?

The first time was a year ago when wheel 2 failed. The spacecraft wasn’t holding fine point on its science target [staying precisely pointed at the target] because that wheel had stopped spinning.

Reaction wheels are notorious for breaking, aren’t they?

Reaction wheels do fail all the time, some fail quite early, some go on for decades without any problem.

What was the status of Kepler after wheel 2 broke?

It was still pointing well, but not precisely. Within a few days, we were able to take the failed wheel offline, reset some parameters on the spacecraft; that essentially told the spacecraft that you only have three wheels, use those wheels. It snapped right back into fine point, but now our concern was, we couldn't afford to lose another one. We had to ask ourselves, what were we doing wrong, what could we be doing differently.

When did you find out that another wheel could be at risk as well?

We had had no indications that wheel 2 had been in distress, but once it had failed, we went back and looked at all our data and put it through different algorithms. We were able to find advance indications saying that here's what would indicate that a wheel is having problems. Had we known what to look for, we could have seen that 6 months before wheel 2 failed. Armed with that knowledge, we could now look back at all the wheels and ask: Are any other wheels showing early signs of stress? And indeed, what we found is that wheel 4 was showing intermittent signs of stress.

Did you do things differently to protect the wheel?

We came up with a list of eight or 10 things we did differently. We increased the operating temperature of the wheel; we raised the speed of the wheel to keep it out of the region below 300 revolutions per minute, where the nature of the friction and the internal lubrication changes state. We also operated the wheel bidirectionally: We spent some of the time going clockwise, and some of the time going counterclockwise.

And yet the wheel gave out?

In early January, we saw the wheel exhibit some elevated friction that didn’t come down. We knew we only had a few months left, so we began quickly developing a thruster control mode so that we could use the thrusters on the spacecraft to stabilize it and point it. In May, wheel 4 seized up and stopped turning.

What happened next?

In July, we tested the broken wheels to see if they could still function. We spun them up only for a few minutes, and only to 300 revolutions per minute, to see if they would move. The short answer is that even though they spun, the friction was higher than we had seen before. We knew that these are clearly damaged wheels.

Did you have a plan for repairing the wheels?

We looked at the internal structure of these wheels’ bearings. There is a plastic separator between the balls that the wheel rolls on. There was a reservoir of lubricant within this plastic cage, and we thought if we raised the temperature, we could force some of that internal lubricant from the reservoir to the surface of the cage, and then we could run the wheels, spread the lubricant around, and that would enable the wheels to run at a lower friction. So that’s what we did. We turned the heaters up to a higher temperature than it had seen in flight before to get the lubricant out. But the friction did not go down, which suggested that the lubricant wasn’t forced to the surface or that wasn’t the problem to begin with.

Was there anything left to do?

Functionally, the wheels were still working, and we wanted to find out how well they could point the spacecraft. We did that test last week.

How did that go?

We returned the spacecraft to wheel control, and it held control pretty good. We still had this super-elevated friction level, but we have relatively large motors on these wheels, and the thought was, we could overcome the friction with the motor force we have available. For the next 2 hours, we watched how the wheels were performing. Things were working pretty good. Normally, we would like to see the wheels hold the spacecraft to within 1 or 2 arc seconds of the target; it was holding at within 10 arc seconds.

Did that allow you to do something useful?

We pointed Kepler’s high-gain antennae at Earth and downlinked the data that has been on its solid state recorder for the past 3 months. There were two aspects of the data that we were most interested in. There were 4 or 5 days of science data that we had collected for 4 or 5 days prior to the wheel failure. We had also been storing high-rate engineering data that would describe exactly how the wheel failed, what the timeline was. We had only very low-resolution insight into the wheel failure, but downlinking that data, we have high-resolution insight. That may not be important to Kepler, but it’ll be important to other missions who want to be able to know how a wheel fails and what to look for.

Why couldn’t you just keep going?

As we were pointing the antennae to Earth, wheel number 2 started experiencing very high friction. A good wheel should have something like 2 or 3 milli-Newton meters. In July, it went up to 30. Now, the friction jumped up to over 100. During the 4 hours, we saw the friction stay in that very high regime, it was getting closer and closer to the 200 milli-Newton meter limit that our motors have on them.

That’s like watching an emergency room patient’s pulse rate going up.

Yes, and there’s nothing you can do about it, other than hoping that it comes to some sort of stable state. We had several of us in one office at Ames looking at the data in real time. At first, when it jumped from 30 to 100, we thought ‘it’s going to die almost immediately.’ But then it didn’t. It started to rise slowly. We thought, ‘well, it’s going to die in an hour or two.’ Then it peaked out and started coming down again, and it just sort of teased us. Then the friction exceeded the limit, we could no longer spin the wheels with our motor, the spacecraft autonomously put itself in safe mode. We commanded the spacecraft back to its point and rest state, our holding pattern.

Are you still hoping that the Kepler spacecraft will be able to do some science?

We will now try to define what kind of pointing can we get with two good wheels and thrusters. We are going to look at what we can do, and we’re going to look at the cost of that. And we’re going to see what kind of science can we do with it. We’ve issued a call for white papers, inviting the scientific community to submit ideas. We’re not limiting it to extrasolar planets. Some ideas would be to do microlensing and to look for near-Earth asteroids. We’re not throwing in the towel.

*Correction, 19 August, 11:06 a.m.: The friction forces specified in this story are in milli-Newton meters, not million Newton meters, as erroneously reported before. Thanks to the reader who brought it to our attention.