Hi Ian, as you know PDF is sometimes a pain. Only because something is published there it doesn't mean it is correct. Moreover, if there are minerals who really know the trace elements? As far as I understand maghemite it is more a defect structure, i.e. also the chemistry is probably not correct. It is a diffusional process to come from Fe3O4 to Fe2O3... I mean, who knows how much material is still Fe3O4 and how much is already Fe2O3? I would even say there is something chemically which is better described by Fe5O7, i.e., also the lattice parameter are changing like in solid solutions. I only know from out TEM guys that they are rolling with their eyes if they hear maghemite. I am sure you can differentiate them if you know that there is 100% maghemite in the sample but this one should make sure. Otherwise you are comparing apples and oranges (your experimental data with PDF data).   

Most structure data indicate a smaller cell for maghemite, so precisely measured peak positions at higher two theta may be indicative, and reference patterns in the powder diffraction file show same trend, but clearly this is a case where you will be best advised to rely not just on XRD but other data as well.

This paper suggests that a very careful look at the powder pattern just might do the trick, but it is subtle as Stephen Hiller says above.  http://www.degruyter.com/dg/viewarticle.fullcontentlink:pdfeventlink/$002fj$002fmsp.2013.31.issue-2$002fs13536-012-0100-6$002fs13536-012-0100-6.pdf?t:ac=j$002fmsp.2013.31.issue-2$002fs13536-012-0100-6$002fs13536-012-0100-6.xml

Hi Joseph, from the XRD pattern published in this paper I wouldn't make any serious conclusion. Which peak should this be indexed as 400 of maghemite? Why the noise is so different? 

Maghemite is a defect structure "formed" during transformation from magnetite to hematite. The only difference is that the oxygen arrangement is still fcc-like and not yet hcp. Maybe the process is more distinct during calcination. I am far away from beeing an expert there, but from such XRD data you cannot seriously expect that a discrimination between these two phases is possible, can you? 

Hello, Stephen, Joseph and Gert, thank you all for your good answers. I do agree with your opinions that single XRD analysis can not provide sufficient evidence to the identification of these two phases. I asked this question because I got confused when I saw a paper marked the peaks of two phases in a  XRD pattern (shown below in which "M" represents "magnetite" and "m" is the maghemite). In fact, I don's agree with this phase analysis, do you?

Gert and Congaing, I do agree that the differences are too slight.  In general when I make suggestions or refer a paper, I am just providing a starting point for further discussion or research. I do think that better diffraction patterns with greater resolution might show some slight differences, but the one in the paper itself if pretty inconclusive. I think, theoretically, there may be a way to distinguish, but it would take a lot more thought.

Congcong, the data in the image you show is not of good enough quality to tell, so I agree with you in that. They have very few counts, poor signal to noise and without a check on the calibration for peak positions, it would be impossible to distinguish on this.

I have looked at a number of patterns of both phases in the PDF, and it seems the higher the angle of diffraction, the more widely separated are the peaks between maghemite and magnetite, but then the intensity will be much lower. You could use variable count times and run the pattern to 100+ degrees 2-theta.

Hi Ian, as you know PDF is sometimes a pain. Only because something is published there it doesn't mean it is correct. Moreover, if there are minerals who really know the trace elements? As far as I understand maghemite it is more a defect structure, i.e. also the chemistry is probably not correct. It is a diffusional process to come from Fe3O4 to Fe2O3... I mean, who knows how much material is still Fe3O4 and how much is already Fe2O3? I would even say there is something chemically which is better described by Fe5O7, i.e., also the lattice parameter are changing like in solid solutions. I only know from out TEM guys that they are rolling with their eyes if they hear maghemite. I am sure you can differentiate them if you know that there is 100% maghemite in the sample but this one should make sure. Otherwise you are comparing apples and oranges (your experimental data with PDF data).   

Congcong, besides the fact that you need the smallest °2Th step size (or recording interval) and the reasonably highest counting time, on a properly pulverized sample (based on the pattern you showed, these conditions are very far from given), you also should try adding synthetic or purified natural hematite and magnetite to your sample, and observe if there are peaks not modified by these minerals. Also using some internal standard for peak position accuracy.

Hi, your result has too many noise and low in intensity. Would you mind to tell us the scan parameters, so that some of the researchers here can suggest a better solution. 

Dear Ian and Ferenc, thanks for your suggestions. I was wondering what if we tried to know the main phase compositions of a natural mineral sample, how should we conduct phase analysis then only through XRD technique? As far as I know, the case of a natural mineral (e.g., magnetite) is more complex when this sample naturally occurs with impurities like MgO, Al2O3, etc. It is not easy to identify the main phases accurately as Fe3O4, or gamma-Fe3O4 or MgFe2O4 or some other spinel phases. A preferential  way I suggest is to apply combined analysis methods including XRD analysis, optical microscopy, SEM-EDS to have a comprehensive understanding of the sample texture.

Sincerely.

Dear Ng, actually, this XRD pattern was derived from a paper finished by some other researchers. I can not provide more experimental parameters because authors didn't mention it in their work. Thanks for your attention.

Hi, Gert. I do agree with your opinion. XRD provides a qualitative method for phase analysis. But we worried about the accuracy of its quantitative results, especially in the case of phases that sharing similar crystal structure or solid solution phases. Thank you for your exciting answers.

Hi Yang Congcong, I wouldn't go that far. XRD is quite quantitative, but one needs to recognize the limits. If you see two people you can clearly differentiate them as long as they don't stand more or less exactly behind each other. If you know that these are slim people a wide silhouette will tell you that there are two people but if you don't know anything about it, there could be also only one fat guy, or even three or more. However, to say who is who, or to say anything about the individual shape is simply cheating and not based on serious results...although software will definitely give you some "optimum". And...there is still another case: it is simply dark or at least too dark. This describes the case if the signal is too bad. Even the best software cannot extract reliable data if the input data are too bad.

In fact, XRD manufacturers can do something against this. They simply have to take errors into account. If the signal is too low, stop any analysis. If a clear seperation is too bad, don't give any result in order to impress people how good the software is. You also wouldn't claim this if you cannot clearly differentiate two people. But then you are using your life-long trained experience. In XRD users are so fascinated about any result that they simply don't ask themselve: Can this result be serious? This is the major problem, not that XRD is only qualitative. In fact, the question maghemite or magnetite is already qualitative not quantitative. There are several technical questions with a similar impact caused by pseudosymmetry: ZrO2, perovskites,  martensite/ferrite, gamma/gamma' etc. The funny thing is that any simulation will already tell you, are there problems to expect? If the signal is too similar increase the error. If there are many different entries in the database describing the same phase, compare them all and compare how big is your confidence range? Is the different signal now still that different?

This is a general problem where people always do the same: they start to apply a software on a probem which is not made for such problematic cases. And manufacturers do not point out that their product possibly can fail. But it is better to know that one cannot trust this result than to publish meaningless things and claim something only because you are forced to publish, or it fits to your theory. I would estimate that a quite big amount of publications nowadays have exactly this problem. Don't trust anything simply because it has been published. Reviewing only excludes absolutely "stupid" errors. The authors who worked seriously in this field spend a lot of time to write this paper and the reviewer is a) perhaps not that specialized in this field, and b) under time pressure, that you will alsways find mistakes, misinformation, forgotten information, typing errors in equations, useless images (they only look nice) etc. that a paper can inspire you but shouldn't make you blind. Like databases...great job from these companies or communities in COLLECTING data, but they are not responsible for misentries.