hw05 Sols
hw05 Sols
1. We (supposedly, but see quiz 2) know that if the derivative of a function f exists at z0 then the
Cauchy-Riemann equations must hold, but that the converse is not necessarily true (additional
conditions are needed, such as continuity of partials).
Show that for the function
( 2 2
x −y −2xy i
x+iy if z 6= 0
f (z) = f (x + iy) =
0 if z = 0
the Cauchy-Riemann equations hold at z = 0 but f is not differentiable at z = 0.
Hint: consider z → 0 along the real axis and along the line y = x.
To check the Cauchy-Riemann equations, we must check that ux = vy and uy = −vx , where
2 z3
f (x + iy) = u(x, y) + iv(x, y). Observe that for z 6= 0, f (z) = zz = |z| (this isn’t strictly necessary,
but is easier for me to think of.) So for z = x + iy, we have
(x − iy)2 (x − iy)3 x3 − 3xy2 y3 − 3x2 y
f (z) = = 2 = + i .
x + iy x + y2 x2 + y2 x2 + y2
Now we calculate
u(x, 0) − u(0, 0) x3 /x2
ux (0, 0) = lim = lim = lim 1 = 1.
x→0 x x→0 x x→0
Similarly,
v(0, y) − v(0, 0) y3 /y2
vy (0, 0) = lim = lim = 1,
y→0 y y→0 y
so ux (0, 0) = 1 = vy (0, 0). The other two partials are even easier:
u(0, y) − u(0, 0) 0 v(x, 0) − v(0, 0) 0
uy = lim = lim 3 = 0 and vx = lim = lim 3 = 0
y→0 y y→0 y x→0 x y→0 x
and hence uy (0, 0) = 0 = −vx (0, 0), showing the Cauchy-Riemann equations hold.
To see that f (z) is not differentiable at z = 0, first look at the derivative (as a limit) along
the real axis (which is actually just ux + ivx ), where we have
f (x + 0i) − f (0) x2 /x
lim = lim = 1.
x→0 x x→0 x
But if we look along the line x = y (that is, Re z = Im z), we get
f (x + ix) − f (0) x2 − x2 − 2x2 i −2i
lim = lim lim = −1 − i.
x→0 x + ix x→0 (x + ix) · x x→0 1 + i
Obviously, these are not the same. The complex derivative of f does not exist at z = 0 (since
we get different values of the limit by different approaches to 0), despite the Cauchy-Riemann
equations being satisfied there.
2
2. Explain why Re(e1/z ) is harmonic everywhere except at the origin.
On any domain not containing the origin, 1/z2 is analytic. Since the exponential function is
2
entire, the composition e1/z is analytic on any domain avoiding the origin.
The real part of any analytic function is always a harmonic function.
If you are a masochist, you can write the function out in real and imaginary parts, and then
compute that uxx + vyy = 0. But that is a lot of work, and I certainly don’t want to do that.
MAT342.02, Spring 2019 Homework 5 Solutions, Page 2
3. (a) Assume that w ∈ C with α < Im w < α + 2π for some (fixed) α ∈ R. Show that for z = r eiθ ,
when the branch of logarithm
log z = ln r + iθ , with r > 0, α < θ < α + 2π
is used, we always have log(ew ) = w.
Let’s write w = x + iy, where x ∈ R and α < y < α + 2π. Then ew = ex+iy = ex eiy , and
`og(ew ) = { ln(ex ) + (y + 2πn)i } , n∈Z ,
where `og represents the multivalued logarithm and (as usual) ln represents the logarithm
from R+ to R. But the branch of the logarithm taken in this problem corresponds to n = 0
(since α < y < α + 2π), and we have
log(ew ) = ln(ex ) + iy = x + iy = w .
(b) Give an branch of the logarithm that ensures that for β = 1 + i we have
log(β 8 ) = 8 log(β ).
√ √
Observe that β = 1 + i = 2 eiπ/4 and β 8 = ( 2 eiπ/4 )8 = 16. Consequently, if the branch
of the logarithm chosen has log β = ln22 + i π4 , we also need log β 8 = log(16) = 4 ln 2 + 2π i.
This means we need to take a branch cut of argument α where 0 < α < π/4. For example,
we may choose α = π/8. Then
log (1 + i)8 = log(16) = 4 ln 2 + 2π i = 8( ln22 + π4 i) = 8 log(1 + i) ,
where log is the branch of the logarithm with log(z) = ln |z| + i arg z, π/8 < arg z < 17π/8.
(c) For the same β as in the previous part, give a branch of the logarithm for which
log(β 8 ) 6= 8 log(β ).
Again, assuming we take a branch so that Im log β = π/4, any branch cut of argument α
where α ≥ π/4 or α ≤ 0 will do. For example, for the principal branch of the logarithm
(that is, with α = −π), we have
Log β 8 = Log 16 = ln 16 = 4 ln 2 6= 8 Log(1 + i) = 8( ln22 + π4 i) = 4 ln 2 + 2π i .
4. Calculate each of the following. Keep in mind that these expressions can be multivalued.
√ 3/2 1/2 √ √
(a) (−1 + i 3)3/2 = 2e2π i/3 = 8e2π i = (8)1/2 = { 2 2, −2 2 } .
Note that this is the same result that you would get by writing
√ √
exp( 2 log(−1 + 3i)) = exp ln(2 ) + 2 (2π/3 + 2nπ)i = 8 e(π+3nπ)i
3 3/2 3
for n ∈ Z
(b) iπ = eπ log i = exp π(iπ/2+2nπ i) = exp i(π 2 /2+2nπ 2 ) = cos 4n+1 2 4n+1 2
2 π +i sin 2 π
for n ∈ Z .
(c) π i = ei log π = ei(ln π+2nπ i) == ei ln π−2nπ) = e2nπ (cos(ln(π)) + i sin(ln(π))) for n ∈ Z .
π
(d) i−2i i−2i = exp(−2i log i) = exp(−2i(i + 2nπ i) = e(4n+1)π for n ∈ Z .
2
Observe that the answer to (a) has two values, as you should expect from a square root. By
contrast, (b) has infinitely many values distributed densely around the unit circle, the answer
to (c) is an infinite set of values along a ray of argument ln π ≈ 1.1447 and part (d) has infinitely
many values, but they are all real.
5. Find all roots of the equation sin z = cosh 4 by equating the real parts of both sides, then equating
the imaginary parts.
1 iz −iz
Recall that sin(z) = 2i (e − e ), and writing z = x + iy gives the equation
ix−y
e − e−ix+y = 2i cosh(4)
which we can rewrite as
e−y (cos x + i sin x) − ey cos(−x) + i sin(−x) = 2i cosh(4) .
Note that cos(−x) = cos(x) and sin(−x) = − sin(x). Then equating real parts of both sides,
and then imaginary parts yields the two equations
cos x (e−y − ey ) = 0 sin x (e−y + ey ) = 2 cosh(4) .
The equation on the left tells us that either y = 0 or x = π2 + 2nπ for some n ∈ Z. But
from the right-hand equation, we cannot have y = 0 (since cosh(4) = 12 (e4 + e−4 ) 6= 0). Hence
sin x = 1.
Using this, we can rewrite the right-hand equation as e−y + ey = e4 + e−4 . Consequently,
y = ±4. There can be no other solutions, since e−y + ey is monotonically decreasing for y < 0
and increasing for y > 0.
Hence, any solution to the equation is of the form
π
z= 2 + 2nπ ± 4i, for n ∈ Z
(and every such number is a solution).
6. Show that sinh z = 0 if and only if z = i nπ with n ∈ Z. You may use facts we already established
about ez , sin z and cos z without reproving them explicitly.
Recall that sinh z = −i sin(iz). We already know that all the zeros of sin z are real numbers of
the form nπ with n ∈ Z, and hence the zeros of sinh z are exactly the points on the imaginary
axis of the form nπ i for n ∈ Z.