Fatou's lemma states that for nonnegative measurable ,
Indeed is measurable, increases to , and satisfies for every . The monotone convergence theorem gives , proving the claim, including infinite values.
The dominated convergence theorem states that if measurable almost everywhere and almost everywhere for one integrable nonnegative , then is integrable and ; in fact . Integrability follows from . For real functions, Fatou applied to gives , and applied to gives the reverse inequality against the upper limit. Thus the integrals converge. To obtain the stronger conclusion, Fatou applied to gives , forcing the latter limit to be zero. Complex functions may be treated through their real and imaginary parts or the same absolute-difference argument.
Use the intrinsic boundary of a simply connected domain, so different approaches to the two sides of a slit remain distinct. The mapping-out function extends as a homeomorphism from this intrinsic compactification to that of the complex upper half-plane. Set . The imaginary coordinate of planar Brownian motion hits zero in finite time almost surely, and is no larger than that time. Thus and continuity gives a finite Euclidean exit point.
By conformal invariance of planar Brownian motion, is a planar Brownian motion in the complex upper half-plane, run with clock
The terminal clock cannot be infinite: that would make the transformed Brownian motion stay in the upper half-plane forever. It cannot stop while the transformed path is in the interior either, since continuity of would then put the original exit point inside . Hence the terminal clock is precisely the transformed Brownian exit time. The transformed path converges to a real boundary point, and applying the extended inverse proves almost sure convergence to a point of the intrinsic boundary.
Write , and let . The point at infinity has zero harmonic measure. Conformal invariance of planar Brownian motion and the Poisson kernel for the upper half-plane give
The hydrodynamic normalization at infinity gives , so along the specified approach and . For each fixed real ,
If , dominated convergence applies because is eventually bounded. If , Fatou's lemma makes the limit infinite. This proves the harmonic-measure asymptotic at infinity
with the equality understood in the extended nonnegative reals.
For , define the set functional
The ROF level-set formulation states that the unique minimizer of has superlevel sets minimizing for almost every . Conversely, an admissible with minimizing superlevel sets is the ROF minimizer. Thus the concise characterization is
The signed layer-cake identity for quadratic fidelity and the coarea formula for BV functions give
Indeed pointwise. Its absolute integral is bounded by , so Fubini's theorem applies for . The negative-level baseline is essential; integrating the unadjusted would generally diverge.
Each attains its minimum. A minimizing sequence of indicator functions has bounded norm, and comparison with the empty set bounds its perimeter by an forcing bound. Bounded-variation compactness supplies a limiting indicator function. The forcing integral converges by dominated convergence, while relative perimeter is sequentially lower semicontinuous.
For , the supplied comparison lemma applied to makes every selected minimizer contained in up to a null set. The comparison of perimeter minimizers with ordered forcing also follows directly: compare with , compare with , add and use submodularity of relative perimeter to obtain .
Select minimizers at rational levels, remove their countably many exceptional null sets, and reconstruct . The strict superlevel set is . This union also minimizes : take , use monotone convergence of its indicator functions and sequential lower semicontinuity, and note that satisfies .
To justify finite energy before assuming it, clip this reconstruction to . Comparison with the empty set gives uniformly on a bounded interval of levels. Since , pairing with compactly supported test-field divergences bounds its total variation seminorm by . Thus belongs to the BV space before applying the coarea formula for BV functions. The layer-cake argument on the finite interval shows for every finite-energy competitor , where is clipping. In particular bounds the norms and variations uniformly. Fatou's lemma excludes infinite values of on a positive-measure set. Bounded-variation compactness and weak convergence in identify an admissible limit . For any fixed competitor, in and its variation tends to that of , by bounded-variation contraction under clipping and sequential lower semicontinuity. Thus .
The quadratic fidelity is strictly convex, so every ROF minimizer equals almost everywhere and has the selected minimizing superlevel sets. Conversely, for any whose levels minimize , integrate the levelwise inequality against any competitor's levels in the displayed identity to obtain . This proves both directions for signed as well as nonnegative data.
A one-dimensional Sobolev representative is absolutely continuous. For , the fundamental theorem of calculus and Holder inequality give
Thus has a representative in . The qualification about representatives matters because a Sobolev space element is an almost-everywhere equivalence class.
In two dimensions, the Sobolev fundamental theorem of calculus on lines and Fubini's theorem imply that almost every horizontal and vertical slice belongs to and has this one-dimensional Hölder continuity. The slice seminorm depends on the slice; this does not give one uniform pointwise estimate on the square. For , Morrey's inequality additionally gives a globally Hölder continuous representative of exponent . For , global continuity need not hold. For example, with a smooth cutoff around an interior point, lies in when , but is unbounded. At , the cutoff version of is unbounded while its gradient has finite squared integral, since
These examples distinguish Sobolev slicing and planar continuity from a false two-dimensional application of the interval exponent.
Put . A BV space is a function whose distributional derivative is a finite vector-valued Radon measure. Equivalently its total variation seminorm is finite:
The BV space has norm . For , integration by parts against the compactly supported field gives . Conversely the measurable choice on nonzero gradients attains the pointwise bound. Approximating this bounded field by smooth fields, using interior cutoffs and the finite measure , justifies the supremum and gives
It is a norm of the derivative measure, rather than a pointwise derivative at jump discontinuities.
There is a genuine mismatch in the printed definition of the next functional. Its constraints on and are independent. Hence its stated supremum, denoted , separates as
The scalar supremum is , by cutoffs approaching one, and the vector supremum is the variation. For an affine image signal with , this gives , whereas the displayed square-root area would give . The intended relaxed graph-area functional instead uses the coupled pointwise constraint , giving
where is the singular part of . Both readings have a minimizer, but their equations are different.
Here is the direct method in the calculus of variations for either reading. Let be the literal or the corrected , and define the energy on , assigning infinity elsewhere. A minimizing sequence has bounded energy by comparison with . Both , so its variation is bounded, and the fidelity bounds , hence also and . By bounded-variation compactness, a subsequence converges strongly in to , and after another subsequence almost everywhere. Fatou's lemma proves
The regularizer is a supremum of affine functionals continuous in , since the test-field divergence is bounded. It is therefore lower semicontinuous. Combining the two lower bounds proves existence of a minimizer. In fact the convex regularizer and the strictly convex squared fidelity make the minimizer unique up to null sets. This does not assert that the minimizer must belong to .
For the intended graph area, conditionally assume that the minimizer is in . For , differentiate at . The derivative of the integrand is bounded by , so dominated convergence applies. The weak equation is
that is,
This is the graph-area Euler-Lagrange equation. Compactly supported variations impose no boundary condition in this statement.
For the literal printed supremum, the constant drops out and one obtains total variation denoising. Its total variation calibration form is
The distributional equation means . In particular wherever the gradient is nonzero; writing this quotient without handling zero gradients would be incomplete. Formally the one-sided derivative of is
Minimality in the directions and bounds the remaining linear functional by the second integral. The Hahn-Banach theorem extends it on that zero-gradient set to a bounded vector field of magnitude at most one, furnishing and the displayed weak equation. Thus the literal definition has a nonsmooth subgradient equation, not the square-root equation above.
The weak topology of probability measures is the smallest topology making every map continuous for bounded continuous real on . Define the bounded-Lipschitz metric by
Using the maximum of the two bounds instead of their sum changes the metric by at most a factor of two and gives the same topology. Symmetry and the triangle inequality follow from the supremum formula. Bounded Lipschitz functions distinguish probability measures by the open-set approximations below, so is a metric. On a separable metric space, induces the weak topology.
Here are the essential details. If , integrals of every bounded Lipschitz function converge after rescaling its Lipschitz bound. For an open set , the functions
increase to ; for take . Thus . It follows both that is lower semicontinuous for the weak topology, as a supremum of continuous maps, and that beta convergence implies
This is the open-set Portmanteau criterion. Its closed-set equivalent, by complements, is
Either family of inequalities is necessary and sufficient for weak convergence of probability measures. For sufficiency of the open inequalities, apply the layer-cake formula and Fatou's lemma to the open superlevel sets of a continuous function , obtaining . Applying the same argument to gives the reverse bound. Scaling handles every bounded continuous function.
Conversely, weak convergence implies beta convergence. Choose a compact set of arbitrarily large -mass, using tightness of a probability measure on a Polish space. The unit bounded-Lipschitz class is uniformly bounded and equicontinuous on , so finitely many of its members approximate all others uniformly there. On the open -neighborhood , approximation errors increase by at most . The open-set inequality makes large for all sufficiently large . Weak convergence for the finitely many selected functions, combined with these approximation errors and the small mass outside , bounds the supremum defining by an arbitrarily small number. The same finite-test and open-mass bounds define weak neighborhoods, so the argument gives equality of the topologies, not only their convergent sequences. It does not first assume uniform tightness of the entire sequence.
The map is injective and continuous: for , is bounded continuous on , and . To prove inverse continuity on its image, suppose . For every open , there is an open with . Hence and . The open-set criterion on now gives the criterion on , so . Both spaces are metrizable, and this sequential argument proves
This is the probability pushforward embedding theorem. The printed claim needs the words “onto its image”: it is generally not onto all , since a Dirac mass at a point outside is not in its image. By Question 1, the completely metrizable subspace is a Borel G-delta set, and
Now let be weakly closed and uniformly tight. Take any sequence in . The space is compact for the weak topology. For completeness, this compactness follows by choosing a countable uniformly dense family in , extracting a diagonal subsequence of its bounded integrals, and extending the limits to a positive normalized functional on ; the Riesz-Markov-Kakutani representation theorem gives the limiting probability measure. Thus, along a subsequence, .
For every , choose compact with for every . The set is compact and closed in , so the closed-set criterion gives
Therefore , and for some . Inverse continuity gives , and closedness puts . Hence is sequentially compact, and metrizability makes it compact. This proves the required closed uniformly tight compactness criterion, rather than assuming it from Prokhorov's theorem.
Such a compact metrizable extension is always available: for a countable dense set , the coordinates embed homeomorphically into . Injectivity and inverse continuity follow by choosing close to a specified point and using the triangle inequality. The closure of the image gives .
To show separability of , use finite sums of Dirac measures on a countable dense subset of with nonnegative rational weights summing to one. This is a countable family. Given , cover a large-mass compact set by finitely many small balls centered in that subset, move the mass in each piece to its center, and move the remaining small mass to one fixed center. The bounded-Lipschitz error is at most the ball radius plus twice the exceptional mass. Approximate the resulting weights by rational weights. Thus these atomic measures are beta-dense.
Finally use a complete compatible metric on the Polish space , as required for the granted total-boundedness-to-tightness assertion. A beta-Cauchy sequence is beta-totally bounded, and so is its closure in . By the allowed assertion this closure is uniformly tight; it is weakly closed because the weak and beta topologies agree. The compactness result just proved gives a convergent subsequence, and the Cauchy property forces the whole sequence to converge in beta. Consequently beta is complete for this choice of , and
If the originally displayed compatible metric is incomplete, replace it by a complete compatible metric for this last argument; the weak topology itself is unchanged.
The dominated convergence theorem says that if measurable almost everywhere and for an integrable , then is integrable, , and consequently .
First derive Fatou's lemma from monotone convergence. For nonnegative , the measurable functions increase to , and . Monotone convergence therefore yields .
Now is measurable and almost everywhere, so is integrable. Apply Fatou's lemma to , which converges almost everywhere to :
The last integrals are nonnegative, so they tend to zero. Finally . This argument also works for complex-valued functions.
No such sequence exists. By Fatou's lemma on ,
But the proposed nonnegative measurable integrand is positive at every point of , so its integral cannot be zero. Explicitly, the sets where it is at least cover ; at least one has positive measure, forcing a positive integral. This contradicts the displayed bound.
No such sequence exists. By Fatou's lemma on ,
But the proposed nonnegative measurable integrand is positive at every point of , so its integral cannot be zero. Explicitly, the sets where it is at least cover ; at least one has positive measure, forcing a positive integral. This contradicts the displayed bound.
The dominated convergence theorem states: if measurable converge almost everywhere to and almost everywhere for an integrable , then is integrable and . In fact .
To derive it from the monotone convergence theorem, first prove Fatou's lemma. For nonnegative measurable , the functions increase to and satisfy . Monotone convergence yields .
Now almost everywhere, so is integrable. Apply Fatou's lemma to the nonnegative functions , which converge almost everywhere to :
Hence , and proves the assertion. Common exceptional null sets may be removed at the beginning since the sequence is countable.
For , write with . The chosen power map sends this to , so
At this is the plane slit along the nonnegative real axis. For every , the map is conformal on the upper half-plane.
To check finiteness of the physical exit time, start a Brownian motion at and let be its real-axis exit. The time change of a continuous process in conformal invariance of planar Brownian motion is
We know almost surely. The Cauchy exit law from a Brownian half-plane assigns probability zero to . On such a path, throughout the compact interval , because it is in the open upper half-plane before exit and has nonzero endpoint. Its modulus therefore has a positive minimum and finite maximum. The displayed integrand is bounded on that interval, proving . The image Brownian motion exits at this finite changed time.
The endpoint printed in the source is degenerate: , so its image is the singleton , not an open domain, and the map is not conformal. If exit from that set is defined for Brownian motion started at one, its exit infimum is zero. It cannot be included in a nondegenerate conformal-domain argument.
Here is the martingale argument for maximal comparisons, including the starting-value issue. For , the Cauchy-Riemann equations and Itô's formula show that
are orthogonal continuous local martingales with the same quadratic variation. Localize before large disks and before exit. The Burkholder-Davis-Gundy inequalities then give
with a constant depending only on ; Fatou's lemma removes the localization. These are centered martingales. For the uncentered processes this implies
since .
The source normalizes at a boundary point, rather than the value of at the starting point. Consequently its displayed constant cannot be uniform in with only dependence. To see this, take and . Then and . Inside this narrow wedge,
Both th moments are finite and the right one is positive; finiteness follows from the wedge radial-maximum tail of order , justified below. The source inequality would force , unbounded as . At , the immediate-exit interpretation gives the direct contradiction for the same polynomial.
For a fixed nondegenerate wedge, an uncentered inequality does hold with a constant allowed to depend on as well as . One rigorous outline for the polynomial class in this question is as follows. Separation of variables in a sector of radius gives, for exit through its circular arc, the expansion
Thus at the specified starting point the radial-maximum tail is comparable to . Hitting any fixed interior subarc has that same power, with positive leading coefficient. If a nonzero polynomial has degree , its real part has magnitude at least on some fixed interior subarc for sufficiently large . Therefore its maximal real-part th moment is infinite whenever . In that case the desired extended-real inequality is automatic.
The remaining polynomials have , so belong to a finite-dimensional vector space space. On that space
is a genuine norm: if it is zero, Brownian motion's positive probability of reaching any interior neighborhood forces throughout , so analyticity makes a purely imaginary constant, and makes it zero. The corresponding imaginary-part functional is bounded on this finite-dimensional vector space normed space; its finiteness follows from and the radial tail bound. equivalence of norms in finite dimensions gives
This establishes the fixed-domain interpretation and identifies exactly why the universal centered martingale inequality cannot silently be applied to the printed uncentered boundary normalization. It is the distinction in starting-value obstruction in wedge conjugate maximal inequalities.
Figure 1.
The power map changes the wedge angle and degenerates to a constant at zero exponent
. Details.
Reverse Fatou lemma 2026-10-31
If are random variables bounded above by an integrable function , with all relevant expectations well defined, Fatou's lemma applied to the nonnegative variables gives . In particular, bounded indicator random variables that tend almost surely to zero have probabilities tending to zero.
For each , enumerate the half-open intervals partitioning and use the functions times their indicator functions. Concatenate these finite blocks in increasing . Every function in block has L1 norm , which tends to zero. Every is in one interval of every block, so the values at that point have a subsequence increasing to infinity. Thus small integral norm does not bound the pointwise upper limit, even when every function is nonnegative. In contrast, Fatou's lemma prevents a strictly positive lower limit everywhere on an interval when the norms tend to zero.