Omnian is a standardless analysis package and like many others it uses pure or multi-element standards to measure the sensitivity of the instrument for each element and also for calculating line overlaps. Based on this information it uses a fundamental parameters matrix correction model to calculate the concentrations. All standardless programs are based on a universal approach and require input from the user on your specific materials and sample preparation to obtain the best results. The more information you can give it the better it will perform. information about whether the elements in your sample are present as oxides or elements and your sample preparation parameters( fused bead, pressed pellet etc.). Sample preparation is critical to getting the best results. Samples prepared as loose powders will generally give poorer results than fused beads.

All that being said having worked with Omnian and other standardless applications for many years, most people will add a few type standards to the standardless calibration to get the best results. Claims about standardless applications should be taken with a grain of salt as "good results" are subject to interpretation. Standardless applications are really helpful if you do not have a lots of calibration materials or if you want a quick screening of materials, but your best results will be obtained by using a proper calibration with CRMs.

if you are preparing your samples as fused beads then you have opportunity to make a universal calibration with traceable pure oxides. Hope this helps but feel free to reach out to me through he group or directly as I have some experience in this area.

If you are analyzing complex materials the actual valence state of the oxides present in your sample, or the mineralogical make up of your sample the better the data will get.

Pressed powder samples of geological materials are unlikely to deliver the highest accuracy when measured with a 'standardless' analysis package because the samples are not homogeneous. The elemental 'mix' will be dependent on the different phases present in the sample itself. That's why, when cement works want to use pressed powders as their specimen type for analysis, for the accuracy they need, the standards have to come from the same source as the routine samples, i.e. they must all come from the same quarry ideally.

If you want to use a standardless package for general geological majors analysis, then you'll need to prepare your samples as fused beads. The two main advantages to this are that the fusion process will both make the specimen homogeneous and flat, a prerequisite for fundamental parameter based corrections, and will also remove any differences caused by elements being in different phases by effectively dissolving the sample into a solid lithium borate solution.

Unfortunately there is no way to avoid the mineralogical effects with the Omnian program. Analysing a loose powder or even a pressed powder does nothing to reduce this effect. Its unique to the specific mineralogy of the sample and specifically which elements are bound in compounds/minerals. The mineralogical effects are largest for the light elements like Al and Si so when analysing geological materials the errors can be very large. However, if you prepare a fused bead and then use Omnian you can expect very good results (as Stephen highlighted). I've done this and the results are satisfactory. Its definitely not as accurate as a custom application and usually takes considerably longer, but the big advantage is you don't need to source large numbers of standards and CRMs.
Where I find Omnian particularly useful is for comparative work where I'm not necessarily interested in highly accurate analysis - for example where a sample has been leached or modified/beneficiated in some way, then the change in the composition can be reliable so a metallurgist can estimate how much the material has been upgraded for example.
Where the nature of the sample is well understood Omnian can yield very good results even on pressed powders. Something to consider is whether or not your sample has a large LOI? If that is not accounted for - the H2O and CO2 content then you can have trouble as you can't quantify these on the XRF and may need to estimate them using TGA or something else. If the LOI is fairly high and the analysis is normalised to 100%, you will have large errors.

Thanks to all of you. I will get back to some of your points, if I may. I will contact privately. I'm not ignoring such good advice and comments but I am hard-pressed to finish a couple of reports before getting back to my analysis.

We've been struggling through many similar issues to yours for several years now, with complex environmental solid samples and metal water pipe scales. We're going to be moving towards using fused beads and Omnian in the near future, as soon as the installation and training is done, but we feel we have had reasonable success for most elements at 100 mg/kg and above using pressed pellets and an augmented IQ+ approach, with additional standards added to improve matrix matching to the extent that we can, and extending the calibration range for some elements that might be much higher in concentration in some samples (e.g. iron, lead, copper, zinc). We tried using conventional XRF calibrations, but we never had good success with the span of elements of potential interest and the diversity of sample compositions that we have to deal with, plus limited sample material (fractions of a gram). We express analytical results as elements rather than oxides because we found we could almost "make up" results by changing what assumptions were made with respect to different valence states of metals and the distribution and speciation of C, S and P in our samples. XRF is supplemented with elemental analysis (particularly total and inorganic C, plus cross-check on S) and TGA as a surrogate for LOI (OH, H2O, CO3). And we never normalize, it just distributes errors into unknown places. I also agree, we get the most variable recoveries compared to SRMs with Si and Al. Certainly, expectations have to be realistic with many different kinds of samples.

All excellent points! I think your point about normalization is really a key one that people who are using standardless analysis programs must consider. Blind normalization is always dangerous. Many sample types contain organics or elements not measure by the standardless analysis programs and normalization just propagates error. There are also situations with sample types containing high concentrations of some elements where un-normalized totals exceed 100%. This could be due to a problem with the standardless calibration or a breakdown of some aspects of the matrix correction model. Blind normalization here might mask other issues leading to erroneous results.

Standardless analysis an analytical tool like any other and should not be considered a black box universal solution.

In working with Omnian, I did try adding additional standards prepared as fused beads and discovered that for the glass and geological materials I ran, the results didn't significantly change. PANalytical advised that this was likely due to the fact that the Omnian standards are mainly fused beads. It was great to demonstrate that the Omnian results were quite good, although the custom calibrations will always be better - our lab mainly analyzes geological materials and glass. Two things that one needs to do to get 'good' Omnian results - touched on by previous comments - is to make sure you set up the software to account for the sample preparation used and (for my sample types) account for %LOI, Li2O, B2O3 and any other component not determined by XRF and normalize to 100% - (%LOI, Li2O, B2O3, etc.). Perhaps XRD data would be a good complement technique for these materials. In most of our materials, it is pretty 'safe' to assume the analyzes are present as oxides. TC, TIC, S speciation, FeO/Fe2O3 analyses are done as needed and these cover most of our sample types.

A strategy I have used is to make Li2B4O7 the balance compound for glass beads and analyse them as "solids" (it's suitable in my case as lithium tetraborate is my flux). This allows the sum to be 100% to accommodate FP, the advantage though is that the analytes are not normalised. But you need to ensure that all the analytes are correctly identified in the spectra. Its also a great check, because if you fuse 1g of sample with 9g of lithium tetraborate, you expect the Li2B4O7 composition of the solid to be approximately 90%. When all that checks out you can have a lot more confidence in the analysis.

I am using Omnian for 3 years. Approximately reports every type of samples but i have set certain set of conditions related to sample preparation. Some samples calcined before making bead or Pressed pellet.. Always i add Loss on ignition (L.O.I) in every analysis to get the actual result. Sometimes i prefer without normalisation, in some specific samples. MgO bearing compounds always show leds MgO in omnian. Similarly SiO2 Al2O3 spinels needs bead preparation other wise it may give 2-3 high Al2O3 and Less SiO2. Ca is drifted on weekly basis so Ca containing samples ahould be analysed carefully. And many more experiences. Share your sample types for further details.