0

I am attempting to numerically solve a BVP for a solid core solenoid field. I am starting with:

$\vec{\nabla{}^2}\vec{A_r}=0$

$\vec{\nabla{}^2}\vec{A_\phi}=\frac{\mu}{\mu_0}\vec{J_\phi}$

$\vec{\nabla{}^2}\vec{A_z}=0$

in cylindrical coordinates where the current $J$ is traveling along a cylindrical surface in the $\phi$ direction. I am assuming that the current has been turned on for a time $t>>t_0$ so we ignore any hysteresis affects of ferromagnetic materials that produce non-linearity. So focusing on just the $\phi$ component, the $\mu$ is a function of space since we assume that the solid core of the solenoid is a cylinder of radius $R$ and length $L$. I can imagine then that $\mu(r,z)$ can be written in one of two ways:

$\mu(r,z) = \mu_sH(L-|z|)H(R-r) + \mu_0H(|z|-L)H(r-R)$

$\mu(r,z) = \left\{ \begin{array}{lr} \mu_s, & \text{if } r < R \text{ or } |z|<L\\ \mu_0, & \text{if } r > R \text{ or } |z|>L \end{array} \right\}$

Regardless of the way this can be written, although I would prefer to use the first method as that would be easier to implement within code, I notice that there is a discontinuity occurring. We know the magnetic field lines must be continuous across the boundary that describes the metal core and the vacuum around it. I saw on the wiki article for Heaviside that at $H(0)$ it looks like it's value becomes $0.5$. However, this still doesn't suffice continuity. Neither would implementing an extra parameter in the piecewise for $r = R \text{ or } |z|=L $. The only idea I had was using the analytic approximation for the Heaviside since that is continuous while taking $k\rightarrow\infty$. However, it still changes value over a finite domain and so it doesn't represent the reality of the system in which essentially the material's surface and vacuum are creating. How then can I create a $\mu$ function that forces continuity and is properly describing physical reality?

Sophia
  • 61
  • It does not matter what value you assign to $H(t)$ at $t=0$ and the vector potential is continuous everywhere, even at the interface where $\mu$ is discontinuous. this may help – hyportnex Apr 12 '23 at 01:13
  • @hyportnex So essentially so long as my conditions include continuity that means that my $\mu$ function can be in either form I wrote. Why doesn't continuity matter in terms of the function that describes permeability? That would mean the BVPs trump any $\mu$ function I create, yet I would think that defining $\mu$ forces discontinuity since it's in the original Poison equation. I am using conditions 7, 8 and 10 from that post along with two more conditions for my specific problem. – Sophia Apr 12 '23 at 07:31
  • take a simple diff eq involving say, the signum function $sgn(x)=\pm 1$: $\frac{d}{dx}(sgn(x)\frac{dy}{dx})=0$ and solve it to find $y(x)$. Solve it and ask your self what does it matter if $sgn(0) = 0, 0.1 or -109.072$ when you integrate the equation for $y(x)$? Does the value of an integral change if you vary a single value of the integrand? – hyportnex Apr 12 '23 at 08:10
  • I worry that your conceptualisation of how you want to solve this problem is itself misguided. However, I will ignore that and mention another issue instead.

    Your magnetic ``constant" $\mu$ is specified wrongly. And your cylinder is length $2L$ if you have been using $|z|<L$ as the condition. I would suggest that you have $\mu = \mu_0 + (\mu_s-\mu_0)H(R-r)H(L-|z|)$ instead, because the specification of the conditions becomes much easier for you.

     – naturallyInconsistent Apr 13 '23 at 06:56
  • I am doubting your starting equation. Was it derived from $\text{rot}\left(\mu(r,z)^{-1}\text{rot}\vec{A}\right)=\vec{J};;$? – HEMMI Apr 13 '23 at 09:04

0 Answers0