( 605 ) 
IL ce = 148413,  G, (c) = 38.50953, GG, (c) = 2,18005, 
log M (c) = 141.66097. 
[Gy (0) — 6 — log 5] = — 0.00661, 
e— [G, (c) + log 2 — Li (e°)] = — 0.00662. 
[log M (c) + log An — €] = — 4.914, 
log e® [Go (c) + log 2 — Li (e°)]= — 4.913. 
LV; 61 = 1096.035, G, (ce) = FON 79963,.. Glo) = 2.52401, 
[Gj (ce) — C — log 7] = 0.00088, 
e-T[G, (©) + log 2 — Li (e?y] — 0.00090. 
If we have s>1, it is fairly evident that for large values of c 
already the expression 
[Gs (c) — log 5 (s) — Li (c~s*1)] 
itself may be considered as evanescent, and that by equating it to 
zero we are sure to obtain, quite independent of the value of Gp (ec), 
a result for Gs(c) that involves only a very small error. 
1 
But if we had s=1, or even ri <s< 1, this result would become 
unreliable, so that for s< 1 a previous knowledge of the value of 
Go (c) remains indispensable. 
This is connected with the fact that in the latter case Gs (c) 
diverges as c becomes infinite. 
Indeed we have for s = 1 
n—=o | 
G()= Er! HE Ep", 
p<e n=2 N pn<e 
and it is the first term >p! that increases beyond all limits for 
p<e 
aoe 1); 
We will examine more clocely the behaviour of = p—'. 
pce 
1 
As s still surpasses 3 We may write. 
Lim [G‘ (ce) — C — log log ce] = 0, 
CSS @ 
or 
‘) This was already stated by Euner (“Introductio in Analysin infinitorum,” I, 
§ 279). 
