— 1. Weighted inequalities —
In the previous notes we have taken a linear or sublinear operator on a measure space
and established
mapping properties, such as
One can view the estimate (1) as an assertion that is at least as large as some sort of “nonlinear adjoint” of
applied to
. For instance:
Exercise 1 Letbe an integral operator with bounded non-negative kernel
, and let
be bounded weights. Show that the inequality
holds for all simple functions
with finite measure support if and only if one has the pointwise estimate
.
We know that boundedness properties of an adjoint can be used to establish boundedness properties of the original operator. There is a counterpart for weighted estimates:
Proposition 2 (Equivalence of weight bounds and vector-valuedestimates) Let
, let
, let
be an arbitrary operator (possibly nonlinear!), and let
be such that
. Then the following are equivalent:
Proof: By dividing by
we may normalise
. We first show that (ii) implies (i). From the converse Hölder inequality we have
Now we establish that (i) implies (ii). Fix ; we may normalise
. We need to find a
obeying (2). From the above arguments we know that a necessary condition for
is that
Observe from Hölder that
Remark 3 By combining this with Exercise 51 from last week’s notes, we obtain a remarkable fact: ifis a linear operator, then there is a unique continuous linear extension from
to
for
if and only if for every
there exists
with
for which one has the weighted estimate (1), where
. Thus in the case of linear operators there is an equivalence between higher unweighted
estimates and weighted
estimates.
We thus see that weighted estimates can be quite powerful; they not only imply unweighted estimates, but also vector-valued versions of those estimates. Indeed, more is true: if we have a sequence of operators which obey weighted estimates with the same set of weights, then we automatically obtain vector-valued estimates which involve all the
at once. For instance, if
are as in the above proposition, and for each
there was a
(independent of
) with
such that
It is thus of interest to obtain weighted estimates, preferably those in which the final weight is given explicitly in terms of
. A particularly pleasant situation occurs when
and the weights
are in fact equal, although having estimates with asymmetric weights can still be very useful.
One particularly common (and useful) class of weights are power weights such as or
for Euclidean spaces
, especially in PDE applications where such weights tend to appear quite naturally (for instance, via the introduction of Sobolev spaces). However, we will work here with more abstract weights. Here it turns out that one can often mimic the proof of an unweighted estimate to deduce the weighted counterpart.
— 1.1. Weighted estimates for the Hardy-Littlewood maximal function —
To illustrate weighted estimates, let us establish a basic estimate for the maximal function.
Proposition 4 (Weighted Hardy-Littlewood maximal inequality) Let. Then we have the weak-type estimate
for all locally integrable
and all
, and the
estimate
for all locally integrable
and
.
Proof: By monotone convergence we may assume that both and
are bounded with compact support. The Stein-Weiss interpolation theorem (and the boundedness of
on
no matter what the weights are) ensure that it suffices to prove the weak-type
estimate. For this we mimic the proof of the unweighted estimate. We normalise
, and let
be any compact set inside
; it suffices to show that
As a corollary we obtain the following useful vector-valued generalisation of the Hardy-Littlewood maximal inequality.
Theorem 5 (Fefferman-Stein vector-valued maximal inequality) Ifand
are any sequence of locally integrable functions then
and the weak-type estimate
for all locally integrable
.
Proof: By the usual monotone convergence arguments we may restrict only finitely many of the to be non-zero, and to have all
bounded and finite measure support. We can then restrict the supremum in the maximal function
to only be over finitely many radii.
When the first claim follows from interchanging the norms and using the usual Hardy-Littlewood maximal inequality. For
the claim follows by combining Proposition 4, Proposition 2, and the fact that
is bounded on
. For
, it suffices by vector-valued real interpolation to prove the weak-type
estimate. We may take each
to be non-negative. By linearisation we can replace
by
for some measurable radius function
(note that we must permit
to depend on
), which is one reason why this weak-type estimate does not immediately follow from the Hardy-Littlewood inequality). Actually it is more convenient to replace this “rough” average by a smoother average
The theory of weighted estimates (and the weight class) has been further developed intensively, but we will not cover it here; see Stein’s “Harmonic analysis” for a thorough exposition of the topic.
Exercise 6 (Hardy-Littlewood maximal inequality forweights) Let
be strictly positive, and let
. Show that the following three statements are equivalent up to changes in the implied constant:
- (i) For every ball
, we have
.
- (ii) For every ball
, we have
.
- (iii) For every locally integrable function
and
, we have
(Note: Weights
with the above properties are known as
weights. Somewhat counter-intuitively, the weighted weak-type
estimate in (iii) is in fact equivalent to its strong-type counterpart, due to a certain “openness” property of
weights, but this will not be shown here.)
Exercise 7 (Calderón-Zygmund theory for power weights) Letbe a CZO. Show that for any
and
that
and
for all bounded, compactly supported
. (Note that the former can be obtained from the latter by a scaling argument.) Hint: use duality to reduce matters to establishing a bilinear weighted estimate for
, and then decompose
and
into dyadic pieces. For the “diagonal” interactions use the fact that
is bounded on
, while for the off-diagonal interactions use some form of Schur’s test.
(Note: the above theory works when
or
is replaced with a general
weight, but this is harder to prove.)
— 2. Pseudodifferential operators —
In the previous set of notes we established constructed a reasonably large class of CZOs, namely the Hörmander-Mikhlin multipliers. These multipliers suffice for the analysis of translation-invariant settings, such as studying operators arising from constant-coefficient differential operators such as the Laplacian. However they are not general enough to handle non-translation-invariant situations. For this, we must turn to the pseudodifferential operators, which generalise the variable-coefficient differential operators when the coefficients
are smooth.
To motivate these operators, we begin with the Fourier inversion formula
Remark 8 Formally, if we applyto a plane wave
, we see that
Thus we can (in principle) reconstruct the symbol
from the operator
by testing it against plane waves. This computation also shows that, formally, every linear operator is of the form
for some
; but in practice most operators will give horrible symbols
which obey no useful estimates.
The operators generalise both spatial multipliers
and Fourier multipliers
, at least when the symbol is smooth. The map from symbol
to operator
is linear. However, in contrast to the homomorphism laws
There are two main aspects to the theory of pseudodifferential operators: the analytic aspects, in which the boundedness properties of operators are analysed, and the (mainly) algebraic aspects, in which one learns how to compose, commute, or invert operators
(usually modulo lower order terms). We shall begin with the analytic aspects.
— 2.1. Symbol classes —
Of course, to get any sort of usable bound on the operators , we shall need some quantitative control on the symbol
. This requires placing
in one of the symbol classes. There are quite a variety of symbol classes to use, which determine the growth and regularity of the symbol
in both the spatial variable
and the frequency variable
. For simplicitly, we shall stick here with the standard symbol class
, which suffices for the local theory of elliptic variable coefficient operators with smooth coefficients. (These are also called the classical symbol classes, but this is confusing, given that the distinction between classical and quantum mechanics in this subject.) Each symbol class represents a decomposition of phase space into regions respecting the Heisenberg uncertainty principle, with the symbol required to be smoothly adapted to each such region and have a certain upper bound on amplitude. In this case, the relevant regions of phase space are the regions where the spatial variable
is localised to a ball of radius
, and the frequency variable
is localised to a dyadic annulus
.
Definition 9 (Standard pseudodifferential operators) A smooth symbolis a (standard) symbol of order
for some
if we have the estimates
for all multi-indices
and
(in particular,
has at most polynomial growth), and we write
in this case. We refer to the operator
as a (standard) pseudo-differential operator (or
DO for short) of order
.
Remark 10 Again, this is a quasi-definition rather than a precise definition, due to the presence of implicit constants; in any conclusion involving these symbols, the implied constants in the conclusion will almost certainly depend on the constants in this definition. In practice, one does not need (4) for all multi-indices, but only for finitely many (e.g. for all
with
), but for notational reasons we shall assume infinite regularity instead of finite. The indices
in
refer to the fact that each derivative in the
variable reduces the growth in
by
orders, whereas each derivative in the
variable reduces the growth in
by
order.
If is a variable-coefficient differential operator with the coefficients
smooth, and obeying the bounds
for all multi-indices
, then
is a pseudodifferential operator of order
. For any fixed
, a fractional differentiation operator
is a pseudodifferential operator of order
. A Littlewood-Paley operator
is a pseudodifferential operator of order
for any
(but not as
– why?), and if
then
is a pseudodifferential operator of any fixed order.
Remark 11 By copious application of Lemma 36 one can show that pseudodifferential operators of any order map Schwartz functions to Schwartz functions.
Remark 12 (Classical multiplication of symbols) Numerous applications of the Leibnitz rule reveal that ifare symbols of order
respectively, then
is a symbol of order
. Also, observe that if
is a symbol of order
, it is automatically a symbol of order
for any
. Also,
will be a symbol of order
.
The basic estimate here is
Theorem 13 (Calderón-Vaillancourt theorem) Every pseudodifferential operator of orderis a CZO. In particular, it is bounded on
with
with norm
.
Remark 14 The theorem of Calderón and Vaillancourt is in fact more general, allowing other types of symbol classes than that presented here, but we will not attempt to describe the strongest version of the theorem in these notes.
Proof: We shall allow all implied constants to depend on , and henceforth omit this dependence explicitly from our notation.
Let be a symbol of order
. We have to check two different things: firstly, that
is bounded on
with operator norm
, and secondly that
has a singular kernel.
Observe that if we smoothly truncate to be compactly supported in
, by multiplying by some cutoff function
for some large
, that
remains a symbol of order
(this is a special case of Remark 12). Because of this, and because of various limiting arguments (restricting all functions to the Schwartz class, of course), we will be able to restrict to the case where
is compactly supported in
, as long as our bounds do not depend on the size of this support.
Let us check the singular kernel property first. If are test functions with disjoint supports, we observe from Fubini’s theorem (and the compact support of
in
) that
By duality, it suffices to show that
To establish this, let us first look at the operator norm of . This is an integral operator with kernel
The case is very similar and is left to the reader. This proves the theorem.
Remark 15 It is possible to deduce a weak version of the Hörmander-Mikhlin theorem (in which thederivatives of the symbol
are assumed to be
for all
, not just
) as a consequence of the Calderón-Vaillancourt theorem and a rescaling argument. Firstly, observe that if
was a symbol of order
, then
is already a pseudodifferential operator of order
and the claim is immediate. If instead
is merely a homogeneous symbol of order
, we first truncate
smoothly in an
-neighbourhood of the origin for some small
to create a truncated symbol
. Then one checks that
is a symbol of order
and hence
is a CZO. Since the class of CZOs is scale-invariant we conclude that
is a CZO uniformly in
. The claim then follows by a limiting argument taking
.
— 2.2. The pseudodifferential calculus —
Now we start performing algebraic manipulations on these pseudodifferential operators, collectively referred to as the pseudodifferential calculus. We shall first need a technical lemma.
Lemma 16 (Oscillatory integral estimate) Letbe a compactly supported function obeying the estimates
for some
, all
, and all multiindices
. Then the function
is a symbol of order
(with implied constants depending on
).
Proof: We begin with some reductions. The main task will be to establish the zeroth order bound
By dividing by
we may take
. By translating
by
we may take
. Our task is now to show that
We shall in fact prove (5) for all compactly supported functions obeying (6). Let us first deal with the case when
is supported on the region
. For fixed
, we may apply Lemma 36 to estimate the
integral by
for any
, which by the triangle inequality is equal to
. Integrating in
we obtain the claim.
By a smooth partition of unity it remains to deal with the case when is supported on the region
. Here what we do is take a large integer
, write
We now use this lemma to establish that the (formal) adjoint of a pseudodifferential operator is also a pseudodifferential operator.
Lemma 17 (Adjoints ofDOs) Let
be a pseudodifferential operator of order
for some
. Then
has an adjoint
which is also a pseudodifferential operator of order
. In fact we have
i.e.
differs from
by a symbol of order
.
Proof: We assume that is compactly supported in frequency; this assumption can be removed by limiting arguments which we leave as an exercise to the reader (the point being that our estimates are uniform in the size of this support). Then the (
) adjoint of
is given by
Remark 18 A corollary of this lemma is that if aDO of order
has real symbol, then it is self-adjoint modulo a
DO of order
, and conversely if a
DO of order
is self-adjoint, then it has a real symbol modulo a symbol of order
. If one switches to the Weyl calculus instead of the Kohn-Nirenberg calculus then one can drop the lower order terms here, and assert simply that a
DO is self-adjoint if and only if it has real symbol.
Next, we study compositions of pseudo-differential operators.
Lemma 19 (Composition ofDOs) Let
and
be pseudodifferential operators of order
,
respectively. Then
is a pseudodifferential operator of order
, and
is a pseudodifferential operator of order
.
Proof: We assume are compactly supported; one can remove these hypotheses by a limiting argument which we omit here. For Schwartz functions
, we expand
It is instructive to establish the above lemma by hand in the case where and
are variable-coefficient differential operators of order
and
respectively. The defect between
and
can be interpreted physically as the difference between quantum and classical mechanics.
The above lemma implies in particular that the commutator
Lemma 20 (Commutator ofDOs) Let
and
be pseudodifferential operators of order
,
respectively. Then
is a pseudodifferential operator of order
.
Proof: From the proof of the previous lemma, we see that is a pseudodifferential operator with symbol
We thus see that the algebraic structure of pseudodifferential operators is intimately connected to the Poisson geometry structure of the phase plane , which in turn is induced by the symplectic geometry of that plane. There is a rich theory of how symplectic geometry (which represents Hamiltonian or “classical” mechanics) interacts with pseudodifferential operators (which represents quantum mechanics); this is the subject of semiclassical analysis and geometric quantisation. However, these topics would take us too far afield.
Exercise 21 Show that ifis a symbol of order
, then
and
, where
and
are symbols of order
. Thus the Kohn-Nirenberg and Weyl quantisations are equivalent modulo lower order operators.
— 3. Sobolev spaces —
With all the various inequalities at our disposal (the Hörmander-Mikhlin multiplier theorem, the Calderón-Vaillancourt theorem, and the Hardy-Littlewood-Sobolev inequality) we can now quickly develop some of the classical theory of homogeneous and inhomogeneous Sobolev spaces on Euclidean spaces ; of course we will not attempt an exhaustive survey of this vast subject here. In particular, we omit one important analytical aspect of these spaces, namely their local stability under smooth diffeomorphisms, which allows one to transplant these spaces onto manifolds and thus have application to various problems in differential geometry. We begin with the inhomogeneous spaces which contain fewer technicalities.
Definition 22 (Inhomogeneous Sobolev norms) Letand
. If
is a Schwartz function, we define the inhomogeneous Sobolev norm
by the formula
We let
be the closure of the Schwartz functions under this norm.
Remark 23 When, the spaces
are often denoted
(but not to be confused with the Hardy spaces
). From the Fourier inversion formula we see that
Clearly is isometric to
, so Sobolev spaces include Lebesgue spaces as special cases.
One can easily verify that Sobolev spaces are Banach spaces, and that the dual of can be identified with
. We have the following basic embeddings.
Proposition 24 (Sobolev embeddings) Letand
and
.
- (i) (Monotonicity in
) If
then
, and
- (ii) (Behaviour with respect to pseudodifferential operators) If
is a pseudodifferential operator of order
for some
, then
extends continuously from
to
:
- (iii) (Behaviour with respect to Hörmander-Mikhlin multipliers) If
is a Hörmander-Mikhlin multiplier, then
extends to a bounded linear operator on
, with
- (iv) (Characterisation using derivatives) If
for some integer
(note from (ii) that
is already well-defined as an element of
at least), then
and
In particular,
- (v) (Sobolev embedding theorem) If
is such that
, then
and
If instead
, then
is bounded and continuous with
Proof: To prove (i), observe that is certainly a Hörmander-Mikhlin multiplier (or a pseudodifferential operator of order
and hence of order
), and hence bounded on
. One can then quickly verify (7) for Schwartz functions and then extend by density.
To prove (ii), observe from the pseudodifferential calculus that is a pseudodifferential operator of order
, and hence bounded on
. The claim then follows as in (i).
To prove (iii), use the Hörmander multiplier theorem coupled with the observation that commutes with
(working in the Schwartz category to begin with).
To prove (iv), we can use Fourier multiplier calculus to write where
are Hörmander-Mikhlin multipliers, and hence
To prove (v), we may lower to assume that
. We need to show that
maps
to
in the first case, or
to
in the second case. But by Lemma 37, the convolution kernel
of this Fourier multiplier is
for any
. In particular it is
, at which point the first claim follows from the Hardy-Littlewood-Sobolev inequality; if
then the kernel lies in
, at which point we get the second claim (the continuity follows by starting with Schwartz functions and taking limits).
Next, we establish a Littlewood-Paley characterisation of Sobolev spaces.
Theorem 25 (Littlewood-Paley characterisation) Letand
. Let
be a bump function adapted to the ball
, and for each
let
be a bump function adapted to the annulus
, such that one has the pointwise estimate
. Then for any
we have
Proof: Let us just prove the upper bound. From Proposition 24(ii) we already know that maps
to
, so it suffices to show that
The lower bound can be proven in a similar manner using the full Littlewood-Paley inequality, as well as a further frequency decomposition into low frequencies and high frequencies
, and we leave this as an exercise to the reader.
Sobolev spaces can be interpolated with each other fairly easily using the complex interpolation method. We present here a sample result to give the flavour.
Proposition 26 Letbe a linear operator on Schwartz functions which is bounded on both
and
for some
and
. Then
is also bounded on
for
, where
and
.
Proof: By hypothesis, the operator is bounded on
for
. By the Hörmander-Mikhlin multiplier theorem, we also see that
is bounded on
for
whose operator norm grows at most polynomially in
. The claim then follows from the Stein interpolation theorem.
We now briefly discuss homogeneous Sobolev spaces , defined using
instead of
; these have the advantage of behaving well under dilations
, although they have some other drawbacks to compensate for this. Here one can run into some technicalities for the very low regularities (
), because the operator
is not well defined distributionally in that case. This problem can be avoided by restricting the Schwartz functions to avoid the frequency origin:
Definition 27 (Homogeneous Sobolev norms) Letand
. If
is a Schwartz function whose Fourier transform vanishes near the origin, we define the homogeneous Sobolev norm
by the formula
We let
be the closure of all such functions under this norm.
Because does not always contain all Schwartz functions, one has to sometimes take a little care with arguments. Nevertheless, one can still establish a reasonable theory for these spaces. For instance,
is an isomorphism between
and
, and we have the homogeneous Sobolev embedding
whenever
and
. For positive
we also have the relationship
, with
Exercise 28 (Endpoint Sobolev embedding) Letand
. Show that if
, then
with
In fact, slightly more is true: show that
has vanishing mean oscillation in the sense that for any
we have
Exercise 29 (Hölder-Sobolev embedding) Let,
, and
. Show that if
, then we have the Hölder estimate
for all
.
— 4. Appendix: the geometric Hahn-Banach theorem —
Here we give a geometric formulation of the Hahn-Banach theorem, sometimes also called the Dieudonné hyperplane separation theorem.
Definition 30 (Algebraic openness) A setin a (real) vector space is algebraically open if the set
is open for all
(i.e. the intersection of
with any line is an open subset of that line).
Algebraic openness is a very mild property; indeed, the topological vector space structure given by weak openness is finer than any other topological vector space structure. For instance, in a normed vector space, any open set will be algebraically open.
Theorem 31 (Geometric Hahn-Banach theorem) Letbe disjoint convex subsets of a real vector space
, with
algebraically open. Then there exists a linear functional
and
such that
on
, and
on
.
Remark 32 In finite dimensions, it is not difficult to drop the algebraic openness hypothesis onas long as one now replaces the condition
by
. However in infinite dimensions one cannot do this. Consider for instance the space
of sequences
with only finitely many
non-zero, let
consist of those sequences whose last non-zero element is strictly positive, and
consist of those sequences whose last non-zero element is strictly negative. Then there is no hyperplane separating
from
.
Remark 33 Ifis algebraically open, and
are non-empty, then
is non-zero, and it is not hard to see that the condition
can be upgraded to
. If
contains the origin, then
must be positive, and can then be rescaled to be
.
Proof: We first observe that it suffices to verify the homogeneous case, when are convex cones and
. Indeed, to then establish the general case, one applies the homogeneous case to the convex cones
defined by
Consider all the pairs of disjoint convex cones, with
algebraically open. We can order these pairs by set inclusion, so that
whenever
and
, and observe that every chain has an upper bound. By Zorn’s lemma, we thus see that to prove the claim it suffices to do so under the additional assumption that
is maximal. (This is the one and only place where we use (crucially) the axiom of choice.)
We can of course assume that neither nor
is empty. We now claim that
is the complement of
. For if not, then there exists
which does not lie in either
or
. By the maximality of
, the convex cone generated by
must intersect
at some point, say
. By dilating
if necessary we may assume that
lies on a line segment between
and some point
in
. By using the convexity and disjointness of
and
one can then deduce that for any
, the ray
is disjoint from
. Thus one can enlarge
to the convex cone generated by
and
, which is still algebraically open and now strictly larger than
(because it contains
), a contradiction. Thus
is the complement of
.
Let us call a line in monochromatic if it is entirely contained in
or entirely contained in
. Note that if a line is not monochromatic, then (because
and
are convex and partition the line, and
is algebraically open), the line splits into an open ray contained in
, and a closed ray contained in
. From this we can conclude that if a line is monochromatic, then all parallel lines must also be monochromatic, because otherwise we look at the ray in the parallel line which contains
and use convexity of both
and
to show that this ray is adjacent to a halfplane contained in
, contradicting algebraic openness. Now let
be the space of all vectors
for which there exists a monochromatic line in the direction
(including
). Then
is easily seen to be a vector space; since
are non-empty,
is a proper subspace of
. On the other hand, if
and
are not in
, some playing around with the property that
and
are convex sets partitioning
shows that the plane spanned by
and
contains a monochromatic line, and hence some non-trivial linear combination of
and
lies in
. Thus
is precisely one-dimensional. Since every line with direction in
is monochromatic,
and
also have well-defined quotients
and
on this one-dimensional subspace, which remain convex (with
still algebraically open). But then it is clear that
and
are an open and closed ray from the origin in
respectively. It is then a routine matter to construct a linear functional
(with null space
) such that
and
, and the claim follows.
To illustrate the power of this theorem, let us give a famous consequence:
Corollary 34 (Hahn-Banach theorem) Letbe a normed vector space, and let
be any subspace of
. Then any bounded linear functional
has an extension
with the same norm.
Proof: We may normalise . Then the algebraically open convex sets
and
are disjoint, with the former containing
, and so by the geometric Hahn-Banach theorem we can find
such that
on the first set and
on the second. The former fact establishes that
is bounded on
with norm at most
, and the latter implies that the null space of
contains the null space of
. Restricting to
we then quickly conclude that
when restricted to
equals
, and the claim follows.
In the converse direction, one can deduce the geometric Hahn-Banach theorem from its more familiar formulations, but this requires some work. The first key observation is that to prove the geometric Hahn-Banach theorem it suffices to do so when is the origin
, since the general case will eventuall follow from the trick of replacing the pair
by
, where
is the Minkowski difference of
and
. The remainder of the argument proceeds either by mimicking the usual proof of the Hahn-Banach theorem or by building a norm somehow out of
.
— 5. Appendix: Schwartz functions —
To facilitate computations we devise some notion regarding bump functions.
We already have defined a notion of what it means for a function to be a bump function adapted to a region such as a ball or an annulus; now we modify the notion to also cover Schwartz functions adapted to similar regions.
Definition 35 Letbe a ball and
. We say that a function
is a Schwartz function of height
adapted to the ball
if we have a representation
for some
which is boundedly Schwartz in the sense that
for all
.
Observe that one may equivalently define a Schwartz function of height adapted to
to be a function
which obeys the bounds
For Schwartz functions adapted to balls centred at the origin, one can compute their Fourier transform easily:
Lemma 36 Letbe a Schwartz function of height
adapted to the ball
. Then
and
are a Schwartz functions of height
adapted to the ball
.
Proof: We can normalise ; by rescaling we can normalise
. The claim then follows from the usual proof of the fact that the Fourier transform (or inverse Fourier transform) of a Schwartz function is Schwartz.
This allows us to take the Fourier transform of symbols:
Lemma 37 Letbe a Schwartz function which is also a symbol of order
for some
, thus
for all
and all multiindices
. Then
obeys the dual symbol estimates
for all
, all multiindices
, and all
.
Remark 38 The qualitative hypothesis thatto be Schwartz is needed to ensure that the inverse Fourier transform of
makes sense classically. It can be removed, but at the cost of interpreting
as a distribution rather than as a function (and excluding the origin
, which may now be singular).
Proof: Using smooth dyadic cutoffs we may split
Exercise 39 Establish a partial converse to Lemma 37: ifis Schwartz and
obeys (8) with
, then
is a symbol of order
.





