4.10 The Logarithm Function

The main goal of this section is to study the inverse of the exponential function i.e for a given \(z\), the solution of the equation \(e^w = z\).

Let us write \(z = r e^{i\theta } \,,\quad -\pi \leq \theta \leq \pi \) and \(w = u + iv\). Then the equation \(e^w = z\) becomes \[e^u\cdot e^{iv} = e^{u + iv} = z = r e^{i\theta },\,\text {i.e}\quad e^u = r, \, v = \theta + 2n\pi \] where \(n\in \mathbb {Z}\). This implies that \(u = \ln (r), \quad v = \theta + 2n\pi \) i.e \(w = u + iv = \ln (r) + i(\theta + 2n\pi ),\) where \(\ln (r) = \log _e(r)\). Since \(||z|| = r\) and \(Arg(z)= \theta \), this we can write the solution \(w\) of the equation \(e^w = z\) as follows \[w = \ln \left |z\right | + i \big (Arg(z) + 2n\pi \big )\] Because of \(n\in \mathbb {Z}\), there are infinitely many \(w\)’s. If we write \[\log (z) = \ln \left |z\right | + i \big (Arg(z) + 2n\pi \big ),\] \[\text {we have}\quad e^{\log (z)} = e^{\ln |z| + i\big (Arg(z) + 2n\pi \big )} = z\] i.e \(e^{\log (z) } = z\). Hence we have the following definition.

Definition 4.60. The multiple-valued function \(\log (z)\) defined by \begin {align*} \log (z) & = \ln \left |z\right | + i Arg(z)\\ & = \ln \left |z\right | + i\big (Arg(z) + 2\pi n\big ) \end {align*}

is called the complex logarithmic function.

Example 4.61. Find \(\log (z)\) for \(z = -1 - i\sqrt {3}\)

Solution
We compute \(\left |z\right | = 2\) and \(Arg(z) = \frac {-2\pi }{3}\). Putting them into the definition, we get \begin {align*} \log (z) & = \ln (2) + i \Big (\frac {-2\pi }{3} + 2n\pi \Big )\\ & = \ln (2) + i\Big (\frac {-2}{3} + 2n\Big )\pi , \quad n\in \mathbb {Z}\\ \end {align*}

Example 4.62. Find all complex solutions of each of the following equations. \[\text {(i)}\quad e^w = i\qquad \text {(ii)}\quad e^w = 1 + i\qquad \text {(iii)}\quad e^w = -2\]

Solution. Write \(w=x+iy\), so that \(e^w=e^x(\cos y+i\sin y)\), giving \(\left |e^w\right |=e^x\) and \(\arg e^w=y\). Matching modulus and argument turns each equation into two real ones, and because the argument is fixed only up to \(2n\pi \) every equation has infinitely many solutions.

(i)
\(\left |i\right |=1\), \(\arg i=\frac {\pi }{2}\), so \(x=0\) and \(y=\frac {\pi }{2}+2n\pi \): \[w=\left (\frac {\pi }{2}+2n\pi \right )i,\qquad n\in \mathbb {Z}.\]
(ii)
\(\left |1+i\right |=\sqrt 2\), \(\arg (1+i)=\frac {\pi }{4}\), so \(x=\frac {1}{2}\ln 2\) and \(y=\frac {\pi }{4}+2n\pi \): \[w=\frac {1}{2}\ln 2+\left (\frac {\pi }{4}+2n\pi \right )i,\qquad n\in \mathbb {Z}.\]
(iii)
\(\left |-2\right |=2\), \(\arg (-2)=\pi \), so \(x=\ln 2\) and \(y=(2n+1)\pi \): \[w=\ln 2+(2n+1)\pi i,\qquad n\in \mathbb {Z}.\]

Part (iii) is worth pausing on: \(e^w\) takes a negative value, although \(e^x\) never does for real \(x\). The complex exponential is onto \(\mathbb {C}\setminus \{0\}\); only \(0\) is missed.

Caution!!!
\(\log e^z\neq z\)

Proof. We compute that \(\left |e^z\right | = \left |e^{x + iy}\right | = e^x\) and \(Arg(e^z) = y + 2\pi n, \quad n\in \mathbb {Z}\), putting them in the definition, we have \begin {align*} \log (e^z) & = \ln \left |e^z\right | + iArg(e^z)\\ & = \ln e^x + i (y + 2n\pi )\\ & = x + iy + 2n\pi i\\ & = z + 2n\pi i, \quad n\in \mathbb {Z} \end {align*}

The logarithmic function is a multi- valued. We restrict the argument in the logarithmic function and deduce a single valued function. □

Definition 4.63 (Principal Value). The complex function \(\ln (z) , \quad z \neq 0\) defined by \[\ln (z) = \ln \left |z\right | + i Arg(z)\,, \quad -\pi < Arg(z) < \pi \] is called the principal value of the complex logarithm \(\log (z)\).

We observe that

1.
\(\ln (z)\) is a single-valued function
2.
\(\log (z)= \ln \left |z\right | + i\big ( Arg(z) + 2n\pi \big ) = \ln (z) + 2n\pi i\)

A simple computation shows that \begin {align*} \log (1) & = \ln (1) + 2n\pi i,\\\\ \ln ( 1) & = \ln \left |1\right | + i Arg(1) = 0 + 0 = 0 \end {align*}

Thus \(\log (1) = 2n\pi i\,,\quad n\in \mathbb {Z}\)

Example 4.64. Find \(\log (-1)\)

Solution
\(\log (-1) = \ln (-1) + 2n\pi i,\) \begin {align*} \implies \quad \ln (-1) & = \ln \left |-1\right | + i Arg(-1)\\ & = 0 + i \pi \\ & = \pi i \end {align*}

So \(\quad \log (-1) = \pi + 2n \pi i = (2n + 1)\pi i\)

Example 4.65. Compute the principal value \(\operatorname {Ln}(z)\) for \[\text {(i)}\quad z = i\qquad \text {(ii)}\quad z = 1 + i\qquad \text {(iii)}\quad z = -2 .\]

Solution. The principal value takes the principal argument, \(-\pi <\operatorname {Arg}z\leq \pi \), so \(\operatorname {Ln}z=\ln \left |z\right |+i\operatorname {Arg}z\). Only the argument needs care.

(i)
\(\operatorname {Ln}i=\ln 1+\frac {\pi }{2}i=\frac {\pi }{2}i\).
(ii)
\(\operatorname {Ln}(1+i)=\frac {1}{2}\ln 2+\frac {\pi }{4}i \approx 0.3466+0.7854\,i\).
(iii)
\(\operatorname {Ln}(-2)=\ln 2+\pi i\approx 0.6931+3.1416\,i\).

Compare with the previous example: there the answers carried a free \(2n\pi \); here the principal branch selects one of them. The price of single-valuedness is the cut along the negative real axis, across which \(\operatorname {Arg}\) jumps by \(2\pi \).

Since \(\operatorname {Ln}(z)\) is a single-valued function drawn from the solutions of \(e^w = z\), we may expect it to invert the exponential. In fact we have the following theorem.

Theorem 4.66. If the complex exponential function \(f(z ) = e^z\) defined on the region \(-\infty < x < \infty , \,\\ -\infty < y < \pi \), the function is one-one and the inverse of the function \(f\) is the principal values of the complex logarithm \(\boxed {f^{-1}(z) = \ln (z)}\) By the theorem, we have \(\boxed {e^{\ln (z)} = z = \ln e^z}\)

Proof. Two things must be checked: that the restricted exponential is one-to-one, so that an inverse exists at all, and that its inverse is the principal logarithm.

Injective on the strip

Suppose \(e^{z_1}=e^{z_2}\) with both \(z_j=x_j+iy_j\) in the strip \(-\pi <y\leq \pi \). Then \(e^{z_1-z_2}=1\), so \(e^{x_1-x_2}=1\) and \(y_1-y_2=2n\pi \) for some integer \(n\). The first gives \(x_1=x_2\), since the real exponential is injective. For the second, both \(y_1\) and \(y_2\) lie in an interval of length \(2\pi \), so \(\left |y_1-y_2\right |<2\pi \), forcing \(n=0\) and \(y_1=y_2\). Hence \(z_1=z_2\).

The inverse is \(\operatorname {Ln}\)

Given \(w\neq 0\), write \(w=\left |w\right |e^{i\operatorname {Arg}w}\) with \(-\pi <\operatorname {Arg}w\leq \pi \), and set \(z=\ln \left |w\right |+i\operatorname {Arg}w\). Then \(z\) lies in the strip and \(e^{z}=w\), so the map is onto \(\mathbb {C}\setminus \{0\}\) and its inverse sends \(w\) to \(\operatorname {Ln}w\).

The value \(0\) is omitted because \(\left |e^{z}\right |=e^{x}>0\) always, which is why the logarithm has no value there. □

Branches of Derivatives
Let us convert \(\log (1 + i)\) in the form of \(x + iy\) \begin {align*} \log ( 1 + i) & = \ln \left |1 + i\right | + i Arg(1 + i)\\ & = \ln (\sqrt {2}) + i ( Arg(1+ i) + 2n\pi ),\quad n\in \mathbb {Z}\\ & = \frac {\ln (2)}{2} + i \Big (\frac {\pi }{4} + 2n\pi \Big )\\ & = \frac {\ln (2)}{2} + i \Big (\frac {1}{4} + 0\Big )\pi \\ & = \frac {\ln (2)}{2} + i \Big (\frac {1}{4} \pm 2\Big )\pi \\ & = \frac {\ln (2)}{2} + i \Big (\frac {1}{4} \pm 4\Big )\pi \\ & = \frac {\ln (2)}{2} + i \Big (\frac {1}{4} \pm 6\Big )\pi \\ \end {align*}

If we fix an arbitrary real number \(\alpha \), say \(\alpha = 7\) then can we find one value of \(\log (1 + i)\) such that \[\alpha < Arg(1 + i) < \alpha + 2n\pi \] The answer is YES. In fact, since \begin {align*} \alpha = 7 < \Big (\frac {1}{4} + 2\Big ) \pi \approx 7.0 < 8 \end {align*}

\[7 + 2\pi = \alpha + 2\pi \]

So the complex number \(\,\log ( 1 + i) = \frac {\ln (2)}{2} + i\Big (\frac {1}{4} + 2n\Big )\pi \,\) satisfies the condition
\(\, \alpha < Arg(1 + i) < \alpha + 2n\).
Now we generalise the argument to any complex number \(z \neq 0\) instead of \(1 + i\).

Theorem 4.67. Analyticity of the principal branch of \(\ln (z),\,\boxed {\ln (z) = \log _e (r) + Arg(z)}\,\) is given by \(f'_1(z) = \frac {1}{z}\).

Proof. Note that \(f_1\) is defined on the domain \(\left |z\right | > 0,\, -\pi \leq \operatorname {arg}(z) \leq \pi \quad \cdots \quad (*)\). In polar co ordinates, \(f_1 = \log _e r + \theta i\). So we can set \(u(r,\theta ) = \log _e r\quad ,\quad v(r,\theta ) = \theta \).
Note that, the Cauchy-Riemann equations in polar coordinates are given by \[\frac {\partial u}{\partial r} = \frac {1}{r}\,\frac {\partial v}{\partial \theta }\qquad ,\qquad \frac {\partial v}{\partial r} = \frac {-1}{r}\,\frac {\partial u}{\partial \theta }\] Now since \(u(r,\theta ) = \log _er\) and \(v(r,\theta ) = \theta \) we have \begin {align*} \frac {\partial u}{\partial r} & = \frac {1}{r}\qquad ,\qquad \frac {\partial v}{\partial \theta } = 1\\\\ \frac {\partial v}{\partial r} & = 0\qquad ,\qquad \frac {\partial u}{\partial \theta } = 0 \end {align*}

So we see that \(f_1\) satisfies the Cauchy-Riemann equations in polar form. Also \(u, v\) and the first partial derivatives of \(u\) and \(v\) are continuous at all points in the domain \((*)\).
It follows that \(f_1\) is analytic in this domain. We now compute the derivative of \(f_1\). Note that \(\, z = re^{i\theta }\). So \[\frac {\partial f_1}{\partial r}= \frac {\partial f_1}{\partial z}\cdot \frac {\partial z}{\partial r}\implies \frac {\partial f_1}{\partial r}=\frac {\partial f_1}{\partial z}\cdot e^{i\theta }\implies \frac {\partial f_1}{\partial z} = \frac {e^{i\theta }\,\partial f_1}{\partial r}\]

Recall that \(f'_1(r) = \frac {\partial u}{\partial r} + i\frac {\partial v}{\partial r}.\,\) This means that \[\frac {\partial f_1}{\partial z} = e^{-i\theta }\Big (\frac {1}{r} + 0\Big ) = \frac {1}{re^{i\theta }} = \frac {1}{z}\] □

Questions on this section

Stuck on something here? Ask below and it stays attached to this topic.