It is not strictly true that there is no physical or geological meaning for negative values of magnetic susceptibility, as diamagnetic minerals have this property (quartz is one). However, the magnitude of this is invariably so low as to be insignificant for magnetic surveys. The one situation where negative apparent magnetic susceptibilities can be used in modelling is as a proxy for reversely magnetised material. This can be valid only if the remanent magnetisation vector is exactly opposed to the main Earth field.

My research in magnetism has zero values as final values, but if you model and can explain negative values effects ,then your contribution will be valued.

Presumably the non-physical values are produced by inversion of data.

My experience with geoacoustic (seismic) inversion suggests the following:

It is odd that the search space explored during the inversion is permitted to venture into non-physical regimes. If you can control the search space (set bounds to its parameters), then you should probably bound it so that non-physical parameters are not included in the search.

The fact that the inversion is driven by the objective function into non-physical regimes suggests that there remains significant mismatch between the model returned by the inversion processes and the real world, not merely in the media parameters values, but in the layout of layers and/or inclusions in the earth model.

Inversion typically aims for the best fit to the data, where “fitness” is assessed through the objective function. But what happens if there is noise or uncertainty in the data? Then the very best fit may actually be fitting noise and error in the data. This kind of fitting (over fitting) could of course lead to non-physical best-fit results from the inversion process.

Ideally you should be using inversion to give you a set of plausible solutions whose fitness all fall within a narrow boundary of each other defined by the degree of noise and uncertainty in your data. This is the set of equally fit inversion solutions for (1) the objective function you have chosen to use, and (2) for the noise and error in your data set.

But this requires that you know the expected impact that noise and error in the data will have on the objective function. You would have to do some analysis into that, presumably Monte Carlo style simulation, using simulated data with a known amount of noise and error added into it.

The point to be made is that you should never blindly accept the very best-fit solution provided by inversion, because it is very likely over-fitting the data, finely adjusting model parameters to replicate the noise and errors in the data, and very possibly driven into a non-physical parameter space in order to do that.

In fact there is no geological sense for negative susceptibility. If you are getting negative values, your inversion is not solving your anomalies. There are several reasons that can lead to this. Have you tried using the Magnetization Vector Inversion (MVI)?

It is not strictly true that there is no physical or geological meaning for negative values of magnetic susceptibility, as diamagnetic minerals have this property (quartz is one). However, the magnitude of this is invariably so low as to be insignificant for magnetic surveys. The one situation where negative apparent magnetic susceptibilities can be used in modelling is as a proxy for reversely magnetised material. This can be valid only if the remanent magnetisation vector is exactly opposed to the main Earth field.

Please give us the magnitude of negative values. If it is low (less than 1) then can be possible that you are investigating an area with diamagnetic materials, as Mark Duffett mentioned. High negative values is again an instability indication due to not suitable conceptual model or lack of reliable acquisition data of your magnetic survey .

Quartz, calcite, dolomite and salt are diamagnetic minerals with negatives values of susceptibility.

The outcome of inversion always has to be viewed skeptically.

When inverting data for environmental properties, you cannot rule out the possibility that the resulting environment is purely functional; i.e., that the environmental properties returned just happen, owing to numerical accident amid double-precision computations with noisy data and approximate models, to maximize a particular fit between a model and observations. You ask inversion for the best fit to noisy data using an approximate model, and inversion will gave it to you.

The expectation that the result is also an accurate representation of the real-work environment boils down to a matter of faith, based on the belief that the combination of your data and forward model, and the combination of their errors, uniquely constrain the outcome to be the "true" physical environment.

There is no assurance that this is true for inversion because:

Change your objective function, or change your computer model, and you generally get a different environment out of the inversion process.

Or leave the search free to wander into non-physical parameters spaces, and it generally will, owing to its in-built relentless obsession to find the very best fit in double precision, wander into those non-physical domains where it may find accidental non-physical compensation against noise and uncertainty in the data, and non-physical compensation against approximations in the forward model. This is over-fitting the data.

The points to be made are:

Much depends on the complexity of the model and objective function (many local maxima or minima, and sensitivity to data noise and model error), and on the quality and the amount of the data.

negative values of magnetic susceptibility is -0.26 to 3.72 (in SI unit). The susceptibility values increase from surface to deep layers and also from south to north. .

I following............

I have to appreciate all the geophysicists who commented here on my question.

I could find interesting information on magnetic surveying at Geosoft.com.

I would recommend you to follow geosoft webinars and forums if you are interested in geomagnetic data processing and interpretation.

Thanks again.

Normally, negative Suscs shouldn't appear in 3D inversion images. Since traditional magnetic inversion is a scalar method, and suppose that the magnetic remanence is zero (usually this is a rough simplification specially in igneous environments), therefore the 3D inversion algorithms sometimes give rise to negative Suscs. Another reason for negative Susc. in 3D imaging is that the primary trend removal was not good. Try with other ways to remove the regional trend like second order or third order removal. Though, the ultimate solution to get rid of negative Suscs. is to run MVI (Magnetization Vector Inversion). MVI recovers susceptibility as the magnitude of the magnetization vectors, so it is never negative.

Good luck

magnetic suscepibility measurements of slurry materials such as some of carbon riched sediments ( with low clay content ) gives negative results.

This is a good indicator for reverse or twisted positioning of responce magnetic field for unstable settled Clay , and high electric conversive or magnetic field defeater behaviour of Carbon and Water (in order of magnetic field load) to sensor's magnetic field