“…2 There are, indeed, examples of terms in the string effective actions which are either not renormalised by string higher-loop corrections (like second-derivative terms and 'anomalycancelling' terms, see, e.g., [12,13,14,15]) or receive contributions only from specific orders of string perturbative expansion (see [16,17]). In [18] we considered a term quartic in the gauge field strength trF 4 which, in heterotic string theory, is absent at the tree level [19,20] but appears at the one-loop level [21,22,23]. We have argued that it does not receive corrections from higher loops since D = 10 supersymmetry relates [24,25] it to the 'anomaly-cancelling' term BtrF 4 [26] which is not renormalised at higher orders [27,15] (see also below).…”