Back to home

Abnormal correlations in the LIGO data

Hao Liu, AEI Hannover

September 2017

On September 7th, 2017 Hao Liu presented a seminar in AEI Hannover. This talk contained a significant number of new results regarding the analysis of LIGO data. Since we believe that these results constitute an interesting and important contribution to the ongoing discussion, we have decided to post the text of his talk. A few preliminary remarks may be useful:

Pages 5 to 7 remind the reader of the strong and unexpected correlation in the residuals that result when the GW template is subtracted from the strain data. These correlations, originally found in the immediate vicinity of the event, have since also been seen as a "precursor effect" well before the GW event (where the template is small) and as an "echo effect" after the event (where the template is strictly zero). These correlations have now been shown to be present both for the original published template and for LIGO's unpublished "best template".

Pages 9 to 17 describe the construction of a "Best Common Signal" obtained without the use of a template. The strategy adopted is to determine that common signal that provides a maximal reduction of residual correlations. The results of this calculation clearly indicate that the LIGO template is not a satisfactory estimator of this best common signal. The consequence is that one must choose between a GW template with strong residual correlations or a best common signal that is inconsistent with a GW template.

Pages 21-24 describe a serious de facto degeneracy in LIGO's templates. Obviously, the cross-correlation in pure GW templates is small when the black hole masses are different. These differences are most pronounced at times well before their merger when the frequencies of the associated strain signal are low. However, such distinctions are effectively eliminated when band-pass filtering is performed to eliminate detector noise. We present an example in which the mass of the heavier black hole is increased from 36 to 43 solar masses without any significant change in the resulting matched and band-passed template. This seriousness of this degeneracy is further emphasized on page 24 where it is shown that the template for GW150914 (with masses of 29 and 36 solar masses) provides an excellent description of the strain data for GW170104 (with best fit masses of 19.4 and 31.2 solar masses). In other words, LIGO's estimates of the uncertainties associated with black hole masses appears to be seriously underestimated.

The results described here provide additional support for our conviction that LIGO's current data analysis techniques are not sufficiently reliable to support their claims. These extraordinary discoveries require far better proof.

A manuscript providing a more detailed description of these new results will soon be available. In the meantime, please feel free to contact us if you have any questions.

Link to presentation (preliminary version)