A spin-exchange relaxation-free (SERF) atomic co-magnetometer achieves angular velocity measurement by determining the probe laser's optical rotation angle. The detection system's polarization error, as a direct interference quantity, is coupled with the output signal and leads to the signal drifting, which sufficiently decreases the inertial measurement system's long-term stability. In the SERF atomic co-magnetometer detection system, the polarization error primarily originates from the non-desirable elliptical polarization component and polarization azimuth error. Most studies focus on the analysis and suppression of the non-desirable elliptical polarization component, but the polarization azimuth error caused by the residual rotation angle has not been widely researched. Thus, the signal drift caused by this polarization error needs to be resolved immediately, and the accuracy of the SERF atomic co-magnetometer needs to be improved. In this study, a polarization error suppression approach based on the optimization of polarization fluctuation sensitivity coefficient is suggested. The accuracy of the atomic co-magnetometer is enhanced by modifying the probe laser wavelength corresponding to the scale coefficient's maximum value and decreasing the atomic vapor cell's temperature. Furthermore, this approach does not add additional devices that can introduce other noises and is simpler to scale down.

In this study, by optimizing the polarization fluctuation sensitivity coefficient, the polarization error of the SERF atomic co-magnetometer is suppressed. First, the equation for the output signal of the SERF atomic co-magnetometer under ideal conditions is modified, considering that the actual atomic spin precession detection system will introduce inevitable residual optical rotation angle error due to the imperfect manufacturing process of optical components or the changes of the ambient temperature and mechanical stress. The influence of the spurious angular velocity caused by the residual optical rotation angle on the co-magnetometer drift is examined according to the modified output signal expression, and the polarization fluctuation sensitivity coefficient is obtained. This coefficient is related to the probe laser wavelength and the atomic cell's temperature. These two parameters are further adjusted to the position where the co-magnetometer is least sensitive to polarization fluctuations, therefore realizing the polarization error suppression. Furthermore, an SERF atomic co-magnetometer is constructed and the suggested approach is experimentally confirmed. The probe laser's wavelength is varied by adjusting the laser source's temperature control parameters. The atomic cell's temperature is controlled using a non-magnetic electric heating system combined with a proportion integration differentiation (PID) closed-loop controller. Under different conditions, the scale coefficient and static drift tests of the atomic co-magnetometer are conducted. Finally, the Allan standard deviation is applied to examine the test findings.

The spurious angular velocity caused by the residual optical rotation angle introduces polarization error to the SERF atomic co-magnetometer. The correlation between the residual optical rotation angle and spurious angular velocity is expressed by the polarization fluctuation sensitivity coefficient. When the scale coefficient of the co-magnetometer is the maximum, the polarization fluctuation sensitivity coefficient has the minimum value [Fig. 2(a)]. Thus, the co-magnetometer is the least sensitive to polarization fluctuation at that time. The scale coefficient varying with the proble laser wavelength has a Lorentzian lineshape (Fig. 4) and peaks at one wavelength detuned from the atomic resonance wavelength. Furthermore, with the increase of atomic cell temperature, the polarization fluctuation coefficient increases monotonically [Fig. 2(b)]. The polarization error is also suppressed by appropriately reducing the atomic cell temperature under the condition that the atomic co-magnetometer is maintained at the SERF region. The drift error analysis of the test data of the SERF atomic co-magnetometer is conducted by Allan variance, and the bias instability is decreased from 0.012 (°)/h to 0.008 (°)/h (Fig.6). The polarization error of the co-magnetometer is efficiently suppressed.

In this study, the influence of the polarization error of the SERF atomic co-magnetometer detection system on the output signal drift is examined. And we quantitatively investigate the co-magnetometer drift caused by the residual optical rotation angle using the polarization fluctuation sensitivity coefficient. This polarization fluctuation sensitivity coefficient converts the change of unit residual optical rotation angle into the resulting spurious angular velocity's change. A parameter optimization method is then suggested to reduce the polarization fluctuation sensitivity coefficient by adjusting the probe laser wavelength and atomic vapor cell temperature, which can minimize the effect of the polarization error caused by the residual optical rotation angle on the co-magnetometer drift. The experimental confirmation is performed on the designed SERF atomic co-magnetometer. The finding demonstrates that the polarization error of the SERF atomic co-magnetometer is suppressed by modifying the probe laser wavelength corresponding to the scale coefficient's maximum value and suitably reducing the atomic vapor cell's temperature. The drift error analysis of the SERF atomic co-magnetometer test data is performed by Allan variance, and the bias instability is reduced from 0.012 (°)/h to 0.008 (°)/h, confirming the effectiveness of the suggested polarization error suppression method.