The first important thing is that you have made sure by all means that this result is not flawed by any experimental misshap. The minimum requirement for this would be that the finding has been reproduced for several specimens (of various sizes well within the limits of acceptable dimensions in the magnetometer) and that the result is surely outside of any statistics in both measurement and evaluation procedures.

How large is the difference in saturation magnetization you observe?

Is your sample single crystalline (or textured) or polycrystalline?

What is/are the magnetic elements and what is the atomic structure?

When all this is being known, then one could think of spin-orbit coupling related effects might produce such a result. They can be appreciable in rare earth containing single crystalline samples and *usually* (but not necessarily) are smaller in transition metal systems.

One example is hexagonal (single crystalline) Co. Here, the origin of the (not very large) effect is largely the anisotropy of the orbital magnetic moment (the spin moment being nearly constant). Rare earth atoms may possess appreciably anisotropic magnetic moments when subject to crystal fields, without changing the ratio of orbital to spin contribution to the total moment. The latter is called "single ion anisotropy".

As for what literature might be useful, I would wait until we have more information on the specimen. Not all effects will be easily found in the same papers/books. But in any case, the Stöhr/Siegmann book on magnetism (Springer, 2006) will most probably prove useful [though maybe not terribly specific].

Thank you for your answer!

The sample is a polycrystalline and not textured, granular Fe/Co film (ratio about 50:50), grown on SiO2.

The difference in saturation magnetization is 10%. The behaviour was observed for several such films.

What are the maximum fields you apply? Are the films being capped for measurements or do they surface oxidize? (and w.r.t. that, how about oxidiation at the interface?)

Does the ratio between the two magnetization directions change with film thickness?

I am asking since for very thin films it could be an effect of some kind of interface induced anisotropy, requiring higher fields for alignment of some fraction of the film. In this case the difference between the MS values should be smaller at larger film thickness.

How about the (apparent) (shape) anisotropy field (the field at which you reach out of plane saturation)? Does it correspond to saturation magnetization, i.e. Hsat=Msat? Should be around 2.5 T then?

I'll give a short feedback: if really the material is non-textured then there is no reason for a difference in saturation magnetization of your film, provided you're sure to measure in saturation in both geometries.

If the film is really thin, then I suspect that you do subtract some diamagnetic contribution to obtain the ferromagnetic signal alone. This is a procedure one can do well and less well... If your sample size complies with the restrictions defined by the instructions for the magnetometer then what you subtract should be identical in both geometries. Is it?

To eliminate any doubt concerning the magnetometer and your use of it, it might be worthwhile to measure an empty substrate (just a piece of wafer from the same batch as those used for the films) and measure their in/out of plane M(H) curves. They should be the same in both geometries, and temperature independent.

The real question to ask first is what fields are you applying here? Screen shot the loops?

Before to answer your question it is important to know how you measure the saturation magnetization?

Thank you very much for your feedback everyone.

You have given me a lot to think about and to try to figure out.

I have indeed been subtracting the diamagnetic contribution from the substrate, and as Kai Fauth pointed out it is possible that an error might have occurred there. A first step will indeed be to measure a piece of substrate of the same size and shape to determine that background more accurately

For the sake of completeness I will now try to give some answers on the inquiries you made:

- the samples are not capped. There is definitely oxidation, since we also observe exchange bias in the out of plane direction

- the maximum applied field is 2T in both orientations

- the thickness of the films is about 20nm. For these kinds of non textures granular films it is actually quite tricky to determine the thickness using XRD, or RBS (I do not have access to TEM) so the thickness is determined by the deposition times;

- since we have a relatively big error on the thickness, it is hard to clearly observe a trend as a function of thickness, since the films we measured were of similar thicknesses. The signal of some of the significantly thinner films have however not been high enough to clearly determine saturation

Johanna, if you can access a SEM, you can fracture the film and inspect the edge. It should be possible to estimate the film thickness in this way. TEM sample preparation is much more cumbersome and I would only do it for single crystalline materials.

As a theorist I had to take part in this kind of experiments.

It is important very accurately to calibrate the vibrometer, to choose the right size of the magnetizing coils ..... But there is a very simple answer to your question: during out of plane magnetization the field 2T is not enough to reach saturation. You must use not less than 3T.

Hi Johanne,

So in order of your points:

1) Are you certain of exchange bias. The Neel temperature of CoO is 291K and that is ignoring thermal activation of the grains. So the following question is are these measurements at room temperature or ~4K? That is not to say they aren't oxidised (they most likely are) but to draw the conclusion from loop shift only is not really suitable - particularly as this could be an indication of an offset in the applied field. This is an issue in superconducting magnets where the 'known' field is based off a calibration table and not a direct measurement of the field.

2) As M. L. Shur said 2T might not be sufficient to saturate the film out of plane. You could very well be measuring a minor loop out of plane which is why your Ms is lower.

3) If you have access to XRD you should be able to easily measure the thickness of the films by doing reflectivity - that is if you have the fitting software. Reflectivity depends solely on changes in densities and not on crystallinity so unless your films are extremely rough (1.5nm roughness +) it should be quite easy. Also you will get some indication of the level of oxidation as the oxide cap will come out quite clearly.

Finally if you do not know the thickness then you have a large error in your volume - how are you comparing the saturation magnetisation? I assume this is somewhat the end goal?

@Robert, with respect to your last point: for one single specimen, there shouldn't be a difference in saturation moment (of the film, in absence of a mechanism producing it). Magnetization is a separate issue, as usual with thin films. Among the crucial points is to make sure you really measure in saturation. This is one of the points several of us have been stressing in different ways. From Johanna's reaction I guess that she's now busy in the lab checking the different things that were addressed here.

@Johanna, once you're done it would be nice to hear about the results...

Cheers

@Kai -  I am aware for a single sample it is the moment that shouldn't change. At the end of the day if she is comparing across a number of samples the magnetisation should be calculated - that or the loops normalised wrt. ms. The behaviour of a single sample is rarely of interest!

you're right, Robert, but here Johanna has a problem, doesn't she. For that a single sample would suffice for the moment... But since the incompatibility between in- and out-of-plane saturation moments seems to persist across samples stakes are high (imho) that something systematic is going or has been going wrong.

a quick update:

- The films are unfortunately too rough for XRR;

- there is no access to TEM or SEM

so my means of determining the thickness accurately are limited

@Robert Carpenter:

the sample consists of a mixture of Fe/Co that has been oxidized not pure Co/CoO.

Exchange Bias was measured at 2K.

 send me your samples!!! We have FIB and TEM, a Jeol JEM 2200Fs+Cs