s1 + s2 = τ(l1 + l2) are differentiated with respect to ti which yields in matrix notation:
b b''(gl) 0
0 b"(gj)
∖ 1 1
∂ ιi ∖
∂τ li + τlir
∂ l
dτ l + τljτ
- ∑n.1(ln + Tin ) )
∣∣" ( gi )( tikii + ki ) dti
-b' ( gj ) tj kji dti
0
Applying Cramer’s Rule:
∂si ∣A11 ∂sj ∣A21 ∂τ _ ∣A3∣
∂⅛i = 7^T, ∂β = 7^T an ∂ti = 7^T,
(24)
where
; = b'' ( gi ) (-b" ( gj ) Σ< ln+τln ) + — 1+j ]+b" ( gj ) dτ 1+7 <0.
∣A11 = -b"(g')(tikii + ki) ( -b"(gj) X(ln + τl"τ) + ^+ ,j )
∖ n=1 η /
r) 1 . ■
+∂τ τ+η?b''( gj >tj .
.. . .. . . ,. ʌ .. ... . ∂ 1
O21 = b''(g')b''(gj)tjkj £(ln + τl';)+ b''(g`)(tikii + k`)dτ —
n =1 '
-— τ+√b''(gj)tjkji, and
∣A31 = b''(gi)b''(gj)tjkj + b''(gi)b''(gj)(t`kβ + ki) > 0. (25)
Given by the second-order conditions of the federal optimization problem ∣A∣ < 0. The sign of
∣A31 is strictly positive since tjkji > 0 and by assumption (A) tik'i + ki > 0. Helpful in signing
∣A11 and ∣A21 we rewrite both determinants as:
∣A11 = β(tik'i + ki) + α and ∣A21 = β(tjkji) - α, (26)
with
..... . ∂ 1 .. . ∂ ∂ 1
α := -b '( g )( tikii + kl ) — —- + b' '( gj ) tj kj — —- and
t ∂τ 1 + ηj t ∂τ 1 + ηi
2
β := b' '( cf ) b' '( cf )£( ln + τlnτ ).
n =1
26