The University of Chicago, Department of Physics Page 1 of 3

Problem set 2 solutions

Michael A. Fedderke

Standard disclaimer: while every effort is made to ensure the correctness of these
solutions, if you find anything that you suspect is an error, please email me at
[email protected].

All problem references to Arfken and Weber, 7th edition.

I. 11.3.3

(a)

The line joining 3 + 4i and 4 − 3i can be parametrised by x(t) = 3 + t and y(t) = 4 − 7t such that t runs from
0 to 1. Then z(t) = (3 + 4i) + (1 − 7i)t, so that dz = (1 − 7i)dt. We then have
Z 4−3i Z 1   2 
2
 76 707
(4z − 3iz)dz = dt(1 − 7i) 4 (3 + 4i) + (1 − 7i)t − 3i (3 + 4i) + (1 − 7i)t = · · · = − i,
3+4i 0 3 3

where I have omitted the tedious yet straightforward algebra in going from the integral over t to the final answer.

(b)

Both the points 3 + 4i and 4 − 3i lie on a circle |z| = 5, at angles θ1 = arctan(4/3) and θ2 = − arctan(3/4),
respectively. Parametrise z = 5eiθ ⇒ dz = 5ieiθ dθ. We then have
Z 4−3i Z − arctan(3/4)
2 76 707
4 × 25e2iθ − 3i × 5eiθ 5ieiθ dθ = · · · =
 
(4z − 3iz)dz = − i.
3+4i arctan(4/3) 3 3

II. 11.3.5

Let f (z = x + iy) = x2 − iy 2 , and consider

H
f (z)dx on the two different contours: (a) the unit circle, (b) the
square with vertices ±1 ± i.

(a)

Parametrise x = cos θ and y = sin θ so z = eiθ ⇒ ieiθ dθ. Then, since the contour runs clockwise, with θ running
from 2π down to 0, we have
I Z 0 Z 2π
(x2 − iy 2 )dz = cos2 θ − i sin2 θ ieiθ dθ = −i cos3 θ + sin3 θ + i cos2 θ sin θ − sin2 θ cos θ dθ = 0,




2π 0

since the odd number of trig functions in each term will integrate out to zero by symmetry / periodicity.

Mathematical Methods of Physics v1.0 Autumn 2015

20 October 2015, 2:20am
The University of Chicago, Department of Physics Page 2 of 3

(b)

For this contour, the

I Z −1 Z −1 Z 1 Z 1
(x2 − iy 2 )dz = idyf (1, y) + dxf (x, −1) + idyf (−1, y) + dxf (x, 1) = 0,
1 1 −1 −1

because f is even in both its arguments, so the x-integrals and y-integrals each cancel in pairs independently owing
to the reversal of the integration limits.

There is no deep analyticity reason for these results to be equal. Note that f (x, y) = u(x, y) +
iv(x, y) ⇒ u(x, y) = x2 , v(x, y) = −y 2 . So clearly (u,x = 2x) 6= (−2y = v,y ) (except on the line x = −y),
and so the function f (z) is definitely not analytic on the entire interior of either integration contour, so path-
independence is not guaranteed (nor, indeed, is path-dependence, as this result clearly shows!). The deepest
reason one can invoke to explain that these functions integrate to zero is that f (z) has a parity symmetry under
z 7→ −z (i.e., (x, y) 7→ (−x, −y)), and that the contours respect this symmetry, so that the line integral around the
contour receives contributions equal in magnitude and opposite in sign for infinitesimal displacements at points on
the contour related by reflection through the origin.

III. 11.3.6
R 1+i R 1+i
Consider 0 z̄dz = 0 (x − iy)(dx + idy).
On the first path, we have
Z Z 1 Z 1
1 1
z̄dz = (dx)x + (idy)(1 − iy) = x2 |10 + iy|10 + y 2 |10 = 1 + i.
C1 0 0 2 2

On the second path, we have

Z Z 1 Z 1
1 1
z̄dz = (idy)(−iy) + (dx)(x − i) = y 2 |10 + x2 |10 − ix|10 = 1 − i.
C2 0 0 2 2

The path dependence indicates that z̄ is not analytic (in particular, it shows that it must be non-analytic at
least one point in the box {z = x + iy|x ∈ [0, 1], y ∈ [0, 1]}; it happens to be non-analytic in the whole box, but
that does not follow directly from the path-dependence shown in this specific result).

IV. 11.3.7

Consider a counter-clockwise circular contour C with radius R. Then

I I   I   I  
dz 1 1 1 1
= dz − = dz − = 2πi − 2πi Θ(R − 1) = 2πi [1 − Θ(R − 1)] ,
C z(z + 1) C z z+1 C z C z+1

where we have for each term used the result that, for a counter-clockwise contour C enclosing the point z0 , the
H
integral C dz/(z − z0 ) = 2πi, and is zero if z0 lies outside the contour. Θ(x) is the Heaviside function, with
Θ(x > 0) = 1 and Θ(x < 0) = 0. So for R > 1, the result is zero; for 0 < R < 1, the result is 2πi.
P
Equivalently, the integral counter-clockwise around the circle is given by 2πi (enclosed residues) = 2πi(Res(0)+
Res(−1)); then since Res(0) = 1 and Res(−1) = −1, the result is zero for R > 1, and 2πi for 0 < R < 1.
Note that a ‘first principles’ derivation here would use Cauchy’s Theorem to deform the circular contour C1 to
the contour C 0 = C2 ∪ C5 ∪ C3 ∪ C4 in Fig. 1. But in the limit where the vertical separation between C4 and C5

Mathematical Methods of Physics v1.0 Autumn 2015

20 October 2015, 2:20am
The University of Chicago, Department of Physics Page 3 of 3

C1
C2 C3
C4

C5

FIG. 1

R R
goes to zero, C4 = − C5 because the function f (z) is analytic (and thus continuous) in the region between these
curves; these contributions thus cancel out. In this same limit, C2,3 go to closed circular contours, which can then
R R R
evaluated directly in the limit where the radius of each contour goes to zero, so C1 = limR2 →0 C2 + limR3 →0 C3 .

Mathematical Methods of Physics v1.0 Autumn 2015

20 October 2015, 2:20am