Contour Integration

Contour integration is the evaluation and transformation of complex integrals along oriented paths in the complex plane. In quantum mechanics it is the practical language behind Fourier-integral evaluation, Green-function boundary prescriptions, scattering poles, dispersion relations, and saddle-point estimates.

This page explains the method-level idea: what a contour integral is, when contours can be deformed, how singularities obstruct deformation, and how large-arc estimates decide which contours are allowed. The Residue Theorem and Branch Cuts receive their own canonical pages; here they appear only as the mechanisms that make contour deformation useful.

What a Contour Is

A contour $C$ is an oriented path in the complex plane. If it is parametrized by

$$z=z(t), \qquad a\le t\le b,$$

then the contour integral is

$$\int_C f(z)\,dz = \int_a^b f(z(t))z'(t)\,dt.$$

Orientation matters. Reversing the contour changes the sign:

$$\int_{-C} f(z)\,dz = - \int_C f(z)\,dz.$$

For a closed contour, one often writes

$$\oint_C f(z)\,dz.$$

The symbol $\oint$ is not a theorem; it only says that the contour closes.

First Example: A Circle

Let $C_R$ be the counterclockwise circle

$$z(t)=Re^{it}, \qquad 0\le t\le2\pi.$$

For $f(z)=1/z$,

$$\oint_{C_R}\frac{dz}{z} = \int_0^{2\pi} \frac{iRe^{it}}{Re^{it}}\,dt = 2\pi i.$$

The value is independent of $R$ because every counterclockwise circle around the origin has the same winding around the singularity.

For $f(z)=z^n$ with $n\ne -1$,

$$\oint_{C_R} z^n\,dz=0.$$

The coefficient of $(z-z_0)^{-1}$ is special; it is the residue. Residue Theorem develops that statement systematically.

Cauchy Deformation Principle

If $f$ is analytic throughout a region swept out while continuously deforming one contour into another, then the integral does not change, provided endpoints and boundary conditions are respected.

For a closed contour $C$ in a simply connected region where $f$ is analytic,

$$\oint_C f(z)\,dz=0.$$

This is the core reason analytic functions are so powerful: an integral can be changed by changing the path, as long as no singularity is crossed.

If a deformation crosses a pole or branch cut, the integral changes. The change is not a vague failure; it is controlled by local data at the singularity or by the discontinuity across a branch cut.

Poles as Obstructions

Consider two closed contours, $C_1$ and $C_2$, with the same orientation. If the annular region between them contains no singularities of $f$, then

$$\oint_{C_1} f(z)\,dz = \oint_{C_2} f(z)\,dz.$$

If the region between them contains a simple pole, the equality fails by a residue contribution. For example, with

$$f(z)=\frac{g(z)}{z-z_0},$$

where $g$ is analytic near $z_0$, the local coefficient

$$\operatorname{Res}_{z=z_0}f = g(z_0)$$

is what contributes when a contour encloses $z_0$.

This local viewpoint is often enough to understand the physics: poles are where analytic deformation stops being free.

Closing Fourier Contours

A common quantum-mechanics integral has the form

$$\int_{-\infty}^{\infty} e^{ikx}F(k)\,dk.$$

When $k=a+ib$,

$$e^{ikx} = e^{iax}e^{-bx}.$$

For $x\gt0$, the exponential decays in the upper half-plane. For $x\lt0$, it decays in the lower half-plane. This determines which large semicircle is useful, provided $F(k)$ does not grow too quickly.

This is why the sign of an exponential and the sign of a time or distance variable decide which poles contribute to a Fourier integral. The rule is not “always close above” or “always close below”; it is dictated by decay.

Worked Example: Massive One-Dimensional Kernel

Evaluate

$$I(x) = \int_{-\infty}^{\infty} \frac{e^{ikx}}{k^2+\kappa^2}\,dk, \qquad \kappa>0.$$

The denominator factors as

$$k^2+\kappa^2 = (k-i\kappa)(k+i\kappa),$$

so the poles are at $k=i\kappa$ and $k=-i\kappa$.

For $x>0$, close the contour in the upper half-plane. The pole inside is $k=i\kappa$, and the large arc vanishes. The residue at $i\kappa$ is

$$\frac{e^{ikx}}{k+i\kappa} \bigg\rvert_{k=i\kappa} = \frac{e^{-\kappa x}}{2i\kappa}.$$

Thus

$$I(x) = 2\pi i\, \frac{e^{-\kappa x}}{2i\kappa} = \frac{\pi}{\kappa}e^{-\kappa x}, \qquad x>0.$$

For $x\lt0$, close in the lower half-plane. The orientation is clockwise, and the pole is $k=-i\kappa$. The result is

$$I(x) = \frac{\pi}{\kappa}e^{\kappa x}, \qquad x<0.$$

Together,

$$I(x) = \frac{\pi}{\kappa} e^{-\kappa\lvert x\rvert}.$$

With a factor of $1/(2\pi)$, this is the Fourier-space construction of the decaying Green function for $-d^2/dx^2+\kappa^2$ on the line.

Large-Arc Estimates

Closing a contour is valid only if the added arc does not contribute in the limit. A typical estimate is

$$\left\lvert \int_{\Gamma_R} f(z)\,dz \right\rvert \le \operatorname{length}(\Gamma_R) \max_{z\in\Gamma_R}\lvert f(z)\rvert.$$

If $\Gamma_R$ is a semicircle of radius $R$, the length is $\pi R$. The integrand must therefore decay faster than $1/R$ on average, or have exponential damping strong enough to overcome algebraic growth.

Jordan-lemma arguments are refinements of this idea for oscillatory factors such as $e^{ikx}$. In practice, one must check:

• which half-plane makes the exponential decay;
• how fast the nonexponential factor grows;
• whether there are branch cuts or poles on the proposed path;
• whether small semicircles around real-axis singularities contribute.

Poles on the Real Axis

When a pole sits directly on the integration path, the integral is ambiguous until a prescription is specified. Common prescriptions include:

• indenting the contour above or below the pole;
• using a Cauchy principal value;
• shifting the pole by an $i0$ prescription;
• specifying retarded, advanced, incoming, or outgoing boundary conditions.

The distributional identity

$$\frac{1}{x\pm i0} = \operatorname{PV}\frac{1}{x} \mp i\pi\delta(x)$$

is not a decorative convention. It records which side of the pole is approached. For the distribution-level statement and sign checks, see Principal Value Distributions.

Branch Cuts and Deformation

Some functions, such as $\sqrt z$ and $\log z$, cannot be made single-valued on the punctured complex plane. A branch cut is a chosen barrier that makes a single branch possible on the remaining domain.

Contour deformation must respect that choice. If a contour is pushed across a branch cut, the integral changes by the discontinuity between the two sides of the cut. In scattering and spectral theory, branch cuts often represent continua or thresholds rather than isolated states.

This is the same logic as for poles, but with extended singular structure instead of isolated singularities.

Green-Function Interpretation

Fourier-space Green functions frequently have denominators such as

$$\frac{1}{E-\hbar^2k^2/(2m)\pm i0}.$$

The pole locations and the $i0$ prescription determine how the contour passes the singularities. In coordinate space this becomes an outgoing, incoming, retarded, or advanced boundary condition, depending on the problem.

This is why Green functions are never just algebraic inverses. The contour or boundary prescription is part of the inverse.

Common Mistakes

• Forgetting contour orientation, especially when closing in the lower half-plane.
• Closing in a half-plane where the exponential grows.
• Applying a residue calculation before checking that the large arc vanishes.
• Moving a contour through a pole without adding the corresponding contribution.
• Treating a real-axis pole as defined before choosing a principal-value or $i0$ prescription.
• Ignoring branch cuts when deforming contours.
• Confusing the contour chosen for one sign convention with another convention’s outgoing or incoming prescription.

Cross-Links

• Analytic Functions
• Complex Exponentials
• Residue Theorem
• Branch Cuts
• Complex Analysis Essentials
• Fourier Transform
• Principal Value Distributions
• Green Functions
• Asymptotic Analysis
• Scattering Amplitude

References

• L. V. Ahlfors, Complex Analysis, 3rd ed., McGraw-Hill, 1979.
• E. M. Stein and R. Shakarchi, Complex Analysis, Princeton University Press, 2003.
• J. W. Brown and R. V. Churchill, Complex Variables and Applications, 9th ed., McGraw-Hill, 2014.
• G. B. Arfken, H. J. Weber, and F. E. Harris, Mathematical Methods for Physicists, 7th ed., Academic Press, 2013.
• C. M. Bender and S. A. Orszag, Advanced Mathematical Methods for Scientists and Engineers, Springer, 1999.

Exercises

1. Parametrize the counterclockwise circle $C_R:\ z=Re^{it}$ and compute $\oint_{C_R} z^2\,dz$.

Solution

With $z=Re^{it}$ and $dz=iRe^{it}\,dt$,

$$\oint_{C_R}z^2\,dz = \int_0^{2\pi} R^2e^{2it} iRe^{it}\,dt = iR^3 \int_0^{2\pi}e^{3it}\,dt =0.$$

The result also follows from analyticity of $z^2$ inside the closed contour.

2. For

$$I(x)= \int_{-\infty}^{\infty} \frac{e^{ikx}}{k-i\alpha}\,dk, \qquad \alpha>0,$$

which half-plane should be used when $x>0$, and which pole is enclosed?

Solution

For $x>0$, the factor $e^{ikx}=e^{iax}e^{-bx}$ decays when $b>0$, so the contour should be closed in the upper half-plane. The pole $k=i\alpha$ lies in the upper half-plane and is enclosed.

3. Why does closing a contour in the lower half-plane introduce a minus sign in a residue calculation?

Solution

The usual residue formula assumes positive, counterclockwise orientation. A real-axis contour closed by a lower semicircle is oriented clockwise around the enclosed region. Therefore the closed-contour integral equals $-2\pi i$ times the sum of residues inside.

4. Explain why a pole on the real integration axis requires extra information.

Solution

The integral is not defined by the formula alone because the path passes through a singularity. One must specify whether the contour avoids the pole above or below, whether a Cauchy principal value is meant, or whether an $i0$ prescription selects a boundary value. Different choices can differ by imaginary delta-function or residue contributions, and in physics they often encode different boundary conditions.