Yamagishi Complexity Torus LabYamagishi Lab

[Paper 1123] The Surface Potential and the Outer Magnetic Field Read Different Numbers of the Current Inside, and Currents Invisible to Both Ignore the Conductor’s Shape

Abstract

Paper 1107 showed numerically, for a source without direction (a density), that when the field equation is of second order what can be read from outside a closed skin is one number per angular pattern, and that the rest can be rearranged without changing the field outside at all. This paper counts the case where the source is a current, which has a direction, and two things are measured outside: the potential on the surface and the magnetic field outside, both produced by a current in a uniform conductor (conductivity 1, nothing conducts outside). Electroencephalography (EEG) and magnetoencephalography (MEG) measure the potential on the surface of the head and the magnetic field outside it, so the model is one example of their mathematical form (medical uses are not discussed). That the current inside is not determined uniquely by the potential outside is cited under the name of Helmholtz (1853). The potential in the conductor was solved by placing fundamental solutions outside the conductor and fitting the boundary condition; for a sphere of radius 1 it agreed with the closed formula for the surface potential and with the formula for the magnetic field outside treated by Sarvas (1987) within 6.36×10⁻⁷ for the surface potential and 2.60×10⁻⁹ for the magnetic field outside (the size of the solver error). This paper claims no new theorem or law.

Six currents were compared in a spherical conductor of radius 1. A dipole at distance 0.6 from the centre appears in both the surface potential (maximum 0.579944) and the magnetic field outside when tangential, but when radial it appears in the surface potential (maximum 15/(4π) = 1.193662) and not in the magnetic field outside (ratio 4.71×10⁻⁹ to the field of the primary current alone, the size of the solver error). A ring current of radius 0.15 does not appear in the surface potential (ratio 1.15×10⁻¹⁶ to the magnitude without cancellation) and appears in the magnetic field outside. The gradient of a smooth bump ψ confined to a ball inside the conductor, and a current wound on a torus, appear in neither (ratios at most 3.88×10⁻¹⁵). When the winding density is made to vary as 1 + 0.5 cos φ around the torus, the current still does not appear in the surface potential but appears in the magnetic field outside (ratio 1.62×10⁻²).

In a spherical conductor, for a harmonic polynomial h of degree ℓ, the component of the surface potential for that pattern is (1/ℓ)∫J·∇h, and that of the magnetic field outside is proportional to ∫J·(y×∇h) (the projections added up numerically agreed for 10 polynomials h within relative differences of 3.06×10⁻¹⁵ and 5.07×10⁻¹⁵). ∇h and y×∇h are perpendicular at every point, and y×∇h is perpendicular to the radius and has no divergence. The surface potential reads the divergence side of the current, and the magnetic field outside reads the circulating side, one number each per pattern. From 36 unknowns, the current placed on 12 shells in three directions, one number is read from the surface potential, one from the magnetic field outside and two from both for every pattern with ℓ ≥ 1, and the third singular value was at most 1.41×10⁻¹⁵ times the largest. For a conductor of any shape the surface potential satisfies ∮V ∂h/∂n dS = ∫J·∇h (h harmonic in the conductor). When a current in a bounded region is split into three parts, the gradient of a function vanishing on the boundary, the gradient of a harmonic function, and a part without divergence that does not cross the boundary, the surface potential reads only the harmonic gradient (the place that Paper 517 did not go to in its scope note).

The currents invisible to both do not depend on the shape of the conductor: when the surface potential is constant, the surface integral in Geselowitz’s formula vanishes and the magnetic field outside is that of the primary current alone, and the surface potential is constant only when the potential of the primary current vanishes outside the conductor, so both are fixed by the current alone. The gradient current is cancelled everywhere by the volume current it drives, and the toroidal current drives no volume current and keeps its field inside. Shape enters only the magnetic field outside of currents that appear in the surface potential, through the volume current. A dipole pointing outward from the centre appears in the magnetic field outside for the ellipsoids with semi-axes (1, 0.95, 0.9), (1, 0.9, 0.8) and (1, 0.8, 0.6), with ratios 0.1575, 0.2728 and 0.4031 to the tangential dipole. A shell current with no radial component (of tangential-gradient form) is also invisible to the magnetic field outside a sphere (ratio 8.07×10⁻¹² to the field of the primary current alone) and visible for the ellipsoid (1, 0.8, 0.6) (7.34×10⁻²). A dipole on the axis of a spheroid, pointing along the axis, does not appear outside because of the symmetry about the axis. When the current inside is cut into 20 radial steps and patterns up to degree 10, 240 of the 7220 unknowns are seen by the surface potential and the magnetic field outside (fraction 0.0332), which is 0.6648 times the fraction 0.0500 for the source without direction in Paper 1107. The separator is whether one measures the surface potential or the magnetic field outside, and which part of the current (the part with divergence, the circulating part, the radial direction).

Keywords: inverse source problem, EEG and MEG models, volume conductor, silent currents, Helmholtz decomposition, spherical symmetry, separator.

Scope (what is not claimed)

This paper claims no new theorem or law. That the current inside a conductor is not determined uniquely by the potential or the magnetic field measured outside (often cited under the name of Helmholtz 1853; the original has not been read here, and its title and journal were checked in a public bibliographic catalogue), the formula for the magnetic field outside a spherically symmetric conductor and the fact that a radial current gives no magnetic field outside it (treated in detail by Sarvas 1987; here the form of the formula was matched numerically with the method of fundamental solutions), Geselowitz’s formula writing the contribution of the volume current as a surface integral of the surface potential (treated in Sarvas 1987), that a closed ring current produces no surface potential, that a coil wound on a torus gives no magnetic field outside, the splitting of a vector field in a bounded region into three parts and Green’s identity, and that the numbers per pattern in a sphere take the forms of electric and magnetic multipoles, ∫J·∇h and ∫J·(y×∇h), all appear to be standard. Which part of the current is fixed by the surface potential and which by the magnetic field outside also appears to have been studied in the literature on inverse problems (not checked here). The review of MEG by Hämäläinen et al. (1993) also writes in its abstract that the activated sites can be found from the measured field distribution provided that appropriate assumptions about the source render the solution of the inverse problem unique. Only the abstracts of Sarvas 1987 and of this review were read for this paper. What is added is lining up the six currents on the same scale for the two measurements, the surface potential and the magnetic field outside; checking the numbers per pattern in a sphere by projections and singular values, which extends the counting per pattern of Paper 1107 to currents; rewriting the numbers of the magnetic field outside as overlaps with the divergence of a magnetization; writing briefly, from Geselowitz’s formula and Green’s identity, that the currents invisible to both do not depend on the shape of the conductor (whether this too is in the literature has not been checked); separating with ellipsoids where shape matters and where it does not; and comparing the fraction that can be seen with Paper 1107. This paper does not discuss medical uses. EEG and MEG are used only as the names of methods that measure the surface potential and the magnetic field outside, as examples of a mathematical form. Which directions of current actually dominate inside the head, the performance or merits of examinations, diseases and diagnosis, layers of different conductivity such as the scalp and the skull, noise and instruments, and methods for solving the inverse problem are not treated. The conductor is uniform, and only spheres and ellipsoids are considered. Paper 1107 showed numerically, for a source without direction, that what can be read from outside is one number per angular pattern and that a rearrangement of Laplacian form is invisible from outside, and took in its scope note the non-uniqueness of the inverse source problem as known. This paper sits beside it and replaces the source by a current; the currents that do not appear in the surface potential are exactly those whose divergence has the form of that rearrangement (Section 5). Paper 1111 extended the same counting to waves. Paper 258 looked at the potentials of point, line and sheet sources in a volume conductor in the forward direction and wrote in its scope note that it does not enter the inverse problem. This paper sits in that seat and counts the reverse direction, reading the current in a volume conductor from outside. Paper 517 wrote that splitting a field into longitudinal and transverse parts is a property of the field, and wrote in its scope note that it does not go to bounded regions, where a harmonic component appears and the splitting has three parts. This paper also sits in that seat. Paper 99 dealt with Calderón’s problem of reading the conductivity inside from the correspondence between voltage and current on the boundary. This paper takes the conductivity as uniform and asks only about the source.

1. What of the current inside can be read from the field outside

Paper 1107 showed numerically that what can be read from outside a closed skin is one number per angular pattern when the field equation is of second order. Adding a Laplacian form $\Delta\varphi$ to the density inside ($\varphi$ is a smooth bump confined to the inside) did not change the field outside. Paper 1111 extended the same counting to waves. In both, the source was a density without direction, and one thing was measured outside.

A current flowing inside a body has a direction, and two things are measured outside: the potential on the surface of the conductor and the magnetic field outside the conductor. EEG measures the potential on the surface of the head and MEG the magnetic field outside it. This paper counts, in a model of a uniform conductor, what of the current inside each of the two measurements reads. That the current inside is not determined uniquely by the potential outside is cited under the name of Helmholtz (1853). This paper does not discuss medical uses, and EEG and MEG are used only as the names of what is measured, as examples of a mathematical form.

The model is a uniform conductor of conductivity 1, with nothing conducting outside. With a primary current $\mathbf J$ inside, the potential $V$ of the conductor satisfies $\Delta V=\nabla\cdot\mathbf J$ inside and $\partial V/\partial n=0$ on the surface (no current leaves). The whole current flowing is $\mathbf J-\nabla V$, and $-\nabla V$ is called the volume current (volume as in the volume conductor of Paper 258). The magnetic field outside follows from the whole current by the Biot–Savart law. With the permeability set to 1, it can be written as the field of the primary current plus a surface integral of the surface potential,

$$\mathbf B(\mathbf x)=\frac{1}{4\pi}\int\mathbf J\times\frac{\mathbf x-\mathbf y}{|\mathbf x-\mathbf y|^3}\,d^3y-\frac{1}{4\pi}\oint V\,\mathbf n\times\frac{\mathbf x-\mathbf y}{|\mathbf x-\mathbf y|^3}\,dS$$

(Geselowitz’s formula; $\mathbf n$ is the outward normal).

The potential was solved by placing 1200 points outside the conductor, adding a sum of $1/|\mathbf x-\mathbf y|$ centred at them to the potential of the primary current, and fitting $\partial V/\partial n=0$ at 7200 points of the surface by least squares (the method of fundamental solutions). For a sphere of radius 1 there are a closed formula for the surface potential and a formula for the magnetic field outside (treated by Sarvas 1987). For three dipoles they agreed within $6.36\times10^{-7}$ of the maximum for the surface potential and within $2.60\times10^{-9}$ of the field of the primary current alone for the magnetic field outside; this is taken as the size of the solver error. When a current is said to appear in neither the surface potential nor the magnetic field outside, the ratio is taken, as in Paper 1107, to the magnitude without cancellation $\frac{1}{4\pi}\int|\mathbf J|/|\mathbf x-\mathbf y|^2 d^3y$. The magnetic field outside was looked at on the sphere of radius 1.2. The maxima written in this paper were found by locating candidates on Fibonacci points of the surface and refining from the points with the largest values. The maxima of (a) and (b) on the sphere were refined with the closed formula and the formula for the field outside the sphere, and agreed with those refined with the method of fundamental solutions within the solver error. Only the ratios of the size of rounding are maxima over fixed points (400 or 800 points of the surface and 400 points of the sphere of radius 1.2).

2. Six currents

Six currents were placed in a spherical conductor of radius 1. (a) and (b) are dipoles at the point $(0,0,0.6)$, (a) tangential and (b) radial. (c) is a ring current of radius 0.15 centred at the same point (current 1; the normal of the ring is along $x$). (d) is the gradient $\mathbf J=\nabla\psi$ of a smooth bump $\psi=(1-|\mathbf u|^2)^4(1+0.8u_x+0.5u_yu_z)$ confined to the ball with centre $(0.1,-0.15,0.2)$ and radius 0.3 ($\mathbf u$ is the position from the centre of the ball divided by 0.3; $\psi$ is not spherically symmetric). (e) is a current wound on a torus, $\mathbf J=\nabla\times(M\hat{\boldsymbol\varphi})$, whose cross-section has its centre at distance 0.3 from the axis and height 0.3 and radius 0.15, with $M$ a smooth bump $(1-d^2/0.15^2)^4$ confined to the cross-section ($d$ is the distance from the centre of the cross-section). (f) is (e) with the winding density made to vary as $1+0.5\cos\varphi$ around the torus.

currentsurface potentialmagnetic field outside
(a) tangential dipolemax 0.579944max 0.061181
(b) radial dipolemax 1.193662ratio $4.71\times10^{-9}$ (solver error)
(c) ring currentratio $1.15\times10^{-16}$max 0.029295
(d) gradient currentratio $3.84\times10^{-15}$ratio $3.88\times10^{-15}$
(e) current wound on a torusratio $1.39\times10^{-16}$ratio $4.16\times10^{-16}$
(f) torus with winding density varying around itratio $1.41\times10^{-16}$ratio $1.62\times10^{-2}$

A maximum is a value, and a ratio is relative to the magnitude without cancellation; but the magnetic field of (b) is relative to the maximum of the field of the primary current alone, the surface potential of (c) is a ratio of maxima, and the ratios of (d) to (f) are the maxima of the pointwise ratios to the magnitude without cancellation.

The tangential dipole (a) appears in both. The radial dipole (b) gives 2.058238 times the surface potential of (a) but does not appear in the magnetic field outside. The maximum of the surface potential of (b) is at the north pole, where the series $\sum(2\ell+1) 0.6^{\ell-1}=15$ gives $15/(4\pi)$. By the formula for the field outside a sphere (Sarvas 1987), the magnetic field outside is proportional to the cross product $\mathbf q\times\mathbf r_0$ of the direction and position of the current, which vanishes for a radial direction. The field of the primary current alone is not zero (maximum 0.116681), and the field of the volume current cancels it exactly outside.

The ring current (c) has no divergence, so it produces no volume current and does not appear in the surface potential. It appears in the magnetic field outside as a magnetic dipole of strength $\pi\times0.15^2=0.070686$ (the difference between the whole field and the field of the primary current alone is $2.39\times10^{-16}$ times).

For the gradient current (d), the potential of the conductor becomes $V=\psi$, the primary current and the volume current cancel everywhere, and the whole current is zero. Nothing appears in the surface potential or in the magnetic field outside. The divergence of this current, $\nabla\cdot\mathbf J=\Delta\psi$, has the same form as the invisible rearrangement $\Delta\varphi$ of Paper 1107.

The toroidal current (e) is $\mathbf J=\nabla\times\mathbf M$, has no divergence, and does not appear in the surface potential. The magnetization $\mathbf M=M\hat{\boldsymbol\varphi}$ also has no divergence, so the magnetic field is confined inside the torus and vanishes outside. In (f), $\nabla\cdot\mathbf M=(1/\rho) \partial M_\varphi/\partial\varphi$ is no longer zero, and the current appears in the magnetic field outside (maximum $8.97\times10^{-4}$) while still not appearing in the surface potential.

The invisibility of (d) and (e) does not use the sphere. Looking at the potential of the primary current and its normal derivative on both the surface of the sphere and the surface of the ellipsoid $(1, 0.9, 0.8)$, the ratios to the magnitude without cancellation were at most $3.84\times10^{-15}$ for the potential and at most $1.01\times10^{-14}$ for the normal derivative. When both the potential of the primary current and its normal derivative vanish on the surface, the boundary adds only a constant, and the surface potential and the surface integral in Geselowitz’s formula vanish whatever the shape (Section 5 writes this beyond the two examples).

3. Per pattern, one number from the potential and one from the magnetic field

In a spherical conductor the numbers read by the two measurements can be written per angular pattern. Let $h$ be a homogeneous harmonic polynomial of degree $\ell$. Taking the overlap with $h$ on the surface of radius 1 and on an outer sphere of radius $r$ (on both spheres $h$ is taken as its value $h(\hat{\mathbf x})$ at the direction $\hat{\mathbf x}$) gives

$$\oint V\,h\,d\Omega=\frac{1}{\ell}\int\mathbf J\cdot\nabla h\,d^3y,\qquad \oint rB_r\,h\,d\Omega=-\frac{r^{-\ell-1}}{2\ell+1}\int\mathbf J\cdot(\mathbf y\times\nabla h)\,d^3y$$

($B_r$ is the radial component of the magnetic field; in a spherical conductor the volume current does not contribute to $B_r$). With six random dipoles, the surface potential and the magnetic field outside ($r=1.3$) were made from the closed formula and the formula for the field outside the sphere, and adding up the left-hand sides numerically for 10 polynomials $h$ ($\ell=1$ to 3) agreed with the right-hand sides within relative differences of $3.06\times10^{-15}$ and $5.07\times10^{-15}$.

$\nabla h$ and $\mathbf y\times\nabla h$ are perpendicular at every point, and $\mathbf y\times\nabla h$ is perpendicular to the radius and has no divergence (checked symbolically). Integration by parts gives $\int\mathbf J\cdot\nabla h=-\int(\nabla\cdot\mathbf J) h$, so the surface potential reads the numbers per pattern of the divergence of the current. This has the same form as the one number per pattern of Paper 1107. The magnetic field outside reads the numbers on the circulating side perpendicular to it, and a radial current, being perpendicular to $\mathbf y\times\nabla h$, is not read. Writing the current as the curl of a magnetization, $\mathbf J=\nabla\times\mathbf M$, gives

$$\int\mathbf J\cdot(\mathbf y\times\nabla h)\,d^3y=(\ell+1)\int(\nabla\cdot\mathbf M)\,h\,d^3y,$$

so the magnetic field outside reads the numbers per pattern of the divergence of the magnetization. For the current (f), the two sides agreed within a relative difference of $1.20\times10^{-15}$ for the 3 patterns that are not zero, and $\int\mathbf J\cdot\nabla h$ was at most $4.98\times10^{-16}$ times the magnitude without cancellation. (e) is invisible from outside because the divergence of its magnetization is zero.

The number per pattern was counted with singular values. The current was placed on 12 shells with radii from 0.1 to 0.7 in three directions (radial, along the tangential gradient, circulating) in the form of one pattern $(\ell,m)$, giving 36 unknowns (the current on a shell was divided into point dipoles, 96 points in $\theta$ and 192 in $\varphi$). The potential at 120 points of the surface and the magnetic field at 160 points on radii 1.2 and 1.6 were lined up, and the singular values above $10^{-6}$ times the largest were counted.

pattern $(\ell, m)$surface potentialmagnetic field outsideboththird singular value (relative to the largest)
$(0, 0)$000(appears in neither)
$(1, 0)$112$3.99\times10^{-16}$
$(2, 0)$112$6.73\times10^{-16}$
$(3, 0)$112$9.87\times10^{-16}$
$(4, 0)$112$1.41\times10^{-15}$
$(2, 2)$112$4.46\times10^{-16}$
$(3, 1)$112$7.09\times10^{-16}$

For the patterns with $\ell\geq1$, one number is read from the surface potential, one from the magnetic field outside and two from both; the second singular value kept is at least $8.25\times10^{-1}$ times the largest, the third at most $1.41\times10^{-15}$ times (the size of rounding), with a cliff of about fifteen orders of magnitude in between. For the pattern $\ell=0$ nothing appears in either (a shell current all pointing radially only shifts the surface potential by a constant). Of the 36 unknowns, 2 can be read from outside.

4. Three parts in a bounded region

Paper 517 wrote that splitting a field into longitudinal (the divergence side) and transverse (the circulating side) parts is a property of the field independent of the coordinates, and wrote in its scope note that it does not go to bounded regions, where a harmonic component appears and the splitting has three parts (this splitting is called the Helmholtz decomposition, a result different from the Helmholtz 1853 of Section 1). The current in a conductor is a field in exactly such a bounded region.

For a conductor of any shape the surface potential satisfies, for every function $h$ harmonic in the conductor,

$$\oint V\,\frac{\partial h}{\partial n}\,dS=\int\mathbf J\cdot\nabla h\,d^3y.$$

It follows by putting $\Delta V=\nabla\cdot\mathbf J$, $\partial V/\partial n=0$ and $\Delta h=0$ into Green’s identity $\int(V\Delta h-h\Delta V) d^3y=\oint(V \partial h/\partial n-h \partial V/\partial n) dS$ and using that $\mathbf J$ does not cross the surface. On the sphere $\partial h/\partial n=\ell h$, which gives the formula of Section 3. With a dipole at the point $(0.2, 0.15, 0.3)$ in the sphere and the ellipsoids $(1, 0.9, 0.8)$ and $(1, 0.8, 0.6)$, and with 6 harmonic polynomials and three functions $1/|\mathbf y-\mathbf p|$ centred at points $\mathbf p$ outside the conductor as $h$, the two sides agreed within relative differences of $2.47\times10^{-14}$, $7.80\times10^{-12}$ and $1.93\times10^{-8}$. Since $\partial h/\partial n$ can be any function of mean zero on the surface (the Neumann problem can be solved), the surface potential is fixed by the right-hand side up to a constant.

A current inside the conductor splits into three parts,

$$\mathbf J=\nabla\psi_0+\nabla u+\mathbf w,$$

where $\psi_0$ vanishes on the surface, $u$ is harmonic ($\Delta u=0$), and $\mathbf w$ has no divergence ($\nabla\cdot\mathbf w=0$) and does not cross the surface ($\mathbf w\cdot\mathbf n=0$). For a function $h$ harmonic in the conductor, integration by parts gives $\int\nabla\psi_0\cdot\nabla h=0$ and $\int\mathbf w\cdot\nabla h=0$, so the surface potential reads only $\int\nabla u\cdot\nabla h$, the harmonic gradient. This integration by parts does not use the shape of the conductor.

The field $\mathbf y\times\nabla h$ read by the magnetic field outside a spherical conductor has no divergence and does not cross the surface, so it belongs to the third part. The magnetic field outside reads the part of the third part of that form, which does not overlap with what the surface potential reads. The current (d) consists of the first part alone and is invisible to both. (e) is the part of the third part perpendicular to the form read by the magnetic field outside.

5. Shape matters only for the magnetic field of currents that appear in the surface potential

Beyond the two examples, the currents invisible to both do not depend on the shape of the conductor. When the surface potential is a constant $c$, the surface integral in Geselowitz’s formula becomes ($\mathbf x$ outside the conductor)

$$-\frac{c}{4\pi}\oint\mathbf n\times\nabla_{\mathbf y}\frac{1}{|\mathbf x-\mathbf y|}\,dS=-\frac{c}{4\pi}\int\nabla_{\mathbf y}\times\nabla_{\mathbf y}\frac{1}{|\mathbf x-\mathbf y|}\,d^3y=0,$$

so the magnetic field outside is that of the primary current alone, which does not know the shape of the conductor. At 50 points outside the ellipsoid $(1, 0.8, 0.6)$, the size of the surface integral $\oint\mathbf n\times(\mathbf x-\mathbf y)/|\mathbf x-\mathbf y|^3 dS$ was at most $4.45\times10^{-15}$ times the sum of the sizes of its terms. The surface potential is constant only when the potential $V_0$ of the primary current (the potential in an unbounded conductor) vanishes outside the conductor. If $V_0$ vanishes outside, both $V_0$ and $\partial V_0/\partial n$ vanish on the surface and the boundary adds only a constant. Conversely, if the surface potential is constant, the left-hand side of the formula of Section 4 is zero, and putting $h=1/|\mathbf y-\mathbf p|$ ($\mathbf p$ outside the conductor) makes the right-hand side $4\pi V_0(\mathbf p)$, so $V_0(\mathbf p)=0$. Since $V_0$ is harmonic outside the current, this condition too does not depend on the shape as long as the current lies inside the conductor. The set of currents invisible to both therefore does not depend on the shape. (d) and (e) are examples: (d) is cancelled everywhere by the volume current it drives, and (e) drives no volume current and keeps its field inside. Shape enters only the magnetic field outside of currents that appear in the surface potential, through the volume current.

At the point $(0.2, 0.15, 0.3)$ of an ellipsoidal conductor, a dipole pointing outward from the centre (radial for a sphere) and a tangential dipole perpendicular to it were placed, and the maxima of the magnetic field outside on the sphere of radius 1.2 were compared.

conductor (semi-axes)outward dipoletangential dipoleratio
sphere $(1, 1, 1)$$4.00\times10^{-13}$$2.50\times10^{-2}$$1.60\times10^{-11}$
ellipsoid $(1, 0.95, 0.9)$$4.31\times10^{-3}$$2.73\times10^{-2}$0.1575
ellipsoid $(1, 0.9, 0.8)$$8.38\times10^{-3}$$3.07\times10^{-2}$0.2728
ellipsoid $(1, 0.8, 0.6)$$1.58\times10^{-2}$$3.93\times10^{-2}$0.4031

For the sphere, the magnetic field of the outward dipole is of the size of the solver error; for the ellipsoids it grows with the departure from the sphere. Solving the ellipsoid $(1, 0.8, 0.6)$ again with a different placement of the fundamental solutions and a different fineness of the surface points changed the ratio by only $2.24\times10^{-8}$ in relative terms. For a sphere, the factor $\mathbf n\times(\mathbf x-\mathbf y)$ in the surface integral of the volume current does not contribute to the radial component of the magnetic field outside, because the normal is radial. The field of the primary radial current has no radial component either, and outside the conductor there is no current, so the magnetic field is a gradient there; when its radial component vanishes, the whole field vanishes. For an ellipsoid the normal departs from the radial direction, and the volume current cannot fully cancel the field of a radial current.

It is not the direction that matters. On a shell of radius 0.4, currents of the pattern $Y=\hat{\mathbf x}\cdot\hat{\mathbf r}$ were placed radially, $R$ ($Y\hat{\mathbf r}$), along the tangential gradient, $T$ ($\nabla_1Y=\hat{\mathbf x}-(\hat{\mathbf x}\cdot\hat{\mathbf r})\hat{\mathbf r}$, with no radial component), and circulating, $C$ ($\hat{\mathbf r}\times\hat{\mathbf x}$), and $T-2R$ was also placed (the shell was divided into point dipoles, 20 points in $\theta$ and 40 in $\varphi$). The magnetic field outside is given relative to the maximum of the field of the primary current of $R$ alone, and the surface potential as the difference between its maximum and minimum on 800 points of the surface.

shell currentfield outside (sphere)field outside (ellipsoid)potential range (sphere)potential range (ellipsoid)
$R$ (radial)$3.08\times10^{-12}$$7.34\times10^{-2}$$3.20\times10^{-1}$$5.29\times10^{-1}$
$T$ (tangential gradient)$1.61\times10^{-11}$$1.47\times10^{-1}$$6.39\times10^{-1}$1.06
$C$ (circulating)$6.67\times10^{-1}$$6.67\times10^{-1}$$4.02\times10^{-10}$$6.24\times10^{-7}$
$T-2R$$1.61\times10^{-11}$$2.09\times10^{-11}$$1.54\times10^{-9}$$6.47\times10^{-7}$

$T$ has no radial component, yet it is invisible to the magnetic field outside the sphere (ratio $8.07\times10^{-12}$ to the field of its own primary current alone) and visible for the ellipsoid ($7.34\times10^{-2}$). For every harmonic $h$, $\int\mathbf J\cdot\nabla h$ is exactly 2 to 1 for $T$ and $R$ ($\ell+1$ to 1 for the pattern $\ell$), so both the surface potential and the magnetic field outside of $T$ are exactly twice those of $R$ (the difference in the magnetic field outside the ellipsoid is $2.80\times10^{-10}$ times that of $R$). $T-2R$ makes every right-hand side of the formula of Section 4 zero, does not appear in the surface potential, and does not appear in the magnetic field outside either the sphere or the ellipsoid. $T-2R$ has both radial and tangential components and looks neither like a gradient nor like a torus, but as the reasoning above says, it does not depend on the shape. $C$ does not appear in the surface potential, and its magnetic field outside is that of the primary current alone for both the sphere and the ellipsoid (the difference is at most $7.36\times10^{-11}$ times). What breaks with the shape is not the radial direction as such, but the cancellation, by the volume current outside a sphere, of the field of a current that appears in the surface potential.

With another symmetry, the field vanishes even if the shape is not a sphere. For a dipole on the axis of the spheroid $(1, 1, 0.8)$ at the point $(0,0,0.4)$, pointing along the axis, the magnetic field outside was $2.56\times10^{-9}$ times the field of the primary current alone (the size of the solver error). The magnetic field of a current symmetric about an axis only circles the axis, and the total current through a circle outside the conductor is zero, so it vanishes outside.

6. Counting

Paper 1107 counted that when the source inside is cut into $M$ radial steps and angular patterns up to degree $L$, $(L+1)^2$ of the $M(L+1)^2$ unknowns are seen from outside, a fraction $1/M$. For a current, each shell has three directions for every pattern with $\ell\geq1$ and only the radial direction for $\ell=0$, so there are $M+3M((L+1)^2-1)$ unknowns. The surface potential and the magnetic field outside see two numbers for every pattern with $\ell\geq1$, that is $2((L+1)^2-1)$.

For $M=20$ and $L=10$, 240 of the 7220 unknowns are seen, a fraction 0.0332 (0.0166 for the surface potential alone). For the source without direction in Paper 1107 it is 121 of 2420, 0.0500, so the fraction here is 0.6648 times that. For $M=100$ and $L=30$ it is 1920 of 288100 (0.0067), 0.6664 times the 0.0100 of Paper 1107. Even with the two measurements together, the fraction that can be seen stays at about two thirds of that for a source without direction.

7. Shared root or distinct roots

The quantities of this paper are separated by the criterion of Paper 300.

(1) Not appearing in the surface potential, and the invisible rearrangement of Paper 1107: the divergence $\Delta\psi$ of the gradient current has the same form as the rearrangement $\Delta\varphi$. More than that, not appearing in the surface potential is the same condition as the divergence $\nabla\cdot\mathbf J$ having the form of the invisible rearrangement of Paper 1107 (the form in which the potential $V_0$ of the primary current is confined inside the conductor; Section 5). The conditions coincide, not merely the forms, so they share a root.

(2) The invisibility of the currents invisible to both (gradient, toroidal, $T-2R$), and the invisibility to the magnetic field outside a sphere of currents that appear in the surface potential (the radial dipole, the tangential-gradient shell): the former is fixed by the current alone and does not change for an ellipsoid; the latter comes from the symmetry of the sphere and breaks for an ellipsoid through the volume current. Their responses to moving the shape of the conductor differ, so their roots are distinct.

(3) The one number per pattern of the surface potential and the one number per pattern of Paper 1107: both read the overlaps per pattern of a divergence through the second-order Laplace equation. They share a root.

(4) The numbers read by the surface potential and those read by the magnetic field outside: $\nabla h$ and $\mathbf y\times\nabla h$ are perpendicular, a radial current moves only the surface potential and a ring current only the magnetic field outside. Their responses to what is moved differ, so their roots are distinct.

8. The separator

The separator is whether one measures the surface potential or the magnetic field outside, and which part of the current. The surface potential reads, for a conductor of any shape, the harmonic gradient (the divergence side) of the current and does not see currents without divergence. The magnetic field outside a spherical conductor reads one form on the circulating side and sees neither radial currents nor shell currents of tangential-gradient form. Even together, two numbers per pattern can be read. The currents invisible to both do not depend on the shape of the conductor. Shape matters for the magnetic field outside of currents that appear in the surface potential, and that a radial current is invisible to the magnetic field outside is due to the symmetry of the sphere.

Closing

The surface potential and the magnetic field outside a uniform spherical conductor read different single numbers per angular pattern from the current inside. The surface potential reads the divergence side and the magnetic field outside the circulating side; together they read two numbers per pattern, and the fraction relative to the unknowns inside is about two thirds of that for a source without direction. For a conductor of any shape the surface potential reads ∫J·∇h, and the currents invisible to both are fixed by the primary current alone and do not depend on the shape of the conductor. Shape matters only for the magnetic field outside of currents that appear in the surface potential; that a radial current is invisible to the magnetic field outside is due to the symmetry of the sphere, and for an ellipsoid it is seen. That the current inside is not determined uniquely by the potential outside can thus be counted in this form: what is invisible differs with what is measured, and what is invisible to both does not depend on the shape of the conductor.

References

  1. Hämäläinen, M., Hari, R., Ilmoniemi, R. J., Knuutila, J., and Lounasmaa, O. V. (1993). Magnetoencephalography: theory, instrumentation, and applications to noninvasive studies of the working human brain. Reviews of Modern Physics.
  2. Helmholtz, H. (1853). Ueber einige Gesetze der Vertheilung elektrischer Ströme in körperlichen Leitern mit Anwendung auf die thierisch-elektrischen Versuche. Annalen der Physik.
  3. Sarvas, J. (1987). Basic mathematical and electromagnetic concepts of the biomagnetic inverse problem. Physics in Medicine and Biology.
  4. Yamagishi, Y. (2026). Reading the Interior from the Boundary [Paper 99]. YOI: https://www.yamagishiyuuki.com/yoi/99; DOI: 10.5281/zenodo.21758790
  5. Yamagishi, Y. (2026). $4\pi$ Appears Only When the Source Is a Point [Paper 258]. YOI: https://www.yamagishiyuuki.com/yoi/258; DOI: 10.5281/zenodo.22169011
  6. Yamagishi, Y. (2026). The Three Hundredth Census: Whether Two Things Share a Root Is Decidable [Paper 300]. YOI: https://www.yamagishiyuuki.com/yoi/300; DOI: 10.5281/zenodo.22171914
  7. Yamagishi, Y. (2026). Longitudinal and Transverse Belong to the Field; $x$ and $y$ Belong to the Coordinates [Paper 517]. YOI: https://www.yamagishiyuuki.com/yoi/517; DOI: 10.5281/zenodo.22862396
  8. Yamagishi, Y. (2026). When the Field Equation Is of Second Order, the Interior Can Be Read from Outside a Closed Skin Only as One Number per Angular Pattern [Paper 1107]. YOI: https://www.yamagishiyuuki.com/yoi/1107
  9. Yamagishi, Y. (2026). The Inverse Square Hides the Interior from a Static Field and Carries the Echo of a Wave Undistorted [Paper 1111]. YOI: https://www.yamagishiyuuki.com/yoi/1111

Reference

Yamagishi, Y. (2026). The Surface Potential and the Outer Magnetic Field Read Different Numbers of the Current Inside, and Currents Invisible to Both Ignore the Conductor’s Shape [Paper 1123]. YOI: https://www.yamagishiyuuki.com/yoi/1123
ORCID: 0009-0006-8587-3450

PDF (English) · PDF (Japanese) · Date proof of the English PDF (OpenTimestamps .ots) · SHA256SUMS

The record page for this paper (in Japanese, with the PDFs in each language and the date proof): YOI 1123. This paper is a preprint and has not been peer reviewed.