Principal Value Distributions

Principal value distributions make singular expressions such as $1/x$ meaningful under an integral. They are essential in Fourier analysis, dispersion formulas, Green functions, and scattering theory because real-axis singularities often appear before a boundary condition has been specified.

The principal value is not an instruction to ignore the singular point. It is a distributional prescription: remove a symmetric neighborhood of the singularity, integrate what remains, and take the limit after pairing with a test function.

Definition

On a test-function space such as $\mathcal S(\mathbb R)$, the Cauchy principal value distribution $\operatorname{PV}(1/x)$ is defined by

$$\left\langle \operatorname{PV}\frac{1}{x}, \varphi \right\rangle = \lim_{\epsilon\to0^+} \left[ \int_{-\infty}^{-\epsilon} \frac{\varphi(x)}{x}\,dx + \int_{\epsilon}^{\infty} \frac{\varphi(x)}{x}\,dx \right].$$

The limit exists for smooth test functions because near $x=0$ one can write

$$\frac{\varphi(x)}{x} = \frac{\varphi(x)-\varphi(0)}{x} + \frac{\varphi(0)}{x}.$$

The first term is locally regular, while the second term cancels under symmetric exclusion around the origin. This cancellation is the point of the principal value prescription.

For a pole at $a$, one similarly defines

$$\left\langle \operatorname{PV}\frac{1}{x-a}, \varphi \right\rangle = \lim_{\epsilon\to0^+} \left[ \int_{-\infty}^{a-\epsilon} \frac{\varphi(x)}{x-a}\,dx + \int_{a+\epsilon}^{\infty} \frac{\varphi(x)}{x-a}\,dx \right].$$

The notation is compact, but the prescription is part of the object.

Ordinary Improper Integrals versus Principal Values

An ordinary improper integral treats the two sides of a singularity separately. For example, it would ask whether

$$\int_{-\infty}^{0} \frac{\varphi(x)}{x}\,dx \qquad \text{and} \qquad \int_{0}^{\infty} \frac{\varphi(x)}{x}\,dx$$

converge independently. The principal value instead removes matching intervals around the pole and only then takes the limit. The two procedures can give different answers, and the principal value can exist even when the ordinary improper integral does not.

This is why a symbol such as

$$\operatorname{PV} \int_{-\infty}^{\infty} \frac{f(x)}{x-a}\,dx$$

is not decorative. It specifies the limiting procedure.

The i0 Prescription

The principal-value distribution is tied to boundary values of complex functions. The central identity is

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

This is a distributional equality. It means that, for every suitable test function $\varphi$,

$$\lim_{\epsilon\to0^+} \int_{-\infty}^{\infty} \frac{\varphi(x)}{x\pm i\epsilon}\,dx = \left\langle \operatorname{PV}\frac{1}{x} \mp i\pi\delta, \varphi \right\rangle.$$

The sign can be checked from

$$\frac{1}{x+i\epsilon} = \frac{x}{x^2+\epsilon^2} - i\frac{\epsilon}{x^2+\epsilon^2}.$$

As $\epsilon\to0^+$, the real part tends to $\operatorname{PV}(1/x)$ and

$$\frac{\epsilon}{x^2+\epsilon^2} \to \pi\delta(x)$$

distributionally. Thus

$$\frac{1}{x+i0} = \operatorname{PV}\frac{1}{x} -i\pi\delta(x),$$

while the opposite sign gives

$$\frac{1}{x-i0} = \operatorname{PV}\frac{1}{x} +i\pi\delta(x).$$

This identity is often called the Sokhotski–Plemelj formula in its simplest real-line form. For the contour-level picture of avoiding poles above or below the real axis, see Contour Integration.

Physical Interpretation

The decomposition

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

separates two roles:

• the principal-value part describes the off-singular, dispersive contribution;
• the delta part describes the on-shell or resonant contribution localized at the zero of the denominator.

In scattering language, “on shell” means that an intermediate state satisfies the energy-conservation condition selected by the delta function. The principal-value term keeps the contribution of virtual or off-shell intermediate states. The exact physical meaning depends on the operator and convention, but the distributional split is the common mathematical skeleton.

Integrals with a Smooth Density

Suppose $f$ is smooth near $E$ and decays enough for the integral to be controlled. Then

$$\int \frac{f(\lambda)}{E-\lambda+i0}\,d\lambda = \operatorname{PV} \int \frac{f(\lambda)}{E-\lambda}\,d\lambda -i\pi f(E),$$

provided $E$ lies in the integration range. With the opposite prescription,

$$\int \frac{f(\lambda)}{E-\lambda-i0}\,d\lambda = \operatorname{PV} \int \frac{f(\lambda)}{E-\lambda}\,d\lambda +i\pi f(E).$$

The sign comes from the denominator as written. If the denominator is instead $\lambda-E\pm i0$, the sign must be recomputed rather than copied.

If the zero occurs at $g(x)=0$, the delta term also carries the usual Jacobian:

$$\delta(g(x)) = \sum_i \frac{\delta(x-x_i)} {\lvert g'(x_i)\rvert}, \qquad g(x_i)=0, \quad g'(x_i)\ne0.$$

The safest habit is to reduce the expression locally to the variable that actually vanishes.

Green Functions and Resolvents

For a self-adjoint Hamiltonian, the resolvent

$$R(z)=(z-H)^{-1}$$

is defined away from the spectrum. On a real energy in a continuous spectrum, the boundary values are written schematically as

$$G^{(\pm)}(E) = \lim_{\epsilon\to0^+} \frac{1}{E-H\pm i\epsilon}.$$

The corresponding spectral identity is often summarized as

$$\frac{1}{E-H\pm i0} = \operatorname{PV}\frac{1}{E-H} \mp i\pi\delta(E-H).$$

This is shorthand for a statement under spectral integrals. If $dP_\lambda$ is the spectral measure of $H$, then matrix elements of the resolvent have the form

$$\left\langle \psi, (E-H\pm i0)^{-1}\psi \right\rangle = \lim_{\epsilon\to0^+} \int \frac{d\mu_\psi(\lambda)} {E-\lambda\pm i\epsilon},$$

where $d\mu_\psi(\lambda)=\langle\psi,dP_\lambda\psi\rangle$. The principal-value and delta terms are therefore statements about the boundary value of this spectral integral.

Green Functions develops the operator and boundary-condition side. The present page supplies the distribution identity that explains why the symbol $i0$ carries both a principal-value contribution and an on-shell delta contribution.

Scattering Context

In the Lippmann–Schwinger construction, the free resolvent is written

$$G_0^{(\pm)}(E) = \frac{1}{E-H_0\pm i0}.$$

The $+i0$ and $-i0$ prescriptions are not small physical damping terms. They are boundary-value prescriptions. In coordinate space, they select outgoing or incoming asymptotic waves. In spectral form, the same signs give

$$G_0^{(+)}(E) = \operatorname{PV}\frac{1}{E-H_0} -i\pi\delta(E-H_0),$$

and

$$G_0^{(-)}(E) = \operatorname{PV}\frac{1}{E-H_0} +i\pi\delta(E-H_0).$$

The Lippmann–Schwinger Equation uses this prescription to build scattering states. The Scattering Convention Dictionary records the outgoing/incoming sign convention.

Fourier Transform Pair

Using the ordinary $k$ convention

$$\mathcal F[f](k) = \int_{-\infty}^{\infty} f(x)e^{-ikx}\,dx,$$

one has the distributional transform

$$\mathcal F\left[ \operatorname{PV}\frac{1}{x} \right](k) = -i\pi\,\operatorname{sgn}(k).$$

This identity is related to the Hilbert transform and is a useful warning: a singular real-space kernel can transform into a bounded but discontinuous distributional multiplier. Convention changes can move factors of $2\pi$, so compare with the Fourier Transform Table before inserting the formula into a calculation.

Common Mistakes

• Treating $\operatorname{PV}(1/x)$ as an ordinary function with a value at $x=0$.
• Replacing $1/(x\pm i0)$ by only the principal-value part and dropping the delta term.
• Copying the sign in the Sokhotski–Plemelj formula without checking whether the denominator is $x\pm i0$, $E-\lambda\pm i0$, or $\lambda-E\pm i0$.
• Forgetting the Jacobian when the singular denominator is a function $g(x)$ rather than a coordinate.
• Interpreting $+i0$ as a small physical absorption in every context; in scattering it usually encodes an outgoing boundary condition.
• Multiplying principal-value distributions with other singular distributions without a definition.

Cross-Links

• Distributions
• Distributional Derivatives
• Delta Function
• Fourier Transform
• Contour Integration
• Complex Analysis Essentials
• Green Functions
• Lippmann–Schwinger Equation
• Scattering Convention Dictionary
• Fourier Transform Table

References

• I. M. Gel’fand and G. E. Shilov, Generalized Functions, Volume 1, Academic Press, 1964.
• L. Schwartz, Théorie des distributions, Hermann, 1966.
• G. B. Folland, Fourier Analysis and Its Applications, American Mathematical Society, 1992.
• M. Reed and B. Simon, Methods of Modern Mathematical Physics, Volume I: Functional Analysis, Academic Press, 1980.
• M. Reed and B. Simon, Methods of Modern Mathematical Physics, Volume III: Scattering Theory, Academic Press, 1979.
• J. R. Taylor, Scattering Theory: The Quantum Theory of Nonrelativistic Collisions, Dover, 2006.

Exercises

1. Let $\varphi$ be an even Schwartz test function. Show that

$$\left\langle \operatorname{PV}\frac{1}{x}, \varphi \right\rangle =0.$$

Solution

For even $\varphi$, the function $\varphi(x)/x$ is odd away from $x=0$. Therefore, for each $\epsilon>0$,

$$\int_{-\infty}^{-\epsilon} \frac{\varphi(x)}{x}\,dx + \int_{\epsilon}^{\infty} \frac{\varphi(x)}{x}\,dx =0.$$

Taking the limit gives zero.

2. Derive the sign in

$$\frac{1}{x+i0} = \operatorname{PV}\frac{1}{x} -i\pi\delta(x).$$

Solution

For $\epsilon>0$,

$$\frac{1}{x+i\epsilon} = \frac{x}{x^2+\epsilon^2} - i\frac{\epsilon}{x^2+\epsilon^2}.$$

The real part tends to $\operatorname{PV}(1/x)$. The family

$$\frac{1}{\pi} \frac{\epsilon}{x^2+\epsilon^2}$$

is a delta sequence, so

$$\frac{\epsilon}{x^2+\epsilon^2} \to \pi\delta(x).$$

The imaginary term has a minus sign, giving

$$\frac{1}{x+i0} = \operatorname{PV}\frac{1}{x} -i\pi\delta(x).$$

3. If $f$ is smooth near $E$, evaluate the distributional boundary value

$$\int \frac{f(\lambda)} {E-\lambda-i0} \,d\lambda.$$

Solution

Using

$$\frac{1}{x-i0} = \operatorname{PV}\frac{1}{x} +i\pi\delta(x),$$

with $x=E-\lambda$, one obtains

$$\int \frac{f(\lambda)} {E-\lambda-i0} \,d\lambda = \operatorname{PV} \int \frac{f(\lambda)} {E-\lambda} \,d\lambda +i\pi f(E),$$

assuming $E$ is inside the integration range.

4. In scattering, why does the identity for $G_0^{(+)}(E)$ not mean that the outgoing boundary condition is “just” a delta function?

Solution

The decomposition

$$G_0^{(+)}(E) = \operatorname{PV}\frac{1}{E-H_0} -i\pi\delta(E-H_0)$$

is a spectral distribution identity. The principal-value part and the delta part together make the outgoing boundary value of the resolvent. The delta term isolates the on-shell component, but the full outgoing Green function also contains the off-shell principal-value contribution and its coordinate-space radiation behavior.