Memory Effects in Compound-Specific D/H Analysis by GC-P-IRMS

Compound-specific D/H measurements of lipids often encounter huge ranges in D/H ratio. Moreover, culture experiments using D-enriched water to study fractionations often extend the working range to 1000 permil or more. While seeking improved methods for standardization of such measurements, we observed that the D/H ratio of each peak in a chromatogram is subtly influenced by that of the preceding peak. Such 'memory' effects have not been previously reported, but can have a significant impact on certain types of measurements. Here we describe experiments that fully characterize this memory phenomenon. To quantitatively evaluate memory effects, we analyzed ethyl palmitate (C16E) with D/H values ranging between - 230 and +800 permil co-injected with n-propyl palmitate (C16P, D/H = -225 permil) and with methane reference gas (D/H = -148 permil) at variable time separations. Five factors are observed to affect the memory effects between two successive peaks: 1) difference in D/H; 2) difference in relative abundance; 3) time separation; 4) age and condition of pyrolysis reactor; 5) difference in D/H between peak and background. For a time separation of 100s, approximately 2.4 to 4.3% of the measured hydrogen in C16P appears to derive from the preceding peak (C16E). This factor decreases to 1.2% if C16P is replaced by methane reference gas that is injected downstream of the GC column. Memory effects decrease exponentially with increasing peak separation with a decay half-time of 433s, and increase as the pyrolysis reactor ages. Our results can be explained quantitatively by two pools of exchangeable H in the system, one in the GC column and one in the pyrolysis reactor. The latter appears to increase in size as the reactor ages. The size of the memory effects make them significant for measurements in which successive peaks differ in D/H by more than 100 permil. Culture experiments using D-enriched waters are particularly susceptible, and will be systematically biased by these effects. Also, the widely used normalization procedure based on n-alkanes with alternating D/H around -50 and -200 permil can be shown to derive almost entirely from memory effects, rather than from true scale compression. A simple strategy to separately assess isotopic memory and scale compression is currently lacking.