I downloaded the DTSA-II software package and ran the Monte Carlo simulations of XEDS spectra, and I can confirm that when I inputted the Oystein assumptions for "the elemental composition of LaClede Standard Primer", I replicated the spectrum as shown above. (I do have some queries about these assumptions.)
...
Harrit et al found that the "paint solvent" methyl ethyl ketone (MEK) failed to dissolve any of the red layer of their chips even after 55 hours of soaking with frequent agitation. In comparison, [normal] "paint chips" partly dissolved when subjected to the same treatment. But the fact is, a cured (crosslinked) epoxy resin does not dissolve at all in MEK or paint solvents / strippers; it merely softens.
The Laclede primer included cured epoxy resin and kaolin! Although if we suppose Harrit's MEK-soaked chip was Tnemec primer, does that also fail to dissolve? The "vehicle" is soya alkyd resin solids (16.5%), hard resin (2.8%), raw linseed oil (35.1%), thinners (32.3%), etc. Harrit's (Tnemec) swelled; Millette's (Laclede) softened as expected.
...
The heat of combustion of epoxy is 25 kJ/g according to Quintiere, Fundamentals of Fire Phenomena, Appendix, Table 9.
...
There remains the fact that the exotherm for epoxy peaks below 400 C, whereas Harrit's chips exhibited narrow exotherms above 400 C. I see Ivan believes the increase in peak temperature is due to additional oxidation; it would be nice if some reference were available.
...
Then when we consider the gray layers, it has to be mill scale formed as Laclede rolled the shapes for the trusses. Structural steel has an ultimate strength of at least 400 MPa, compared to only 12 to 30 MPa for epoxy adhesive, which is claimed to achieve a lap shear strength to aluminum of 2,300 psi or 15.8 MPa. However, mill scale's adhesion strength ranges from 1 to 18 MPa. Thus, slivers of this could fracture away whilst epoxy remained attached. Moreover, mill scale is mostly magnetite, which is consistent with the red-gray chips being attracted by a magnet and with the gray layer appearing gray rather than the "red-brown-orange" of rust. And mill scale has a total layer thickness of around 50 microns. (Cont.)
...
Those who claim there is no resemblance between the simulated spectra for Laclede primer and Harrit's red chips are being a bit silly. But here's some differences I have: First the nitrogen. Even if we take your 7% (or 6.7%) Oystein, there is a small but unmistakeable N peak observed between the large C and O peaks. No such peak is observed in Harrit's or Millette's spectra.
From what I can tell, 7% for both N and H seems to be towards the low end of estimates for N. Here is an estimate for elemental composition that I calculated: C 39.25%, O 20.8%, Fe 10.96%, H 2.59%, N 18.44%, Si 2.54%, Al 2.44%, Sr 0.49%, Cr 0.29%, Cl 2.19%. (We are in near enough agreement for six elements, and differ for C, H, N and Cl.)
I took an example of TETA (C6HN4) as a "typical" hardener and added this to C21H25ClO5 (epoxy) to get C27H26N4O5Cl as the 45% "unmodified epoxy amine" (MW=521.5), and C6HN4.H2O or C6H3N4O (MW=147) as the 55% "deionized water plus amine", these adding up to the 71.5% "vehicle".
I accept my chlorine figure is likely to be too high - that would assume none was dissipated as a gas during the curing process. Patents for "extremely low" chlorine epoxies (e.g. for electronics applications) state 100 ppm or lower, and claim 1,000 ppm and up is typical. 0.1% Cl shows up as a small peak on XEDS; not consistent with most Harrit spectra, but is similar to their Fig. 18 of a high-iron region on MEK-soaked chip. At 0.3% the peak is clearly larger, and gets very significant at 2%.
And so different assumptions about the H - N ratio in the amine could lead to an increase in N above 6.7% and a larger N peak.
...
The Laclede primer was specified to a film thickness of 1 mil which is 25.4 microns. Its density was 9.3 lbs/US gal which is 1114 kg/m^3. For a compromise estimate, let's suppose that the "unmodified epoxy amine", 32.175% of the total, yields 25 kJ/g, or 8.04 kJ/g for the dried paint. So for every 1 square meter of steel, there is 0.0000254 m^3 of paint = 28.3 grams of paint, yielding 227 kJ.