Gaia is a cornerstone mission in the science programme of the European Space Agency (ESA). The spacecraft construction was approved in 2006, following a study in which the original interferometric concept was changed to a direct-imaging approach. Both the spacecraft and the payload were built by European industry. The involvement of the scientific community focusses on data processing for which the international Gaia Data Processing and Analysis Consortium (DPAC) was selected in 2007. Gaia was launched on 19 December 2013 and arrived at its operating point, the second Lagrange point of the Sun-Earth-Moon system, a few weeks later. The commissioning of the spacecraft and payload was completed on 19 July 2014. The nominal five-year mission started with four weeks of special, ecliptic-pole scanning and subsequently transferred into full-sky scanning mode. We recall the scientific goals of Gaia and give a description of the as-built spacecraft that is currently (mid-2016) being operated to achieve these goals. We pay special attention to the payload module, the performance of which is closely related to the scientific performance of the mission. We provide a summary of the commissioning activities and findings, followed by a description of the routine operational mode. We summarise scientific performance estimates on the basis of in-orbit operations. Several intermediate Gaia data releases are planned and the data can be retrieved from the Gaia Archive, which is available through the Gaia home page.

The Gaia satellite will survey the entire celestial sphere down to 20th magnitude, obtaining astrometry, photometry, and low resolution spectrophotometry on one billion astronomical sources, plus radial velocities for over one hundred million stars. Its main objective is to take a census of the stellar content of our Galaxy, with the goal of revealing its formation and evolution. Gaia's unique feature is the measurement of parallaxes and proper motions with hitherto unparalleled accuracy for many objects. As a survey, the physical properties of most of these objects are unknown. Here we describe the data analysis system put together by the Gaia consortium to classify these objects and to infer their astrophysical properties using the satellite's data. This system covers single stars, (unresolved) binary stars, quasars, and galaxies, all covering a wide parameter space. Multiple methods are used for many types of stars, producing multiple results for the end user according to different models and assumptions. Prior to its application to real Gaia data the accuracy of these methods cannot be assessed definitively. But as an example of the current performance, we can attain internal accuracies (rms residuals) on F, G, K, M dwarfs and giants at G = 15 (V = 15-17) for a wide range of metallicites and interstellar extinctions of around 100 K in effective temperature (T eff ), 0.1 mag in extinction (A 0 ), 0.2 dex in metallicity ([Fe/H]), and 0.25 dex in surface gravity (log g). The accuracy is a strong function of the parameters themselves, varying by a factor of more than two up or down over this parameter range. After its launch in December 2013, Gaia will nominally observe for five years, during which the system we describe will continue to evolve in light of experience with the real data.

Context. The Gaia catalogue will contain observations and physical parameters of a vast number of objects, including ultra-cool dwarf stars, which we define here as stars with a temperature below 2500 K. Aims. We aimed to assess the accuracy of the Gaia T eff and log (g) estimates as derived with current models and observations. Methods. We assessed the validity of several inference techniques for deriving the physical parameters of ultra-cool dwarf stars: Gaussian processes, support vector machines, k-nearest neighbours, kernel partial least squares and Bayesian estimation. In addition, we tested the potential benefits of data compression for improving robustness and speed. We used synthetic spectra derived from ultracool dwarf models to construct (train) the regression models. We derived the intrinsic uncertainties of the best inference models and assessed their validity by comparing the estimated parameters with the values derived in the bibliography for a sample of ultra-cool dwarf stars observed from the ground. Results. We estimated the total number of ultra-cool dwarfs per spectral subtype, and obtained values that can be summarised (in orders of magnitude) as 400 000 objects in the M5−L0 range, 600 objects between L0 and L5, 30 objects between L5 and T0, and 10 objects between T0 and T8. A bright ultra-cool dwarf (with T eff = 2500 K and log (g) = 3.5) will be detected by Gaia out to approximately 220 pc, while for T eff = 1500 K (spectral type L5) and the same surface gravity, this maximum distance reduces to 10−20 pc. We found the cross-validation RMSE prediction error to be 10 K for regression models based on the k-nearest neighbours and 62 K for Gaussian process models in the faintest limit (Gaia magnitude G = 20). However, these values correspond to the evaluation of the regression models with independent test sets of synthetic spectra of the same model families as used in the training phase (internal errors). For the k-nearest neighbours model, this seems an overly optimistic error estimate due to the use of a dense grid of examples in the training set, together with a relatively high signal-to-noise ratio for the end-of-mission data. The RMSE of the prediction deduced from ground-based spectra of ultra-cool dwarfs simulated at the Gaia spectral range and resolution, and for a Gaia magnitude G = 20 is 213 K and 266 K for the models based on k-nearest neighbours and Gaussian process regression, respectively. These are total errors in the sense that they include the internal and external errors, with the latter caused by the inability of the synthetic spectral models (used for the construction of the regression models) to exactly reproduce the observed spectra, and by the large uncertainties in the current calibrations of spectral types and effective temperatures. We found maximum-likelihood methods (minimum χ 2 , k-nearest neighbours, and Bayesian estimation with flat priors) to be biased in the L0-T0 range in that they systematically assign a temperature around 1700 K. Finally, the likeliho...