Polarization-resolved laser-induced breakdown spectroscopy (PRLIBS) has important significance in reducing the detection limit of LIBS with its advantage of background suppressing. However, the controversy in theory of polarization and the unstable improvement of the signal-to-back ratio limit its application prospects. In order to investigate the theory of polarization characteristics and the improvement of the signal-to-back ratio, this study used 1 064 nm nanosecond pulsed laser and fiber optic spectrometer to conduct an exploratory experiment on the improvement of polarization-resolved LIBS signal-to-back ratio and polarization mechanism of aluminum-iron alloy samples. According to estimation of the energy of bremsstrahlung radiation, it’s confirmed that proportion of bremsstrahlung radiation in background reduces with time. By varying the energy density, detection angle, analyzer angle, delay, integration time and other factors, the spectral intensity and wavelength data were collected, and polarization degree and the signal-to-background ratio were calculated. It was observed that the background and the discrete spectrum of the plasma spectrum of the aluminum-iron alloy was partial polarized. There were phenomena of polarization of spectrum and there were differences in polarization degree and directions of polarization. It was found that the effect of polarized LIBS to improve the signal-to-back ratio is related to experimental parameters including energy density, delay time, detection angle and wavelength: signal-to-back ratio of PRLIBS varies with respect to energy density similar to that of LIBS. SBR becomes low and flat when the energy density is large. The analyzer angle affects the SBR, which is related to the polarization direction and degree of polarization of the spectrum. The formula for improvement about the polarization degree, the angle of detection and the polarization direction is derived. The polarization degree of continuum is flat at all wavelengths and polarization degree of the discrete spectrum decreases as the spectral intensity increases. Though not obvious, the polarization degree changes with the delay increasing. The reason is that the amount of change in the delay time is too little compared to the integration time. The trend of the signal-to-back ratio is consistent with non-polarized LIBS. The representative explanations of the PRLIBS mechanism at home and abroad are summarized and discussed. It’s proved that laser field, Fresnel reflection effect, anisotropic electron velocity distribution and other factors play an inconclusive role in the polarization properties of the plasma. The conclusion is that in the ns-LIBS experiment, most of the background in the visible and ultraviolet spectrum comes from the recombination radiation. The polarization characteristics are mainly due to the anisotropic recombining process during the recombination stage of plasma. During this process, the number of magnetic sublevel of the excited atoms is imbalanced, and the difference between the polarization degree and direction of the background and the atomic spectrum mainly depends on the mechanism of polarization. Studies have shown that PRLIBS does not always improve the signal-to-back ratio of elements, especially for weak discrete spectrum. When the energy densities, analyzer angle, delay and integration time of PRLIBS are controlled, a better background effect can be obtained. With the energy density of 20 J·cm-2 and the integration time at 30 μs, the SBR at Fe 407.12 nm increased from 4.86 to 12.97. It was found out that polarization degree is less correlated with the variable detection angle, for which the reason is probably that the Fresnel reflection effect of the conductor is too weak. The research results provided an effective theoretical basis for the research and application of PRLIBS.