A null result is not a “failure”

How Initial LIGO paved the way for one of the year’s biggest scientific discoveries.

amber-stuverSometimes negative results are unexpected, but contribute to advances in science. Sometimes they’re by design, with experiments created not to produce results in the traditional sense, but to fine-tune methods and processes for the future. Amber Stuver, a Data Analysis and EPO Scientist at the LIGO Livingston Observatory, explains how the LIGO project put this approach into practice, leading to the groundbreaking announcement of the first detection of gravitational waves earlier this year.


By Amber Stuver

I joined the LIGO Scientific Collaboration in 1999 as a graduate student studying physics at Penn State. It was an opportunity that I couldn’t pass up; the construction of the detectors was just completed, so I had a chance to get in on the ground floor and the science was as exotic as it was promising. If we were able to detect gravitational waves, then we could use them as a new medium to observe the universe and the promise of novel discoveries was awe-inspiring. This was a great intersection of physics and astronomy and fulfilled all my childhood fantasies about being a scientist and studying the universe.

But no one told me to be optimistic that a detection would happen in the near future. From the beginning, the founders of LIGO estimated that the first operation of the detector, what is now called Initial LIGO, would not make a detection. It could make a detection, and that was a better opportunity than we ever had before, but the first detection wasn’t expected to come until we took what we learned from Initial LIGO and integrated that and other new technological developments into upgrades.

So for years I (along with many others) worked on making better ways to separate out the enormous amount of noise we measure from real signals. At the first collaboration meeting I attended after we took our first science-quality data with Initial LIGO, I remember a talk along the lines of, “We are almost certain that everything we recorded is garbage. So how do we set out to scientifically establish that? How do we get better at being sure that a glitch isn’t a real signal?” That really struck me with how far we had to go.  And we have come far indeed!

Let me tell you about three meaningful null results we had with Initial LIGO:

1. GRB 070201

On 1 February 2007, there was a short-duration gamma-ray burst observed that could have originated within the Andromeda galaxy. Since this is the Milky Way’s nearest neighbor, as far as we are concerned it’s in our “back yard.” Not a lot is known about the progenitors of this kind of gamma-ray burst, but most hypotheses involve binary mergers of dense objects, which result in the generation of gravitational waves. By observing gravitational waves, which gives us information about the bulk motion of source mass, we would be able to at least begin to solve this mystery.

Careful study of our data around the time of the gamma-ray burst was done and no gravitational waves were found. While this was disappointing, there was still science to be had! We were able to determine that: if the gamma-ray burst really did originate in the Andromeda galaxy, it was not caused by a compact binary merger of massive objects within constrained mass ranges; we determined that if the gamma-ray burst was caused specifically by the merger of two neutron stars, they had to have been much farther away than the Andromeda galaxy; and we determined the maximum amount of energy that could have been radiated in gravitational waves if the source was the Andromeda galaxy. Clearly, even though gravitational waves were not observed in this instance, we still learned more about the event than we knew before.

implications-2008

2. Ruling out early models of the universe

Another kind of gravitational waves we search for are called stochastic gravitational waves, which are simply a superposition of many unresolved gravitational waves. These are the kind of gravitational waves we expect to see left over from the Big Bang, although any time you have many sources all added together you will have a stochastic gravitational wave.

By using observations from Initial LIGO, we were able to constrain some theories about the early universe, that is, the universe less than one minute after the Big Bang. We were able to rule out theories that have a large equation of state (ratio of pressure to energy density in the perfect fluid) and some cosmic (super)string models.

Given how sensitive we knew our detector to be, and how long we made observations, if any of the theories mentioned above were plausible, we would have seen the relic gravitational waves from the Big Bang. Again, a null result has contributed to our overall knowledge of the universe.

upper-limit

3. The “Big Dog Event”

Finally, how do you know you are doing something correctly if you have never done it before?  That was a concerning question during Initial LIGO since we had never detected a gravitational wave before. How do we know our data analyses are not missing them? And, when we do detect one, how do we know that the science we have extracted from the signal is reliable?

The answer is to do a blind injection test where only a select few expert administrators are able to put a fake signal in the data, maintaining strict confidentiality. They did just that in the early morning hours of 16 September 2010. Automated data analyses alerted us to an extraordinary event within eight minutes of data collection, and within 45 minutes we had our astronomer colleagues with optical telescopes imaging the area we estimated the gravitational wave to have come from. Since it came from the direction of the Canis Major constellation, this event picked up the nickname of the “Big Dog Event”.

For months we worked on vetting this candidate gravitational wave detection, extracting parameters that described the source, and even wrote a paper. Finally, at the next collaboration meeting, after all the work had been cataloged and we voted unanimously to publish the paper the next day. However, it was revealed immediately after the vote to be an injection and that our estimated parameters for the simulated source were accurate.

Again, there was no detection, but we learned a great deal about our abilities to know when we detected a gravitational wave and that we can do science with the data. This became particularly useful starting in September 2015.

paramater-2013Detection!

16 years, 3 months, 14 days after I started working with LIGO as a graduate student, I was (and still am) a scientist at the LIGO Livingston Observatory. I was sound asleep in my bed when the first detection of gravitational waves happened on 14 September 2015 at 04:50 CDT. Even when results were null, we used them to learn more than we knew before about our universe and about our abilities. All of the years of work of thousands of scientists, engineers, and technicians had finally paid off, and the new field of gravitational-wave astronomy was born.

I’m just as proud now to be a LIGO scientist as I was when all we had were years of null results.

Amber Stuver is a Data Analysis and EPO Scientist at the LIGO Livingston Observatory and an instructor of physics and astronomy at Louisiana State University. You can find her on ResearchGate and on Twitter (@livingligo).

Featured image: Artist impression of gravitational waves generated by binary neutron stars. Courtesy of Penn State.