Analysis

Singular integrals and Calderón–Zygmund theory

The study of integral operators with singular kernels, central to modern harmonic analysis and PDE.

IntuitionA kernel too singular to be an ordinary integral

A photo-editing 'sharpen' filter, an audio phase-shifter, and a crack in an elastic material all hide the same mathematical object: a convolution against a kernel that blows up at the origin, like $1/x$ in one dimension. Such a kernel is not absolutely integrable near 0, so $\int K(x-y)f(y)\,dy$ does not converge in the ordinary sense — yet the positive and negative parts of the kernel on either side of the singularity cancel almost perfectly, and a principal-value limit exists for reasonable $f$ . Singular integral operators are exactly the operators built this way, and Calderón–Zygmund theory is the machinery that makes sense of them and proves they behave as well as one could hope.

The Hilbert transform arises as the boundary values of a holomorphic function on the upper half-plane; explore how this complex map behaves as you zoom near the boundary.

UndergraduateThe prototype: the Hilbert transform

Definition: Hilbert transform

For a nice function $f$ on $\mathbb{R}$ , its Hilbert transform is the principal-value convolution $Hf(x) = \text{p.v.}\ \dfrac{1}{\pi}\int_{-\infty}^{\infty} \dfrac{f(y)}{x-y}\,dy$ : the singularity at $y=x$ is excised symmetrically before taking the limit, which is exactly what lets the cancellation between $y<x$ and $y>x$ produce a finite value.

Hf(x) = \text{p.v.}\ \dfrac{1}{\pi}\int_{-\infty}^{\infty} \dfrac{f(y)}{x-y}\,dy

On the Fourier side the Hilbert transform is remarkably simple: it multiplies each frequency by a pure phase, $\widehat{Hf}(\xi) = -i\,\operatorname{sgn}(\xi)\, \hat f(\xi)$ , flipping sign at $\xi=0$ . Every frequency is rotated by $90^\circ$ (with the sign of the rotation depending on whether the frequency is positive or negative) but not rescaled — which already hints that $H$ should be an isometry on $L^2$ .

\widehat{Hf}(\xi) = -i\,\operatorname{sgn}(\xi)\, \hat f(\xi)

Definition: Calderón–Zygmund operator

More generally, a Calderón–Zygmund operator is a convolution $Tf(x) = \text{p.v.}\int_{\mathbb{R}^n} K(x-y) f(y)\,dy$ on $\mathbb{R}^n$ whose kernel $K$ satisfies a size bound and a smoothness (Hörmander) bound away from the origin, $|K(x)| \le \dfrac{C}{|x|^n}, \qquad |\nabla K(x)| \le \dfrac{C}{|x|^{n+1}}$ , together with the assumption that $T$ is already known to be bounded on $L^2$ . The Hilbert transform is the case $n=1$ , $K(x)=1/(\pi x)$ ; the Riesz transforms $R_j f = \text{p.v.}\, c_n \int \frac{x_j-y_j}{|x-y|^{n+1}} f(y)\,dy$ are the higher-dimensional analogues used below in elliptic PDE.

From one dimension to many
Operator	Dimension	Kernel $K(x)$	Typical use
Hilbert transform	$n=1$	$1/(\pi x)$	Analytic signal, phase shifting
Riesz transforms	$n\ge 2$	$(x_j-y_j)/\|x-y\|^{n+1}$	Second-derivative estimates for elliptic PDE
General Calderón–Zygmund $T$	$n\ge 1$	Size + Hörmander smoothness	Common framework for all of the above

ResearchKey theorems

Theorem: L² boundedness of the Hilbert transform

For every $f \in L^2(\mathbb{R})$ , $\|Hf\|_{L^2} = \|f\|_{L^2}$ : the Hilbert transform is not merely bounded but an isometry on $L^2$ .

Why is it true?

This is the base case every general Calderón–Zygmund theorem builds on: before controlling a singular integral operator on $L^1$ or $L^p$ , one first needs it to be well-behaved on $L^2$ , where Plancherel gives direct access via the Fourier transform.

Proof

Step 1 (find the multiplier via a regularization). The kernel $1/(\pi x)$ is not integrable, so approximate it: for $\epsilon>0$ let $k_\epsilon(x) = \frac{1}{\pi}\frac{x}{x^2+\epsilon^2}$ , an odd, integrable approximation that converges to the principal-value kernel as $\epsilon\to 0$ . A direct (contour or tables) computation gives $\hat k_\epsilon(\xi) = -i\,\operatorname{sgn}(\xi)\, e^{-2\pi\epsilon|\xi|}$ .

Step 2 (pass to the limit). As $\epsilon \to 0^+$ , $e^{-2\pi\epsilon|\xi|} \to 1$ for every fixed $\xi \ne 0$ , and $k_\epsilon * f \to Hf$ for nice $f$ . Passing to the limit inside the convolution theorem gives the multiplier identity $\widehat{Hf}(\xi) = -i\,\operatorname{sgn}(\xi)\, \hat f(\xi)$ : the Hilbert transform acts on the frequency side by multiplication by $-i\,\operatorname{sgn}(\xi)$ .

Step 3 (apply Plancherel). Since $|-i\,\operatorname{sgn}(\xi)| = 1$ for every $\xi \ne 0$ (a single point has measure zero and does not affect the integral), $|\widehat{Hf}(\xi)|^2 = |\hat f(\xi)|^2$ for a.e. $\xi$ . Integrating and invoking Plancherel's theorem on both sides gives $\|Hf\|_{L^2} = \|f\|_{L^2}$ : the Hilbert transform is an isometry on $L^2(\mathbb{R})$ , in particular bounded.

Theorem: The Calderón–Zygmund theorem

If a convolution operator $T$ is bounded on $L^2(\mathbb{R}^n)$ and its kernel satisfies the Calderón–Zygmund conditions $|K(x)| \le \dfrac{C}{|x|^n}, \qquad |\nabla K(x)| \le \dfrac{C}{|x|^{n+1}}$ , then $T$ is of weak type (1,1), $\big|\{x : |Tf(x)| > \lambda\}\big| \le \dfrac{C}{\lambda}\|f\|_{L^1}$ , and consequently (by the Marcinkiewicz interpolation theorem, plus duality) $T$ is bounded on $L^p(\mathbb{R}^n)$ for every $1<p<\infty$ .

Why is it true?

This single theorem explains why singular integrals — despite their kernels not being absolutely integrable — behave like ordinary bounded operators on almost every $L^p$ space: the only casualty is the endpoint $p=1$ , where boundedness weakens from strong to merely weak type.

Proof

Step 1 (Calderón–Zygmund decomposition). Fix $f \in L^1$ and $\lambda>0$ . Using a stopping-time argument on dyadic cubes (equivalently, the Hardy–Littlewood maximal function $Mf$ ), split $f = g + b$ where the 'good' part satisfies $\|g\|_\infty \le C\lambda$ and $\|g\|_1 \le \|f\|_1$ , and the 'bad' part $b = \sum_j b_j$ is a sum of pieces supported on pairwise disjoint dyadic cubes $Q_j$ with mean zero, $\int_{Q_j} b_j = 0$ , and total measure $\sum_j |Q_j| \le \frac{C}{\lambda}\|f\|_1$ (this is exactly the weak-(1,1) bound for $M$ applied to the stopping cubes).

Step 2 (the good part). Since $T$ is bounded on $L^2$ and $\|g\|_2^2 \le \|g\|_\infty \|g\|_1 \le C\lambda\|f\|_1$ , Chebyshev's inequality gives $|\{|Tg| > \lambda/2\}| \le \frac{4}{\lambda^2}\|Tg\|_2^2 \le \frac{C}{\lambda}\|f\|_1$ : the good part alone already satisfies the weak-(1,1) bound.

Step 3 (the bad part, off the doubled cubes). Let $Q_j^*$ be the cube with the same center as $Q_j$ and twice the side length; the union $\bigcup_j Q_j^*$ has measure at most $\frac{C}{\lambda}\|f\|_1$ as well. Away from $Q_j^*$ , the mean-zero condition on $b_j$ lets one subtract a constant from the kernel: $Tb_j(x) = \int_{Q_j} \big(K(x-y)-K(x-c_j)\big) b_j(y)\,dy$ for the center $c_j$ of $Q_j$ , and the Hörmander smoothness bound $|K(x)| \le \dfrac{C}{|x|^n}, \qquad |\nabla K(x)| \le \dfrac{C}{|x|^{n+1}}$ shows this difference is small enough that $\int_{(Q_j^*)^c} |Tb_j(x)|\,dx \le C\|b_j\|_1$ .

Step 4 (combine). Summing Step 3 over $j$ and applying Chebyshev off the doubled cubes gives $\big|\{x \notin \bigcup_j Q_j^* : |Tb(x)|>\lambda/2\}\big| \le \frac{C}{\lambda}\sum_j \|b_j\|_1 \le \frac{C}{\lambda}\|f\|_1$ . Adding the measure of $\bigcup_j Q_j^*$ from Step 1, the bad part contributes $O(\|f\|_1/\lambda)$ to the level set as well, and combining with Step 2 for the good part gives the full estimate $\big|\{x : |Tf(x)| > \lambda\}\big| \le \dfrac{C}{\lambda}\|f\|_{L^1}$ .

UndergraduateReal-World Applications and Worked Examples

Calderón–Zygmund theory earns its central place in analysis by controlling two very different-looking objects with the same tool: the second derivatives of solutions to elliptic PDE, and the 'quadrature' signal engineers need to extract instantaneous amplitude and phase from a real-world waveform.

Example: Second derivatives of the Newtonian potential

In the theory of elliptic PDE, one wants to solve $\Delta u = f$ on $\mathbb{R}^n$ and control the second derivatives of $u$ in terms of $f$ . If $f \in L^p(\mathbb{R}^n)$ for some $1<p<\infty$ , why does $\|D^2 u\|_{L^p} \le C_p \|f\|_{L^p}$ hold?

Solution

The solution is given by convolution with the Newtonian potential, $u = f * N$ where $N(x) \sim c_n|x|^{2-n}$ for $n\ge 3$ . Differentiating twice under the integral sign formally gives $\partial_i\partial_j u = f * (\partial_i\partial_j N)$ , but $\partial_i\partial_j N(x)$ is homogeneous of degree $-n$ with mean zero on spheres around the origin — exactly a Calderón–Zygmund kernel, up to a Dirac mass term that only matters for $i=j$ .

So away from that constant multiple of $f$ itself, $\partial_i \partial_j u = R_{ij}f$ where $R_{ij}$ is (a Riesz-transform-type) Calderón–Zygmund operator with a kernel satisfying the size and Hörmander smoothness conditions.

By the Calderón–Zygmund theorem, $R_{ij}$ is bounded on $L^p$ for every $1<p<\infty$ , so $\|\partial_i\partial_j u\|_{L^p} \le C\|f\|_{L^p}$ for each pair $i,j$ , which is exactly $\|D^2 u\|_{L^p} \le C_p \|f\|_{L^p}$ . This is the classical Calderón–Zygmund estimate that underlies elliptic regularity theory: it is precisely what fails at $p=1$ and $p=\infty$ , forcing PDE theorists to work in $L^p$ , $1<p<\infty$ , or in Hölder spaces instead.

Example: Building the analytic signal for a radio engineer

A communications engineer has a real carrier signal $f(t) = \cos(2\pi\nu t)$ with $\nu>0$ , and wants its 90°-shifted 'quadrature' companion to build the analytic signal $f(t) + iHf(t)$ used to extract instantaneous amplitude and phase (as in single-sideband radio and envelope detectors). Compute $Hf$ .

Solution

Write $\cos(2\pi\nu t) = \tfrac12 e^{2\pi i \nu t} + \tfrac12 e^{-2\pi i \nu t}$ , whose Fourier transform is a pair of spikes $\tfrac12\delta_\nu + \tfrac12\delta_{-\nu}$ at $\xi = \pm\nu$ .

Apply the Hilbert transform multiplier $\widehat{Hf}(\xi) = -i\,\operatorname{sgn}(\xi)\, \hat f(\xi)$ : at $\xi=\nu>0$ it multiplies by $-i$ , and at $\xi=-\nu<0$ it multiplies by $+i$ . So on the frequency side the transformed signal is $-\tfrac{i}{2}\delta_\nu + \tfrac{i}{2}\delta_{-\nu}$ .

Inverting, this is $-\tfrac{i}{2}e^{2\pi i\nu t} + \tfrac{i}{2}e^{-2\pi i\nu t} = \tfrac{i}{2}\big(e^{-2\pi i\nu t}-e^{2\pi i\nu t}\big) = \sin(2\pi\nu t)$ (using $e^{-i\theta}-e^{i\theta}=-2i\sin\theta$ ), giving $H(\cos(2\pi\nu t)) = \sin(2\pi\nu t)$ .

So the analytic signal is $\cos(2\pi\nu t) + i\sin(2\pi\nu t) = e^{2\pi i \nu t}$ — a single positive-frequency spinning exponential, whose modulus (here constantly 1) and phase directly give the instantaneous amplitude and phase the engineer wants, with the redundant negative-frequency component discarded entirely.

Common mistake. Two expectations from ordinary calculus fail here. First, it is tempting to think

\int K(x-y)f(y)\,dy

can just be computed as an ordinary (Lebesgue) integral; in fact the principal value — the symmetric cancellation across the singularity

y=x

— is not optional, and dropping it makes the 'integral' meaningless (it is not even absolutely convergent). Second, students often assume boundedness on

L^2

upgrades for free to boundedness on

L^1

L^\infty

; it does not. The Calderón–Zygmund theorem only gives weak type (1,1) at

p=1

, and the Hilbert transform of a bounded function need not even be bounded — the endpoints

p=1

and

p=\infty

are genuinely different regimes (governed instead by weak-

L^1

and BMO).

Research frontier as of 2026

The theory keeps sharpening. In 2012, Tuomas Hytönen resolved the long-standing $A_2$ conjecture: every Calderón–Zygmund operator $T$ satisfies the linear weighted bound $\|T\|_{L^2(w)\to L^2(w)} \le C_T [w]_{A_2}$ for Muckenhoupt weights $w$ , and this linear dependence is sharp — closing a question open since the 1970s work of Hunt–Muckenhoupt–Wheeden. Other directions remain genuinely open: the 'Zygmund conjecture' on whether the directional Hilbert transform along a general (merely measurable) vector field is bounded on $L^2$ is unresolved in full generality, with Michael Bateman and Nets Katz settling only special classes of vector fields (2013); sharp two-weight characterizations, and rough (Dini, rather than Hörmander) kernel analogues of the $A_2$ theorem, are still active areas of research.

The Fourier multiplier of the Hilbert transform, defined by $\widehat{Hf}(\xi) = -i\,\operatorname{sgn}(\xi)\, \hat f(\xi)$ , is:

The Calderón–Zygmund theorem says a singular integral operator bounded on $L^2$ with a suitable kernel is, at the endpoint $p=1$ :

In single-sideband (SSB) radio and envelope detection, engineers build the 'analytic signal' $f(t) + iHf(t)$ because:

For the Poisson equation $\Delta u = f$ with $f\in L^p(\mathbb{R}^n)$ , $1<p<\infty$ , the Calderón–Zygmund estimate controls:

References

Tuomas P. Hytönen (2012). The sharp weighted bound for general Calderón–Zygmund operators · arXiv:1007.4330
Elias M. Stein (1970). Singular Integrals and Differentiability Properties of Functions