Fig. 5 : Normalization error of multipoint normalizations based on the isotope range of the standards. A: matrix matched and bounded three-point normalizations do not exhibit significant differences with isotope range; B: matrix matched and bounded two-point normalizations do not exhibit significant differences with isotope range; C: matrix mixed and extrapolated three-point normalizations have significantly higher normalization errors when the isotope range is less than 15‰; D: matrix mixed and extrapolated two-point normalizations have significantly higher normalization errors when the isotope range is less than 15‰. Significance is shown in the compact letter display, where any two isotope ranges with the same letter are not significantly different as per Dunn’s post-hoc testing. Nitrogen (uppercase) and carbon (lowercase) letters are not compared with each other.

Interlaboratory comparison of instrument linearity

To assess the linearity effect across laboratories and instrument manufacturers, five different working standards were each analyzed across a range of sample weights (n=10) while the sample dilution was held constant, thus producing a range of peak amplitudes. Both instruments exhibited a non-linear deviation in reportedδ 15N as beam amplitude decreased (Fig. 6) – at U.S. EPA this deviation occurred at amplitudes below 6nA, and at UNM CSI it occurred below 2V (peak amplitude is measured in units of current (nA) on Elementar instruments and voltage (V) on ThermoFinnigan instruments). The linearity effect of reportedδ 15N occurred regardless of sample matrix and resulted in substantial (>1.5‰) decrease from the medianδ 15N at both facilities. Instrument linearity effects for δ 13C were not evident at UNM CSI across the observed range of peak amplitudes. At U.S. EPA, instrument linearity effects for δ 13C were not observed below 20nA, but linearity effects were observed for beam amplitudes above 20nA (Fig. 6).
Within 48 hours of analyzing the solid samples, diagnostic reference gas linearity tests were performed for N and C at U.S. EPA and for N at UNM CSI with the same tuning and configuration. Both facilities displayed a small reference gas linearity effect for N across the tested range of peak amplitudes (Fig. 6), with a total isotope range of 0.20‰ and 0.27‰ at UNM CSI and U.S. EPA, respectively. At the U.S. EPA, the reference gas C linearity diagnostics suggested a consistent inverse relationship between reported δ 13C and beam height across the tested range of peak amplitudes (2-12nA), with a total isotope range of 0.33‰.
To investigate whether the amount of combustion in the elemental analyzer varies as a function of sample weight, and thus influences the linearity results, we assessed how the ratio between peak amplitude and sample weight varied as a function of sample weight (Figure S1). Combustion effects appeared to be matrix-dependent, with high-organic matrixes exhibiting a higher variation in the ratio between peak amplitude and sample weight than other matrixes.