Gravitational waves and data analysis

I. Coordinator:
  • Feng-Li Lin (National Taiwan Normal University)

II. Core Members:

  • Core members
  • Chian-Shu Chen (Tamkang University), Ting-Wai Chiu (National Taiwan University/National Taiwan Normal, University/Institute of Physics, Academia Sinica), Sadakazu Haino (Institute of Physics, Academia Sinica), Yuki Inoue (National Central University), Chun-Yu Lin (National Center for High-performance Computing), Guo-Chin Liu (Tamkang University)
  • Postdocs
  • Ling-Wei Luo (Institute of Physics, Academia Sinica), Alessandro Parisi (Tamkang University), Hong Zhang (National Taiwan Normal University)


III. Research Themes:

  • Theoretical studies of dynamics for the sources of gravitational waves
  • Gravitational wave (GW) signals encode the dynamical information for the black holes, neutron stars and exotic objects such as dark stars, as well as the possible modification of Einstein gravity. Therefore, it is important to have theoretical studies of source dynamics. Currently, we are working on
  • (i) the equation of state (EOS) of neutron stars from AdS/QCD, and its astronomical implication to gravitational wave physics (ASIOP/NTNU)
  • (ii) the tidal effect of dark stars or boson stars (TKU)
  • (iii) the effective field theory for post-Newtonian/post-Minkowski approach for modified gravity (NTNU/University of Science and Technology, China)
  • (iv) weak cosmic censorship for modified gravity (Caltech/NTNU/Sichuan University)
  • (v) self-force of ring string around black holes (NTNU)
  • Data analysis of gravitational waves for LIGO/Virgo/KAGRA (LVK)
  • We have been working on data analysis for KAGRA for more than one year, and mainly focus on GPU acceleration for both low- and high-latency searches. We also have tested the machine learning techniques on GW detections, which is comparable with state of art. In the coming year, we would like to focus on some specific tasks for O3 of LVK, which will share their data to LVK members. These include

  • (i) For low latency part, we will work on channel noise calibration and deglitching, with two possible methods: (a) by adopting Independent Component Analysis (ICA) or Principal Component Analysis (PCA) with GPU acceleration with the real data from KAGRA. (b) Using deep learning techniques such as Generative Adversarial Nets (GAN) or Variational AutoEncoder (VAE) to train with the current glitch classes (such as the ones from Gravity Spy) and generate random glitches, which should be the most general glitch patterns. We can then mix them with the gravitational wave templates to train Convolution Neural Network (CNN) for de-glitching. The method (a) could be used for linear mixings of signals and glitches, and (b) for nonlinear mixings.We can also try to incorporate new sky location algorithm developed by one of our student members (collaborating with Linqin Wen’s group at University of West Australia) into the low-latency search.
  • Currently, AS group is providing a law-latency calibration machine for KAGRA (with CPU base) and h(t) reconstruction pipeline with gstlal. We
    possibly can extend this scheme with a few GPU clusters and try the superlow latency de-glitching or signal search at KAGRA online data . Then once we can demonstrate it with KAGRA data we can negotiate with LIGO/Virgo to apply our approach to world wide network. (ASIOP/NTNU/NCHC/TKU)
  • (ii) For high latency part, we will work on the Test of General Relativity with GPU-based Parameter Estimation (P. E.). To do this, we need to parametrize the deviation of waveform from the standard one then make posterior distributions with P.E. framework. Or, we can work on some specific source dynamics or modified gravity, which could be provided by our theory group as mentioned in (a), e.g., we can modify the dispersion relation of graviton to extract luminosity distance, which can then be used for standard siren for measuring the Hubble parameters; we can test EOS for neutron stars or boson stars studied by our theory group.
  • Acceleration of P.E. is essential for the many-detection (50~100 BBH) era at O3 and O4. Currently, Reduced Order Model (ROM) and Reduced Order Quadrature (ROQ) are used e.g. https://journals.aps.org/prl/abstract/10.1103/PhysRevLett.114.071104 which can accelerate by 100~1000 for P.E. with EOB waveform and it is essential for current LIGO/Virgo analysis. However, we need to prepare for ROM in advance for a specified waveform and it cannot be easily applied for variety of modified waveforms. Thus, GPU-based P.E. acceleration has some advantages for the modified waveform analysis for many (50~100) BBH data analysis in O3, O4 era. (ASIOP/NTNU/NCHC/TKU)
  • (iii) Machine Learning is getting more and more common and competitive in GW community, so we want to think our originality. For the moment, besides what have mentioned in (i), we may also think about the possibility of using machine learning for P.E. in (ii), or generate the model GW waveforms, which are very costly in numerical relativity.
  • Future upgrade plan for KAGRA
  • We also try to push a collaboration between Taiwan, KAGRA and Caltech on the future upgrade plan for KAGRA based on a proposal by Caltech team by using entangled light for broad-band squeezing. This scheme is especially fitted for KAGRA as there is no need of building additional filtering cavity arms and can relieve KAGRA’s tight space for its underground design. We hope that a Taiwan local team working on quantum optics can join the upgrade plan if it is finally approved. Our core members will also join the theoretical evaluations and studies of the proposal.
IV. Tentative plan of activities:
  • • We will hold weekly group meetings for all members to push our core projects listed above, and guide the students their progresses.
  • • We will hold an annual international 3 to 4-day workshop/school on the recent issues in GW physics.
  • • The core members will attend the regular KAGRA meetings and the related workshops, and visit the institutes in the world working on the GW physics.
  • • We will hold irregular seminar or one-day workshop by inviting speakers from other institutes inside or outside Taiwan.

VExpected achievements:

  • We expect to join O3 of LVK to accomplish the listed tasks in the above, and participate the discoveries of GW events, and then check some of our theoretical predictions of gravity theory. This will also enhance our visibility and collaborations in the GW community of the world. In the meantime, we also hope to promote the local research activities on GW physics, and encourage more young peoples to join this field, for which we expect there is a flourish future in the next 20 or 30 years.

cron web_use_log