In 2004, liquid chromatography coupled with isotope ratio mass spectrometry (LC/IRMS) has been developed. Today, LC/IRMS system has been used for various compounds. Stable isotope analysis has proved to be a powerful tool for source apportionment of various compounds. We analysed delta C-13 of 6 sugar, 10 amino acid and 2 caffeine samples using direct injection mode and online LC/IRMS mode. In this study, the ion chromatography system was coupled with IRMS instrument. On direct injection mode, the standard deviation for all samples were < 0.26 parts per thousand. The difference between EA/IRMS and direct injection mode was 0.41 parts per thousand (sugar), 2.01 parts per thousand (amino acid), 5.09 parts per thousand (caffeine, Wako Pure Chemical Industries), and 7.41 parts per thousand (caffeine, IAEA). The differences for amino acid and caffeine samples were bigger than sugar samples. The reason of the difference is still unclear. In the case of online LC/IRMS mode, the differences between the EA/IRMS and online LC/IRMS without phosphoric acid for amino acid were from 2.54 parts per thousand to 9.93 parts per thousand. In contrast, when 1 parts per thousand phosphoric acid was added, the differences between the EA/IRMS and online LC/IRMS was ranged from 0.09 parts per thousand to 2.09 parts per thousand. The eluent for LC was 5mM sodium tetraborate decahydrate (pH=9.51). When the 1 parts per thousand phosphoric acid was not added, the oxidation potential of sodium peroxodisulfate was decreasing. Thus, when the eluent of alkalinity eluent was used, it is necessary to add phosphoric acid. However, the difference of 2 caffeine samples between EA/IRMS and online LC/IRMS without phosphoric acid and with phosphoric acid were 5.30 parts per thousand and 5.82 parts per thousand, and 5.68 parts per thousand to 7.32 parts per thousand, respectively. For caffeine samples, the delta C-13 did not change with or without phosphoric acid. Other reasons were considered e.g. incomplete combustion. (C) 2015 Published by Elsevier B.V.