UC logo


    We would like to thank all people and organisations that make this site possible. We are very grateful to both the MOA and OGLE collaborations that freely allow us to use their data, as well as all other contributing observers that help provide follow up observations. All observational data remains the property of the original owner, and they should be the first contact if you wish to use their data.
    We would like to acknowledge the generous donation by NVIDIA of a Tesla GPU, that was gifted to us for the purpose of enhancing our gravitational microlensing research.


    I am a PhD student at the University of Canterbury New Zealand researching gravitational microlensing. My research is focused on designing a fast new modelling approach to analyse binary microlensing events, with the goal of detecting new exoplanets. Typically, this analysis takes a long time and large computational resources. I aim to produce a new routine that is fast and does not require the same scale of computational resources. This website has been created to display the latest results from the automated single lens modelling process and all binary analysed events.


    All gravitational microlensing events on this website are modelled using our own modelling routines. Relying on the dedicated microlensing survey telescopes our system automatically keeps up to date with the latest microlensing events and rapidly produces single lens models to fit the data. It does not use conventional simplex routines to fit the data and is unique in that it includes bad data points in it's modelling (no sigma/kappa clipping). Using a specially adapted likelihood function it models the bad data and provides an estimate of the probability of the data quality.
    From the single lens modeling it is possible to detect deviations from the model. These are flagged as a possible binary event which are then processed using our new unique modelling method. Our method utilizes the processing power of GPU devices, by specially adapting the modelling procedure it is possible to perform a complete grid search of a large parameter space within a minimal time. The outcome of the grid search procedure can then put into a markov chain monte carlo process that narrows down the solution and is also able to include higher order effects, as appropriate.
    All models and the progress during the modelling process are published via this website, to keep all visitors up to date with the latest analysis of current events.


    The single lens modelling procedure uses a cluster of standard desktop computers to run a markov chain monte carlo procedure via the python module emcee. This process fits the single lens model including the probability of bad data (one per data set) for every detected microlensing event published on the MOA and OGLE alert pages. The data retrieval and analysis system is fully automated to allow for the fastest possible analysis of data.
    The binary part of the website if achieved in a different way. Once an event has manually been identified as a possible binary, an administrator can mark it as a possible binary candidate and initialise the modelling procedure. This process involves analytically solving the binary lens equation to produce a magnification map for various combinations of the d (mass seperation), q (mass ratio) parameter space, these maps are then convolved for various ρ values (source sizes). This set of convolved magnification maps can then be interpolated across to produce an analytically solved light curve for a given binary lens source trajectory.


  • MOA - A microlensing survey group.
  • OGLE - A microlensing survey group.
  • NVIDIA - Gifted us a TESLA GPU for our research.
  • EMCEE - A python Markov Chain Monte Carlo module.
  • D.Hogg, J.Bovy, D.Lang - Modelling data, using likelihood functions.