Varying alpha, blinding, and bias in existing measurements

Abstract

ABSTRACT The high resolution spectrograph ESPRESSO on the VLT allows measurements of fundamental constants at unprecedented precision and hence enables tests for space–time variations predicted by some theories. In a series of recent papers, we developed optimal analysis procedures that expose and eliminate the subjectivity and bias in previous quasar absorption system measurements. In this paper, we analyse the ESPRESSO spectrum of the absorption system at zabs = 1.15 towards the quasar HE 0515-4414. Our goal here is not to provide a new unbiased measurement of Δα/α in this system (that will be done separately). Rather, it is to carefully examine the impact of blinding procedures applied in many previous measurements of the fine structure constant in quasar absorption systems. To do this, we emulate previous procedures, using supercomputer Monte Carlo AI calculations to generate a large number of independently constructed models of the absorption complex. Each model is obtained using ai-vpfit, with Δα/α fixed until a ‘final’ model for the absorption system is obtained, at which point Δα/α is then released as a free parameter for one final optimization. The results show that the value of Δα/α obtained in this way is systematically biased towards the initially fixed value i.e. this process produces measurements that are unrelated to the true value of Δα/α. The implication is straightforward: to avoid bias, all future measurements must include Δα/α as a free parameter from the beginning of the modelling process.