de Investigación AboutGrupo the Mathematical Foundation of Quantum Mechanics M. Victoria Velasco Collado Departamento de Análisis Matemático Universidad de Granada (Spain) Operator Theory and The Principles of Quantum Mechanics CIMPA-MOROCCO research school, Meknès, September 8-17, 2014 Lecture nº 2 12-09-2014 de Investigación AboutGrupo the Mathematical Foundation of Quantum Mechanics Lecture 1: About the origins of the Quantum Mechanics Lecture 2: The mathematical foundations of Quantum Mechanics. Lecture 3: About the future of Quantum Mechanics. Some problems and challenges Lecture 2: The mathematical foundations of Quantum Mechanics Hilbert spaces: orthogonality, summable families, Fourier expansion, the prototype of Hilbert space, Hilbert basis, the Riesz-Fréchet theorem, the weak topology. Operators on Hilbert spaces: The adjoint operator, the spectral equation, the spectrum, compact operators on Hilbert spaces. CIMPA-MOROCCO MEKNÈS, September 2014 Operator Theory and The Principles of Quantum Mechanics From Quantum Mechanics to Functional Analysis Definition: A normed space is a (real or complex) linear space π equipped with a norm, i. e. a function β : π β β satisfying i) π₯ = 0 β π₯ = 0 (separates points) ii) πΌπ₯ = πΌ π₯ (absolute homogeneity) iii) π₯ + π¦ β€ π₯ + π¦ triangle inequality (or subadditivity). Definition: A Banach space is a complete normed space. Definition: A Hilbert space is a inner product Banach space π». That is a Banach space π» whose norm is given by π₯ = π₯, π₯ , for every π₯ β π», where β,β is an inner product, i. e. a mapping β,β : π» × π» β K such that: i) π₯, π₯ = 0 β π₯ = 0 ii) π₯, π₯ β₯ 0 iii) πΌπ₯ + π½π¦, π§ = πΌ π₯, π§ + π½ π¦, π§ iv) π₯, π¦ = π¦, π₯ Note: If π» is a real space, then the conjugation is the identity map. Hilbert spaces Examples of Banach spaces and Hilbert spaces: a) b) c) d) e) f) g) β, β (Hilbert) Espacios de Banach βπ , βπ (Hilbert) Espacios de Hilbert Sequences spaces ππ (Hilbert si π = 2) Spaces of continuous functions πΆ,π, πSpaces of integrable functions πΏπ ,π, π- (Hilbert if π = 2) Matrices ππ×π (Hilbert) Spaces of bounded linear operators: πΏ π, π , πΏ(π») Theorem (Cauchy-Schawarz inequality): If π» is a linear space equipped with a inner product β,β then π’, π£ β€ π’, π’ v π£, π£ (π’, π£ β π») From the Cauchy-Schawarz inequality we obtain straightforwardly the Minkowsky inequality: π’ + π£, π’ + π£ β€ π’, π’ + π£, π£ π’, π£ β π» Theorem: If π» is a linear space equipped with a inner product normed space with the norm π’ = π’, π’ (π’ β π»). β,β , then π» is a Hilbert spaces An inner product space is linear space π» equipped with an inner product β,β . Let us rewrite Cauchy-Schwarz inequality in terms of the associated norm: Fact: If π» is a inner product space, then π’, π£ β€ π’ π£ (π’, π£ β π») Cauchy-Schwarz Corollary: If π» is a inner product space, then β,β is continuous. This means that if π’ = πππ π’π and π£ = πππ π£π , then π’, π£ = lim π’π , vπ . I fact, in terms of the associated norm, the inner product is given by: Polarization identities: Let π» be a inner product space over K 1 If K = β, then π’, π£ = ( π’ + π£ 2 β π’ β π£ 2 ). If K = β, then π’, π£ = 1 4 4 ( π’+π£ 2 β π’βπ£ 2 )+ π 4 ( π’ + ππ£ 2 β π’ β ππ£ Consequently, if π» is a inner product space over K (= β or β) then Parallelogram identity: π’+π£ 2 + π’βπ£ 2 = 2( π’ 2+ π£ 2). 2 ). Hilbert spaces Question: Given a normed space π», howv to know if π» is an inner product space? If this is the case, then such a norm needs to satisfy the paralelogram indetity Theorem (Parallelogram theorem): If (π», β ) is a normed space then β is given by an inner product if and only of if, for every π’, π£ β π» we have that v π’+π£ 2 + π’βπ£ 2 = 2( π’ 2+ π£ 2) (Parallelogram identity) Therefore we know, for instance that: π£ ππ is Hilbert space β π = 2. π’ βv π’ Theorem: Any inner product space may be completed to a Hilbert space. Proof (sketch): If π’ = πππ π’π and π£ = πππ π£π (for π’, π£ β π») then define π’, π£ : = πππ π’π, π£π . It happens that this not depend of the choice of the sequences. Moreover, π’ = πππ π’π = πππ π’π , π’π = π’, π’ If follows that π» is Hilbert space from the parallelogram identity. Hilbert spaces: orthogonality Law of cosines (Euclides): π2 = π 2 + π 2 β 2ππ πππ π π π π π Consequently: πππ π = π’ 2 + π£ 2 β π’βπ£ 2 2 π’ π£ = π’,π’ + π£,π£ β π’βπ£,π’βπ£ 2 π’ π£ = 2 π’,π£ 2 π’ π£ = π’ π’ , π£ π£ π’βπ£ π£ π Fact : π’ β₯ π£ βΊ πππ π=0 βΊ π’, π£ =0 π£ Definition: Let π» be an inner space. It is said that π’, π£ β π» are orthogonal vectors if π’, π£ =0. If this is the case, then we write π’ β₯ π£. Definition: Let π» be an inner space. The orthogonal set of S β π» is defined by π β₯ β π’ β π»: π’ β₯ s, βπ β π . Proposition: If π» is an inner space, and if S, W β π» then: i) 0 β π β₯ iv) π β π βΉ π β₯ β π β₯ ii) 0 β π βΉ π β© π β₯ = 0 v) π β₯ is a closed linear subspace of H iii) *0+β₯ = π»; π»β₯ = 0 vi) π β π β₯β₯ Hilbert spaces: orthogonality Definition: Let π be a linear space, and let π and π be linear subspaces. Then: π = πβ¨π βΊ π = π + π ; Y β© π = 0 . (Direct sum) Let π be a normed space such that π = πβ¨π. Let π₯ = π¦ + π§ and π₯π = π¦π + π§π in πβ¨π. We hope that : π₯π β π₯ βΊ π¦π β π₯ y π§π β π§. If this is the case, then we say that π = πβ¨π is a topological direct sum. Fact: If π is a Banach space then, π = πβ¨π is a topological direct sum βΊ the subspaces π and π are closed Theorem (best approximation): Let π» be a Hilbert space, π’ β π», and π a closed subspace of π». Then, there exists a unique π β π such that π’ β π = π π’, π . Notation: ππ π’ = π β π: π’ β π = π π’, π = best approximation from π’ to π. Orthogonal projection theorem: Let π» be Hilbert space and M a closed subspace. (Topological direct sum) Then, π» = π β¨ π β₯ . Morever if ππ : π» β π is the canonical projection then ππ = ππ . Also ππ = 1 and π’ = ππ (π’) + π’ β ππ (π’) = ππ (π’) + π’ β ππ (π’) . Remmark: Results also know as Hilbert projection theorem Hilbert spaces: orthogonality Fact: Every closed subspace of a Hilbert space admits a topological complement. This is a very important property. Indeed it characterizes the Hilbert spaces Theorem (Lindestrauss-Tzafriri, 1971): Assume that every closed subspace of a Banach space π is complemented. Then π is isomophic to a Hilbert space. To be complemented means to admit a topological complement . Hilbert spaces: orthogonality Definition: Let *ππ +πβπΌ be a family of nonzero vectors of π». We say that *ππ +πβπΌ is a orthogonal family if ππ , ππ = 0 βπ β π. If in addition ππ = 1, βπ β πΌ, then we say that *ππ +πβπΌ is a ortonormal family. Proposición: Every orthogonal family of vectors in π» is linearly independent. The GramβSchmidt process is a method for orthonormalising a set of vectors in an inner product space. It applies to a linearly independent countably infinite (or finite) family of vectors in a inner space. Proposition (GramβSchmidt process): Let *π£π +πββ be a linearly independent family of vectors in a inner space π». Then the family *π’π +πββ given by π’1 = π£1 πβ1 π’π = π£π β π=1 is a orthogonal family in π». Morever the family *ππ +πββ given by ππ = π’π π’π π’π , π£π π’π , π’π π’π is orthonormal. Hilbert spaces : summable families Definition: A family *π₯π +πβπΌ in a normed space π is summable if there exists π₯ β π such that βπ > 0 there exists a finite set π½π β πΌ, such that if J β πΌ is finite and π½π β π½ then πβπ½ π₯π β π₯ < π. If this is the case we write π₯ = πβπΌ π₯π . Proposition: A family of positive real numbers *π₯π +πβπΌ is summable if and only if the set * πβπ½ π₯π : J β πΌ ; π½ πππππ‘π+ is bounded. If this is the case then, π₯= π₯π : J β πΌ ; π½ πππππ‘π+ . π₯π = sup* πβπΌ πβπ½ Definition: A family *π₯π +πβπΌ in a normed space π satisfies the Cauchy condition if β π > 0 there exists a finite π½π β πΌ, such that if J β πΌ is finite, and if π½ β© π½π = β then, πβπ½ π₯π < π. Theorem: Let π be a Banach space and *π₯π +πβπΌ a family in π. Then, i) *π₯π +πβπΌ is summable βΊ *π₯π +πβπΌ satisfies the Cauchy condition. ii) *π₯π +πβπΌ summable βΉ π β πΌ: π₯π β 0 is countable. 1 π Remark: Apply the Cauchy condition with π = . Then we obtain π½ 1 such that πβπ½ π₯π then, < 1 π π₯π < if π½ β© π½π = β . Hence, πΌ0 β 1 , π βπ, so that π₯π = 0. π π½ 1 is a countable set. Moreover if π β πΌ β πΌ0 π Hilbert spaces : summable families Theorem: Let *ππ +πβπΌ be a orthogonal family *ππ +πβπΌ in a Hilbert space π». Then, *ππ +πβπΌ summable (in π») βΊ * ππ 2 +πβπΌ is summable en β, in whose case: πβπΌ ππ 2 = πβπΌ ππ 2 . This result addresses the problem of the summability in π» to the problem of the summability in β. On the other hand, it generalizes the Pythagoras's theorem. Remark: The canonical base of β3 is orthogonal. Moreover if π₯ β β3 is such that π₯ = (π₯1 , π₯2 , π₯3 ), that is π₯ = π=3 π=1 π₯π ππ , then π=3 π₯= π₯, ππ ππ π=1 The idea is to generalize this fact to an arbitrary π» to have βcoordinatesβ. Remark: Let *ππ +πβπΌ be a othonormal family in a Hilbert space π». Then *π(ππ )ππ +πβπΌ summable in π» βΊ * π(ππ ) 2 +πβπΌ es summable in β . If this is the case, then: πβπΌ π(ππ )ππ 2 = πβπΌ π(ππ ) 2 . Particularly: * π₯, ππ ππ +πβπΌ summable in π» βΊ * π₯, ππ 2+ πβπΌ summable β . Hilbert spaces : summable families When * π₯, ππ 2+ πβπΌ is summable in β ? When is π₯ = πβπΌ π₯, ππ ππ ? Let π½ β πΌ, such that π½ is finite. Then 2 0β€ π₯β π₯, ππ ππ = π₯β πβπ½ Therefore, πβπ½ π₯, ππ Consequently πβπΌ π₯, ππ. ππ , π₯ β πβπ½ 2 π’, ππ This means that * π₯, ππ β€ π₯ 2 2 2 β πβπ½ π₯, ππ 2 πβπ½ for every π½ β πΌ, with π½ finite. = sup* 2+ πβπΌ π₯, ππ ππ =β β β = π₯ πβπ½ π’, ππ 2: J β πΌ ; π½ πππππ‘e+ β€ π₯ 2 is summable in β and hence we have the following: Theorem (Bessel's inequality): If *ππ +πβπΌ is a orthormal family in a Hilbert space π», then for every π’ β π», the family πβπΌ π’, ππ ππ is summable. Moreover π’, ππ ππ β€ π’ πβπΌ Hilbert spaces : summable families Corollary (characterization of the maximal ortonormal families): Let *ππ +πβπΌ be an orthonormal system in a Hilbert space π». The following conditions are equivalent: i) π’ = πβπΌ π’, ππ ππ ii) π’, π£ = πβπΌ π’, ππ ππ , π£ (Parsevalβs identy) 2 iii) π’ = . πβπΌ π’, ππ iv) *ππ +πβπΌ is a maximal orthonormal maximal v) π’ β₯ ππ βπ β π’ = 0 vi) π» = πππ ππ : π β πΌ . Proof: TPO+Bessel: π’ β π» = πππ ππ : π β πΌ βΊ π’ = The other assertions are easy to prove. πβπΌ π’, ππ ππ . Fact: By Zornβs lemma there exist maximal ortonormal basis. Definition: A Hilbert basis (or an orthonormal basis) in π» is a maximal orthonormal family of vectors in π». Example: *ππ +πββ is an orthonormal family in π2 = *π’ = πΌπ πββ βΆ πΌπ 2 < β+ such that π’ β₯ ππ βπ β π’ = 0. Therefore π΅ β *ππ : π β β+ is a Hilbert basis of π2. Note that π΅ is not a Hamel basis. Inndeed the linear span of π΅ is π00 (sequences with finite support). Similarly with π2(I) where π2 = π2(β). Hilbert spaces : Fourier expansion Theorem: Every Hilbert space π» has an orthonormal basis. Moreover all the orthonormal bases of π» have have the same cardinal. Definition: The Hilbert space dimension of a Hilbert space π» is the cardinality of an orthonormal basis of π». Remark: Denote by dimπ» the algebraic dimension of π». If dimπ» < β , then π΅ = π1 ,β¦, ππ is a Hilbert basis βΊ π΅ is a Hamel basis. Indeed, π» = πππ*π1 ,β¦, ππ + = πππ*π1 ,β¦, ππ +. If dimπ» = β, and if π΅ = *ππ +πβπΌ is a Hilbert basis then we have π» = πππ*ππ : π β πΌ+ . But *ππ +πβπΌ do not need to be a spanning set of π». Therefore Hilbert basis β Hamel basis Definition: Let π» be a Hilbert space and *ππ +πβπΌ a Hilbert basis. For every π’ β π», π’ = πβπΌ π’, ππ ππ (Fourier expansion) Fact: The Fourier coefficients * π’, ππ +πβπΌ are the unique family of scalars *πΌπ +πβπΌ such that π’ = πβπΌ πΌπ ππ vector coordinates in β-dim!! Hilbert spaces : Fourier expansion Example: Let be the Lebesgue π-algebra of βπ, π and let π be the π π Lebesgue measure. Let π = if K = β, and π = if K = β . Then π 2π The family π πππ‘ : π β β€ is a Hilbert basis of πΏ2 ( βπ, π , , π)β . It is known as the trigonometric system. Therefore, for π β πΏ2 βπ, π we have π π‘ = πββ€ π (π) π πππ‘ (Fourier expansion) convergence in β 2, where 1 π π (π):= β«βπ π(π‘)π βπππ‘ ππ‘ 2π 1 Similarly, the family , cos ππ‘ , sin(ππ‘) : π β β is a Hilbert basis of 2 πΏ2 ( βπ, π , , π)β. Therefore, for π β πΏ2 βπ, π we have π π(π‘) = 0+ β π=1(ππ cos ππ‘ + ππ sen ππ‘ ) (Fourier expansion) 2 Convergence in β 2, where: 1 π π0 = β«βπ π(π‘) ππ‘ ; π 1 π ππ = β«βπ π(π‘) cos ππ‘ ππ‘ ; π 1 π ππ = β«βπ π(π‘) sen ππ‘ ππ‘ π Hilbert spaces : The prototype Fact: If π» is a Hilbert space and π΅ β *ππ : π β π+ is a Hilbert basis then, every π’ β π» has a unique expansion given by π’ = πβπΌ π’, ππ ππ and 2. π’ = πβπΌ π’, ππ Consequently, the mapping π β π, ππ πβπ° defines an isometric isomorphism from π― onto ππ(π°). Therefore, π2(πΌ) becomes the Hilbert space prototype. Example: The mapping πΏ2 βπ, π isometric isomorphism. β β π2(β€), given by π β π (π) nββ€ defines an For this reason the Hilbert spaces have βcoordinatesβ, as well as the euclidean space (which is also a Hilbert space) has its own coordinates. Theorem: Two Hilbert spaces are isometrically isomorphic, if and only if, they have the same Hilbert dimension. (that is that an orthonormal basis of one of this spaces has the same cardinality of an orthonormal basis of the other one). Hilbert spaces: the Riesz-Fréchet theorem Let π» be a Hilbert space over K. Who is π» β ? This is to determine the set of continuous linear functionals π: π» β K. For general Banach spaces, this was open problem along more of 20 years. The answer was the Hahn-Banach theorem (a cornerstone theorem of the Functional Analysis). Fréchet, Maurice 1878-1973 If dim π» = π then π» = Kπ . Fix an orthonormal basis π΅ = *π1 ,β¦, ππ +. If π: π» β K is linear then, for π’ = ππ=1 πΌπ ππ , we have π π’ = π π=1 πΌπ π(ππ ) = π’, π£ where π£ = ππ=1 π(ππ )ππ . Therefore π = β, π£ . Conversely, π = β, π£ is a continuous linear functional, obviously. Thus, π» β = β, π£ : π£ β π» . If ππππ» = β then, β, π£ : π£ β π» β π» β trivially. What F. Riesz and M. Fréchet showed (in independent papers published in Comptes Redus in 1907) is that, in fact, β, π£ : π£ β π» = π» β . Note also that if ππ£ : = β, π£ , then ππ£ = π£ . Espacios de Hilbert: El teorema de Riesz-Fréchet Riesz-Fréchet theorem (1907): Let π» be a Hilbert space and let π β π» β . Then, there exists a unique π£ β π» such that π = ππ£ = β, π£ . Moreover, π = π£ . Proof: Let π β π» β . If π = 0, then obviously π = π0 = β, 0 . If π β 0, then kerπ is a proper closed subspace of π», so that kerπ β₯ is nonzero. Therefore, by the OPT there exists π β kerπ β₯ such that π(π) β 0. Now it is easy to check that π£ = π(π) π 2 π is the vector that we look for, and the result follows. Corollary: If π» is a Hilbert space, then there exists an isometric conjugate-linear bijection between π» and π» β . Proof: Consider the map π» β β π» given by ππ£ = β, π£ β π£. Corollary: If π» is a Hilbert space, then π» β is also a Hilbert space. Proof: Define ππ’ , ππ£ β π’, π£ and check it. Corollary: Every Hilbert space is reflexive (i. e. π» = π» ββ ). Corollary: Any inner product space may be completed to a Hilbert space. Hilbert spaces: The weak topology Corollary: If π» is a Hilbert space and π is a subspace, then every bounded linear functional π: π βK admits a bounded linear extension π : π» βK with π = π . Bounded linear functionals in a Hilbert space do exist in abundance, as we have shown. Consider the smallest topology that makes that all of them are continuous. This is the so-called weak topology (the strong topology is the norm topology). Unless that dim π» < β , the weak topology is not normable. Definition: A sequence π’π , in a Hilbert π», converges weakly to π’ β π» (and we π write π’π β π’ ) whenever π’π , π£ β π’, π£ , for every π£ β π». From the continuity of the inner product we deduce the following result. Corollary: If in a Hilbert space, π’π β π π’ β π’π β π’. The converse does not hold. For instance, in π2 we have that ππ , π£ β0 = 0, π£ , π so that ππ β 0, meanwhile ππ β ππ = 2. Corollary: The (norm)-closed unit ball in a Hilbert space is weakly compact. Operators in Hilbert spaces: The adjoint operator Let π»1 and π»2 be Hilbert spaces and let π β πΏ π»1 , π»2 . As a consequence of the Riesz-Fréchet theorem, there exists a unique bounded linear operator π β : π»2 β π»1 such that ππ’, π = π’, π β π , (π’ β π»1 , π β π»2). Note that πβ π 2 = πβ π , πβ π = ππ β π , π β€ π πβ π π , Therefore πβ π Consequently, π β is continuous and β€ π π . πβ β€ π . Definition: Let π»1 and π»2 be Hilbert spaces and let π: π»1 β π»2 be a bounded linear operator. The adjoint operador of π is defined as the unique bounded linear operator π β : π»2 β π»1 such that ππ’, π = π’, π β π , (π’ β π»1 , π β π»2). Fact: π ββ = π. Indeed, if π’ β π»2 , and π β π»1, then: π β π’, π = π, π β π’ = ππ, π’ = π’, ππ = π’, π ββ π . Since π ββ π is unique, we obtain that π ββ π = ππ. Operators in Hilbert spaces: The adjoint operator Proposition: Let π»1 and π»2 be Hilbert spaces and let π, π β πΏ(π»1 , π»2 ). Then: i) π ββ = π ii) πΌπ β = πΌπ β (πΌ β K) iii) π + π β = π β + π β Moreover, if π»3 is a Hilbert space and π β πΏ(π»2 , π»3 ), then (π π)β = π β π β . Since π ββ = π and Indeed: π ββ β€ π β , we obtain that πβ = π . Proposition: Let π»1 and π»2 be Hilbert spaces and let π β πΏ(π»1 , π»2 ). Then, πβ = π = πβπ = ππ β (Gelfand-Naimark). Fact: The completeness is essential for the existence of the adjoint. Example: Let π: π00 β π00 be the operador ππ₯ = π₯ = π₯1 , π₯2 β¦ . . π₯π , β¦ β ππ₯ = * β π₯π π=1 π π1 . That is β π₯π π=1 π , 0 β¦ , 0, β¦ +. Let π¦ = *π¦π + with π¦1 β 0. Let π β π¦ = π§π . Then π β π¦ β π00. Operators in Hilbert spaces: The adjoint operator From the above result we obtain straightforwardly the following proposition. Proposition: Let π»1 and π»2 be Hilbert spaces and let π β πΏ(π»1 , π»2 ). Then, i) π β is injective βΊ π has dense range. (Ker π β = (Im π)β₯ ) ii) π is injective βΊ π β has dense range. (Ker π = (Imπ β )β₯ ) iii) π is biyective βΊ T β is bijective, in whose case (T β )β1 = (T β1 )β . Notation: πΏ(π») = πΏ(π», π»). Examples: Let π» be a Hilbert space and let B = *ππ +πβπΌ be a Hilbert basis. i) If π β πΏ π» is diagonal (i.e. πππ = ππ ππ ), then π β ππ = ππ ππ . ii) If π β πΏ π» is the orthogonal projection on a closed subspace, then π = π β . Examples: i) If, repect to the basis B, the matrix associated to π β πΏ π2 is πππ , then the matrix associated to T β is πππ = πππ . ii) If π β πΏ πΏ2 π, π is an integral operator with kernel π β πΏ2 ( π, π × π, π ), then π β is an integral operator with kernel π β π‘, π = π π , π‘ . Operators in Hilbert spaces: The adjoint operator Fact: From now on, we will consider operators from a Hilbert space into itself, that is the space πΏ π» : = πΏ π», π» of bounded linear operators π: π» β π». Recall that if π β πΏ π» , then π β is the unique operator πΏ(π») such that ππ’, π = π’, π β π , (π’, π£ β π»). Definition: π β πΏ(π») is a self-adjoint operator if π = π β . Proposición: Let S, π β πΏ π» be selft-adjoint operators. Then, i) π + π is selft-adjoint. ii) ππ selft-adjoint βΊ ππ = ππ. Proposition: Let π β πΏ π» . If π = π β , then π» = ker π β¨ πΌππ. Proposition: Let π» be a complex Hilbert space and let π β πΏ π» . Then: π = π β βΊ ππ’, π’ β β, for every π’ β π». Moreover, if π is selft-adjoint, then π = sup ππ’, π’ : π’ β€ 1 = sup*|β¨Tu, uβ©|: βuβ = 1+. Operators in Hilbert spaces: The adjoint operator Proposition: Let π» be a complex Hilbert space and let π β πΏ(π»). Then, there exist self-adjoint operators R, S β πΏ H (which are unique) such that π = π + ππ. Proof: π = π+π β 2 + πβπ β π . 2π The unique self-adjoint operators R, S β πΏ π» , such that π = π + ππ, are called the real part and the imaginary part of π, respectivelty. Definition: π β πΏ H is called normal if ππ β = π β π. π β πΏ π» self-adjoint βΉ π normal π β πΏ(π») diagonal βΉ π normal Proposition: Let π β πΏ H . Then π normal βΊ ππ’ = π β π’ π’βπ» . Proposition: Let π» be a complex Hilbert space. Then π β πΏ π» is normal if, and only if, its real and imaginary parts commute. Proposition: If π β πΏ(π») is such that π 2 = π (i.e. π is a projection), then, T normal βΊ T is a orthogonal projection βΊ π = 1. Operators in Hilbert spaces: The spectral equation Goal: To solve the spectral equation π β ππΌ π₯ = π¦. Let π: βπ β βπ be a linear operator. If π΄ is the matrix associated to π respect to the canonical basis, then we like to solve the equation π΄ β ππΌ π₯ = π¦, where π΄ is an π × π matrix, πΌ is the identity matrix, and π₯ and π¦ are the coordinates of the corresponding vectors (where π¦ is known and π₯ is not). If det π΄ β ππΌ β 0, then π΄ β ππΌ is an invertible matrix and the given equation has a unique solution given by π₯ = π΄ β ππΌ β1 π¦. If det π΄ β ππΌ = 0, then π β ππΌ is not injective, nor surjective. Therefore the system has a solution (not unique) if, and only if, π¦ β πΌπ π β ππΌ . Thus, determining the set of all π β β such that π β ππΌ is not invertible is relevant: *π β β βΆ det π΄ β ππΌ = 0+. Definition: π β πΏ π» is invertible if there exists π β πΏ π» such that ππ = π π = πΌ. If turns out that π is unique (whenever it exists). We denote π = π β1. Operators in Hilbert spaces: The spectrum Definition: Let π» be a complex Hilbert space. The spectrum of π β πΏ H is the set π π = *π β β βΆ π β ππΌ is not invertible+ If π: βπ β βπ is a linear operator, and if π΄ is its matrix respecto to the canonical basis, then π π = *π β β βΆ π β ππΌ is not invertible+ = ππ π = *π β β βΆ det π΄ β ππΌ = 0+. In fact: π invertible β π injective β π surjective Thus, ππ π = *π β β βΆ det π΄ β ππΌ = 0+ = *π β β βΆ π β ππΌ is not injective+. Definition: Let π» be a complex Hilbert space. The pointwise spectrum of π β πΏ(π») is the set given by ππ π = *π β β βΆ π β ππΌ is not injective+. The elements of ππ π are called eigenvalues of π. If π₯ β ker π β ππΌ then π₯ is an eigenvector of π associated to the eigenvalue π. Fact: Let π» be a finite-dimensional complex Hilbert space. Then, π π = ππ π , for every π β πΏ π» (i. e. the spectrum and the pointwise spectrum coincide). Operators in Hilbert spaces: The spectrum Fact: If π» is an infinite-dimensional complex Hilbert space, then ππ π β π π and these sets do not need to coincide. Example: If π: π» β π is the projection of π» over a closed subspace π, then ππ π = 0,1 . Moreover π is the invariant subspace associated to the eigenvalue 0 and πβ₯ the one associated to the eigenvalue 1. Example: The Volterra operator π: πΏ2 ,0,1- β πΏ2 0,1 is such that ππ π = β . Fact: Let π β πΏ π» . By the Banach isomorphism theorem we have that there exists π β1 β πΏ(π») such that ππ β1 = π β1 π = πΌ if, and only if, π is bijective. Definition: Let π» be a complex Hilbert space. The surjective spectrum of π β πΏ H is defined by ππ π’ π = *π β β βΆ π β ππΌ is not surjective+. Proposition: Let π» be a complex Hilbert space. For every π β πΏ H we have π π = ππ π βͺ ππ π’ π . Operators in Hilbert spaces: The spectrum Theorem: Let π» be a complex Hilbert space, and let π β πΏ H . Then, π π is a non-empty compact subset of β. Remark: If π» is a real Hilbert space, then the set *π β β βΆ π β ππΌ is not bijective+ may be empty (in whose case, no information is provided). Example: π β πΏ ββ given by π π₯ = ππ₯. In fact, (π β ππΌ)π₯ = π β π π₯ always is bijective because if π β β, then π β π β 0. Definition: Let π» be a real Hilbert space. The complexification of π» is defined as the complex Hilbert space π»β β π»β¨ππ». Moreover, the spectrum of π β πΏ(π») is defined as the spectrum of the operator πβ β πΏ(π»β) given by πβ π’ + ππ£ = π π’ + ππ π£ . Theorem: Let π» be a Hilbert space (either real or complex). If π β πΏ(π») then π π is a non-empty compact subset of β. Agreement: From now on, only complex Hilbert spaces will be considered. Operators in Hilbert spaces: The spectrum Proposition : If π β πΏ H then π π β = *Ξ» βΆ π β π π + (the conjugate set of π π ). Proof: π β ππΌ invertible β π β ππΌ π = π π β ππΌ = πΌ β π β π β ππΌ β π β π β β ππΌ = π β π β β ππΌ = πΌ β π β β ππΌ invertible. β = π β ππΌ β π β = πΌ β Corollary: If π β πΏ H is self-adjoint then π π β β. Proposition: If π β πΏ H is normal then, π β π(π) βΊ there exists π₯π with π₯π = 1 such that Note: This is equivalent to the following fact: π β β\π(π) βΊ there exists π > 0 such that π β ππΌ π₯ (π β ππΌ)π₯π β 0. β₯π π₯ , Corollary: If π β πΏ H is normal, then π = max π : π β π π . Proof: This is a consequence of the fact that if π β πΏ H is normal, then π = π π’π π₯ =1 ππ₯, π₯ (the numerical radius). π₯βπ» . Compact operators on Hilbert spaces Let (Ξ©, Ξ£, π) be a measure space and let π(π , π‘) β πΏ2(Ξ© × Ξ©, Ξ£ × Ξ£, π × π). Then, the integral operator with kernel π(π , π‘) is the operator π: πΏ2 (π) β πΏ2 (π) given by ππ₯ π = β«Ξ© π π , π‘ π₯ π‘ ππ(π‘) (π β Ξ©, π₯ β πΏ2 (π)) In Quantum Mechanics the integral equation π β ππΌ π₯ π = π¦ π , where π is the integral operator with kernel π, is essential. Definition: Let π and π be normed spaces. An operator π β πΏ X, Y is compact if, for every bounded sequence π₯π in π, the sequence ππ₯π has a convergent subsequence in π. Example: The integral operator π: πΏ2(π) β πΏ2 (π) with kernel π is compact. Proposition: Let π and π be normed spaces and let π β πΏ X, Y . Then, the following assertions are equivalent: i) π is compact, ii) For every sequence π₯π in π with π₯π = 1, the sequence ππ₯π has a convergent subsequence in π, iii) π(π΅π ) is compact in π (where π΅π denotes the closed unit ball of π). Compact operators on Hilbert spaces Example: If π β πΏ X, Y is a finite rank opertor, then π is compact. If π₯π = 1, then ππ₯π is a bounded sequence in a finite dimensional normed space π, and hence it has a convergent subsequence. Example: If π» is an infinite dimensional Hilbert space, then the identity operator πΌ: π» β π» is not compact. In fact, if ππ is an orthonormal sequence, then πΌππ does not have a convergent subsequence. Notation: Let π and π be normed spaces. Let denote πΎ(π, π)= *π β πΏ π, π : π is compact+. Proposition: πΎ X, Y is closed whenever π is a Banach space. Proposition: Let π and π be normed spaces and let π β πΎ X, Y . Then, i) π(π) is a separable subspace of π. ii) If π is a Hilbert space, and if π΅ = *ππ : π β β+ is a Hilbert basis, then π = πππππ π where ππ is the orthogonal projection on πΏππ π1 , β¦ , ππ . Compact operators on Hilbert spaces Notation: Let π» and πΎ be Hilbert spaces. Let denote πΉ π», πΎ = *π β πΏ π», πΎ : π has finite rank+. Corollary (Jordan): Let π» and πΎ be Hilbert spaces. Then, πΎ π», πΎ = πΉ(π», πΎ). Note that the integral operator π with kernel π is the limit of the sequence of finiterank operators ππ , where ππ is the integral operators with kernel ππ for a sequence ππ of simple functions with ππ β π. Corollary (Theorem of the adjoint): Let π» and πΎ Hilbert spaces. Then, π β πΎ π», πΎ βΊ π β β πΎ π», πΎ . Theorem: Let π β πΏ(π») be a compact operator and let π β 0. Then, i) dim ker π β π πΌ < β (Theorem of the kernel) (Riesz) ii) (π β ππΌ)(π») is closed (Theorem of the rank). Corollary: Let π β πΏ(π») a compact operators and let π β 0. Then, (π β ππΌ)(π») = π» βΊ ker π β π πΌ = *0+. Consequently, ππ π’ (π)\*0+ = ππ (π)\*0+ = π(π)\*0+. In finite dimension ππ π’ (π) = ππ (π) = π(π). Compact operators on Hilbert spaces Theorem (Fredholm alternative): Let π β πΏ(π») be a compact operator and let π β 0. Consider the following equations: (a) π β π πΌ π₯ = π¦ (b) π β β ππΌ π₯ = π¦ (c) π β π πΌ π₯ = 0 (d) π β β ππΌ π₯ = 0 Then either i) the equations (a) and (b) has a solution π₯ and π₯, for every π¦, π¦ β π», resp. ii) or the homogeneous system of equations (c) and (d) have a non-trivial solution. In the case (i), the solutions π₯ and π₯ are unique and depend continously on π¦ and π¦ respectively. In the case (ii), the equation (a) has a unique solution π₯ if, and only if, π¦ is orthogonal to all the solutions of (d). Similarly (b) has a unique solution π₯ if, and only if, π¦ is orthogonal to all the solutions of (a). Proof: If π β π(π) then we have (i) obviously, because π(π) = π(π β ). In fact, π₯ = (π β ππΌ)β1 π¦ meanwhile π₯ = (π β ππΌ)β1 π¦. Otherwise, π β π(π) β *0+ = ππ (π) β *0+, and hence (c) and (d) have a non-trivial solution. Moreover, (a) has a solution βΊ π¦ β π β π πΌ π» = ker π β β ππΌ Similarly (b) has a solution βΊ π¦ β π β β ππΌ π» = ker π β π πΌ β₯ . β₯ . Compact operators on Hilbert spaces Corollary: If π β πΏ(π») is a self-adjoint compact operator, then π or β π eigenvalue of π. Indeed it is known that if π is normal, then is an π = max π : π β π π . Remarks: i) Recall that if π is compact, then ππ (π)\*0+ = π(π)\*0+. ii) The restriction of a self-adjoint compact operator to an invariant subspace is a self-adjoint compact operator. Diagonalization of a selft-adjoint compact operator Let π β πΏ π» a self-adjoint compact operator (π β 0). Let π1 β ππ (π) such that π1 = π . Let π’1 with π’1 = 1 such that π β π1 πΌ π’1 = 0. Let π»1 = π» and π»2 = π’ β π»1 : π’ β₯ π’1 = π’1 β₯ . If π’ β π»2 , then π’, π’1 = 0 and hence, 0 = π1 π’, π’1 = π’, π1 π’1 = π’, ππ’1 = π β π’, π’1 = π1 π’, π’1 = ππ’, π’1 , Thus, π(π»2)= π»2, so that πβπ»2 is self-adjoint and compact. Compact operators on Hilbert spaces Repeat the process with πβπ»2 . Let π2 β ππ (πβπ»2 ) β ππ (π) such that π2 = πβπ»2 . Note that π2 β€ π1 because πβπ»2 β€ π . Let π»3 = π’ β π»2 : π’ β₯ π’2 . Then πβπ»2 is self-adjoint and compact and πβπ»2 (π»3 ) = π»3 . Note that π»2 = π’ β π»1 : π’ β₯ π’1 = π’1 β₯ π»3 = π’ β π»2 : π’ β₯ π’2 = π’1 β₯ β© π’2 β₯ = π’1 , π’2 β₯ Reiterating the process we obtain nonzero eigenvalues π1 , π2 ,β¦, ππ such that ππ β€ β― β€ π2 β€ π1 , with unital eigenvectors π’1 , π’2 ,β¦, π’π , and closed invariant subspaces π»1 , π»2,β¦, π»π , where π»π+1 = π’ β π»π : π’ β₯ π’π is such that π»π β β― β π»2 β π»1 and ππ = πβπ»π . The process stops only when πβπ»π+1 = 0. Since π»π+1 = π’1 , π’2 ,β¦, π’π β₯ and π» = π’1 , π’2 ,β¦, π’π β¨ π’1 , π’2 ,β¦, π’π β₯ (OPT) we have: if π’ β π», then π’ = ππ=1 π’, π’π π’π + π£ with π£ β π»π+1 . Thus, if πβπ»π+1 = 0 then ππ’ = (this is the case, particularly, if ππππ»<β). π π=1 ππ π’, π’π π’π Compact operators on Hilbert spaces If the process does not stop, then we obtain a sequence of eigenvalues ππ β 0. 1 1 Indeed, if ππ β π > 0 then π’π is bounded so that π( π’π )= π’π has a ππ ππ covergent subsequence which contradicts that π’π β π’π = 2. Hence, in this case, for every π β β we have that π» = π’1 , π’2 ,β¦, π’π β¨ π’1 , π’2 ,β¦, π’π β₯. π π=1 π’, π’π π’π + π£π with π£π β π’1 , π’2 ,β¦, π’π β₯ = π»π+1 , then we obtain that ππ£π = πβπ»π+1 π£π β€ πβπ»π+1 π£π = ππ+1 π£π β€ ππ+1 π’ β 0, And hence, If π’ = β ππ’ = ππ π’, π’π π’π π=1 Note that if π is a nonzero eigenvalue of π, then π β ππ : π β β . Otherwise, if π’ is an associated eigenvalue then, ππ’ = ππ’ = 0, which is impossible. Compact operators on Hilbert spaces: spectral theorem Theorem (spectral theorem for compact self-adjoint operators): Let π β πΏ(π») be a compact self-adjoint operator. Then π is diagonalizable. Indeed, one of the following assertions holds: i) There exist eignevalues π1 , π2 ,β¦, ππ and a system of associated orthonormal eigenvectors π’1 , π’2 ,β¦, π’π such that, for every π’ β π», π ππ’ = ππ π’, π’π π’π π=1 (uniform convergence over the compact subsets of π»). ii) There exists a sequence ππ of eigenvalues such that ππ β 0, and a sequence of associated orthonormal eigenvectors π’π such that, βπ’ β π», β ππ’ = ππ π’, π’π π’π π=1 (uniform convergence over the compact subsets of π»). In (i) as well as in (ii), if π is a non-zero eigenvalue, then π β π1 , π2 β¦ . Moreover, the dimension of the invariant subspace associated to π coincides with the number of times that π appears in π1 , π2 β¦ . Compact operators on Hilbert spaces: spectral theorem Corollary: π β πΏ π» is a compact self-adjoint operator βΊ π is diagonalizable, i.e. π= ππ β, π’π π’π π for a countable family of π’1 , π’2 β¦ . real numbers π1 , π2 β¦ and a orthonormal system Rearranging the above sum, fix π and denote πππ = ππ=ππ β, π’π π’π . Since the linear subspace generated by π’π , with ππ = ππ , is precisely ker π β ππ πΌ , we obtain that ππ is nothing but the projection of π» over ker π β ππ πΌ . Theorem (spectral resolution of a compact self-adjoint operator): Let π β πΏ(π») be a compact self-adjoint operator. For every eignevalue π let ππ be the orthogonal projection on ker π β ππΌ . Then the family πππ πβππ (π) is summable in the Banach space πΏ π» , and Remark: Now each Ξ» π= πππ . appears only once. πβππ(π) Moreover, for every π β ππ (π)\*0+, the corresponding projection ππ has finite rank and these projections are mutually orthogonal, i.e. if Ξ» and π are nonβequal eigenvalues, then ππ ππ = ππ ππ = 0. Compact operators on Hilbert spaces: spectral theorem Recall that π β πΏ(π») is normal β π = π + ππ where π and π are self-adjoint operators such that π π = ππ . Theorem (spectral resolution of a compact normal operator): Let π β πΏ π» be a compact normal operator. For every eigenvalue π, let ππ be the orthogonal projection on the invariant subspace ker π β ππΌ . Then, the family πππ πβππ (π) is summable in the Banach space πΏ π» , and π = πβππ (π) πππ . Moreover, for every π β ππ (π)\*0+, the corresponding projection ππ has finite rank, and for every non-equal eigenvalues Ξ», and π, we have that ππ ππ =ππ ππ =0. Corollary: If π β πΏ(π») is a compact normal operator, then π is diagonalizable. In fact, π= ππ β, π’π π’π π where π1 , π2 β¦ is the set of non-zero eigenvalues of π and orthonormal system of associated eigenvectors. π’1 , π’2 β¦ is an Compact operators on Hilbert spaces: spectral theorem Corollary: Let π β πΏ(π») be a compact normal operator. If π = π ππ β, π’π π’π then, for every π¦ β π» and π β 0 we have that: i) If π β π1 , π2 β¦ , then the equation π β ππΌ π₯ = π¦ has a unique solution for every π¦ β π». This solution is given by 1 π π₯= ( π¦,π’π πβ ππ ππ π βπ π π’π β π¦). ii) Otherwise, the equation π β ππΌ π₯ = π¦ has a solution βΊ π¦ β₯ ker π β ππΌ . In this is the case, the general solution is given by 1 π π₯= ( π¦,π’π πβ ππ ππ π βπ π π’π β π¦) + π§ (π§ β ker π β ππΌ ). Proof: If π β ππΌ π₯ = π¦, then ππ₯ = ππ₯ + π¦, so that ππ₯, π’π = ππ₯ + π¦, π’π = π π₯, π’π + π¦, π’π Since ππ₯, π’π = π ππ π₯, π’π π’π , π’π = ππ π₯, π’π we obtain that π π₯, π’π + π¦, π’π = ππ π₯, π’π , and hence (ππ β π) π₯, π’π = π¦, π’π . 1 Therefore if (ππ β π)β 0, then π₯, π’π = π¦, π’π . Consequenly 1 π 1 π ππ βπ π₯ = (ππ₯ β π¦) = ( ππ π₯, π’π π’π β π¦) = 1 π ( πβ ππ ππ π¦,π’π ππ βπ π’π β π¦).
© Copyright 2026 Paperzz