for the operatorM(t), defined by
M(t) =−m
d^2
dt^2−V′′(q 0 (t)) (14.64)and subject to the boundary conditionsy(ta) =y(tb) = 0. We will now show that, if the complete
spectrum of the self-adjoint operatorM(t) (subject to the above boundary conditions) is known,
then this functional integral of Gaussian fluctuations can be computed. Let the eigenfunctions
φn(t) satisfy
M(t)φn(t) =λnφn(t) (14.65)subject toφn(ta) =φn(tb) = 0. SinceM(t) is self-adjoint, the eigenvaluesλnare real and the
eigenfunctionsφn(t) span an orthonormal basis of the function space fory(t). SinceM(t) is actually
real, we may choose a basis of real functionsφn(t), obeying
∫tb
tadtφn(t)φn′(t) = δn,n′
∑
nφn(t)φn(t′) = δ(t−t′) ta≤t,t′≤tb (14.66)The completeness ofφn(t) allows us to expand
y(t) =∑
ncnφn(t) (14.67)for real coefficientscn. Since the metric on function space and on thecnare related by
∫tb
tadtδy(t)^2 =∑
n(δcn)^2 (14.68)the change of variablesy(t)→{cn}has unit Jacobian, and we have
Dy=∏
n(dcn) (14.69)It is now also straightforward to evaluate
∫tb
tadty(t)M(t)y(t) =∑
nλn(cn)^2 (14.70)so that the entire functional integral over Gaussian fluctuations becomes,
∫
Dyexp
{
i
2∫tbtadty(t)M(t)y(t)}
=∏
n(∫
dcnexp{
i
2
λn(cn)^2})
=∏
nλ
−^12
n (14.71)Relating the product of the eigenvalues to the determinant,one often writes,
∫
Dyexp{
i
2∫tbtadty(t)M(t)y(t)}
= Det(
−m
d^2
dt^2−V′′(q 0 (t)))−^12
(14.72)