- Physical Review Journals
All JournalsPhysics Magazine

Physical Review E

Reuse & Permissions

It is not necessary to obtain permission to reuse this article or its components as it is available under the terms of the Creative Commons Attribution 4.0 International license. This license permits unrestricted use, distribution, and reproduction in any medium, provided attribution to the author(s) and the published article's title, journal citation, and DOI are maintained. Please note that some figures may have been included with permission from other third parties. It is your responsibility to obtain the proper permission from the rights holder directly for these figures.

Export citation

Export citation

Choose format for download:

Download Citation
  • Open Access

Granger causality maps for Langevin systems

Lionel Barnett*, Benjamin Wahl†, Nadine Spychala, and Anil K. Seth

  • Sussex Centre for Consciousness Science, Department of Informatics, University of Sussex, Falmer, Brighton, United Kingdom

PDF

Phys. Rev. E 113, 064105 – Published 1 June, 2026

DOI: https://doi.org/10.1103/96ch-bxmf

Abstract

Wahl et al. [Phys. Rev. E 93, 022213 (2016)] introduced Granger causality (GC) maps for Langevin systems: Dynamics are localized linearly at each point in phase space as vector Ornstein-Uhlenbeck (VOU) processes, for which GCs may in principle be calculated, thus constructing a GC map on phase space. These maps may, furthermore, be averaged over phase space to yield a systemwide global GC value. The implementation, however, suffered some significant drawbacks: GCs were approximated from models based on discrete-time stroboscopic sampling of local VOU processes, which is not only computationally inefficient but, more seriously, infeasible on regions of phase space where local dynamics are unstable, leaving “holes” in the GC maps. We solve these problems by deriving an analytical expression for GC rates associated with a VOU process which, under quite general conditions, yields a meaningful solution even in the unstable case. Applied to GC maps, this not only “fills in the holes” but also furnishes a computationally efficient method of calculation devolving to solution of algebraic Riccati equations which, in the case of a univariate source, become simple quadratic equations. We show, furthermore, that the GC rate for VOU processes is invariant under rescaling of the overall fluctuations intensity, so that GC maps may effectively be calculated for deterministic nonlinear dynamical systems, with a residual “ghost of noise” represented by a variance-covariance map.

View figure in article

Physics Subject Headings (PhySH)

Article Text

References (49)

  1. N. Wiener, The theory of prediction, in Modern Mathematics for Engineers, edited by E. F. Beckenbach (McGraw–Hill, New York, 1956), pp. 165–190.
  2. C. W. J. Granger, Economic processes involving feedback, Inform. Control 6, 28 (1963).
  3. J. Geweke, Measurement of linear dependence and feedback between multiple time series, J. Am. Stat. Assoc. 77, 304 (1982).
  4. J. Geweke, Measures of conditional linear dependence and feedback between time series, J. Am. Stat. Assoc. 79, 907 (1984).
  5. It is underappreciated that even if a (stationary) stochastic process has a nonlinear generative mechanism, it may nonetheless be amenable to linear modelling. Specifically, under mild conditions it will have a Wold representation as a linear moving-average which may, if minimum-phase conditions hold, be suitable for Granger-causal analysis.
  6. T. Schreiber, Measuring information transfer, Phys. Rev. Lett. 85, 461 (2000).
  7. M. Paluš, V. Komárek, Z. Hrnčíř, and K. Štěrbová, Synchronization as adjustment of information rates: Detection from bivariate time series, Phys. Rev. E 63, 046211 (2001).
  8. L. Barnett, A. B. Barrett, and A. K. Seth, Granger causality and transfer entropy are equivalent for Gaussian variables, Phys. Rev. Lett. 103, 238701 (2009).
  9. R. E. Spinney, M. Prokopenko, and J. T. Lizier, Transfer entropy in continuous time, with applications to jump and neural spiking processes, Phys. Rev. E 95, 032319 (2017).
  10. P. Shahsavari Baboukani, C. Graversen, E. Alickovic, and J. Østergaard, Estimating conditional transfer entropy in time series using mutual information and nonlinear prediction, Entropy 22, 1124 (2020).
  11. B. Wahl, U. Feudel, J. Hlinka, M. Wächter, J. Peinke, and J. A. Freund, Granger-causality maps of diffusion processes, Phys. Rev. E 93, 022213 (2016).
  12. B. Wahl, U. Feudel, J. Hlinka, M. Wächter, J. Peinke, and J. A. Freund, Conditional Granger causality of diffusion processes, Eur. Phys. J. B 90, 197 (2017).
  13. J.-P. Florens and D. Fougère, Noncausality in continuous time, Econometrica 64, 1195 (1996).
  14. L. Barnett and A. K. Seth, Detectability of Granger causality for subsampled continuous-time neurophysiological processes, J. Neurosci. Methods 275, 93 (2017).
  15. G. T. Wilson, The factorization of matricial spectral densities, SIAM J. Appl. Math. 23, 420 (1972).
  16. M. Dhamala, G. Rangarajan, and M. Ding, Estimating Granger causality from Fourier and wavelet transforms of time series data, Phys. Rev. Lett. 100, 018701 (2008).
  17. That is, global GC rates; as VOU processes are linear, there is nothing to gain from local linearization – see Section 3.
  18. L. Barnett and A. K. Seth, Granger causality for state-space models, Phys. Rev. E 91, 040101(R) (2015).
  19. B. Øksendal, Stochastic Differential Equations: An Introduction with Applications (Springer-Verlag, Berlin, 2003).
  20. Throughout, subscripts 1,2,3 on vectors and matrices are multi-indices corresponding to the given partitioning.
  21. L. Barnett and T. Bossomaier, Transfer entropy as a log-likelihood ratio, Phys. Rev. Lett. 109, 138105 (2012).
  22. A. K. Seth, Causal networks in simulated neural systems, Cogn. Neurodyn. 2, 49 (2008).
  23. We assume all requisite derivatives of f(y) and Σ(y) exist.
  24. For Langevin systems, “multiplicative” refers to noise with state-dependent magnitude [49]). It is unclear to us how restrictive the weakly multiplicative assumption might be in the context of the various physical scenarios typically modelled by Langevin equations; while it would be of interest to consider whether the theory might be extended to relax this assumption, we leave that for future work.
  25. E. N. Lorenz, Deterministic nonperiodic flow, J. Atmos. Sci. 20, 130 (1963).
  26. The time unit is arbitrary labelled as “seconds.”
  27. The ode45 solver uses an adaptive step size which may not correspond to the sampling increment h.
  28. L. Bereska and S. Gavves, Mechanistic interpretability for AI safety—A review, arXiv:2404.14082.
  29. F. Comte and E. Renault, Noncausality in continuous time models, Econ. Theory 12, 215 (1996).
  30. K. Kawasaki, Simple derivations of generalized linear and nonlinear Langevin equations, J. Phys. A 6, 1289 (1973).
  31. V. Solo, State-space analysis of Granger-Geweke causality measures with application to fMRI, Neural Comput. 28, 914 (2016).
  32. W. Arnold and A. J. Laub, Generalized eigenproblem algorithms and software for algebraic Riccati equations, Proc. IEEE 72, 1746 (1984).
  33. L. Barnett, J. T. Lizier, M. Harré, A. K. Seth, and T. Bossomaier, Information flow in a kinetic Ising model peaks in the disordered phase, Phys. Rev. Lett. 111, 177203 (2013).
  34. See, for example, NVIDIA's cuSOLVER, which provides GPU platform support for various matrix decompositions (QR, SVD, LU, eigensolvers, etc.) from which CARE solvers may be constructed.
  35. We have encountered one curious edge case for which we do not as yet have an adequate explanation. Here, though the requisite Jacobians are nonsingular and the appropriate detectability criteria are indeed satisfied, QZ-based CARE solvers fail; solution matrices blow up in magnitude, yielding inaccurate results. This may possibly be a frailty of the Schur decomposition algorithm which, while apparently rare, merits future investigation.
  36. R. Hindriks, R. Jansen, F. Bijma, H. D. Mansvelder, M. C. M. de Gunst, and A. W. van der Vaart, Unbiased estimation of Langevin dynamics from time series with application to hippocampal field potentials in vitro, Phys. Rev. E 84, 021133 (2011).
  37. M. Tabar and R. Rahimi, Kramers-Moyal expansion and Fokker-Planck equation, in Analysis and Data-Based Reconstruction of Complex Nonlinear Dynamical Systems: Using the Methods of Stochastic Processes (Springer International, Cham, 2019), pp. 19–29.
  38. P. P. Lin, M. Wächter, J. Peinke, M. Tabar, and R. Rahimi, Assessing the suitability of the Langevin equation for analyzing measured data through downsampling, J. Phys.: Complexity 6, 015016 (2025).
  39. L. Barnett and A. K. Seth, Dynamical independence: Discovering emergent macroscopic processes in complex dynamical systems, Phys. Rev. E 108, 014304 (2023).
  40. https://github.com/lcbarnett/VOUGC.
  41. J. D. Hamilton, Time Series Analysis (Princeton University Press, Princeton, NJ, 1994).
  42. Although Barnett and Seth [14] consider only the stable case, the result may be shown to hold in the unstable case.
  43. A. J. Gutknecht and L. Barnett, Sampling distribution for single-regression Granger causality estimators, Biometrika 110, 933 (2023).
  44. M.-L. Ni, Existence condition on solutions to the algebraic Riccati equation, Acta Automat. Sinica 34, 85 (2008).
  45. J. N. Cooper and C. D. Edgar, A development of continuous-time transfer entropy, arXiv:1905.06406.
  46. Up to an additive constant that depends only on the dimension of the variable y.
  47. This confirms that, as our notation suggests, the expression for Ty3→y1|y2(h) in (B22) does not depend on the time stamp t.
  48. T. Kailath, Linear Systems (Prentice-Hall, Hoboken, NJ, 1980).
  49. P. Arnold, Symmetric path integrals for stochastic equations with multiplicative noise, Phys. Rev. E 61, 6099 (2000).

Outline

Information

Phys. Rev. E 113, 064105– Published 1 June, 2026
  • Received 7 January 2026
  • Accepted 13 May 2026
Crossmark - Check for updates button
Creative Commons Logo

Published by the American Physical Society under the terms of the Creative Commons Attribution 4.0 International license. Further distribution of this work must maintain attribution to the author(s) and the published article's title, journal citation, and DOI.

Published by the American Physical Society

Sign In to Your Journals Account

Filter

Filter

Article Lookup

Enter a citation