Fund Project:Project supported by the National Natural Science Foundation of China (Grant No. 61901086), the Postdoctoral Innovation Talents Support Program, China (Grant No. BX20180057), the China Postdoctoral Science Foundation (Grant No. 2018M640907), and the Fundamental Research Funds for the Central Universities of Ministry of Education of China (Grant Nos. ZYGX2019J101, ZYGX2019Z016)
Received Date:12 October 2020
Accepted Date:11 November 2020
Available Online:30 March 2021
Published Online:05 April 2021
Abstract:With the integration of electromagnetic devices, the modules that make up into the devices and the functions that the devices needed to achieve are becoming more and more diverse. The coupling between the modules is difficult to ignore, the difficulty in designing increases sharply, and the traditional design methods gradually become incompetent. It is urgent to find a new comprehensive electromagnetic design method. This paper is to use the spatiotemporally synchronous focusing characteristics of time-reversed electromagnetic waves to explore the possibility of applying time-reversal technique to device design. First, based on the general device inverse design process, using the time-reversal technique, dyadic Green's function and basic principle of electromagnetics, a method of converting the port field distribution into the internal field distribution of the device is proposed. It is also proved that the continuous equivalent source obtained by the time-reversed field at a certain position in space can produce a field distribution close to the desired field at the port. In the single frequency inverse design process, only the tangential component of the electric field or magnetic field of the port is needed to be known. Then, with the help of the reciprocity of Green's function, the above theory is transformed to facilitate the numerical simulation. This numerical simulation realizes the reconstruction of the amplitude distribution source and the phase distribution source. It should be noted that the amplitude distribution source and phase distribution source are both randomly constructed. The numerical simulation verification is completed in two different cases and a variety of different initial conditions. All the simulation results are consistent with the theoretical results, which proves that it is feasible to apply time-reversal technique to the inverse design of electromagnetic devices. Keywords:time-reversal/ inverse design/ Green's function
观察表1及图4不难发现, 不论使用何种信息, 最终获得的重建源与初始源都十分接近的, 且在相同空间条件下, 对相位源的重建结果总好于幅度源. 使用两个平面的全部场信息能获得较好的重建效果, 与本文的理论分析一致; 使用两个平面的部分场信息仍能实现较好的重建效果, 是因为根据稳态电磁场中的唯一性原理“在闭合面S包围的区域V中, 当边界面S上的电场强度的切向分量E或者磁场强度H给定时, 体积V中任一点的电磁场由Maxwell方程唯一的确定.”序号为3, 6, 9的仿真实验的成功重建说明了这一点. 对于图3(b)所示的条件, 根据唯一性定理只需要知道$ {{\cal{F}}}_{{field}}^{ \xi, \text{up}}, {{\cal{F}}}_{{field}}^{ \xi, \text{down}} $的切向分量即可唯一地确定surface a的场分布, 因此${field}=\{{Ex}, {Ey}\} $. 对于图3(a)所示的空间示意图, 其四周为开放空间, 根据唯一性原理若仅知道端口平面up, down的切向分量在理论上是不能完整重建初始源的, 但数值仿真实验仍然实现了较好的重建效果, 分析原因可能是因为重建场的极化较为单一. 考虑仿真实验的结果较多, 因而只选取部分结果作展示. 对不同的仿真实验情况, 在实验序号为1—6的实验选取一个结果, 在实验序号为7—9的实验选取一个结果, 选取原则为选取重建结果代价函数最大的仿真实验代号作为展示. 代价函数越大代表其复原效果越差, 但即使是最差的复原效果仍得到了较完好的重建效果, 充分证明了理论的正确性. 图6展示了在图3(a)所示自由空间条件下对图5所示初始源$ {{{\mathit{\boldsymbol{f}}}}}_{{Ex}\text{Phas}e}^{a}\left(x, y\right) $和$ {\mathit{\boldsymbol{f}}}_{{Ex}\text{Amplitude}}^{a}\left(x, y\right) $, 选用不同反演信息激励时得到的重建源幅相分布, 其重建结果与图5十分接近. 图6(a)和图6(b)是对$ {\mathit{\boldsymbol{f}}}_{{Ex}\text{Phase}}^{a}\left(x, y\right) $的重建结果展示图, 对应仿真实验序号为5, 8, 代表了仿真实验中获得的最差重建结果, 其中主图为重建源的相位分布, 左上角为幅度分布. 图6(c)和图6(d)是对$ {\mathit{\boldsymbol{f}}}_{{Ex}\text{Amplitude}}^{a}\left(x, y\right) $的重建结果, 对应仿真实验序号为2, 8, 其中主图为重建源的幅度分布, 左上角为相位分布. 图6(d)的重建幅度约为图6(c)的两倍是因为使用两个反演源激励时其输入功率是单个反演源激励的2倍, 除幅度差异外, 图6(c)和图6(d)两图的趋势一致且其重建的相位分布均为同向分布. 不难发现, 不论是选何种反演源信息重建何种源, 最终获得的重建源与初始源的幅相分布都是十分接近的, 其代价函数计算结果均小于0.1, 代表着重建源与初始源十分接近. 图 5 初始源幅相分布图 (a) $ {Ex} $相位分布为特定函数; (b) $ {Ex} $幅度分布为特定函数 Figure5. The amplitude-phase distribution of the original source: (a) The phase distribution is a special function; (b) the amplitude distribution is a special function.
图 6 自由空间条件下重建源的幅相分布 (a)序号为5的重建$ {{{\mathit{\boldsymbol{f}}}}}_{{Ex}\text{Phase}}^{a}\left(x, y\right) $的幅相分布; (b) 序号为8的重建$ {{{\mathit{\boldsymbol{f}}}}}_{{Ex}\text{Phase}}^{a}\left(x, y\right) $的幅相分布; (c)序号为2的重建$ {{{\mathit{\boldsymbol{f}}}}}_{{Ex}\text{Amplitude}}^{a}\left(x, y\right) $的幅相分布; (d)序号为8的重建$ {{{\mathit{\boldsymbol{f}}}}}_{{Ex}\text{Amplitude}}^{a}\left(x, y\right) $的幅相分布 Figure6. The amplitude-phase distribution of the reconstructed source in free space: (a) The amplitude-phase distribution of the reconstructed $ {{{\mathit{\boldsymbol{f}}}}}_{{Ex{\rm{Phase}}}}^{a}\left(x, y\right) $ with experimental number 5; (b) the amplitude-phase distribution of the reconstructed $ {{{\mathit{\boldsymbol{f}}}}}_{{Ex{\rm{Phase}}}}^{a}\left(x, y\right) $ with experimental number 8; (c) the amplitude-phase distribution of the reconstructed $ {{{\mathit{\boldsymbol{f}}}}}_{{Ex}\text{Amplitude}}^{a}\left(x, y\right) $ with experimental number 2; (d) the amplitude-phase distribution of the reconstructed $ {{{\mathit{\boldsymbol{f}}}}}_{{Ex}\text{Amplitude}}^{a}\left(x, y\right) $ with experimental number 8.
图7与图6的不同之处仅在于其所处的仿真实验环境为图5(b)所示的四周为理想电导体条件的环境, 重建结果展示的选取原则与图6一致. 图7(a)和图7(b)是对$ {\mathit{\boldsymbol{f}}}_{{Ex}\text{Phase}}^{a}\left(x, y\right) $的重建结果展示图, 对应仿真实验序号4, 7; 图7(c)和图7(d)是对$ {\mathit{\boldsymbol{f}}}_{{Ex}\text{Amplitude}}^{a}\left(x, y\right) $的重建结果, 对应仿真实验序号为3, 9. 与图5所示的初始源对比不难发现, 图7重建结果虽然与图5较为相似, 但多出了许多规律性斑点, 且斑点间距接近一个波长. 分析原因可能是反演源$ {{\cal{F}}}_{{field}}^{\xi, \beta } $的激励不仅形成了重建源, 同时也激励出腔体自身的本征模式. 事实上, 仿真实验序号为1, 4, 7的重建源斑点分布较其他仿真实验代号更为明显, 分析原因可能是数值仿真中存在一定的仿真误差, 而磁场反演源的误差引入恰好激励出了更强的腔体本征模式. 在图7中同样可以观察到图7(d)图的重建幅度约为图7(c)的两倍, 与图6(c)和图6(d)两图存在差异的原因相同. 图 7 四周为理想电导体条件下重建源的幅相分布 (a)序号为4的重建$ {{{\mathit{\boldsymbol{f}}}}}_{{Ex}\text{Phase}}^{a}\left(x, y\right) $的幅相分布; (b) 序号为7的重建$ {{{\mathit{\boldsymbol{f}}}}}_{{Ex}\text{Phase}}^{a}\left(x, y\right) $的幅相分布; (c)序号为3的重建$ {{{\mathit{\boldsymbol{f}}}}}_{{Ex}\text{Amplitude}}^{a}\left(x, y\right) $的幅相分布; (d)序号为9的重建$ {{{\mathit{\boldsymbol{f}}}}}_{{Ex}\text{Amplitude}}^{a}\left(x, y\right) $的幅相分布 Figure7. The amplitude-phase distribution of the reconstructed source in PEC space: (a) The amplitude-phase distribution of the reconstructed $ {{{\mathit{\boldsymbol{f}}}}}_{{Ex}\text{Phase}}^{a}\left(x, y\right) $ with experimental number 4; (b) the amplitude-phase distribution of the reconstructed $ {{{\mathit{\boldsymbol{f}}}}}_{{Ex}\text{Phase}}^{a}\left(x, y\right) $ with experimental number 7; (c) the amplitude-phase distribution of the reconstructed $ {{{\mathit{\boldsymbol{f}}}}}_{{Ex}\text{Amplitude}}^{a}\left(x, y\right) $ with experimental number 3; (d) the amplitude-phase distribution of the reconstructed $ {{{\mathit{\boldsymbol{f}}}}}_{{Ex}\text{Amplitude}}^{a}\left(x, y\right) $ with experimental number 9.