# American Institute of Mathematical Sciences

February  2019, 13(1): 39-67. doi: 10.3934/ipi.2019003

## Magnetic moment estimation and bounded extremal problems

 1 Team Factas, Inria, 2004, route des Lucioles - BP 93, 06902 Sophia Antipolis Cedex, France 2 Department of Mathematics, Vanderbilt University, Nashville, TN 37240, USA 3 Department of Earth, Atmospheric and Planetary Sciences - MIT, Cambridge, MA 02139, USA 4 Center of Applied Mathematics, École des Mines ParisTech - CS 10 207, 06904 Sophia Antipolis Cedex, France

* Corresponding author: laurent.baratchart@inria.fr

Received  October 2017 Published  December 2018

We consider the inverse problem in magnetostatics for recovering the moment of a planar magnetization from measurements of the normal component of the magnetic field at a distance from the support. Such issues arise in studies of magnetic material in general and in paleomagnetism in particular. Assuming the magnetization is a measure with L2-density, we construct linear forms to be applied on the data in order to estimate the moment. These forms are obtained as solutions to certain extremal problems in Sobolev classes of functions, and their computation reduces to solving an elliptic differential-integral equation, for which synthetic numerical experiments are presented.

Citation: Laurent Baratchart, Sylvain Chevillard, Douglas Hardin, Juliette Leblond, Eduardo Andrade Lima, Jean-Paul Marmorat. Magnetic moment estimation and bounded extremal problems. Inverse Problems & Imaging, 2019, 13 (1) : 39-67. doi: 10.3934/ipi.2019003
##### References:

show all references

##### References:
Approximation error of $b_3^*[\phi_{\mathit{\boldsymbol{e}}_k}(\lambda)]$ with respect to $\mathit{\boldsymbol{e}}_k$ when $\lambda$ varies (the scale for $\lambda$ is logarithmic). The solid line corresponds to the case when $k = 1$, while the dashed line corresponds to the case when $k = 3$. As expected from Lemma 4.1, on the one hand, when $\lambda$ goes to $0$ (i.e., $\log_{10}(\lambda) \to -\infty$), the error tends to $0$. On the other hand, when $\lambda$ goes large, the constraint $M(\lambda)$ goes to $0$, meaning that ${\phi _{{\rm{opt}}}}$ is forced to go to $0$, whence the relative error tends to $1$
Constraint $M(\lambda) = \nabla \phi_{\mathit{\boldsymbol{e}}_k}(\lambda)$ as a function of $\lambda$ (the scale for $\lambda$ is logarithmic). The solid line corresponds to the case when $k = 1$, while the dashed line corresponds to the case when $k = 3$. As expected from Lemma 4.1, these are strictly decreasing smooth functions tending to $+\infty$ when $\lambda \to 0$ (i.e., $\log_{10}(\lambda) \to -\infty$) and tending to $0$ when $\lambda \to +\infty$
The mesh on $Q$, here with $P = 4$. The points of coordinates $(\kappa_p, \kappa_q)$ ($1 \le p, q \le P$) are represented by bullets. The elementary squares $Q_{p, q}$ overlap, as shown in the diagram
"L-curves" showing the approximation error $\|b_3^*[\phi_{\mathit{\boldsymbol{e}}_k}(\lambda)]-\mathit{\boldsymbol{e}}_k\|_{L^2(S, \mathbb{R}^3)}/\|\mathit{\boldsymbol{e}}_k\|_{L^2(S, \mathbb{R}^3)}$ as a function of the constraint $M = \|\nabla \phi_{\mathit{\boldsymbol{e}}_k}(\lambda)\|_{L^2(Q, \mathbb{R}^2)}$. The upper plot corresponds to the case when $k = 1$, and the lower one to the case when $k = 3$. As expected, the error decreases and tends to $0$ as the constraint is relaxed
$\phi_{\mathit{\boldsymbol{e}}_1}(\lambda)$ (top) and $\phi_{\mathit{\boldsymbol{e}}_3}(\lambda)$ (bottom) for $\lambda = 10^{-21}$. On both plots, the rectangles $Q$ (red) and $S$ (blue) are drawn together on the bottom layer to help visualize their respective positions
Synthetic magnetization (from left to right) $m_1$, $m_2$ and $m_3$ on $S$
Field $b_3[\mathit{\boldsymbol{m}}]$ corresponding to the synthetic magnetization shown on Figure 6 and an additive Gaussian white noise generated on the same $P \times P$ mesh on $Q$, with $P = 100$. The computed values are used to approximate the field and the noise as functions of $\text{Span}\{\psi_{p, q}\}_{1 \le p, q \le P}$. Note that the color scales are not the same on both pictures
The components $\langle m_k \rangle$ ($k = 1, 2, 3$) of the net moment $\langle \mathit{\boldsymbol{m}} \rangle$ are approximated thanks to the linear estimator as $\mu_k = \langle b, \, \phi_{\mathit{\boldsymbol{e}}_k}(\lambda) \rangle_{L^2(Q)}$ for several values of $\lambda$ and with $b$ being either the exact synthetic field $b_3[\mathit{\boldsymbol{m}}]$ or the exact field plus some noise.
The quantities $\delta_k$ are the relative errors $\delta_k = \big(\mu_k - \langle m_k \rangle\big)/\langle m_k \rangle$. The quantity $\delta_r$ is the relative error of the amplitude of $\mathit{\boldsymbol{\mu}} = (\mu_1, \mu_2, \mu_3)$ as a vector approximation of $\langle \mathit{\boldsymbol{m}}\rangle$, i.e., $\delta_r = \big(\|\mathit{\boldsymbol{\mu}}\|-\|\langle \mathit{\boldsymbol{m}} \rangle\|\big)/\|\langle \mathit{\boldsymbol{m}} \rangle\|$, where $\|\cdot\|$ denotes the Euclidean norm. Finally, $\theta$ is the angle between the vectors $\mathit{\boldsymbol{\mu}}$ and $\langle \mathit{\boldsymbol{m}} \rangle$, i.e., $\theta = \frac{360}{2\pi}\, \arccos\big(\frac{\mathit{\boldsymbol{\mu}}}{\|\mathit{\boldsymbol{\mu}}\|}.\frac{\langle \mathit{\boldsymbol{m}} \rangle}{\|\langle \mathit{\boldsymbol{m}} \rangle\|}\big)$
 No noise With noise $\lambda$ $\delta_1$ (%) $\delta_2$ (%) $\delta_3$ (%) $\delta_r$ (%) $\theta$ (°) $\delta_r$ (%) $\theta$ (°) $10^{-18}$ $12.56$ $14.77$ $3.02$ $-13.10$ $2.13$ $-12.93$ $2.28$ $10^{-19}$ $7.41$ $9.15$ $2.08$ $-8.05$ $1.23$ $-7.73$ $1.33$ $10^{-20}$ $4.91$ $5.52$ $1.61$ $-5.01$ $0.65$ $-4.44$ $0.70$ $10^{-21}$ $3.50$ $3.17$ $1.25$ $-3.10$ $0.34$ $-0.41$ $1.03$ $10^{-22}$ $2.50$ $1.71$ $0.95$ $-1.86$ $0.26$ $6.66$ $2.37$ $10^{-23}$ $1.71$ $0.86$ $0.73$ $-1.08$ $0.23$ $17.87$ $4.34$ $10^{-24}$ $1.11$ $0.38$ $0.53$ $-0.59$ $0.18$ $31.97$ $5.76$
 No noise With noise $\lambda$ $\delta_1$ (%) $\delta_2$ (%) $\delta_3$ (%) $\delta_r$ (%) $\theta$ (°) $\delta_r$ (%) $\theta$ (°) $10^{-18}$ $12.56$ $14.77$ $3.02$ $-13.10$ $2.13$ $-12.93$ $2.28$ $10^{-19}$ $7.41$ $9.15$ $2.08$ $-8.05$ $1.23$ $-7.73$ $1.33$ $10^{-20}$ $4.91$ $5.52$ $1.61$ $-5.01$ $0.65$ $-4.44$ $0.70$ $10^{-21}$ $3.50$ $3.17$ $1.25$ $-3.10$ $0.34$ $-0.41$ $1.03$ $10^{-22}$ $2.50$ $1.71$ $0.95$ $-1.86$ $0.26$ $6.66$ $2.37$ $10^{-23}$ $1.71$ $0.86$ $0.73$ $-1.08$ $0.23$ $17.87$ $4.34$ $10^{-24}$ $1.11$ $0.38$ $0.53$ $-0.59$ $0.18$ $31.97$ $5.76$
 [1] Lorenzo Zambotti. A brief and personal history of stochastic partial differential equations. Discrete & Continuous Dynamical Systems - A, 2021, 41 (1) : 471-487. doi: 10.3934/dcds.2020264 [2] Gunther Uhlmann, Jian Zhai. Inverse problems for nonlinear hyperbolic equations. Discrete & Continuous Dynamical Systems - A, 2021, 41 (1) : 455-469. doi: 10.3934/dcds.2020380 [3] Yueyang Zheng, Jingtao Shi. A stackelberg game of backward stochastic differential equations with partial information. Mathematical Control & Related Fields, 2020  doi: 10.3934/mcrf.2020047 [4] Anna Anop, Robert Denk, Aleksandr Murach. Elliptic problems with rough boundary data in generalized Sobolev spaces. Communications on Pure & Applied Analysis, , () : -. doi: 10.3934/cpaa.2020286 [5] Mostafa Mbekhta. Representation and approximation of the polar factor of an operator on a Hilbert space. Discrete & Continuous Dynamical Systems - S, 2020  doi: 10.3934/dcdss.2020463 [6] Chongyang Liu, Meijia Han, Zhaohua Gong, Kok Lay Teo. Robust parameter estimation for constrained time-delay systems with inexact measurements. Journal of Industrial & Management Optimization, 2021, 17 (1) : 317-337. doi: 10.3934/jimo.2019113 [7] Zhiting Ma. Navier-Stokes limit of globally hyperbolic moment equations. Kinetic & Related Models, 2021, 14 (1) : 175-197. doi: 10.3934/krm.2021001 [8] Petr Čoupek, María J. Garrido-Atienza. Bilinear equations in Hilbert space driven by paths of low regularity. Discrete & Continuous Dynamical Systems - B, 2021, 26 (1) : 121-154. doi: 10.3934/dcdsb.2020230 [9] Hua Qiu, Zheng-An Yao. The regularized Boussinesq equations with partial dissipations in dimension two. Electronic Research Archive, 2020, 28 (4) : 1375-1393. doi: 10.3934/era.2020073 [10] Neil S. Trudinger, Xu-Jia Wang. Quasilinear elliptic equations with signed measure. Discrete & Continuous Dynamical Systems - A, 2009, 23 (1&2) : 477-494. doi: 10.3934/dcds.2009.23.477 [11] Michiyuki Watanabe. Inverse $N$-body scattering with the time-dependent hartree-fock approximation. Inverse Problems & Imaging, , () : -. doi: 10.3934/ipi.2021002 [12] Yi-Hsuan Lin, Gen Nakamura, Roland Potthast, Haibing Wang. Duality between range and no-response tests and its application for inverse problems. Inverse Problems & Imaging, , () : -. doi: 10.3934/ipi.2020072 [13] Kha Van Huynh, Barbara Kaltenbacher. Some application examples of minimization based formulations of inverse problems and their regularization. Inverse Problems & Imaging, , () : -. doi: 10.3934/ipi.2020074 [14] Xinlin Cao, Huaian Diao, Jinhong Li. Some recent progress on inverse scattering problems within general polyhedral geometry. Electronic Research Archive, 2021, 29 (1) : 1753-1782. doi: 10.3934/era.2020090 [15] Hua Chen, Yawei Wei. Multiple solutions for nonlinear cone degenerate elliptic equations. Communications on Pure & Applied Analysis, , () : -. doi: 10.3934/cpaa.2020272 [16] Pierre Baras. A generalization of a criterion for the existence of solutions to semilinear elliptic equations. Discrete & Continuous Dynamical Systems - S, 2021, 14 (2) : 465-504. doi: 10.3934/dcdss.2020439 [17] Nahed Naceur, Nour Eddine Alaa, Moez Khenissi, Jean R. Roche. Theoretical and numerical analysis of a class of quasilinear elliptic equations. Discrete & Continuous Dynamical Systems - S, 2021, 14 (2) : 723-743. doi: 10.3934/dcdss.2020354 [18] Hongbo Guan, Yong Yang, Huiqing Zhu. A nonuniform anisotropic FEM for elliptic boundary layer optimal control problems. Discrete & Continuous Dynamical Systems - B, 2021, 26 (3) : 1711-1722. doi: 10.3934/dcdsb.2020179 [19] Fabio Camilli, Giulia Cavagnari, Raul De Maio, Benedetto Piccoli. Superposition principle and schemes for measure differential equations. Kinetic & Related Models, 2021, 14 (1) : 89-113. doi: 10.3934/krm.2020050 [20] Hai Huang, Xianlong Fu. Optimal control problems for a neutral integro-differential system with infinite delay. Evolution Equations & Control Theory, 2020  doi: 10.3934/eect.2020107

2019 Impact Factor: 1.373