Conventional single coil two dimensional electromagnetic micromirror used in Light Detection and Ranging （LiDAR） applications suffer from inter axis crosstalk. This crosstalk originates from the single coil topology operating in an approximately 45° oblique magnetic field， which inherently generates torque components about both rotational axes； as a result， driving the fast axis alone produces an uncontrolled slow axis deflection. The crosstalk effect becomes more pronounced for mirrors with larger apertures. Motivated by this mechanism and informed by relevant literature， this work presents a novel two dimensional electromagnetic micromirror that preserves a large mirror aperture while effectively reducing inter axis crosstalk.This work proposes a dual axis independently driven micromirror architecture that prevents generation of slow axis torque when the fast axis is driven. This is achieved by spatially separating the electromagnetic coils and optimizing the local magnetic field orientations： the slow driving coil is mounted on a gimbal and driven by a predominantly horizontal magnetic field for quasi static scanning， while the fast driving coils are symmetrically placed on both sides of the fast axis and driven by a predominantly vertical magnetic field for resonant scanning. Both axes are driven by independent current sources. However， due to manufacturing and assembly tolerances， the electromagnetic forces produced by the symmetrically placed fast axis coils become asymmetric， resulting in a residual net torque about the slow axis. To address this， this work reduced the effective stiffness of the gimbal and optimized the mirror’s backside geometry to improve the transmission efficiency of the fast axis driving torque； this lowers the required drive current for a given fast axis rotation and further reduces crosstalk caused by unbalanced coil torques. Two device variants were implemented for comparison： Model 1 micromirror with optimized stiffeners and Model 2 micromirror without stiffener optimization. And the effectiveness of the stiffener optimization was validated by simplified theoretical derivations and finite element simulations for both models.In conclusion， the results demonstrate， to achieve a resonant optical rotation of 15°， Model 2 micromirror requires a drive current of 138.6 mA， whereas Model 1 micromirror requires only 26.2 mA. With both devices having the same quality factor， Model 1 demonstrates superior electromechanical energy conversion efficiency， consistent with the trends predicted by theoretical derivations and simulation analyses. By comparing the micromirrors designed in this work with a commercially available conventional single coil two dimensional electromagnetic micromirror， analysis of the crosstalk test results shows that at a fast axis optical rotation of 15°， the crosstalk of Model 2 micromirror is approximately one‑half that of the single coil micromirror， while the crosstalk of Model 1 micromirror is roughly one‑eighth of Model 2 micromirror. The coil separated topology therefore already yields lower crosstalk than conventional single coil devices on the market， and the stiffener optimization further reduces crosstalk. At a fast axis optical rotation of 30°， the crosstalk of Model 1 micromirror is only about one‑twenty‑first that of the single coil micromirror. Model 1 micromirror thus demonstrates both good scanning performance and very low inter axis crosstalk.The proposed micromirror architecture， combining independent dual axis actuation with backside stiffener optimization， effectively suppresses inter axis crosstalk while supporting large mirror apertures and large angle fast axis scanning， making it suitable for high resolution， high reliability LiDAR and precision optical scanning systems. However， asymmetry in the electromagnetic forces of the fast axis coils still exists， so crosstalk has not been completely eliminated. Future work will pursue systematic improvements to address the residual crosstalk， including mechanical measures such as reducing the number of fast axis coils while increasing turns per coil to lower sensitivity to geometric errors， and circuit level strategies such as independent current control and precise current compensation for symmetric coil pairs， with the goal of further eliminating residual crosstalk and improving scanning accuracy and consistency.