BSK) use an analytical description of nucleation and growth to explain the observed grain size distribution in an amorphous Si film. It is shown below, however, that the proposed description does not produce the known limit of transient nucleation which arises for a negligible crystallization fraction, X͑t͒ (and which corresponds to t c !`in Eq. (5) of BSK. Proposed generalizations for finite values of X͑t͒ and experimental relevance of the treatment are also discussed.For negligible values of the transformed fraction and for a high nucleation barrier the nucleation equation [which can be taken either in the "Becker-Döring" (BD) or in the "Zeldovich-Frenkel" (ZF) forms] can be treated using a matched asymptotic technique. In both cases the flux of growing nuclei is given by [2,3] J͑g, t͒ J st exp͕2exp͓2e 2x ͔͖,x ϵ ͓t 2 t i ͑g͔͒͞t .(1) In the above, t is identical to the one used in BSK, while t i ͑g͒ is the incubation time which depends on the form of the deterministic growth rate y g , and which in a general form is given by Eq. (7) or (9) of [3]. In the simpler ZF case, one has y g ͑dg͞dr͒u͑1 2 r ء ͞r͒, with u r ء ͞t being the growth velocity for large grains. The incubation time for the ZF equation is given by Eq. (12) of [3]: t i ͑r͒͞t r͞r ء 2 2 1 ln͑r͞r ء 2 1͒ 1 ln͑6W ء ͞kT ͒ .(2) The so-called time lag, which can be obtained exactly from the nucleation equation [4], can also be deduced from the asymptotic solution as t L ͑r͒ t i ͑r͒ 1 0.5772 · · · t [3]. With certain effort the exact and asymptotic results can be identified with each other [5(a)], which provides an independent justification of the solution.The corresponding distribution is given by the standard relation f͑g, t͒ J͑g, t͒͞y g (e.g., Ref.[2] or [6]). Comparison with the distribution of BSK will be limited to the ZF case since the value of l employed in BSK is appropriate for that equation (although Eq. (1) and the logarithmic part of Eq. (2) are of general validity). Compared to Eq. (2), a corresponding time scale, t ء 1 t e , in BSK has the same linear r dependence, but lacks the loga-rithmic term. Also, an extra term exp͓22͑t 2 · · ·͒͞t͔ was introduced in BSK into the double-exponential function (1). These modifications have no analytical justification (see below) and they lead, e.g., to an incorrect time lag. The added constant 2exp͑22e l 2 e 2l ͒ ഠ 22 3 10 28 [which was introduced to ensure at t 0 an exact, rather than an asymptotic, zero], on the other hand, is too small to be discussed [7].The reason for errors in BSK is the selection of g 0 -the size from which particles start deterministic growth (and which should satisfy g 0 2 g ء ¿ d [2,3]) too close to g ء , and the treatment of Eq. (1) of BSK as a "nucleation rate." For small g 0 2 g ء ϳ d, matching of the nucleation and the growth regions will not be asymptotically smooth, and the results will explicitly depend on the chosen value of g 0 , as happened in BSK and in previous works by the authors where g 0 was selected as g ء 1 1.1d or g ء 1 1.5d. A similar g 0 ...