Interview:

Lennart, the 3-day science meeting in Lund, Sweden, to mark your retirement is called „The science of Gaia and future challenges“. Before we come to the current science of Gaia and the future, can you tell us why a mission like Gaia is so important for astronomy?

Lennart Lindegren: As in all science, progress in astronomy is driven by an interplay between theory and observation. This interplay can take many different forms. One is that better observations produce more accurate data, which allow astronomers to test and improve their theoretical models. Gaia will for example give very accurate distances to stars in various stages of their evolution. This will allow astronomers to compute the total energy production in the stars, and compare with theoretical predictions. This will ultimately lead to a better understanding of the life cycles of stars.

In this example, it might be sufficient to study a few thousand well-chosen stars. But here we come to another important aspect of Gaia: how can we know which stars are most interesting to study? For this we first need to look at all the possible candidates, which Gaia will also do. By exploring all the objects within reach of its telescope, Gaia provides data that are not biased by our preconceived ideas about what is important and what is not. A third important aspect is that astronomers will find new, unexpected patterns in the data, discoveries that eventually lead to a deeper understanding of the universe.

You were already involved in Gaia’s predecessor HIPPARCOS, a satellite which measured the position, motions, and distances of more than 100,000 stars between 1989 and 1993. Can you tell us what your work for this mission was?

Lennart Lindegren: In 1976, while still a PhD student in Lund, I became involved in HIPPARCOS thanks to Erik Høg in Copenhagen, who was then my de facto supervisor. My interest was immediately caught by the challenges of data analysis for the future mission, and by questions on the optimisation of the satellite. From 1980, when HIPPARCOS was approved by ESA, I worked with colleagues in the UK and Denmark on preparations for NDAC, one of the three HIPPARCOS data analysis consortia, and in 1990 I took over from Erik as the leader of NDAC. 

What do you think was the biggest impact that the HIPPARCOS catalogue had on astronomy?

Lennart Lindegren: The Dutch astronomer Adriaan Blaauw once likened HIPPARCOS to a giant celestial vacuum cleaner. I rather like the image: in one big sweep the catalogue removed a lot of dust and cobweb that had accumulated in astronomy for more than a century. By this I mean the systematic errors in the positions, motions, and parallaxes of stars that would have prevented astronomers from getting the full benefit of the powerful new telescopes that came into operation at about the same time. The vacuuming left a clean and solid ground for further advances.   

What was your motivation to be involved in another astrometric mission after HIPPARCOS?

Lennart Lindegren: HIPPARCOS was a very unique experiment at the time. Nobody had done anything remotely similar before, and I was very lucky to have been part of it. At the same time we realised that it might be just the beginning of a much more ambitious development. Several of us were eager to explore these possibilities and make sure that HIPPARCOS would not remain a singularity. As always, I was probably more motivated by the technical and methodological challenges of a successor mission than of the science it might produce.

Directly after HIPPARCOS has finished its measurements, you and your colleagues made first proposals for a follow-up mission. Can you explain what your ideas were and how these compare to the actual design of Gaia?

Lennart Lindegren: In the early 1990s, when HIPPARCOS was still sending data to the ground, digital cameras (CCD detectors) were already used in the Hubble Space Telecope, and on ground they were revolutionising how astronomers used their telescopes. By contrast, HIPPARCOS was based on old-fashioned photomultiplier technology, designed a decade earlier when CCDs were simply not good enough for the purpose. Although the photomultipliers did a good job for HIPPARCOS, they are much less efficient than CCDs and can only observe one star at a time.

So around 1993 it was clear that one could do much better than HIPPARCOS simply by replacing the old detectors with the latest technology. But to really boost the accuracy in a follow-up mission it would also be necessary to have a bigger telescope. One of the mistakes we made then was to think that a cheap way of incleasing the size of the telescope is to use only the outer parts of the telescope mirror, in a so-called Fizeau interferometer. Hence the original name GAIA, an acronym for "Global Astrometric Interferometer for Astrophysics". The concept even called for two such interferometers, connected in an extremely stable structure. In retrospect this seems horrendously complicated and would probably not work.

Fortunately a much simpler configuration was soon adopted instead. The mission is still called Gaia, but since it is no longer an interferometer the name is not written in capital letters. The original (interferometric) design aimed to measure about 50 million of the brightest stars, while the actual Gaia observes objects that are a hundred times fainter, reaching 20-30 times as many stars. The projected accuracy, about 20 microarcsec at magnitude 15, is still the same as in the original proposal. I should add that Gaia would not have been possible without the advances in computer technology since HIPPARCOS. Both the data handling on-board and the subsequent data analysis would have been quite impossible with the computers of the 1980s.  

You are the intellectual father of the method to determine the stellar positions, the motions and the parallaxes  (distance measurements) from the individual measurements of the Gaia satellite. Can you explain us in a simple way how the Astrometric Global Iterative Solution AGIS works?

Lennart Lindegren: The idea is to construct first a network of perhaps ten million stars, covering the whole celestial sphere. These are the so-called primary sources. In the end all the stars, asteroids and galaxies observed by Gaia are linked to this network. At any time, Gaia's digital camera will "see" a small part of the sky, containing a few hundred primary sources, and their locations in the camera frame can be determined.

To build the network, all the measured locations of the primary sources need to be combined in the computer. The problem is that these locations are measured in the reference frame of the camera, and to translate them into celestial positions one must know the exact pointing of the camera at any time, and the geometrical layout of the detectors, optical distortions, etc. Mathematically speaking, we have a model of the observations with many unknowns, or parameters, describing not only the primary sources - their positions, parallaxes, and proper motions - but also the camera pointings, distortions, etc.

The Astrometric Global Iterative Solution is a way to find the combination of all the parameters that best "explains" the observations of the primary sources. In very simple terms you can think of it as assembling a gigantic jig-saw puzzle covering the whole surface of a globe - the celestial sphere. The pieces are the individual observations, and they must all fit together to cover exactly the whole globe. To account for camera distortion and other calibration effects you may have to adjust the size and shape of the pieces, but there is only one way to do it that will fit everywhere.

This explains why AGIS is "Global". It is also "iterative", which means that the adjustments are not made in one go, but in many small steps or iterations. With each iteration the fit gets a little bit better, until there is no further improvement after 100 iterations or so. At that point we have the desired primary solution. Linking the observations of the other sources to this network is then simply a matter of applying the camera pointings and calibrations as determined in the primary solution.  

In September last year the first Gaia catalogue was published. That Gaia Data Release 1 not only contains positions of more than one billion stars of our Milky Way and beyond, but also more than 2 million stars for which additionally their motions and parallaxes could be determined. How was it possible to determine these important astrometric parameters using effectively only about 11 months of Gaia data?

Lennart Lindegren: Getting the full information from such a short stretch of data is difficult, because you need a certain minimum number of observations of each star to run AGIS successfully. For the final catalogue we will have at least five years of data, which gives many more observations per star than the required minimum. With 11 months a solution is marginally possible, but it would not be very useful because the proper motions are poorly determined, which in turn makes the parallaxes poor. So what we did was to use the 2 million brightest stars as primary sources and fold in their positions around 1991, as determined by the HIPPARCOS satellite and available in the Tycho-2 catalogue.

The combination of the old HIPPARCOS positions with the new Gaia data gave a good determination of the proper motions for these stars and then also their parallaxes. This so-called Tycho-Gaia Astrometric Solution (TGAS) was however special for Gaia Data Release 1 - future releases will be based exclusively on Gaia data.

You are strongly involved in the production of the second Gaia Data Release planned for April 2018.  What can be expected for this second Gaia catalogue and what are the major challenges you have to deal with?

Lennart Lindegren: The second release will be based on nearly two full years of Gaia observations. This means that there is no longer any need to rely on the earlier data from HIPPARCOS, and we are not limited to the two million brightest stars. Gaia DR2 will give positions, parallaxes, and proper motions for more than a billion stars, including photometric information - magnitudes and colours - and in many cases also spectroscopically determined radial velocities.

Future releases will be much more accurate, and contain many more kinds of data, but DR2 is the first time the community will feel the full extent of Gaia's potential. It is a huge step from DR1.  

Each new release brings new challenges. With the steadily accumulating data the AGIS solutions become better all the time, which puts more demands on both computers and people. The improved solutions reveal new and subtle details about how the instrument behaves, which need to be analysed and understood, and incorporated in future solutions. So the analysis is in continuous development. This is a time-consuming and difficult, but fascinating task.

What new science can be expected from Gaia DR2?

Lennart Lindegren: One of the strengths of Gaia is that it serves so many different areas of astrophysics. This makes it hard to predict what the new science will be. However, I expect some very interesting results from the detailed mapping of stars, and how they move, in "our" part of the galaxy, within a few thousand light years from the Sun. This could shed some light on how the spiral arms work and the distribution of the mysterious dark matter. 

Now that you get officially retired you will still proceed with your work on Gaia. Why is this work so fascinating for you? What will be your future involvement in the challenges of the Gaia mission?

Lennart Lindegren: As mentioned before, I am fascinated by a number of technical aspects of the work, in particular trying to understand how Gaia works and how this understanding can be used to improve the end results. That is why it is such an exciting time now. We are learning to know the instrument on the level needed to reach the intended final accuracy, or perhaps even better. And we can see what a magnificent instrument Gaia is! It is really a privilege to be involved in this work which I hope will go on for many more years.