A Spring in Imaginary Time

Math 241 Homework

John Baez
Answers by Garett Leskowitz

One of the stranger aspects of Lagrangian dynamics is how it turns into statics when we replace the
time coordinate t by it — or in the jargon of physicists, when we ‘Wick rotate’ to ‘imaginary time’ !
People usually take advantage of this to do interesting things in the context of quantum mechanics,
but the basic ideas are already visible in classical mechanics. So, let’s look at them!

1. Suppose you have a spring in Rn whose ends are held fixed, tracing out a curve

q: [s0 , s1 ] → Rn
with endpoints
q(s0 ) = a, q(s1 ) = b.
Suppose the spring is put into a potential

V : Rn → R

(perhaps due to gravity, but not necessarily). What curve will the spring trace out when it is in
equilibrium?

Hint: Hooke’s law says that a stretched spring has energy proportional to the square of how much it
is stretched. Here this is true of each little piece of the spring, so its total energy due to stretching
will be
k s1
Z
q̇(s) · q̇(s) ds
2 s0
for some ‘spring constant’ k. But in addition, each little piece will acquire energy due to the potential
V at that point, so the spring will also have potential energy
Z s1
V (q(s)) ds.
s0

The total energy of the spring is thus:

Z s1  
k
E= q̇(s) · q̇(s) + V (q(s)) ds.
s0 2
Our study of statics has taught us that in equilibrium, a static system minimizes its energy, or at
least finds a critical point. So, set
δE = 0
for all allowed variations δq of the path, and work out the differential equation this implies for q.

Answer: A variation δq(s) produces the variation

Z s1
δE = (k q̇(s) · δ q̇(s) + ∇V (q(s)) · δq(s)) ds
s0

in E. The first term in the integrand can be integrated by parts, and, as δq(s0 ) = δq(s1 ) = 0, this
yields
Z s1 Z s1
δE = (−k q̈(s) · δq(s) + ∇V (q(s)) · δq(s)) ds = (−k q̈(s) + ∇V (q(s))) · δq(s) ds.
s0 s0

1
If δE is to vanish for all variations δq(s), we must have

−k q̈(s) + ∇V (q(s)) = 0,

which is the required differential equation for q(s).

2. Suppose the spring is in a constant downwards gravitational field in R3 , so that

V (x, y, z) = mgz

where m is the mass density of the spring and g is the acceleration of gravity (9.8 meters/second 2 ).
What sort of curve does the spring trace out, in equilibrium?

Answer: As ∇V (x, y, z) = mgẑ, the equations for x, y, and z are

mg
ẍ(s) = 0, ÿ(s) = 0, and z̈(s) = .
k
The solutions are
x(s) = ux s + vx
y(s) = uy s + vy
mg 2
z(s) = uz s + vz + s ,
2k
where ui and vi denote constants easily calculated from s0 , s1 , a, and b. These solutions paramet-
rically define a parabola in R3 .

3. The calculation in problem 1 should remind you strongly of the derivation of the Euler-Lagrange
equations for a particle in a potential. To heighten this analogy, take the energy
Z s1  
k
E= q̇(s) · q̇(s) + V (q(s)) ds.
s0 2

and formally replace the parameter s by it, replacing the real interval [s0 , s1 ] ⊂ R by the imaginary
interval [t0 , t1 ] ⊂ iR, where iti = si . Show that up to some constant factor, the energy of the static
spring becomes the action for a particle moving in a potential.

Answer: Formally replacing s by it in our expression for E yields

Z it1  
k
E= q̇(it) · q̇(it) + V (q(it)) d(it).
it0 2

Using
d
q(it) = iq̇(it),
dt
we can write E in the form
Z it1   Z it1  
k d d k d d
E= − q(it) · q(it) + V (q(it)) d(it) = −i q(it) · q(it) − V (q(it)) d(t).
it0 2 dt dt it0 2 dt dt

This quantity is proportional to the action of a particle of mass k moving in a potential V (q(it)).

4. Fill in the blanks in this analogy:

2
 STATICS DYNAMICS
Principle of Least Energy Principle of Least Action
spring particle
energy −i×action
elastic energy density kinetic energy
potential energy density potential energy
spring constant mass
k m

5. What particular dynamics problem is the statics problem in 2 analogous to? How is the solution
to the statics problem related to the solution of this dynamics problem?

Answer: The statics problem in 2 is analogous to the dynamical problem of a mass moving in a
uniform gravitational field subject to conditions on its position at fixed times t0 and t1 . The solution
to the dynamics problem, also a parabolic curve, can be recovered from that of the statics problem
by formally replacing s by it.

Note that while in the statics problem the parabola ‘hangs down,’ in the dynamics problem the
parabola ‘arches up’ !

6. What does Newton’s law F = ma become if we formally replace t by s = it?

Hint: by ‘formally’, I’m suggesting that you shouldn’t think too much about what this actually means!
It’s a good thing to think about, but don’t let that stop you from solving what’s meant to be a quick
and easy problem.

Answer: This formal replacement only affects the second derivative inherent in a, and the result is
d2
F = −m ds 2 q. One way to view this (at least in the one dimensional case q: [s 0 , s1 ] → R) is as an

analogue of the stretched elastic string free to oscillate longitudinally, wherein a restoring force is
proportional to the second derivative of displacement with respect to length along the string. In
this case m plays the role of an elastic constant.

In short, working in imaginary time replaces F = ma by F = −ma.