Discussion Forum

Hi

I believe that MF and MEF are set up with the hypothesis that transient effects are instantaneous. That is linked to the ACDC hypothesis, versus the RF hypothesis. Coupling both (time current and magnetic effects) is possible but you should consider that the MF part has no time delay.

Now I must admit I got cought on this issue the other day, with a fluid flow, resolved in time to give a velocity field, and then I have magnetic particles in the fluid, acting with a DC magnetic field. Since I need to get the forces from the Lorentz component, I need to calculate the velocity per time and get the MF/MEF solved once per "time" step. I havent managed to couple the two physics correctly yet, but I neither have no time jsut now to dig into this issue, as my priority is with a FSI issue today ;)

So any comment from others (or COMSOL ;) would be appreciated

--
Good luck
Ivar

Hi,

Yes you may manually combine the EC and MF interfaces.

The trick is to set up the EC interface with a suitable excitation for example an applied voltage. Then you set up the MF interface with the total EC current density as an "external current density".

Note that this approach will not allow you to specify a prescribed current in a conductor as the induced current density in the MF interface will initially counteract and in good conductors largely cancel any applied or prescribed current from the EC interface. The current will asymptotically approach the prescribed one.

Best regards,

Magnus Olsson
Product Manager AC/DC and RF Modules
COMSOL

Hello!

2 years ago I had the same problem for V4.0a.
From support I got the answer, that it will not be the correct way, because the second derivative of A is missing...

"Nach detaillierter Untersuchung der Maxwell-Gleichung fehlt es in den
gekoppelten "ec" und "mf" Applikationen einen Term eps0*epsr*Att.
Also könnte es ein Unterschied geben zwischen "mef" einerseits und "ec"
und "mf" anderseits wenn die zweite Ableitung von dem Vektorpotential A
nach der Zeit t nicht gleich null ist."

Is this solved in 4.3?

Ralf

Hi

if you check the formulas in EC, MF and MEF for frequency domain, they look rather overlapping, apart from the volumetric current source Qj in EC that does not exist in the MEF
At least I see second derivatives of A in both MF and MEF (check both main physics and primary nodes)

But I agree in "time dependent" (which does not exist in MEF) you are missing the "eps0*epsr*Att" term in MF

Now why such a difference in time and frequency domain ?
And then why MEF in frequency domain with omega and omega^2 but not time domain with "At, and Att"?

Is it only due to transient instabilities or addition BC to be added, that can be solved still in a harmonic steady state development ?

Perhaps it's worth to take a look at the doc too ;)
--
Good luck
Ivar

Thanks for the comments. I believe they will be helpful.

A. J.

I am 'old' on Maxwell's and new to COMSOL. Finding the framework and documentation ....errr.... challenging.

I have a transient-but-not-RF situation with really poor conduction and highly non-homogeneous dielectric and conductivity. I am charging up something (with volts or amps, don't care) and then discharging it by time-varying a conductance.

Any further wisdom and advice since 2012? Must I mix EC and MF manually as suggested by COMSOL? Or has the situation improved?

Thanks in advance.

Regards
Colin Hales

Hi Magnus,

Thankyou for taking the time on this.

I am afraid reading a user manual isn't going to work this time.

I have written my PhD on the field system origins of the brain. The brain produces, with tiny filamentary (transmembrane ) currents, both an electric field and a magnetic field. It blazes out into the space above the scalp. I aim to build hardware (designed by COMSOL was the intent) that produces _both_ field systems except with inorganic crystalline solids. This is low voltage, low current, low frequency. No RF. Magneto-quasistatics. In future we must physically replicate complex wave mechanics (E&B) interference patterns in space above arrays of these devices, that won't be expressed by EC alone. We are examining the field system around ion channels.

Therefore I am interested in studying the electric and magnetic fields produced by physically tiny filamentary currents. You can characterise the most simple requirement as "Charge up a flat-plate capacitor with a current source in space, disconnect the source, then discharge it (through a separate a special geometry that is effectively a dielectric failure), producing both E and B (just like the brain does) in the really fat (poorly) conductive 'plates'. Just like in the brain.

In reality source currents _can_ have non-zero divergence. The way this has been done in practice was set up in the 1960s by Plonsey and colleagues (references below). Called 'volume conduction'...It works by the careful, mathematically rigorous localised violation of Ohms Law. At distance from the violation the fields are well behaved and this system of equations, for conductive media, quite happily produces E and B. I used the 'volume conduction' formula to great effect in my thesis. I wrote it in MATLAB. I have documented all the maths and verified the derivation myself, including the charge conservation issue.

DIV J does not have to be zero throughout all space.

As a result of COMSOL not using a well established form of EM theory, a whole community of biological workers involved in nervous system biology and related implant design engineering miss out on the benefits of COMSOL. This is called 'bioelectromagnetism' and should be in COMSOL. There is a whole journal based on it.

So ... I cannot use the 'Electric Currents' module alone. I must have E and B together, just as predicted by the appropriate (Plonsey/volume-conduction) formulations of Maxwell's equations, driven by localised 'volume current' sources. Qi = DIV.Ji (A/m^3), where Ji is a (physically tiny) current density source.

Initially I will try and use EC alone for the electric part. In the end, however, I must have E and B. Without that potentiality it stops me and anyone else from using COMSOL. If you want me to help set up the equations then let me know. Contact me privately and we can arrange something.

It would be fantastic of COMSOL could do this.

regards

Colin Hales

REFERENCES
Clark, J. and Plonsey, R. 1966: 'A Mathematical Evaluation of the Core Conductor Model'. Biophysical Journal, 6, pp. 95-&.
Clark, J. and Plonsey, R. 1968: 'Extracellular Potential Field of Single Active Nerve Fiber in a Volume Conductor'. Biophysical Journal, 8, pp. 842-&.
Malmivuo, Jaakko and Plonsey, Robert 1995: Bioelectromagnetism : principles and applications of bioelectric and biomagnetic fields. New York: Oxford University Press.
Plonsey, R. 1964: 'Volume Conductor Fields of Action Currents'. Biophysical Journal, 4, pp. 317-&.
Plonsey, R. and Collin, R. 1961: Principles and Applications of Electromagnetic Fields. New York: McGraw Hill.
Plonsey, R. and Heppner, D. B. 1967: 'Considerations of Quasi-Stationarity in Electrophysiological Systems'. Bulletin of Mathematical Biophysics, 29, pp. 657-&.
Plonsey, Robert and Fleming, David G. 1969: Bioelectric phenomena. New York,: McGraw-Hill.

Hi Colin,

I am by no means a specialist in bioelectromagnetics so correct me if I make some incorrect assumptions.

I would assume that on the time and material parameter scales we are talking about that (magnetically) induced electric fields would be many orders of magnitude smaller than field contributions from charge imbalances and any externally applied electric fields.

In that case, the best approach would be to use the electric currents interface to solve for the current and potential (electric field) distributions. Then, one would solve a time dependent magnetic fields model with the source currents from the electric fields - or rather a slowly varying magnetostatic magnetic fields model (essentially a parametric solution for time).

When stating that magnetic fields must have divergence free source currents, it is important to keep in mind that any displacement currents are included in those (total) source currents. As far as I could tell, the formulation in the paper "A MATHEMATICAL EVALUATION
OF THE CORE CONDUCTOR MODEL"
respects the equation of continuity (that allows for divergence in the conduction current density but not in the total current density).

So, I suppose the bottom line is. Do you really need to account for the supposedly very weak inductive effects in your model? What is the typical relation between magnetic and electric energy? Can you even resolve the induced electric field when solving model in double precision? If the inductive correction to the impedance is negligible, you can just as well solve the much simpler electric currents model and compute the magnetic fields using a one-way coupled approach.

It is not a problem to formulate a strict full Maxwell formulation based on the magnetic vector potential and the electric scalar potential. The problem is that it becomes numerically unstable at low but non zero frequency unless you have a computer with infinite precision. So by throwing out insignificant terms, you get a model that is numerically stable (for your available precision) and the omitted terms would not have changed the results within available numeric precision anyway.

Best regards,

--
Magnus

Hi Magnus,

OK. I think there is some hope for existing COMSOL EC to at least help me to make a start. With a little coaching through the suggested EC/MF mechanism I should be able to see B and E via A and V.

In the bigger picture of motivating the expansion of COMSOL AC/DC (to full bioelectromagnetism) is that this is science, not just engineering. What you call 'inductive effects' are what magnetoencepahlography (MEG) measures at the scalp in people. The B field is of the order pico-Tesla. Their actual causal role in tissue (if any) is unknown. Same with E (end electroencepaholgraphy/EEG). Except in the case of E it is much stronger and we definitely know it has a causal role in tissue (recently confirmed).

If we make artificial (inorganic) brain tissue that is truly bio-mimetic then it should produce similar B/E field systems that can also be measured in the MEG/EEG resp. We build the technology to explore the brain and make predictions in tissue, which then feeds back into the technology.

Therefore: We need a fully functioned E/B field AC/DC module, at least for electro-quasistatics and magneto-quasistatics, without any limits, for _science_ purposes, so it can make predictions for lab experiments, not just for designing technology. It is not up to us to decide any natural phenomenon is just a negligible 'inductive effect'. We need to generate it to find out its significance elsewhere in a natural setting and for experimental purposes. Different perspective.
=======================
ASIDE
Another requirement: COMSOL Electrolyte/battery module. I have not even begun to broach the other version of this problem: actual nervous tissue study. It needs to have fully functioned water-electrolyte ion diffusion (at physiological temperatures only) and Lorentz force with E/B causal effects. We need to know where charges go in space. This is what the brain is actually doing at nm scales. Currently no software does this combined electrodiffusion with full active field expression. Ohm's law does not apply. This is convection. Whole other problem. Very needed. Again science. Again missing.

Nervous tissue diffuso-bioelectromagnetism (just made up a name!) could again be a whole new COMSOL module and it would revolutionise neuroscience. There is an opportunity here. I will write it myself if I have to.
=======================

OK. I'll forge ahead with EC. for the moment on the proviso that I can get a little coaching through the EC/MF linkage....

Thanks for your attention to this.

regards

Colin Hales

Dear Magnus and colleages,

Finally after all this time I have finally managed to get my EC-based capacitor charged up and to be able to discharge it by collapsing the dielectric from within.

So I now have large EC solutions for currents. Lots of time steps.

I want to accept your offer to help me manually feed it into the MF/MEF so that I can see the magnetic field produced by the process.

Please can Magnus or someone take me step by step through how I feed the output of EC into MEF or MF. Alternatively I can empty out the solution and you can do changes. Please advise.

Thanks.
Colin Hales