Hello Qi, thank you for sharing such a great lecture! I have several question regarding the inverse magnonic device. Why you chose the forward volume geometry? Did you try optimization also for in-plane magnetized structures? Could the functionality be even enhanced by using such anisotropic dispersion relation for in-plane magnetization? Thank you in advance Best, Ondrej
Hi Ondrej, Thank you for your interest. I also tried to optimize the in-plan magnetization. However, the results are not good. The main reason is the strong demagnetic field in the nanoscale structure. When the holes have been introduced in the design region, the strong demagnetic field will dramatically change the dispersion curves locally. Here, we chose FV geometry which can ignore this issue. But I think it could also work for the in-plane magnetized structure in a large size where the demagnetic field is not critical. Best wishes, Qi
Hello Qi, I would like to ask you about the detection of the intensity. Can you please explain, how to calculate intensity of different frequency in Mumax? Secondly, how efficient is the random method compare to the other machine learning methods? Thank you in advance.
Hi Daniel, Thanks for your interesting of our work. (1) The mumax simulation will give you mx, my, mz as function of time for all the cells. You can chose the interesting cells (here we chose the cells at two outputs.) and perform FFT for time domain. Then, we get the frequency domain information. Here, we can define the intensity of spin waves for different frequencies. (2) We did not evaluated the other optimisation algorithm. In general, the direct binary search is a sample algorithm which is suitable for our case, just with or without materials. If you want to continually modify the parameters of the materials, maybe the gradient-based algorithm is better.
@@qiwang3939 Thank you very much for your reply. I have tried the method you suggested. However, the time requires to calculate the intensity is more than the mumax simulation itself. Is it the same in your case? in case not, can you please send me the intensity reading script if possible? Thanks and looking forward to your reply.
![The Dark Matter of AI [Mechanistic Interpretability]](http://i.ytimg.com/vi/UGO_Ehywuxc/mqdefault.jpg)
Dear Qi, this was one of the best lectures about magnonics I have ever attended. Xiexie ni! Milad (Southeast University-Nanjing)
Hello Qi,
thank you for sharing such a great lecture! I have several question regarding the inverse magnonic device. Why you chose the forward volume geometry? Did you try optimization also for in-plane magnetized structures? Could the functionality be even enhanced by using such anisotropic dispersion relation for in-plane magnetization?
Thank you in advance
Best,
Ondrej
Hi Ondrej,
Thank you for your interest. I also tried to optimize the in-plan magnetization. However, the results are not good. The main reason is the strong demagnetic field in the nanoscale structure. When the holes have been introduced in the design region, the strong demagnetic field will dramatically change the dispersion curves locally. Here, we chose FV geometry which can ignore this issue. But I think it could also work for the in-plane magnetized structure in a large size where the demagnetic field is not critical.
Best wishes,
Qi
@@qiwang3939 Thank you very much for detailed answer
Hello Qi,
I would like to ask you about the detection of the intensity. Can you please explain, how to calculate intensity of different frequency in Mumax? Secondly, how efficient is the random method compare to the other machine learning methods?
Thank you in advance.
Hi Daniel, Thanks for your interesting of our work. (1) The mumax simulation will give you mx, my, mz as function of time for all the cells. You can chose the interesting cells (here we chose the cells at two outputs.) and perform FFT for time domain. Then, we get the frequency domain information. Here, we can define the intensity of spin waves for different frequencies. (2) We did not evaluated the other optimisation algorithm. In general, the direct binary search is a sample algorithm which is suitable for our case, just with or without materials. If you want to continually modify the parameters of the materials, maybe the gradient-based algorithm is better.
@@qiwang3939 Thank you very much for your reply. I have tried the method you suggested. However, the time requires to calculate the intensity is more than the mumax simulation itself. Is it the same in your case? in case not, can you please send me the intensity reading script if possible? Thanks and looking forward to your reply.