<<

Mathematics and Statistics 9(2): 179-187, 2021 http://www.hrpub.org DOI: 10.13189/ms.2021.090213

A Dirac Delta Operator

Juan Carlos Ferrando

Centro de Investigacion Operativa, Universidad Miguel Hernandez, E-03202 Elche, Spain

Received February 1, 2021; Revised March 18, 2021; Accepted April 9, 2021 Cite This Paper in the following Citation Styles (a): [1] Juan Carlos Ferrando, ”A Dirac Delta Operator,” Mathematics and Statistics, Vol.9, No.2, pp. 179-187, 2021. DOI: 10.13189/ms.2021.090213. (b): Juan Carlos Ferrando, (2021). A Dirac Delta Operator. Mathematics and Statistics, 9(2), 179-187. DOI: 10.13189/ms.2021.090213.

Copyright ©2021 by authors, all rights reserved. Authors agree that this article remains permanently open access under the terms of the Creative Commons Attribution License 4.0 International License

Abstract If T is a (densely defined) self-adjoint opera- operator λ 7→ δ (λI − T ) at T by tor acting on a complex H and I stands for Z +∞ the identity operator, we introduce the delta function opera- f (T ) = f (λ) δ (λI − T ) dλ (1.1) tor λ 7→ δ (λI − T ) at T . When T is a bounded operator, −∞ then δ (λI − T ) is an operator-valued distribution. If T is un- bounded, δ (λI − T ) is a more general object that still retains for each f ∈ C (R), i. e., for each real-valued continuous func- some properties of distributions. We provide an explicit repre- tion f (λ). Here dλ is the Lebesgue measure of R, but the sentation of δ (λI − T ) in some particular cases, derive vari- right-hand side of (1.1) is not a true integral. If T is a bounded ous operative formulas involving δ (λI − T ) and give several operator, we shall see at once that δ (λI − T ) must be regarded applications of its usage in as well as in Quan- as a vector-valued distribution, i. e., as a continuous tum Mechanics. from the space D (R) into the locally convex space L (H) of Keywords Hilbert Space, Self-adjoint Operator, Vector- the bounded linear operators (endomorphisms) on H equipped valued Distribution, Spectral Measure with the strong operator topology [10, 11], whose action on f ∈ D (R) we denote as an integral. If T is unbounded we shall see that δ (λI − T ) still retains some useful distributional-like properties. The previous equation means

Z +∞ δ (λI − T ) hy, f (T ) xi = f (λ) hy, δ (λI − T ) xi dλ (1.2) 1 The delta function −∞

The scalar delta ‘function’ λ 7→ δ (λ − a) along with its for each (x, y) ∈ D (f (T )) × H, where D (f (T )) stands for derivatives were introduced by Paul Dirac in [1], and later in the domain of the self-adjoint operator f (T ). [2, Section 15], although its definition can be traced back to Let us recall that if T is a (densely defined) self-adjoint op- Heaviside. The rigorous treatment of this object in the context erator, there is a unique spectral family {Eλ : λ ∈ R} of self- of distribution theory is due to Laurent Schwartz [6, 12]. In this adjoint operators defined on the whole of H that satisfy (i) paper we extend the definition of δ (λ − a) from real numbers Eλ ≤ Eµ and EλEµ = Eλ for λ ≤ µ, (ii) lim→0+ Eλ+x = to self-adjoint operators on a Hilbert space H. We denote by Eλx, and (iii) limλ→−∞ Eλx = 0 and limλ→∞ Eλx = x in H x ∈ H D (T ) T D (R) = lim−→ D ([−n, n]) the linear space of infinitely differen- for all . The domain of consists of those tiable complex-valued functions of compact support, equipped x ∈ H such that with the inductive limit topology. As usual in physics we shall Z +∞ 2 2 assume that the scalar product in H is anti-linear for the first |λ| d kEλxk < ∞. variable. −∞ If T is a densely defined self-adjoint operator1 on H and I In this case, the spectral theorem (cf. [8, Section 107]) and stands for the identity operator, we define the delta function the Borel-measurable functional calculus provide a self-adjoint operator f (T ) defined by 1In what follows σ (T ) will denote the spectrum of T . Recall that the residual spectrum of a self-adjoint operator T is empty, so that σ (T ) = Z +∞ σp (T ) ∪ σc (T ), where σp (T ) denotes the point spectrum (the eigenvalues) f (T ) = f (λ) dE (1.3) and σ (T ) the continuous spectrum of T . λ c −∞ 180 A Dirac Delta Operator

for each Borel-measurable function f (λ), whose domain If λ 7→ Y (λ − µ) denotes the unit step function at µ ∈ R, given by Y (λ − µ) = 0 if λ < µ and Y (λ − µ) = 1 if λ ≥ µ,  Z +∞  2 2 since Eλ = Y (λI − T ) for each λ ∈ R, formally D (f (T )) = x ∈ H : |f (λ)| d kEλxk < ∞ −∞ 0 dEλ/dλ = Y (λI − T ) . (1.7) is dense in H. Observe that if T is bounded, f (T ) need not So, from (1.6) and (1.7) we get Y 0 (λI − T ) = δ (λI − T ). be bounded. Moreover, since λ 7→ Eλ is constant on the set R \ σ (T ) of T , an open set in R, equation (1.3) tell us that Proposition 1. If T is a bounded self-adjoint operator on H f (λ) need not be defined on \ σ (T ). 1 R and f ∈ C (R), then Thanks to (1.3) the definition of δ (λI − T ) may be extended Z +∞ to Borel-measurable functions by declaring that the equation 0 0 (1.1) holds for (x, y) ∈ D (f (T )) × H and each Borel func- f (λ) δ (λI − T ) dλ = −f (T ) . −∞ tion f. But, by reasons that will become clear later, we shall restrict ourselves to those Borel functions which are continuous The same equality holds if T is unbounded but f ∈ D (R). at each point of σp (T ). Moreover, working with the real and complex parts, no difficulty arises if the function f involved If T is a self-adjoint operator and f ∈ Bp (R), then in the equation (1.1) is complex-valued (except that f (T ) is Z + ∞ Z + ∞ 2 2 no longer a self-adjoint operator whenever Imf 6= 0). Thus, |f (λ)| δ (λI − T ) dλ = |f (λ)| dEλ. unless otherwise stated, we shall assume that both in (1.1) and −∞ −∞ (1.3) the function f is complex-valued. Note that the complex where the latter equality is the definition of |f (T )|2. So, we Stieltjes measure d hEλx, yi need not be dλ-continuous. In have the following result. what follows we shall denote by Bp (R) the linear space over C consisting of all complex-valued Borel-measurable functions Proposition 2. If T is self-adjoint and f ∈ Bp (R), then of one real variable which are continuous on σp (T ). ∞ + ∞ If fn → f in D (R), the sequence {fn}n=1 is uniformly Z hf (T ) y, f (T ) xi = |f (λ)|2 hy, δ (λI − T ) xi dλ bounded and fn (x) → f (x) at each x ∈ . So, if T is R −∞ bounded on H (equivalently, self-adjoint on the whole of H) it turns out that fn (T ) → f (T ) in the strong operator topol- for every x, y ∈ D (f (T )). ogy [3, 10.2.8 Corollary]. Therefore, in this case δ (λI − T ) is an L (H)-valued distribution. Proof. We adapt a classic argument. Indeed, for every x, y ∈ As all integrals considered so far are over σ (T ), we have D (f (T )) we have Z + ∞ δ (λI − T ) = 0 ∀λ∈ / σ (T ) . (1.4) hf (T ) y, f (T ) xi = f (λ) d hf (T ) y, Eλxi . −∞ Also δ (−λI + T ) = δ (λI − T ) for all λ ∈ . On the other R E E = E µ ≤ λ hE y, xi hand, if µ ∈ σ (T ) and y is an eigenvector corresponding to Since µ λ µ whenever , and λ does not p µ the eigenvalue µ, clearly depend on , by splitting the integral we get Z + ∞ Z λ δ (λI − T ) y = δ (λ − µ) y (1.5) f (µ) d hEµy, Eλxi = f (µ) d hy, Eµxi , (1.8) −∞ −∞ for every λ ∈ R. In the particular case when Ta is the linear where clearly the first integral is hf (T ) y, Eλxi. Plugging operator defined on H by Tax = ax for a fixed a ∈ R, then Ta d hf (T ) y, Eλxi into (1.8), we are done. is a self-adjoint linear operator with σ (Ta) = σp (Ta) = {a}. In this case δ (λI − Ta) x = δ (λ − a) x for every x ∈ H, i. e., Corollary 3. Under the same conditions of the previous theo- δ (λI − Ta) = δ (λ − a) I. rem, the equality D E Since equality f (T )† y, x = hy, f (T ) xi holds for all Z + ∞ 2 2 x, y ∈ D (f (T )) and each f ∈ Bp (R), we may infer that kf (T ) xk = |f (λ)| hx, δ (λI − T ) xi dλ (1.9) −∞ hy, δ (λI − T ) xi = hδ (λI − T ) y, xi holds for every x ∈ D (f (T )). ∞ holds (in a ‘distributional’ sense) for all x, y ∈ D (T ). This Proposition 4. If T is self-adjoint and {fn}n=1 is a uniformly suggests that in certain sense δ (λI − T ) may be regarded (pos- bounded sequence in Bp (R) such that fn → f pointwise on sibly for almost all λ ∈ R) as a Hermitian operator on D (T ). R with f ∈ Bp (R), then fn (T ) x → f (T ) x for every x ∈ Let us also point out that as equation (1.1) holds for all f ∈ D (T ). D ( ), in a distributional sense we have R Proof. This is a straightforward consequence of preceding d corollary and the Lebesgue dominated convergence theo- hy, E xi = hy, δ (λI − T ) xi (1.6) dλ λ rem. Mathematics and Statistics 9(2): 179-187, 2021 181

This proposition holds in particular if fn → f in D (R). If we denote by L (H) the linear space of all linear endo- Hence, even in the unbounded case, δ (λI − T ) behaves as a morphisms on H, the next theorem summarize some previous vector-valued distribution-like object. results. Proposition 5. Let (λ, µ) 7→ g (λ, µ) be a function defined on Theorem 7. If T is a densely defined self-adjoint operator on 2 R such that g (λ, ·) ∈ L1 (R) for every λ ∈ R and g (·, µ) ∈ a Hilbert space H, there is an L (H)-valued linear map δT on Bp (R) for every µ ∈ R. If the parametric integral Bp (R), whose action on f ∈ Bp (R) we denote by Z + ∞ Z + ∞ f (λ) = g (λ, µ) dµ hδT , fi = f (λ) δ (λI − T ) dλ, −∞ −∞ is continuous on R and makes sense if we replace λ by a self- such that hδT , fi = f (T ). If {fn} ⊆ Bp (R) is uniformly adjoint operator T , the value of the integral bounded and fn (t) → f (t), with f ∈ Bp (R), for all t ∈ R Z +∞ Z +∞ then hδT , fni x → hδT , fi x for all x ∈ H. If T is bounded, g (λ, µ) δ (λI − T ) dµ dλ δT is an L (H)-valued distribution, so hδT , fi is a bounded −∞ −∞ operator on H. In addition δ (λI − T ) = 0 if λ∈ / σ (T ) and does not depend on the integration ordering. hy, δ (λI − T ) xi = hδ (λI − T ) y, xi for x, y ∈ D (T ).

Proof. Since g (·, µ) ∈ Bp (R) for every µ ∈ R., one has Z +∞ 2 Explicit form of δ (λI − T ) g (T, µ) = g (λ, µ) δ (λI − T ) dλ, −∞ If Q is a vector-valued distribution, the Fourier transform which implies of Q is defined as the vector valued distribution FQ on S (R) hFQ, fi = hQ, Ffi F −1 Z +∞ Z +∞  such that . As usual, we denote by f (T ) = g (λ, µ) δ (λI − T ) dλ dµ. the inverse Fourier transform. −∞ −∞ Theorem 8. If T is a self-adjoint operator, the identity On the other hand, by the definition of δ (λI − T ) we have 1 Z + ∞ Z +∞ Z +∞  δ (λI − T ) = eit(λI−T )dt (2.1) f (T ) = g (λ, µ) dµ δ (λI − T ) dλ, 2π −∞ −∞ −∞ holds for every λ ∈ R, and the action f (T ) of δ (λI − T ) on for (x, y) ∈ D (T ) × H. So, the proposition follows. f ∈ S (R) is given by Theorem 6. If T is a self-adjoint operator on H, then Z +∞ Z +∞  f (λ) it(λI−T ) +∞ f (T ) = e dλ dt. Z −∞ −∞ 2π f (λ) δ λI − T 2 dλ = (1.10) 0 Proof. Setting δT (λ) = δ (λI − T ) observe that Z +∞ 1 n √  √ o √ δ λI − T − δ λI + T f (λ) dλ 1 −itT (FδT )(t) = √ e . 0 2 λ 2π if λ > 0 and f ∈ B ( ), both members acting on D T 2. p R Indeed, if f ∈ S (R) we have Proof. First note that T 2 ≥ 0. Hence σ T 2 ⊆ [0, + ∞), Z + ∞ which implies that δ λI − T 2 = 0 if λ < 0. Since T 2 is a hFδT , fi = hδT , Ffi = (Ff)(λ) δ (λI − T ) dλ −∞ f ∈ B ( ) self-adjoint operator, for p R we have Z + ∞ 1 −itT Z +∞ = (Ff)(T ) = √ f (t) e dt. f (λ) δ λI − T 2 dλ = f T 2 2π −∞ 0 Consequently On the other hand, it is clear that   −1 1 −itT Z +∞ f (λ) √  Z +∞ δT = F √ e . (2.2) √ δ λI − T dλ = f µ2 δ (µI − T ) dµ 2π 0 2 λ 0 Functionally, the action of δT on f ∈ S (R) by means of equa- whereas, using that δ (−µI + T ) = δ (µI − T ), we have tion (2.2) becomes Z +∞ √  Z 0   f (λ) 2 1 −itT −1  − √ δ λI + T dλ = f µ δ (µI − T ) dµ hδT , fi = √ e , F f (t) (2.3) 0 2 λ −∞ 2π So, the right-hand side of (1.10) coincides with Consequently, we have Z +∞ Z +∞ Z +∞  2 2 f (µ) it(µI−T ) f µ δ (µI − T ) dµ = f T hδT , fi = e dµ dt −∞ −∞ −∞ 2π since µ 7→ f µ2 is a Borel function. with the order of the integration as stated. 182 A Dirac Delta Operator

Corollary 9. If e−itT x = x (t), for x ∈ D (T ) one has Example 11. The spectral family of the (up to a sign) one- dimensional momentum operator of the 1 Z +∞ free particle P = iD, where Dϕ = ϕ0, acting on the Hilbert δ (λI − T ) x = eiλtx (t) dt 2π −∞ space H = L2 (R) is given by Z +∞ iλ(s−x) and if x ∈ D (f (T )) and f ∈ S (R), then 1 1 e (Eλϕ)(x) = ϕ (x) + p.v. ϕ (s) ds 2 2πi −∞ s − x Z +∞ Z +∞  1 iλt f (T ) x = f (λ) e dλ x (t) dt. for every regular compactly supported ϕ ∈ D (P ). 2π −∞ −∞ Proof. As is well-known P is a self-adjoint operator with Remark 10. Consider the one-parameter unitary group 2,1 D (P ) = H (R) and σc (P ) = R. Since {U (t): t ∈ R} generated by the self-adjoint operator T , that is, U (t) = exp (−itT ) for every t ∈ R. If F denotes the e−itP ϕ (x) = etDϕ (x) = ϕ (x + t) Fourier transform, equation (2.2) can be written as for a regular enough ϕ ∈ D (P ), by Corollary 9 we have 1 δ (λI − T ) = √ F −1 (U)(λ) . (2.4) 2π 1 Z +∞ {δ (µI − P ) ϕ} (x) = eiµtϕ (x + t) dt. 2π So, equation (2.3) reads as −∞ Note that the integral of the right-hand side does exist because 1 Z +∞ f (T ) = √ F −1f (t) U (t) dt. (2.5) ϕ has compact support. 2π −∞ According to the definition of Eλ for the continuous spec- trum and keeping in mind the order of integration as indicated In what follows we shall compute the spectral family in Corollary 9, one has {Eλ : λ ∈ R} for some useful self-adjoint operators of Quan- Z + ∞ Z + ∞ tum Mechanics by means of the delta δ (λI − T ). Nonetheless, 1 iµt {Eλϕ} (x) = Y (λ − µ) e ϕ (x + t) dµ dt. although Eλ = Y (λI − T ), the identification 2π −∞ −∞ Z + ∞ So, since Y (λI − T ) = Y (λ − µ) δ (µI − T ) dµ −∞ 1 Z + ∞ √ Y (λ − µ) eiµtdµ = F (Y )(t) · eiλt, might be not well-defined because µ 7→ Y (λ − µ) has a jump 2π −∞ µ = λ λ ∈ σ (T ) x discontinuity at . Indeed, if p and is an bearing in mind the distributional relation eigenvector corresponding to λ, then rπ  1 1 Z + ∞ (Z λ ) F (Y )(t) = δ (t) + p.v. , (2.8) Y (λ − µ) δ (µI − T ) x dµ = δ (µ − λ) dµ x 2 iπ t −∞ −∞ we get and the right-hand integral makes no sense (see [4] for a useful 1 1 Z +∞ eiλ(s−x) discussion). If λ∈ / σp (T ) we define {Eλϕ} (x) = ϕ (x) + ϕ (s) ds 2 2πi −∞ s − x Z + ∞ where the last integral must be understood in Cauchy’s princi- Eλ = Y (λ − µ) δ (µI − T ) dµ (2.6) −∞ pal value sense.

If λ belongs to σp (T ), then (µ 7→ Y (λ − µ)) ∈/ Bp (R). In Example 12. The spectral family of the one-dimensional order to define Eλ we enlarge a little the interval of integration Quantum Mechanics kinetic term of the free particle, by considering the integral corresponding to the Laplace operator T = −D2 on H = 2 00 L2 (R), where D ϕ = ϕ , is given by Z λ+ δ (µ − λ) dµ 1 Z +∞ cos (λ (s − x)) − 1 −∞ (Eλϕ)(x) = p.v. ϕ (s) ds iπ −∞ s − x for small  > 0. So, if λ ∈ σp (T ) we define for λ > 0 and Eλ = 0 whenever λ < 0, where ϕ is a regular Z + ∞ function with compact support belonging to D (T ). Eλ = lim Y (λ +  − µ) δ (µI − T ) dµ. (2.7) →0+ −∞ Proof. In this case T is a self-adjoint operator with σ (T ) = [0, +∞). Since T = (iD)2, according to (1.10) we have The limit is well-defined since lim→0+ Eλ+ = Eλ pointwise on H. In the particular case when λ belongs to σd (T ), the 1 n √  √ o δ (λI − T ) = √ δ λI − iD − δ λI + iD discrete part of σp (T ), λ is isolated in σp (T ). 2 λ Mathematics and Statistics 9(2): 179-187, 2021 183

−1/2 −inx regarded as a functional on S (R) through dλ-integration over Z}, with ϕn (x) = (2π) e , are the solutions of the [0, +∞). Plugging eigenvalue problem iϕ0 = λϕ with ϕ (−π) = ϕ (π). So, for L2 P +∞ ϕ ∈ D (S) we have ϕ = cnϕn with 1 Z n∈Z (δ (µI ∓ iD) ϕ)(x) = eiµtϕ (x ± t) dt 2π −∞ Z π 1 inx cn = hϕ, ϕni = ϕ (x) e dx into the previous expression and keeping in mind the correct 2π −π order of integration, we see that for every n ∈ . Since σ (S) = σd (S), recalling the definition ∞ Z Z of the operator E for λ ∈ σ (S), clearly we have f (λ)(δ (λI − T ) ϕ)(x) dλ = λ d 0 √ Z + ∞ Z ∞ Z +∞ i λt 1 e (Eλϕ)(x) = lim Y (λ +  − µ)(δ (µI − S) ϕ)(x) dµ f (λ) √ [ ϕ (x + t) − ϕ (x − t)] dλ dt + →0 −∞ 4π 0 −∞ λ for every λ ∈ . So, the fact that E is a bounded operator for every f ∈ S (R). By the definition of Eλ if λ > 0 and the R λ fact that δ (µI − T ) = 0 whenever µ < 0, we have yields X E ϕ = c E ϕ Z + ∞ λ n λ n n∈ (Eλϕ)(x) = Y (λ − µ)(δ (µI − T ) ϕ)(x) dµ. Z 0 Using that δ (µI − S) e−inx = δ (µ − n) e−inx and that Working out the penultimate integral with µ instead of λ and Y (λ + 0 − n) = Y (λ − n), we get f (µ) = Y (λ − µ), we obtain √ −inx + ∞ +∞ i µt X cn −inx X e Z Z e (Eλϕ)(x) = √ Y (λ − n) e = √ . Y (λ − µ) √ [ ϕ (x + t) − ϕ (x − t)] dµ dt 2π 2π n∈ n∈ , n≤[λ] 0 −∞ µ Z Z ( √ ) Z +∞ Z λ ei µt = √ dµ [ ϕ (x + t) − ϕ (x − t)] dt Remark 14. Since in the previous example S is bounded on −∞ 0 µ H = L2 [−π, π], the delta operator δ (λI − S) should be re- √ garded as a continuous endomorphism as well. In this case for λ > 0. So, by setting u = µ we get X Z +∞ Z λ δ (λI − S) ϕ = cn δ (λ − n) ϕn. dt iut (Eλϕ)(x) = [ ϕ (x + t) − ϕ (x − t)] e du. n∈Z −∞ 2π 0 Example 15. The one-dimensional Quantum Mechanics po- Now we have sition operator on L2 (R). This operator is defined on H = λ 1 Z L2 (R) by (Qϕ)(x) = xϕ (x) for every x ∈ R. Clearly √ eiutdu = 1 − eiλt F −1 (Y )(t) , σc (Q) = and ϕ ∈ D (Q) if (x 7→ x ϕ (x)) ∈ L2 ( ). More- 2π 0 R R over, it is clear that so, using that F −1 (Y (v)) = F (1 − Y (v)) (t) as well as equation (2.8), we get {exp (it (λI − Q)) ϕ} (x) = ei(λ−x)tϕ (x) .

1 Z λ rπ  1 1 √ eiutdu = 1 − eiλt δ (t) − p.v. So we have 2π 0 2 iπ t (δ (λI − Q) ϕ)(x) = δ (λ − x) ϕ (x) . which implies Hence, in this case we can write (Eλϕ)(x) = 1 Z +∞ ϕ (s) 1 Z +∞ cos (λ (s − x)) Z +∞ − ds + ϕ (s) ds {Eλϕ} (x) = Y (λ − µ) δ (µ − x) ϕ (x) dµ πi −∞ s − x πi −∞ s − x −∞ where the integrals are understood in Cauchy’s principal value Therefore, if λ 6= x we get sense. Example 13. Spectral family of the (up to a sign) one- {Eλϕ} (x) = Y (λ − x) ϕ (x) . dimensional Quantum Mechanics momentum operator S for Example 16. Explicit form of δ (λI − M) for the Hermitian a bounded particle on H = L2 [−π, π] with domain matrix of H = C3 0 {ϕ ∈ L2 [−π, π]: ϕ ∈ L2 [−π, π] , ϕ (−π) = ϕ (π)}  0 1 1  As is well-known this is a self-adjoint operator with discrete M =  1 0 1  . spectrum σ (S) = Z whose eigenfunction system {ϕn : n ∈ 1 1 0 184 A Dirac Delta Operator

−1 Proof. In this case M = PJM P with σ (M) = {−1, 2} If T is an unbounded self-adjoint operator then D (T ) 6= H and and D (T n) becomes smaller as n grows. So, the following result, makes sense only if the operator T is bounded.  −1 0 0   1 1 1  JM =  0 −1 0  ,P =  −1 0 1  Theorem 18. In general, if T is a bounded self-adjoint opera- 0 0 2 0 −1 1 tor, one has

∞ (n) Using (2.1) we get X n δ (λ) δ (λI − T ) = (−1) T n (2.9) n!  δ (λ + 1) 0 0  n=0 δ (λI − M) = P 0 δ (λ + 1) 0 P −1   which is the Taylor series of δ (λI − T ) at λI. 0 0 δ (λ − 2) Proof. Developing the operator function exp (itT ), which is Let us compute the spectral family and the projection operator well-defined by the spectral theorem, we get onto the eigenspace ker (M + I). Clearly ∞ n   1 Z +∞ X (it) Y (λ + 1) 0 0 δ (λI − T ) = e−iλt T ndt, −1 2π n! Eλ = P  0 Y (λ + 1) 0  P −∞ n=0 0 0 Y (λ − 2) so that, formally interchanging the sum and the integral, we for every λ ∈ R. If λ1 = −1, the orthogonal projection Pλ1 may write onto ker (I + M) is ∞ n 1 X F{(it) } (λ) n  1 0 0   2 −1 −1  δ (λI − T ) = √ T . 1 1 2π n! P = P 0 1 0 P −1 = −1 2 −1 n=0 λ1 3   3   0 0 0 −1 −1 2 Using the fact that √ P = E − E = E n n (n) since λ1 λ1 λ1−0 λ1 . F{(it) } (λ) = (−1) 2π δ (λ) Example 17. Consider a compact self-adjoint operator K act- for every n ∈ , we obtain (2.9). ing on a separable Hilbert space H which does not admit the N eigenvalue zero. Let {ui : i ∈ N} be a Hilbert basis of H with its corresponding sequence of real eigenvalues {λi : i ∈ N}, 3 The resolvent operator and where |λi+1| ≤ |λi| for every i ∈ N. Let us compute the action of the operator (λI − K)−1 on any x ∈ H and the operator δ (λI − T ) δ (λI − T ). Recall that the spectrum σ (T ) of a (densely defined) self- P∞ H Proof. If x ∈ H, we can write x = i=1 hx, uii ui. Since adjoint operator on a complex Hilbert space is a closed (λI − K)−1 is a bounded operator whenever λ∈ / σ (K), we subset of C contained in R (see for instance [9, 3.2]). If have z ∈ C \ σ (T ), i. e., if z is a regular point of T , and ∞ Z + ∞ −1 −1 X 1 R (z, T ) = (zI − T ) (λI − K) x = hx, uii δ (µI − K) ui dµ −∞ λ − µ i=1 denotes the resolvent operator of T at z (see [7, Definition −1 so we obtain the classic series 8.2]), the function λ 7→ (z − λ) is continuous on σ (T ). The resolvent is well-defined over H, so it is a bounded normal ∞ −1 X 1 z ∈ \ σ (T ) R (z, T ) (λI − K) x = hx, u i u . operator. If R then is even self-adjoint. λ − µ i i From (1.1) it follows that i=1 i Z +∞ 1 For the solution of the equation (I − zK) x = y with z ∈ C R (z, T ) = δ (λI − T ) dλ we get the Schmidt series −∞ z − λ

∞ −1 X 1 which is the integral form of the resolvent of T . So, by con- x = (I − zK) y = hy, uii ui sidering the complex-valued function f (λ) = (z − λ)−1 with 1 − zµi i=1 z ∈ C \ σ (T ) and using the fact that whenever z−1 ∈/ σ (T ). On the other hand, since δ (λI − K)  1  √ acts on H as a continuous endomorphism, equation F −1 (t) = 2π ieiztY (t) λ − z ∞ X δ (λI − K) x = hx, uii δ (λ − µi) ui. then, according to (2.5), for Imz > 0 we have i=1 Z ∞ −1 izt holds for every x ∈ H. (zI − T ) = −i e U (t) dt. 0 Mathematics and Statistics 9(2): 179-187, 2021 185

From here, it follows that Since in the sense of distributions −1 1  1 1  R (z, iT ) = iR (iz, −T ) = LU (z) − → δ (λ − µ) 2πi λ − i − µ λ + i − µ if Imz > 0, where L is the Laplace transform. This is the Hille-Yosida theorem which relates the resolvent with the one- as  → 0+, we have parameter group of unitary transformations {U (t): t ∈ R} 1 Z  Z T f (λ) 2 dλ → f (λ) δ (λ − µ) dλ generated by the self-adjoint operator . π (λ − µ) + 2 If T is a bounded self-adjoint operator, γ is a closed Jordan R R + contour that encloses σ (T ) and f (z) is holomorphic inside as  → 0 . Hence, if gn is defined by the left-hand side the connected region surrounded by the path γ, the Dunford µ-parametric integral with  = 1/n, then gn → f point- integral formula asserts that wise on R. So, if f ∈ D (R) and T is bounded (hence with σ (T ) compact), as can be easily checked {g }∞ is 1 Z n n=1 f (z) R (z, T ) dz = f (T ) . a uniformly bounded sequence of continuous functions, with 2πi γ sup kg k ≤ kfk , that converges pointwise on to f. n∈N n ∞ ∞ R In [13] is pointed out that (2πi)−1 R (z, T ) can be considered Thus, by [3, 10.2.8 Corollary] one has gn (T ) → f (T ) in the as the indicatrix of a vector-valued distribution with values in strong operator topology, that is ( ) L (H). Dunford integral formula is easily obtained by using 1 Z +∞ Z +∞  the δ (λI − T ) operator since, if we apply the Proposition 5 f (λ) 2 dλ δ (µI − T ) dµ −1 π (λ − µ) + 2 with g (λ, µ) = f (z (µ)) (z (µ) − λ) , where z (µ) = γ (µ) −∞ −∞ and 0 ≤ µ ≤ 1, then Z +∞ → f (µ) δ (µI − T ) dµ Z Z +∞ Z f (z)  −∞ f (z) R (z, T ) dz = dz δ (λI − T ) dλ + γ −∞ γ z − λ as  → 0 in the strong operator topology of L (H). Therefore, Z +∞ if T is bounded and f ∈ D (T ) then = 2πi f (λ) δ (λI − T ) dλ = 2πif (T ) . Z +∞ f (λ) n o −∞ ((λ − i) I − T )−1 − ((λ + i) I − T )−1 dλ Example 19. Derivation of the orthogonal projection operator −∞ 2πi onto ker (M + I) of the Hermitian matrix M of the Example goes to f (T ) as  → 0+. This proves that for bounded T 16 by the resolvent technique. We must compute 1 −1 −1 1 Z lim ((λ − i) I − T ) − ((λ + i) I − T ) P = R (z, M) dz. →0+ 2πi λ1 2πi |z+1|=1 coincides with δ (λI − T ) as an L (H)-valued distribution. Clearly, we have  z − 1 1 1  1 5 Unitary equivalence of δ (λI − T ) R (z, M) = 1 z − 1 1 . z2 − z − 2   1 1 z − 1 Theorem 20. If T is a self-adjoint operator defined on the Using that whole of H, there exist a finite measure µ on the Borel sets of the compact space σ (T ) and a linear isometry U from Z {1, z − 1}  2πi 4πi dz = − , L2 (σ (T ) , µ) onto H such that |z+1|=1 (z + 1) (z − 2) 3 3 U −1δ (λI − T ) U = δ (λI − Q) we reproduce the result we got earlier. where (Qϕ)(x) = x ϕ (x) is the position operator. 4 The δ (λI − T ) operator as a limit Proof. According to [5] there exist a finite measure µ on the Borel sets of the compact space σ (T ) and a linear isometry U −1 from L (σ (T ) , µ) onto H such that As µ 7→ (λ ± i − µ) is continuous, for self-adjoint T 2 U −1TU ϕ = Qϕ ((λ − i) I − T )−1 − ((λ + i) I − T )−1 −1 Z +∞  1 1  for every ϕ ∈ L2 (σ (T ) , µ). So, since U TU is a self- = − δ (µI − T ) dµ. adjoint operator on L2 (σ (T ) , µ), we have −∞ λ − i − µ λ + i − µ U −1δ (λI − T ) U = δ λI − U −1TU = δ (λI − Q) If f ∈ D (R), Proposition 5 yields as stated. Z +∞ f (λ) n o ((λ − i) I − T )−1 − ((λ + i) I − T )−1 dλ −∞ 2πi Remark 21. For such linear isometry U the equation ( ) 1 Z +∞ Z +∞  U −1δ (λI − T ) U ϕ (x) = δ (λ − x) ϕ (x) = f (λ) dλ δ (µI − T ) dµ. 2 2 π −∞ −∞ (λ − µ) +  holds for every ϕ ∈ L2 (σ (T ) , µ). 186 A Dirac Delta Operator

6 Commutation relations and integrating by parts, it follows that Z +∞ Let S and T be two self-adjoint operators defined on the λ [δ (λI − S) , δ (µI − T )] dλ whole of H for which equations [S, [S, T ]] = [T, [S, T ]] = 0 −∞ hold. In this case i Z +∞ = − [S, T ] seis(µI−T )ds. 2π [−itS, [−itS, −isT ]] = it2s [S, [S, T ]] = 0 −∞ Observe that a second application of equation (6.1) and a sec- and the Baker-Campbell-Hausdorff formula yields ond integration by parts yield δ (λI − S) δ (µI − T ) = Z +∞ λµ [δ (λI − S) , δ (µI − T )] dλ dµ = 1 ZZ ei(tλ+sµ)e−itSe−isT dt ds = −∞ 2 +∞ +∞ (2π) 2 i Z Z  R − [S, T ] se−isT µeisµdµ ds = 1 ZZ −st 2π i(tλ+sµ) 2 [S,T ] −i(tS+sT ) −∞ −∞ 2 e e e dt ds. (2π) 2 Z +∞ R [S, T ] δ (s) {1 − isT } e−isT ds = [S, T ] Likewise, since [T, [T,S]] = [S, [T,S]] = 0 one has −∞ as expected. δ (µI − T ) δ (λI − S) =

1 ZZ −st i(tλ+sµ) 2 [T,S] −i(tT +sS) 2 e e e dt ds 7 A remark on the Stone formula 2 (2π) R 1 ZZ st i(tλ+sµ) 2 [S,T ] −i(tS+sT ) Let T be a self-adjoint operator densely defined on a Hilbert = 2 e e e dt ds. 2 (2π) R space H. If A is a Borel set in σ (T ), defining So, using that Z +∞ E (A) := χA (λ) dEλ (7.1) −∞ ist   ist  ist  exp i [S, T ] −exp − i [S, T ] = 2i sin [S, T ] 2 2 2 where χA stands for the characteristic function of A (which is a bounded Borel function), then E is an L (H)-valued finitely we have additive and pointwise countably additive measure (i.e., count- able additivity under the strong operator topology of L (H)) on [δ (λI − S) , δ (µI − T )] = the σ-algebra A of Borel subsets of σ (T ). So, if the charac- i ZZ ist  teristic function χA of A with respect to R is continuous on sin [S, T ] ei(tλ+sµ)e−i(tS+sT ) dt ds. 2 σ (T ) then 2π 2 2 p R Z +∞ For position Q and momentum P of a one-dimensional par- E (A) = χA (λ) δ (λI − T ) dλ. ticle, one has H = L2 (R) and [Q, P ] = i~I. Therefore −∞ [Q, [Q, P ]] = [P, [Q, P ]] = 0 and For −∞ < a < b < ∞ and  > 0, we have

[δ (λI − Q) , δ (µI − P )] = Z b Z +∞  1 1   i ZZ  st  − δ (µI − T ) dµ dλ i(tλ+sµ) −i(tQ+sP ) a −∞ λ − i − µ λ + i − µ 2 sin − ~ e e dt ds. 2π 2 2 ( ) R Z +∞ Z b 2i dλ = δ (µI − T ) dµ = 2 2 According to Theorem 8, if [δ (λI − S) , δ (µI − T )] acts on −∞ a (λ − µ) +  f (λ) = λ, formally we have Z  b − µ a − µ 2i arg tan − arg tan δ (µI − T ) dµ.   Z +∞ R λ [δ (λI − S) , δ (µI − T )] dλ = + −∞ If the limit as  → 0 the bracketed function is equal to 0 if i ZZ ist  µ ∈ R \ [a, b], equal to π if a < µ < b and equal to π/2 if sin [S, T ] 2 µ ∈ {a, b}. So, if a, b∈ / σp (T ) so that χ and χ both 2π 2 2 (a,b) [a,b] R B ( ) Z +∞  belong to p R , setting λeitλ dλ eisµe−i(tS+sT ) dt ds. Z b −1 −∞ 1 2in dλ gn (µ) := π (λ − µ)2 + n−2 So, using the distributional equality a for each n ∈ and Z +∞ N itλ 2π 0 λe dλ = δ (t) (6.1) f (µ) := χ (µ) + χ (µ) , −∞ i (a,b) [a,b] Mathematics and Statistics 9(2): 179-187, 2021 187 then g (µ) → f (µ) for every µ ∈ and sup kg k ≤ 1 DOI: 10.1119/1.19283 n R n∈N n ∞ which, according to Proposition 4, implies that gn (T ) x → f (T ) x for every x ∈ D (T ). In other words [5] Halmos, P., What does the spectral theorem say? Amer. Math. Monthly 70, 241–247, (1963). Z b 1  −1 −1 lim (λ − i − T ) − (λ + i − T ) dλ DOI: 10.1080/00029890.1963.11990075 →0+ 2πi a Z +∞ 1  = χ(a,b) + χ[a,b] δ (µI − T ) dµ, [6] Horvath,´ J., Topological Vector Spaces and Distributions, 2 −∞ Vol. I, Addison-Wesley, Reading, Massachusetts, 1966. holds pointwise on the domain D (T ) of T . Hence, by virtue of (7.1) we get [7] Moretti, V., Spectral Theory and Quantum Mechanics, 2nd Edition, Unitext 110, Springer, 2013. 1 Z b   lim (λ − i − T )−1 − (λ + i − T )−1 dλ = + [8] Riesz, F., Nagy, B. Sz.-, Functional Analysis, Dover Pub- →0 2πi a 1 1 1 1 lications, Mineola, 1990. E ((a, b)) + E ([a, b]) = E (a, b) + E (a) + E (b) 2 2 2 2 [9] Schmudgen,¨ K., Unbounded Self-adjoint Operators on which is Stone’s formula. Hilbert Space, Springer, Dordrecht, 2012.

[10] Schwartz, L., Theorie´ des distributions a` valeurs vecto- REFERENCES rielles. I, Anal. Inst. Fourier, 1–141, 7 (1957). DOI : 10.5802/aif.68 [1] Dirac, P. A. M., The Physical Interpretation of the Quan- tum Dynamics, Proc. Royal Soc. London 113, 621–641, [11] Schwartz, L., Theorie´ des distributions a` valeurs vecto- (1927). rielles. II, Anal. Inst. Fourier 8, 1–209, (1958). DOI: 10.1098/rspa.1927.0012 DOI: 10.5802/aif.77 [2] Dirac, P. A. M., The Principles of Quantum Mechanics, [12] Schwartz, L., Theorie´ des Distributions, Hermann, Paris, 4th Edition, Clarendon Press, Oxford, 1988. 1966.

[3] Dunford, N. and Schwartz, J. T., Linear Operators. Part [13] Tillmann, H. G., Vector-valued distributions and the II: Spectral Theory, John Wiley, New York, 1988. spectral theorem for self-adjoint operators in Hilbert space [4] Griffiths, D. and Walborn, S., Dirac deltas and discontin- , Bull. Amer. Math. Soc. 69, 67–71, (1963). uous functions, Am. J. Phys. 67, 446–446, (1999). DOI: 10.1090/S0002-9904-1963-10862-9