We can rewrite
| (C.1) |
with, for ξ0′ > ξ0T ,
| (C.2) |
where
| (C.3) |
In the limit ϵ ≪ 1, we have
Ψ | ≡ | ||
= | |||
= 1 + ϵ + | (C.4) |
so that
ξ′ | = | ||
= ξ0′ |
In fact, the series expansion of the square root must be kept to all orders for the term in ϵ, because the ratio ϵ∕ξ0′2 → 1 for ξ0′→ ξ0T . This gives
where we used
α | = 1 + ∑ n=1∞xn | (C.5) |
1∕2 | = 1 + ∑ n=1∞ | (C.6) |
= 1 + x + ∑ n=2∞ | (C.7) | |
= 1 + x + ∑ n=2∞xn | (C.8) | |
= 1 + ∑ n=2∞xn | (C.9) |
In addition,
= | |||
= 1 - ϵ + | (C.10) |
We get
λ2,-1,0 | = ∫ 02π | (C.11) | |
= ∫ 02π | |||
(C.12) | |||
= 1 -∑ n=1∞λ 2,-1,0(n) + | (C.13) |
with
λ2,-1,0(1) | = ∫ 02π | (C.14) |
= | (C.15) |
and, for n ≥ 2,
λ2,-1,0(n≥2) | = ∫ 02πn | (C.16) |
= n ∫ 02πn | (C.17) | |
= n ∫ 02πn | (C.18) | |
= n ∫ 0πsin2nθ | (C.19) |
We can transform, for n ≥ 2,
Y n | = ∫ 0πsin2nθ | ||
= ∫ 0πsinθ sin2n-1θ | |||
= ∫ 0πcos2θ sin2θ | |||
= | |||
= |
so that
with
so that
Y n | = | ||
= | |||
= |
and we find
λ2,-1,0(n≥2) | = n | (C.20) |
= n | (C.21) |
We obtain
I | = ∫ ξ0Tξ0 | ||
= ∫ ξ0Tξ0 dξ0′ | |||
= ∫ ξ0Tξ0 dξ0′ | |||
= ξ0 -ϵ1∕2 + ∑ m=1∞I + |
with
where we used
= 1 + ∑ m=1∞xm | (C.22) | |
ξ0T | = = | (C.23) |
= | (C.24) |
We have
Clearly, the only term of order ϵ-1∕2 comes from the ξ0′-2 contribution integrated and taken on ξ0T . We need to keep only
I | = ∫ ξ0Tξ0 dξ0′m + | ||
= ∫ ξ0Tξ0 dξ0′m + | |||
= ∫ ξ0Tξ0 dξ0′∑ i1=1∞∑ i2=1∞∑ im=1∞C i1Ci2Cimϵi1+i2++im i1+i2++im + | |||
= ∑ i1=1∞∑ i2=1∞∑ im=1∞C i1Ci2Cimϵi1+i2++im ∫ ξ0Tξ0 dξ0′i1+i2++im + | |||
= ∑ i1=1∞∑ i2=1∞∑ im=1∞ξ0Tξ0 + | |||
= ϵ1∕2 ∑ i1=1∞∑ i2=1∞∑ im=1∞ + |
with
and finally we get
∫ -11dξ 0σξ0HI | |||
= 2∫ ξ0T1dξ 0ξ02 - ξ 0ϵ1∕2 | |||
× + | |||
= - ϵ1∕2 + |
and therefore
| (C.25) |
with
For example, C1 = 1∕2, C2 = 3∕16, C3 = 5∕32,
m | i1 | i2 | correction |
1 | 1 | - | 1∕4 |
1 | 2 | - | 1∕64 |
1 | 3 | - | 1∕256 |
1 | 4 | - | 25∕16384 |
1 | 5 | - | 17∕22737 |
2 | 1 | 1 | 1∕48 |
2 | 1 | 2 | 3∕1280 × 2 |
2 | 2 | 2 | 9∕28672 |
2 | 1 | 3 | 5∕7168 × 2 |
2 | 2 | 3 | 5∕49152 × 2 |
2 | 3 | 3 | 6∕173015 |
The sum of all these coefficient gives K = 1.486, which is already much better that 9∕8 = 1.591
We compute the coefficient K for several values of the highest terms M and N in the series (C.25). The results are shown on Fig. C.1. We first see that K converges for high M and N, towards a value that seems close to 1.46. The best combination seems to be for M = N, as both high M and high N are needed for convergence. Therefore, we take M = N and calculate K as a function of M. Results are shown on Fig. C.2. We can see that K indeed converges, and that the asymptotic limit is about K = 1.46. For M = N = 6, we find K = 1.467.