Search for continuous gravitational waves using a novel coherent tree-based pruning algorithm

Continuous gravitational waves are long-duration, nearly monochromatic signals expected from rapidly rotating, non-axisymmetric neutron stars. Detecting them requires sensitive, phase-coherent integration over long timescales, which becomes computationally prohibitive due to the exponential scaling of parameter space in frequency, sky location, orbital parameters, etc. We present a dynamic, tree-based pruning algorithm designed to overcome this challenge. The method coherently combines Short-Time Fourier Transform segments while recursively exploring and pruning parameter space using a structured branching scheme. Each node in the tree represents a candidate parameter set, which is expanded and statistically tested using phase-tracking models based on Taylor (or Chebyshev) expansions. Only the most promising branches are retained at each stage. A key innovation is a dynamic thresholding strategy that balances detection probability against computational cost. The algorithm adaptively selects pruning thresholds using survival probabilities and complexity metrics, maintaining a manageable subset of hypotheses throughout the search. This pruning algorithm offers a promising approach to improving the computational efficiency of continuous gravitational wave searches. By reducing the cost of exploring large parameter spaces, it may help enable deeper searches over longer observation times and improve sensitivity to low signal-to-noise ratio sources that are currently challenging to detect with existing methods.