Correction: negative Nb anomalies are indeed present.

Hey the following references may help you to find some answers.

1) Geochimica et Cosmochimica Acta, Volume 34, Issue 5, May 1970, Pages 630-634

Homogenized fission track analysis of uranium in some ultramafic rocks of known potassium content--- David E Fisher

2) GSA Memoirs 1966, v. 97, p. 521-542

SECTION 24: ABUNDANCES OF URANIUM, THORIUM, AND POTASSIUM*

Sydney P. Clark Jr., Zell E. Peterman and Knut S. Heier

3) Earth and Planetary Science Letters

Volume 4, Issue 2, April 1968, Pages 155-166

Thorium, uranium and potassium abundances in peridotite inclusions and their host basalts

 D.H. Green, J.W. Morgan *, K.S. Heier

Dear Willem Kruger: you don't specify where does this ultramafic cumulate rocks come from, but being in South Africa I would certainly think they come from the famous Bushweld Complex or similar mafic-ultramafic intrusions nearby. U and Th, and also Nb, are incompatible elements for mafic minerals in cumulates, such as olivine and diopside-augite, as their large ionic radii and high charge don't allow them to enter into the very compact structure of olivine, where only octahedral sites are present, fitted for medium sized cations like Mg, Fe2+ and some Ni and Co; and their high charge also doesn't allow them to sit comfortably in the clinopyroxene eight-fold coordination sites. Therefore it is not rare that those cumulates of olivine clinopyroxenites will have negative U, Th and Nb anomalies, since these elements would tend to stay in the basaltic magma and not form part of the cumulates derived from it. But clinopyroxene can easily host some Ti and Ba in eight-fold coordination sites present in its structure, the first, even being highly charged, is of similar size as Fe2+, and the other is a bivalent large cation, which can susbtitute partially for Ca. Also some Na and K can enter these ample sites, but always in low proportions. Pb and also Ba could be included in plagioclase, providing this mineral is present in these cumulates, but you don't specify this. All this could explain the positive Pb, Ba and Ti anomalies. Of course, if this cumulate is a mantle nodule or xenolith coming from a kimberlite pipe -something quite possible in your country- then the story is completely different. So the first thing you should do is to clear out the provenance of the sample: cumulate in mafic layered intrusion or mantle nodule. Regards. Sebastian.  

Hi Sebastian, thanks for your reply. The rocks all come from small satellite intrusions that are assumed to be comagmatic with the Bushveld igneous event. They primarily consist of ultramafic layers of olivine clinopyroxenite with associated olivine diorites. The ultramafic rocks contain up to 10% modal plagioclase in some layers, while it is completely absent in others. With the exception of Sr, the plagioclase does not appear to have a significant effect on the patterns observed on the spidergrams. 

What is confusing me is that U and Th are unlikely to be incompatible with mantle minerals, and rocks that crystallize during partial melting are usually enriched in these elements in comparison with the primitive mantle. However, the ultramafic cumulates all contain less U and Th relative to the primitive mantle. As far as I know Nb can be relatively depleted in the mantle if it becomes enriched in LILE, and I suspect the same process might be responsible for the depletion in U and Th. However, I am unable to find something in the literature that describes a similar pattern so far.

Dear Wllem: so they are ultramafic cumulates of olivine clinopyroxenite in layered gabbroic intrusions satellites to the Bushweld Complex. But one thing I don't understand is when you state: 

"U and Th are unlikely to be incompatible with mantle minerals, and rocks that crystallize during partial melting are usually enriched in these elements in comparison with the primitive mantle"

This makes no sense, as rocks don´t crystallize during partial melting, they melt! Perhaps you wanted to say: "U and Th are likely to be incompatible with mantle minerals, and the residual rocks of its partial melting are usually depleted in these elements in comparison with the primitive mantle"

Am I right?  Anyway, I believe that when the mantle undergoes partial melting the incompatible elements tend to be fractioned and enriched into the resulting basaltic melts, this is a quite logical conclusion, geochemically sound too. If, thereafter, melts produced in this way crystallize forming ultramafic cumulates, the opposite could happen: the cumulates will be depleted in those incompatible elements, which will be then fractionated into the residual liquid. 

By the way, it is the first time i my life that I've heard of olivine diorites, troctolites are common but contain quite calcic plagioclase, but olivine diorites should be quite rare rocks indeed, because they would contain more sodic plagioclase (%An

Dear Willen: I'm subscribed to the blog Geology Page, and just received an article regarding fractioning of trace elements in basaltic MORB magmas, which could offer an answer to your problem and questions. This is the link:

http://www.geologypage.com/2013/01/oceanic-crust-breakthrough-solving-a-magma-mystery.html

The conclusion is quite ingenious, you'll love it! Regards, Sebastian.

Hi Sebastian,

Yes, that is indeed what I meant to say. I will write you a proper response a bit later as I am quite busy at the moment.

Have a pleasant weekend.