Factor Ordering and Path Integral Measure for Quantum Gravity in (1+1) Dimensions

We develop a mathematically rigorous path integral representation of the time evolution operator for a model of (1+1) quantum gravity that incorporates factor ordering ambiguity. In obtaining a suitable integral kernel for the time-evolution operator, one requires that the corresponding Hamiltonian...

Ausführliche Beschreibung

Gespeichert in:

Bibliographische Detailangaben
Datum:	2017
Hauptverfasser:	Haga, J., Maitra, R.L.
Format:	Artikel
Sprache:	English
Veröffentlicht:	Інститут математики НАН України 2017
Schriftenreihe:	Symmetry, Integrability and Geometry: Methods and Applications
Online Zugang:	http://dspace.nbuv.gov.ua/handle/123456789/148635
Tags:	Tag hinzufügen Keine Tags, Fügen Sie den ersten Tag hinzu!
Назва журналу:	Digital Library of Periodicals of National Academy of Sciences of Ukraine
Zitieren:	Factor Ordering and Path Integral Measure for Quantum Gravity in (1+1) Dimensions / J. Haga, R.L. Maitra // Symmetry, Integrability and Geometry: Methods and Applications. — 2017. — Т. 13. — Бібліогр.: 35 назв. — англ.

Institution

Digital Library of Periodicals of National Academy of Sciences of Ukraine

id	irk-123456789-148635
record_format	dspace
spelling	irk-123456789-1486352019-02-19T01:31:27Z Factor Ordering and Path Integral Measure for Quantum Gravity in (1+1) Dimensions Haga, J. Maitra, R.L. We develop a mathematically rigorous path integral representation of the time evolution operator for a model of (1+1) quantum gravity that incorporates factor ordering ambiguity. In obtaining a suitable integral kernel for the time-evolution operator, one requires that the corresponding Hamiltonian is self-adjoint; this issue is subtle for a particular category of factor orderings. We identify and parametrize a complete set of self-adjoint extensions and provide a canonical description of these extensions in terms of boundary conditions. Moreover, we use Trotter-type product formulae to construct path-integral representations of time evolution. 2017 Article Factor Ordering and Path Integral Measure for Quantum Gravity in (1+1) Dimensions / J. Haga, R.L. Maitra // Symmetry, Integrability and Geometry: Methods and Applications. — 2017. — Т. 13. — Бібліогр.: 35 назв. — англ. 1815-0659 2010 Mathematics Subject Classification: 81V17; 81S40; 83C80 DOI:10.3842/SIGMA.2017.039 http://dspace.nbuv.gov.ua/handle/123456789/148635 en Symmetry, Integrability and Geometry: Methods and Applications Інститут математики НАН України
institution	Digital Library of Periodicals of National Academy of Sciences of Ukraine
collection	DSpace DC
language	English
description	We develop a mathematically rigorous path integral representation of the time evolution operator for a model of (1+1) quantum gravity that incorporates factor ordering ambiguity. In obtaining a suitable integral kernel for the time-evolution operator, one requires that the corresponding Hamiltonian is self-adjoint; this issue is subtle for a particular category of factor orderings. We identify and parametrize a complete set of self-adjoint extensions and provide a canonical description of these extensions in terms of boundary conditions. Moreover, we use Trotter-type product formulae to construct path-integral representations of time evolution.
format	Article
author	Haga, J. Maitra, R.L.
spellingShingle	Haga, J. Maitra, R.L. Factor Ordering and Path Integral Measure for Quantum Gravity in (1+1) Dimensions Symmetry, Integrability and Geometry: Methods and Applications
author_facet	Haga, J. Maitra, R.L.
author_sort	Haga, J.
title	Factor Ordering and Path Integral Measure for Quantum Gravity in (1+1) Dimensions
title_short	Factor Ordering and Path Integral Measure for Quantum Gravity in (1+1) Dimensions
title_full	Factor Ordering and Path Integral Measure for Quantum Gravity in (1+1) Dimensions
title_fullStr	Factor Ordering and Path Integral Measure for Quantum Gravity in (1+1) Dimensions
title_full_unstemmed	Factor Ordering and Path Integral Measure for Quantum Gravity in (1+1) Dimensions
title_sort	factor ordering and path integral measure for quantum gravity in (1+1) dimensions
publisher	Інститут математики НАН України
publishDate	2017
url	http://dspace.nbuv.gov.ua/handle/123456789/148635
citation_txt	Factor Ordering and Path Integral Measure for Quantum Gravity in (1+1) Dimensions / J. Haga, R.L. Maitra // Symmetry, Integrability and Geometry: Methods and Applications. — 2017. — Т. 13. — Бібліогр.: 35 назв. — англ.
series	Symmetry, Integrability and Geometry: Methods and Applications
work_keys_str_mv	AT hagaj factororderingandpathintegralmeasureforquantumgravityin11dimensions AT maitrarl factororderingandpathintegralmeasureforquantumgravityin11dimensions
first_indexed	2025-07-12T19:50:23Z
last_indexed	2025-07-12T19:50:23Z
_version_	1837471975548649472
fulltext	Symmetry, Integrability and Geometry: Methods and Applications SIGMA 13 (2017), 039, 19 pages Factor Ordering and Path Integral Measure for Quantum Gravity in (1+1) Dimensions John HAGA and Rachel Lash MAITRA Department of Applied Mathematics, Wentworth Institute of Technology, 550 Huntington Ave., Boston MA 02115, USA E-mail: hagaj@wit.edu, maitrar@wit.edu Received December 19, 2016, in final form June 01, 2017; Published online June 07, 2017 https://doi.org/10.3842/SIGMA.2017.039 Abstract. We develop a mathematically rigorous path integral representation of the time evolution operator for a model of (1+1) quantum gravity that incorporates factor ordering ambiguity. In obtaining a suitable integral kernel for the time-evolution operator, one re- quires that the corresponding Hamiltonian is self-adjoint; this issue is subtle for a particular category of factor orderings. We identify and parametrize a complete set of self-adjoint extensions and provide a canonical description of these extensions in terms of boundary conditions. Moreover, we use Trotter-type product formulae to construct path-integral rep- resentations of time evolution. Key words: factor ordering in quantum gravity; path integrals in quantum gravity; singu- larity avoidance in quantum gravity; quantization on a half-line 2010 Mathematics Subject Classification: 81V17; 81S40; 83C80 1 Introduction In general relativity, the configuration space for the gravitational field presents two formidable challenges to quantization: first, it is curved, and second, it has a boundary. In terms of Arnowitt–Deser–Misner (ADM) metric variables, the curvature of the configuration space of 3-metrics hab on a spacelike slice is evident in the DeWitt supermetric Gabcd = 1 2 √ h (hachbd + hadhbc − habhcd), (1.1) which appears in the kinetic term of the Hamiltonian constraint H = Gabcdπ abπcd − √ h ( (3)R− 2Λ ) (1.2) generating time-reparametrization invariance. Here πab = √ h(habK − Kab) is the momentum canonically conjugate to hab, defined in terms of the extrinsic curvature Kab of the spacelike slice. The potential term of the Hamiltonian constraint (1.2) involves the scalar curvature (3)R of hab and a cosmological constant Λ. The difficulty for quantization engendered by the DeWitt supermetric lies in the ambiguity as to how noncommuting operators π̂ab and ĥab are to be ordered in the canonically quantized Hamiltonian constraint, i.e., the Wheeler–DeWitt equation ĤΨ = 0. The second problem for quantization, that of the boundary in configuration space, results from the requirement that spatial 3-metrics be positive definite. In general, configuration spaces with boundary require the imposition of boundary conditions on quantum states in order for the Hamiltonian operator to be self-adjoint. In the case of quantum gravity, such boundary conditions take on added interest because of their significance to the quantum behavior of the universe close to the Big Bang/Big Crunch singularity. mailto:hagaj@wit.edu mailto:maitrar@wit.edu https://doi.org/10.3842/SIGMA.2017.039 2 J. Haga and R.L. Maitra The ambiguities of operator ordering and boundary condition are not special to canoni- cal quantization, but manifest themselves in a path integral approach through the choice in definition of a path integral measure and contour of paths to integrate over. The link is most readily seen by considering a (formal) path integral expression for a wave functional Ψ[hab] = N ∫ C e−IE [gαβ ]D[gαβ] (1.3) as in the Hartle–Hawking wave function of the universe [16], where N represents a normalization constant, IE [g] is the Euclidean signature action evaluated on a trajectory through the super- space of metric configurations, and C represents a suitable contour of 4-metrics gαβ bounded by the 3-metric hab at time t = 0. Since the wave functional (1.3) must satisfy a Wheeler–DeWitt equation for given factor ordering together with a boundary condition on superspace, and since its only variability is in the definition of the path integral measure D[gαβ] and contour C, these choices must be made jointly. Various proposals exist to motivate the choice of particular orderings or measures on physical grounds, primarily supported by the rationale that classically reparametrization-invariant theo- ries should retain coordinate invariance upon quantization [2, 4, 15, 23]. Reasoning along these lines has led to the arguments in favor of a Laplace–Beltrami ordering in which the kinetic term of the Hamiltonian is promoted to a covariant Laplace–Beltrami operator on configuration space. More generally, a multiple of the Ricci curvature of configuration space may be added without breaking reparametrization invariance; such an ordering is known as a conformal ordering. An important means of illuminating the question of factor ordering is to investigate the explicit dependence of the path integral measure on the choice of ordering. The subtlety of this correspondence stems in part from the fact that rigorous definition of a path integral measure for a quantum field theory is intrinsically difficult, since one integrates over paths through a space of field configurations. In addition, two complications arise in general relativity: (1) the DeWitt supermetric (1.1) has signature (−+++++) (so that one mode always has an opposite sign in the kinetic term), and (2) the configuration space of 3-metrics gab is restricted by the requirement of positive definiteness. In this paper, we examine the correspondence between factor orderings and path integral measures in the model problem of (1+1) gravity. As shown by Nakayama in [24], a reduction from field theory to a mechanical system is possible through gauge fixing, with no imposi- tion of symmetries. This model facilitates a direct consideration of the boundary problem in configuration space, absent the indefinite signature in the kinetic term. We introduce a 1-parameter family of factor orderings and demonstrate that this is equivalent to the addition of a singular quantum potential in the Hamiltonian. The mathematically rigorous construction of the Feynman path integral involving a singular potential was formulated by Nelson in [25]. Orderings which yield a potential well or sufficiently shallow barrier necessitate a boundary condition on quantum states to determine a time evolution operator; a sufficiently steep positive quantum potential obviates the need for any boundary condition on wave functions (resulting in a unique time evolution operator). Orderings yielding a positive barrier, in concert with a Dirichlet boundary condition on quantum states, permit application of Trotter-type product formulae to express the time evo- lution operator as a path integral. In the critical case between a potential well and a barrier, path integral representations of time evolution under each possible boundary condition may be constructed for those orderings which yield no added quantum potential. In particular, the Laplace–Beltrami ordering (which here coincides with the conformal ordering) belongs to this class. For orderings yielding a well near the singularity, the product formulae under con- sideration do not apply, and the question of path integral representation remains for future investigation. Factor Ordering and Path Integral Measure for Quantum Gravity in (1+1) Dimensions 3 We state some supporting mathematical results in Section 2. In Section 3 a brief exposition of the model for (1+1) gravity is presented. We canonically quantize the model and introduce factor orderings in Section 4. Section 5 contains a rigorous examination of the domains of quantum states corresponding to each factor ordering in the 1-parameter family, and we derive path-integral representations of time evolution in Section 6. 2 Mathematical background In this section we state some fundamental results used in subsequent sections. We take H to be a separable Hilbert space with inner product 〈·, ·〉. For detailed development of the functional calculus for self-adjoint operators on H , we direct the reader to [29, 30]; however we state some important definitions and theorems for completeness. Theorem 2.1 (von Neumann). Let A be a symmetric operator on a Hilbert space H . Then the following are equivalent: (a) A is essentially self-adjoint. (b) Ker(A∗ ± i) = {0}. (c) Ran(A± i) is dense in H . Every symmetric operator A has a closed extension, as A∗ is always closed for densely defined operators. However there may be multiple closed extensions of A as we note below. Definition 2.2. Suppose that A is a symmetric operator. Let K± = Ker(i∓A∗) = Ran(±i+A)⊥. K+ and K− are the deficiency subspaces of A, and the pair of numbers n+ and n− given by n+(A) = dim(K−) and n−(A) = dim(K+) are the deficiency indices of A. As famously proven by von Neumann, the self-adjoint extensions of a closed symmetric opera- tor bijectively correspond to partial isometries between the deficiency subspaces. In particular, considering operators of the form Ĥ = −~2 d2 dx2 +V (x), we make use of a theorem due to H. Weyl. Definition 2.3. The potential V (x) is in the limit circle case at infinity (resp. at zero) if for some λ, all solutions of −ϕ′′(x) + V (x)ϕ(x) = λϕ(x) are square integrable at infinity (resp. at zero). If V (x) is not in the limit circle case at infinity (resp. at zero), it is said to be in the limit point case. Theorem 2.4 (Weyl’s limit point-limit circle criterion). Let V (x) be a continuous real-valued function on (0,∞). Then Ĥ = −~2 d2 dx2 + V (x) is essentially self-adjoint on C∞0 (0,∞) if and only if V (x) is in the limit point case at both zero and infinity. We seek to characterize the self-adjoint extensions of the Hamiltonian in terms of explicit boundary conditions. The following result appears in [13]: Theorem 2.5. If A is a symmetric operator on H with deficiency indices (n, n), n ≤ ∞, if there is an n-dimensional Hilbert space Hb (the “boundary space”), and if there exist two linear maps Γ1,Γ2 : D(A∗)→Hb such that for all f1, f2 ∈ D(A∗), 〈A∗f1, f2〉 − 〈f1, A ∗f2〉 = 〈Γ1f1,Γ2f2〉b − 〈Γ2f1,Γ1f2〉b, 4 J. Haga and R.L. Maitra and for any two Ψ1,Ψ2 ∈Hb there exists g ∈ D(A∗) satisfying Γig = Ψi, i = 1, 2, then there is a one-to-one correspondence between unitary maps U ∈ U (Hb) and the self-adjoint extensions of A, where a map U describes the domain of the corresponding extension AU as D(AU ) = { ψ ∈ D(A∗) ∣∣ (U − 1b)Γ1ψ + i(U + 1b)Γ2ψ = 0 } . The existence of the boundary space for symmetric A is proven in [13]. We apply this theory to the Hamiltonians arising from various factor orderings in Section 5. 3 Action for (1+1)-dimensional gravity Gravity in one space and one time dimension has been considered both as a model problem for full (3+1) geometric theories of gravity and for its own importance to membrane theory. Since Einstein gravity in (1+1) has trivial dynamics, one must devise an alternative formulation for geometric gravity. The model for (1+1) gravity considered here originates in membrane theory (see the seminal paper [28] by Polyakov), where gravity is introduced through an area-minimizing action which in the conformal gauge gab = eφĝab (with respect to the reference metric ĝab) yields the Liouville action SL[φ, ĝ] = ∫ D [ 1 4 ĝab∂aφ∂bφ+ 1 2 φRĝ + 4Λeφ ] d2x. Subsequent work by Verlinde [35] allows this action to be written in an arbitrary gauge. By working in the (Euclidean signature) proper-time gauge g00 = 1, g01 = g10 = 0, g11 = γ(x0, x1), Nakayama [24] obtains a reduced quantum mechanical form of Verlinde’s action for the (1+1) spacetime manifold R × S1 parametrized by (x0, x1), 0 ≤ x1 ≤ π. The Euler– Lagrange equations are solved to yield a factorization of the space-space component of the metric γ(x0, x1) = γ0(x0)γ1(x1). Together with the condition∫ π 0 √ γ1(x1)dx1 = (m+ 1)π, m = 0, 1, 2, . . . , this implies that the normalized spatial arc length reduces to l(x0) = 1 π ∫ π 0 √ γ(x0, x1)dx1 = (m+ 1) √ γ0(x0). Consequently, the Hamiltonian becomes Hred ∝ 1 4l(x0) ( dl dx0 )2 − Λl (still in the Euclidean signature), meaning that in Lorentzian signature one may adopt the Lagrangian action S[l] = ∫ T 0 [ 1 4l l̇2 − Λl ] dx0 with corresponding Hamiltonian H = lΠ2 l + Λl, (3.1) Factor Ordering and Path Integral Measure for Quantum Gravity in (1+1) Dimensions 5 where the momentum conjugate to l is given by Πl = l̇ 2l . We note that since l is invariant under reparametrizations of the coordinate x1, the reduced action S and the subsequent constructions below are naturally preserved by spatial re-coordinatization. Note that in [24], a term proportional to l−1 is included in the reduced Hamiltonian to account for Casimir energy. Since factor ordering of the kinetic term of the quantized Hamiltonian will introduce such a term naturally (see Section 4), we regard (3.1) as the classical Hamiltonian. 4 Schrödinger quantization A prototype for sampling the effects of factor ordering ambiguity has been to consider a family of orderings of the form (e.g., [1, 16, 22, 32]) Ĥ(j1,j2,j3) = −~2lj1 d dl lj2 d dl lj3 + Λl, j1 + j2 + j3 = 1, (4.1) where the quantized circumference operator l̂ : Ψ → lΨ and its conjugate momentum operator Π̂l : Ψ→ −i~dΨ dl are variously composed. Denoting J± = j3 ± j1, (4.1) is formally self-adjoint with respect to the measure lJ−dl on R+ = (0,∞). Transforming states and observables according to L2 ( R+, µ(l)dl ) → L2 ( R+, ν(l)dl ) , Ψ 7→ Ψ̃ = [ µ(l) ν(l) ]1/2 Ψ, Â 7→ Ã = [ µ(l) ν(l) ]1/2 A [ µ(l) ν(l) ]−1/2 , (4.2) with µ(l) = lJ− and ν(l) = l−1/2, (4.1) maps to an operator H̃ which is formally self-adjoint with respect to a standard measure l−1/2dl. Thus we have Ĥ(j1,j2,j3) = l−mH̃lm, where m = 1 4 + J− 2 and H̃ = −~2l1/2 d dl l1/2 d dl + ~2 4 ql−1 + Λl, q = J2 + − 1 4 . Normalization of states and formal self-adjointness of operators is preserved under the trans- formation (4.2). Note that the differential operator l1/2 d dl l 1/2 d dl is the Laplace–Beltrami ope- rator on R+ with respect to the metric g = (l−1), and the new effective potential ~2 4 ql −1 + Λl differs from the original potential by the singular quantum potential ~2 4 ql −1. Thus, the Laplace– Beltrami ordering (j1 = 1 2 = j2, j3 = 0) belongs to the class of orderings \|J+\| = 1 2 , where the quantum potential vanishes. In the parameter space (j1, j3) of orderings, the direction J+ is physically significant in dictating the coefficient of the quantum potential. In contrast, J− is insignificant, because if J+ = J ′+ for two orderings (j1, j2, j3) and (j′1, j ′ 2, j ′ 3), we have Ĥ(j′1,j ′ 2,j ′ 3) = l−αĤ(j1,j2,j3)l α, α = 1 2 (J− − J ′−). A further substitution y = 2l1/2, with dy = l−1/2dl and d2 dy2 = l1/2 d dl l 1/2 d dl , transforms the curved coordinate l to a flat one, yielding H̃ = −~2 d2 dy2 + ~2qy−2 + Λ 4 y2. (4.3) 6 J. Haga and R.L. Maitra In what follows, we denote the kinetic and potential terms of H̃ as T̃ = −~2∆ = −~2 d2 dy2 , Ṽ = ~2qy−2 + Λ 4 y2. The Hamiltonian H̃, including the singular term ~qy−2, resembles radial Hamiltonians arising in nuclear physics; see, e.g., [18]. 5 Self-adjoint extensions of H̃ In defining a time evolution operator, the Hamiltonian (as the infinitesimal generator) must be self-adjoint (cf. Stone’s theorem in [20]). Indeed, if one chooses D(H̃) = C∞0 (R+), standard analysis (see [12, Chapter 4]) yields that D(H̃∗) = { ψ \|ψ,ψ′ a.c. in (0,∞); ψ, H̃ψ ∈ L2(R+) } , (5.1) the so-called natural domain of the differential operation Ȟ = −~2 d2 dy2 + ~2qy−2 + Λ 4 y 2. In Section 5.1 we show that H̃ has a unique self-adjoint extension for a large class of orderings; in Sections 5.2 and 5.3 we parametrize and characterize the self-adjoint extensions corresponding to those orderings admiting multiple extensions. 5.1 Deficiency subspaces We proceed by determining deficiency subspaces for H̃. To compute the deficiency indices of H̃ we search for solutions to (H̃∗ − (±i)I)ψ(y) = 0; (5.2) the solution space will be denoted by K ± (respectively). To transform (5.2), we define the variables z and ϕ as follows z = √ Λ 2~ y2 and ψ(y) = zβe−z/2ϕ(z), where β = 1 4 + \|J+\| 2 . From this (5.2) becomes z d2ϕ dz2 + (1 + \|J+\| − z) dϕ dz − ( 1 + \|J+\| ± i/(~ √ Λ) 2 ) ϕ(z) = 0. (5.3) Defining α = (1 + \|J+\| ± i/(~ √ Λ))/2 and γ = 1 + \|J+\|, (5.3) is recognizable as the confluent hypergeometric equation z d2ϕ dz2 + (γ − z)dϕ dz − αϕ = 0. (5.4) Solutions to (5.4) are confluent hypergeometric functions of the first and second kind (respec- tively) [21]: ϕ1(z) = Φ(α, γ; z) := ∞∑ k=0 (α)k (γ)k zk k! , ϕ2(z) = Ψ(α, γ; z) := Γ(1− γ) Γ(1 + α− γ) Φ(α, γ; z) + Γ(γ − 1) Γ(α) z1−γΦ(1 + α− γ, 2− γ; z), Factor Ordering and Path Integral Measure for Quantum Gravity in (1+1) Dimensions 7 where \|z\| < ∞, \| arg(z)\| < π, γ 6= 0,−1,−2, . . . , and the Pochhammer symbol (a)k = a(a + 1) · · · (a+ k − 1) for a ∈ C and integer k ≥ 0. We obtain the following solutions to (5.2): ψ±1 (y) = zβe−z/2 ∞∑ k=0 ( 1 + \|J+\| 2 ± i 2~ √ Λ ) k zk k!(1 + \|J+\|)k , ψ±2 (y) = Γ(−\|J+\|) Γ ( 1−\|J+\| 2 ± i 2~ √ Λ )zβe−z/2 ∞∑ k=0 ( 1 + \|J+\| 2 ± i 2~ √ Λ ) k zk k!(1 + \|J+\|)k + Γ(\|J+\|) Γ ( 1+\|J+\| 2 ± i 2~ √ Λ )zβ−\|J+\|e−z/2 ∞∑ k=0 ( 1− \|J+\| 2 ± i 2~ √ Λ ) k zk k!(1− \|J+\|)k . With the intent of applying Theorem 2.4 we investigate asymptotic behavior of these solutions for small and large y. As y → 0+, ψ±1 (y) = O ( y2β ) = O ( y 1 2 +\|J+\|), ψ±2 (y) = O ( y2β ) +O ( y2β−2\|J+\|) = O ( y 1 2 +\|J+\|)+O ( y 1 2 −\|J+\|) = O ( y 1 2 −\|J+\|). Thus ψ±1 ∈ L2 near 0, and ψ±2 ∈ L2 near 0 precisely when \|J+\| < 1. As in [21], for z →∞: Φ(α, γ; z) = O ( ezz−(γ−α) ) , Ψ(α, γ; z) = O ( z−α ) , and as y →∞, ψ±1 (y) = O ( e √ Λ 4~ y 2 y − 1 2 ∓ i ~ √ Λ ) , ψ±2 (y) = O ( e− √ Λ 4~ y 2 y 3 2 +2\|J+\|± i ~ √ Λ ) . Thus ψ±1 6∈ L2 near ∞ and ψ±2 ∈ L2 near ∞ for all factor orderings. Consequently for \|J+\| ≥ 1, K ± = {0} and n+(H̃) = 0 = n−(H̃). For \|J+\| < 1, K ± = Span{ψ±2 } and n+(H̃) = 1 = n−(H̃). These analyses support the following result related to the Hamiltonian (4.3): Theorem 5.1. The operator H̃ = −~2 d2 dy2 +~2 ( \|J+\|2− 1 4 ) 1 y2 + Λ 4 y 2 is essentially self-adjoint on C∞0 (0,∞) iff \|J+\| ≥ 1. Proof. We have seen that ψ±2 (y) is square integrable near zero iff \|J+\| < 1. Thus, V (y) = ~2 ( \|J+\|2 − 1 4 ) 1 y2 + Λ 4 y 2 is in limit point case at zero iff \|J+\| ≥ 1. Since ψ±1 (y) is never square integrable at infinity, V (y) is always in limit point case at infinity. Applying Theorem 2.4 now establishes the desired result. � 5.2 Self-adjoint extensions in terms of boundary spaces For the case when \|J+\| < 1, we seek to characterize and classify the self-adjoint extensions of H̃ in terms of boundary conditions as in Theorem 2.5. Our method is similar to that presented in [8]. Suppose that H̃b = C, and ϕ(1), ϕ(2) are solutions to H̃ϕ = 0 with Wronskian W [ϕ(1), ϕ(2)] = 1 (referred to as reference modes). Let Γi : D(H̃∗)→ C be defined by Γiψ = lim y→0+ W [ ϕ(i), ψ ] , for i = 1, 2. A straightforward computation establishes that for all ψ1, ψ2 ∈ D(H̃∗),〈 H̃∗ψ1, ψ2 〉 − 〈 ψ1, H̃ ∗ψ2 〉 = 〈Γ1ψ1,Γ2ψ2〉C − 〈Γ2ψ1,Γ1ψ2〉C, and for any Ψ1,Ψ2 ∈ C, one may construct g ∈ D(H̃∗) (by choosing g1,2 ∼ C1,2y ±2ν+ 1 2 and taking g = g1 + g2) satisfying that Γig = Ψi, i = 1, 2. 8 J. Haga and R.L. Maitra To determine an explicit formula for the reference modes, we write H̃ϕ = −~2ϕ′′(y) + ( ~2q y2 + Λy2 4 ) ϕ(y) = 0, where, as before, q = J2 + − 1 4 . Making the substitutions w(y) = √ Λ 4~ y 2 and ϕ(y) = w1/4u(w) we obtain w2u′′(w) + wu′(w)− ( w2 + ( J+ 2 )2 ) u(w) = 0. This is a modified Bessel equation; letting ν = \|J+\| 2 , solutions to this equation are given by the modified Bessel functions of the first (Iν(w)) and second (Kν(w)) kind; their Wronskian satisfies W [Kν(w), Iν(w)] = 1/w. Choosing ϕ(1)(y) = √ y √ 2 Kν (√ Λ 4~ y2 ) and ϕ(2)(y) = √ y √ 2 Iν (√ Λ 4~ y2 ) , it follows that W [ϕ(1), ϕ(2)] = 1. The asymptotic equivalences (for w → 0) of modified Bessel functions are well-known: Kν(w) ∼  π 2 1 sin(νπ) 1 Γ(1− ν) (w 2 )−ν , ν /∈ Z, − ln (w 2 ) , ν = 0, Iν(w) ∼ 1 Γ(ν + 1) (w 2 )ν . These yield ϕ(1)(y) ∼ { C1y −2ν+ 1 2 , 0 < \|J+\| < 1, − √ 2y1/2 ln y + ky1/2, J+ = 0, (5.5) where C1 = π 2 √ 2 1 sin(νπ) (√ Λ 8~ )−ν 1 Γ(1− ν) , k = − 1√ 2 ln (√ Λ 8~ ) , and ϕ(2)(y) ∼ C2y 2ν+ 1 2 , (5.6) where C2 = 1√ 2 1 Γ(ν + 1) (√ Λ 8~ )ν . The corresponding asymptotic behavior of the derivatives is given by ϕ(1)′(y) ∼ {( −2ν + 1 2 ) C1y −2ν− 1 2 , 0 < \|J+\| < 1, − 1√ 2 y−1/2 ln y + ( k 2 − √ 2 ) y−1/2, J+ = 0, ϕ(2)′(y) ∼ ( 2ν + 1 2 ) C2y 2ν− 1 2 . Factor Ordering and Path Integral Measure for Quantum Gravity in (1+1) Dimensions 9 For detailed treatment of modified Bessel functions and asymptotic behavior we direct the reader to [21]. Let ξ ∈ U(1) and ψ ∈ D(H̃∗). Writing L = tan(arg(ξ)/2) we have, for 0 < \|J+\| < 1: (ξ − 1)Γ1ψ + i(ξ + 1)Γ2ψ = lim y→0+ ψ(y) [ L(2β − 1)C1y −2β − 2βC2y 2β−1 ] + ψ′(y) [ LC1y 1−2β + C2y 2β ] , (5.7) where as before, β = 1 4 + \|J+\| 2 . For \|J+\| = 0 we have (ξ − 1)Γ1ψ + i(ξ + 1)Γ2ψ = lim y→0+ ψ(y) √ y [ L (√ 2 ln(y)− k 2 + √ 2 ) − 1 2C2 ] + ψ′(y) √ y [ L ( − √ 2 ln(y) + k ) + C2 ] .(5.8) The self-adjoint extensions of H̃ are parametrized by L ∈ (−∞,∞] and may thus be characte- rized as the restriction of H̃∗ to functions in D(H̃∗) such that the limits in (5.7) and (5.8) (for respective values of \|J+\|) are zero. 5.3 Self-adjoint extensions in terms of asymptotics An alternate approach to characterizing self-adjoint domains for H̃ involves describing the asymptotic behavior of functions in the natural domain (5.1) of Ȟ and imposing restrictions which force lim ε→0+ lim L→∞ (ψ′ψ − ψψ′)\|Lε = 0. Combined with the polarization identity, such a re- striction causes the integration-by-parts terms obstructing self-adjointness of H̃ to vanish (see, e.g., [12, Section 3.2]). The asymptotic analysis of the natural domain is accomplished by applying methods similar to those used by Gitman et al. [11, 12] for the Calogero problem. The condition Ȟψ ∈ L2(R+) effectively describes each ψ ∈ D(H̃∗) as the solution of some differential equation Ȟψ = χ, χ ∈ L2(R+). (5.9) The general solution to (5.9) can be written in terms of the reference modes ϕ(1), ϕ(2) as ψ = c1ϕ (1) + c2ϕ (2) + ψ̃, c1, c1 ∈ C, ψ̃ = 1 ~2 [ ϕ(1) ∫ y a2 χϕ(2)dy′ − ϕ(2) ∫ y a1 χϕ(1)dy′ ] , (5.10) since a straightforward computation verifies that ψ̃ is a particular solution to (5.9). Utilizing the asymptotic behavior of the reference modes (5.5), (5.6) (valid for all values of \|J+\|, with minor adjustment to coefficients) and applying the Cauchy–Schwarz inequality yields∫ y a1 χψ̃1dy′ = { O ( y−\|J+\|+1 ) , \|J+\| 6= 1, O (√ ln(y) ) , \|J+\| = 1, ∫ y a2 χψ̃2dy′ = O ( y\|J+\|+1 ) . (5.11) For \|J+\| 6= 0, we have that when \|J+\| 6= 1, ψ̃ = O(y3/2) and ψ̃′ = O(y1/2); when \|J+\| = 1, ψ̃ = O(y3/2 ln(y)) and ψ̃′ = O(y1/2 ln(y)). For J+ = 0, we apply analogous asymptotics (relying upon a series relation between K0(w) and I0(w) (see, e.g., [26])): K0(w(y)) = − ln ( w(y) 2 ) I0(w(y)) + ∞∑ k=0 w(y)2k 2k(k!)2 Ψ(k + 1) = −2 [ ln(y)I0(w(y)) +R(y) ] , 10 J. Haga and R.L. Maitra where R(y) = 1 2 [ ln (√ Λ 8~ ) I0(w(y))− ∞∑ k=0 w(y)2k 2k(k!)2 Ψ(k + 1) ] , and Ψ is the digamma function Ψ(x) = d dx [ln(Γ(x))]. Note that R(y) = O(1) as y → 0. Thus we can write (taking a1 = 0 = a2) ψ̃(y) = − √ y ~2 I0(w(y)) [ ln y ∫ y 0 χ(y′) √ y′I0(z(y′))dy′ − ∫ y 0 χ(y′) √ y′ ln(y′)I0(w(y′))dy′ ] − √ y ~2 [ R(y) ∫ y 0 χ(y′) √ y′I0(w(y′))dy′ − I0(w(y)) ∫ y 0 χ(y′) √ y′R(y′)dy′ ] . (5.12) To simplify the asymptotic analysis of (5.12) as y → 0, define Θ(y) = ∫ y 0 χ(y′) √ y′I0(w(y′))dy′, Θ′(y) = χ(y) √ yI0(w(y)), and use the Cauchy–Schwartz inequality as in (5.11) to obtain Θ(y) = O(y) as y → 0. We can now use integration by parts to rewrite (5.12) as ψ̃(y) = √ y ~2 I0(w(y)) [∫ y 0 ln(y′)Θ′(y′)dy′ − ln yΘ(y) ] + √ y ~2 [ I0(w(y)) ∫ y 0 χ(y′) √ y′R(y′)dy′ −R(y) ∫ y 0 χ(y′) √ y′I0(w(y′))dy′ ] = − √ y ~2 I0(w(y)) ∫ y 0 1 y′ Θ(y′)dy′ + 1 ~2 [ A(y) ∫ y 0 χ(y′)B(y′)dy′ −B(y) ∫ y 0 χ(y′)A(y′)dy′ ] , (5.13) where for simplicity A(y) = √ yI0(w(y)) and B(y) = √ yR(y). Using the fact that Θ(y) = O(y) and R(y) = O(1), we can once again apply the Cauchy–Schwartz inequality to derive the asymtoptic estimates∫ y 0 χAdy′ = O(y) = ∫ y 0 χBdy′, so that ψ̃ = O(y3/2). Differentiating (5.13) and applying similar asymptotic analysis yields ψ̃′ = O(y1/2). When \|J+\| ≥ 1, H̃ is essentially self-adjoint. Indeed, in this case, (5.10) implies c1 = 0 since ψ ∈ L2(R+); furthermore, the term c2ϕ (2) is absorbed into the asymptotic term corresponding to ψ̃, permitting no freedom in the choice of self-adjoint domain. When H̃ is not essentially self-adjoint (i.e., when \|J+\| < 1), a 1-parameter family of pairs (c1, c2) indexes the set of self-adjoint extensions. We determine the self-adjoint domains for H̃ by requiring that lim ε→0+ lim L→∞ (ψ′ψ − ψψ′)\|Lε = 0 , or, since lim L→∞ ψ(L) = 0 and lim L→∞ ψ′(L) = 0, that lim ε→0 (ψ′ψ − ψψ′)\|ε = 0. The asymptotic equivalences for 0 < \|J+\| < 1 ψ ∼ c1C1y −\|J+\|+ 1 2 + c2C2y \|J+\|+ 1 2 +O ( y 3 2 ) , ψ′ ∼ c1C1 ( −\|J+\|+ 1 2 ) y−\|J+\|− 1 2 + c2C2 ( \|J+\|+ 1 2 ) y\|J+\|− 1 2 +O ( y 1 2 ) Factor Ordering and Path Integral Measure for Quantum Gravity in (1+1) Dimensions 11 imply that lim ε→0 (ψ′ψ − ψψ′)\|ε = −2\|J+\|C1C2(c1c2 − c1c2), so we must require c1c2 − c1c2 = 0 to obtain a domain on which H̃ is self adjoint. For orderings satisfying J+ = 0, we have instead the asymptotic equivalences ψ ∼ c1 ( − √ 2y 1 2 ln(y)− ky 1 2 ) + c2C2y 1 2 +O ( y 3 2 ) , ψ′ ∼ c1 ( − 1√ 2 y− 1 2 ln(y) + ( k 2 − √ 2 ) y− 1 2 ) + c2 1 2 C2y − 1 2 +O ( y 1 2 ) , where as before, k = − 1√ 2 ln (√ Λ 8~ ) . These similarly yield lim ε→0 (ψ′ψ − ψψ′)\|ε = − √ 2C2(c1c2 − c1c2)/~. Thus c1c2 = c1c2, implying c1c2 ∈ R, and hence c2 is a real multiple of c1, since αc1 = (c2c1)c1 = (c1c2)c1 = (c1c1)c2 = βc2, α, β ∈ R. Note that we can define θ such that tan θ = β α , yielding c1 cos(θ) = c2 sin(θ), and allowing us to write ψ in the form ψ = C ( sin(θ)ϕ(1) + cos(θ)ϕ(2) ) +O ( y 3 2 ) , (5.14) implying the asymptotics ψ ∼ { C ( sin(θ)C1y −\|J+\|+ 1 2 + cos(θ)C2y \|J+\|+ 1 2 ) +O ( y 3 2 ) , 0 < \|J+\| < 1, C ( sin(θ) ( − √ 2y 1 2 ln(y) + ky 1 2 ) + cos(θ)C2y 1 2 ) +O ( y 3 2 ) , \|J+\| = 0, (5.15) and similarly for ψ′. Taking θ = tan−1(L) we observe that this parametrization of the self-adjoint extensions for H̃ may be recast in the form as in (5.7) and (5.8). 6 Formulation of the propagator and path-integral representation of the fundamental solution In this section, we use extensions of the Trotter product formula due to Kato and to Exner, Neidhardt, and Zagrebnov to derive Feynman–Kač formulae (i.e., path integral representations) for the time-evolution operators e−itH̃/~ and e−tH̃/~ evolving initial data according to the real- and imaginary-time Schrödinger equations, respectively. By factoring the time evolution opera- tor Ut = e−atH̃/~ (a = i, 1) as the limit of the concatenation of short-time evolution operators( e−atT̂ /~ne−atV̂ /~n )n , we can write a path-integral expression for the propagator (integral kernel) of e−atH̃ = e−at(T̃+Ṽ ) by using the known integral kernel for e−atT̃ /~n = ea~t∆/n. The Trotter product formula and its extensions require that the operators T̃ and Ṽ be self- adjoint [14], so the domain of T̃ = −~2 d2 dy2 must correspond to one of the self-adjoint extensions of the Laplacian on the half-line; namely, T̃ must be taken to be T̃β, β ∈ (−∞,∞], where T̃β = −~2∆ = −~2 d2 dy2 , D(T̂β) = { ψ ∈ D∗ Ť \| lim ε→0 [ψ(ε)− βψ′(ε)] = 0 } (as usual, β =∞ denotes the Neumann boundary condition ψ′(0) = 0). The familiar Trotter product formula further requires that H̃ = T̃β + Ṽ be essentially self- adjoint on D(T̃β) ∩ D(Ṽ ), a problematic condition in the presence of multiple self-adjoint extensions of H̃ for \|J+\| < 1. The extended results of Kato and of Exner et al do not re- quire essential self-adjointness of T̃β + Ṽ , but instead demand that T̃β and Ṽ be nonnegative. 12 J. Haga and R.L. Maitra Consequently their application is limited to factor orderings such that \|J+\| ≥ 1 2 and hence V (y) = ~2 ( \|J+\|2 − 1 4 ) 1 y2 + Λ 4 y 2 ≥ 0. The following subsections construct Feynman–Kač formulas in real and imaginary time, for the cases \|J+\| ≥ 1 2 . The case \|J+\| = 1 2 is special because for such factor orderings, the quantum potential term ~qy−2 = ~ ( J2 + − 1 4 ) y−2 in V (y) vanishes. For \|J+\| < 1 2 , neither the standard Trotter product formula nor extensions due to Kato and to Exner et al apply. In the case of the extensions, the operator Ṽ fails to be nonnegative, while the standard Trotter product theorem does not apply since D(T̃β) ∩D(Ṽ ) does not determine one of the self-adjoint extensions of H̃. Indeed, functions ψ ∈ D(Ṽ ) satisfy∫ ∞ 0 y−4\|ψ\|2dy <∞, (6.1) while elements of the self-adjoint extension of H̃ corresponding to the angle θ in Span{ϕ(1), ϕ(2)} behave asymptotically as (5.15) near y = 0. For \|J+\| < 1 2 , the condition (6.1) can only be satisfied when C = 0, meaning that D(T̃β) ∩D(Ṽ ) satisfies the boundary conditions of all self- adjoint extensions of H̃, and thus H̃ is not essentially self-adjoint on D(T̃β) ∩ D(Ṽ ). In fact, (6.1) can only be satisfied when C = 0 for all \|J+\| < 1 (where \|J+\| 6= 1 2 , since at \|J+\| = 1 2 (6.1) does not apply), so the standard Trotter product formula is inapplicable for all such orderings. Complementary to the path integral approach to constructing a time evolution operator one may use spectral decomposition of the Hamiltonian to define propagators as in [27]. 6.1 Imaginary-time Feynman–Kač formula 6.1.1 Factor orderings with \|J+\| > 1 2 For factor orderings in this range, both T̃β and Ṽ are nonnegative self-adjoint operators. In addition, the intersection of their form domains Q(T̃β)∩Q(Ṽ ) (equivalently, D(T̃ 1/2 β )∩D(Ṽ 1/2)) is dense (since it contains C∞0 (R+)). Thus the operators T̃β and Ṽ satisfy the hypotheses of the strong Kato–Trotter product formula (see, e.g., [29, Theorem S.21]). We conclude that the quadratic form 〈ϕ, T̃βψ〉L2(R+) + 〈ϕ, Ṽ ψ〉L2(R+) is closed on the domain Q(T̃β)∩Q(Ṽ ), and hence is associated with a self-adjoint operator, denoted T̃β+̇Ṽ (the form sum of T̃β and Ṽ ). The product formula e−t(T̃β+̇Ṽ ) = s-lim n→∞ ( e−tT̃β/ne−tṼ /n )n (6.2) then holds. We first investigate the domain D(T̃β+̇Ṽ ) ⊆ Q(T̃β)∩Q(Ṽ ) for the self-adjoint operator T̃β+̇Ṽ . Observe that since Q(Ṽ ) = { ψ ∈ L2(R+) ∣∣ ∫ ∞ 0 V (y)\|ψ(y)\|2dy <∞ } , every element ψ ∈ D(T̃β+̇Ṽ ) must satisfy∫ ∞ 0 y−2\|ψ(y)\|2dy <∞. (6.3) For \|J+\| ≥ 1, there is only one self-adjoint domain for H̃, and the asymptotic analysis of Section 5.3 implies that the elements satisfy ψ ∼ O(y 3 2 ) near y = 0, consistent with (6.3); indeed, D(T̃β+̇Ṽ ) must coincide with the single self-adjoint extension of H̃ (for all values of β). For 1 2 < \|J+\| < 1, the self-adjoint extensions of H̃ are indexed by one-dimensional subspaces of Factor Ordering and Path Integral Measure for Quantum Gravity in (1+1) Dimensions 13 Span{ϕ(1), ϕ(2)}, with each subspace determined by θ as in (5.14), so that elements of D(T̃β+̇Ṽ ) must be characterized by a particular angle θ = θ′. Combining (6.3) and (5.14), we obtain∫ 1 0 y−2 [ sin(θ′)ϕ(1)(y) + cos(θ′)ϕ(2)(y) ]2 dy <∞. (6.4) Since ϕ(1) ∼ C1y −\|J+\|+ 1 2 , ϕ(2) ∼ C2y \|J+\|+ 1 2 , (6.4) can only be satisfied if θ′ = 0, identifying T̃β+̇Ṽ as the self-adjoint extension of H̃ purely in Span{ϕ(2)} (for all β). For \|J+\| > 1 2 the conditions of the Kato–Trotter product formula identify a single self-adjoint extension of H̃ for which one may construct a time-sliced path integral, regardless of the choice of T̃β. Thus without loss of generality we fix T̃0 (corresponding to the Dirichlet boundary condition on the Laplacian) as the self-adjoint extension for T̃ . We now combine the Kato–Trotter product formula with the well-known Dirichlet heat kernel on R+ (see, e.g., [3, 6, 31]) to construct a Feynman–Kač formula for the imaginary-time evolution operator e−tH̃/~ (guaranteed by the spectral theorem to be well-defined for self-adjoint H̃). The Dirichlet heat kernel on the half-line, given by pD+(t, y, z) = 1√ 4πt [ e−(y−z)2/4t − e−(y+z)2/4t ] , (6.5) is the integral kernel for et∆0 , and may be obtained by applying the ordinary heat kernel on R to an odd extension of data ψ : R+ → C (with ψ(0) = 0). The heat kernel (6.5) preserves the Dirichlet boundary condition for all time. From (6.2) and (6.5), we obtain e−tH̃/~ψ = lim n→∞ ( e−tT̃0/~ne−tV (y)/~n)nψ = lim n→∞ ( e~t∆/ne−tV (y)/~n)nψ = lim n→∞ N ∫ R+ · · · ∫ R+ n∏ j=1 [( e −n(zj+1−zj)2 4~t − e −n(zj+1+zj)2 4~t ) e −tV (zj) ~n ] ψ(z1)dz1 · · · dzn, N = ( n 4π~t )n/2 . (6.6) The path integrals corresponding to different factor orderings are distinguished by the quan- tum potential ~2qy−2 = ~2 ( J2 + − 1 4 ) y−2, so that we may write the factor n∏ j=1 e−tV (zj)/~n = n∏ j=1 e−tΛz 2 j /4~n n∏ k=1 e−t~qz −2 k /n and regard the latter product on the right-hand side as a weight on the path integral measure corresponding to the factor ordering chosen. Notice that this weight is small for histories tarrying near the singularity y = 0. A large value of q in the quantum potential intensifies this effect. The n-fold time-sliced measure( n 4π~t )n/2 n∏ j=1 ( e −n(zj+1−zj)2 4~t − e −n(zj+1+zj)2 4~t ) dz1 · · · dzn represents an avoiding measure supported on paths away from the singularity y = 0. As dis- cussed by Farhi and Gutmann in [6], the first term of (6.5) is the heat kernel for propagation from z > 0 to y > 0 on R, but to construct an avoiding measure on the half-line R+, this must be adjusted by subtracting the weight for paths from y > 0 to z > 0 incident on the origin. The appropriate weight to subtract is precisely the second term of (6.5), because paths from y > 0 to z > 0 intersecting the origin are in one-to-one correspondence with paths from y to −z (for a full treatment, refer to [6, Section 2].) 14 J. Haga and R.L. Maitra We conclude that by introducing a quantum potential, factor orderings in the range \|J+\| ≥ 1 require that the path integral measure be supported on paths avoiding the singularity. A larger value of q results in a decreased contribution by histories for which the circumference of the universe remains mostly small. For orderings with 1 2 < \|J+\| < 1, the extended product formulae limit the construction of a time-sliced path integral to one whose measure is supported on paths avoiding the singularity. 6.1.2 Factor orderings with \|J+\| = 1 2 As in the preceding section T̃β and Ṽ are nonnegative self-adjoint operators with Q(T̃β) ∩Q(Ṽ ) dense, so that once again the hypotheses of the strong Kato–Trotter product formula are sa- tisfied, for β ∈ (−∞,∞]. Thus T̃β+̇Ṽ is self-adjoint on D(T̃β+̇Ṽ ) ⊆ Q(T̃β) ∩ Q(Ṽ ), and the product formula (6.2) holds. We wish to identify D(T̃β+̇Ṽ ) as one of the self-adjoint extensions of H̃, corresponding to θ in Span{ϕ(1), ϕ(2)}. However, the absence of the quantum potential (since q = J2 +− 1 4 = 0) removes the condition (6.3); membership of ψ in Q(Ṽ ) no longer imposes any restriction on asymptotics near y = 0, and instead we utilize that D(T̃β)∩D(Ṽ ) ⊆ D(T̃β+̇Ṽ ) (see, e.g., [7, Proposition 3.1]). Thus the corresponding self-adjoint extension of H̃ contains elements satisfying the boundary condition ψ(0) = βψ′(0). Since the asymptotic analysis of Section 5.3 yields (for \|J+\| = 1 2), ψ ∼ C ( C1 sin(θ) + C2 cos(θ)y ) +O ( y3/2 ) , ψ′ ∼ C ( C2 cos(θ) ) +O ( y1/2 ) , (6.7) we conclude that β determining the self-adjoint extension of the Laplacian is related to θ for the self-adjoint extension of H̃ by β = C1 C2 tan θ. (6.8) The Dirichlet boundary condition β = 0 corresponds to θ = 0, and as in the preceding section, we use (6.5) with the Kato–Trotter product formula to obtain the same time-sliced path integral expression for e−tH̃/~ (cf. (6.6)). The Neumann boundary condition ψ′(0) = 0 (formally β =∞) corresponds to θ = π 2 . In this case e−tT̃∞/n in (6.2) may be expressed in terms of the well-known Neumann heat kernel on R+ (see, e.g., [3, 6, 9, 31]): pN+ (t, y, z) = 1√ 4πt [ e−(y−z)2/4t + e−(y+z)2/4t ] , (6.9) derived by evenly extending data on R+ (with ψ′(0) = 0) to R and evolving via the ordinary heat kernel. Similarly to the case for Dirichlet boundary conditions, we combine (6.9) with the Kato–Trotter product formula to obtain the time-sliced path integral propagator: e−tH̃/~ψ = lim n→∞ ( e−tT̃∞/~ne−tV (y)/~n)nψ = lim n→∞ ( e~t∆/ne−tV (y)/~n)nψ = lim n→∞ N ∫ R+ · · · ∫ R+ n∏ j=1 [( e −n(zj+1−zj)2 4~t + e −n(zj+1+zj)2 4~t ) e −tV (zj) ~n ] ψ(z1)dz1 · · · dzn, N = ( n 4π~t )n/2 . In contrast to the singularity-avoiding measure generated by the Dirichlet heat kernel, the Neumann n-fold time-sliced measure( n 4π~t )n/2 n∏ j=1 ( e −n(zj+1−zj)2 4~t + e −n(zj+1+zj)2 4~t ) dz1 · · · dzn Factor Ordering and Path Integral Measure for Quantum Gravity in (1+1) Dimensions 15 corresponds to a reflecting measure supported on paths which may strike the singularity on multiple occasions. As discussed in [6], the first term in (6.9) yields the weight for paths striking the origin an even number of times, since these paths are in one-to-one correspondence with paths from y > 0 to z > 0 in the whole line. The second term in (6.9) supplements the first by weighting paths which strike the origin an odd number of times, since the same correspondence associates such paths with paths in the whole line from y > 0 to −z < 0. We have derived path integral representations for the imaginary-time evolution operators corresponding to two of the self-adjoint extensions of H̃ (θ = 0 and θ = π 2 ) when \|J+\| = 1 2 , by using the heat kernel for Dirichlet and Neumann boundary conditions, respectively. For the remaining extensions, the heat kernel for the boundary condition ψ(0) = βψ′(0) may be constructed using spectral decomposition of −∆ with corresponding boundary behavior. The resulting eigenfunctions are ψp(y) = ( 2 π )1/2 cos(py + φp), where tanφp = −1/pβ (see [3, 6, 9]), yielding the heat kernel p (β) + (t, y, z) = 1√ 4πt [ e−(y−z)2/4t + e−(y+z)2/4t ] − β−1 exp { t β2 + (y + z) β } erfc [( t β2 )1/2 + ( y + z 2β )( t β2 )−1/2 ] . The term containing the complementary error function erfc(·) decreases as β ranges from 0 to∞, and may be regarded as interpolating between the avoiding and reflecting measures. The time-sliced path integral expression for the imaginary-time evolution operator may thus be constructed as e−tH̃/~ψ = lim n→∞ ∫ R+ · · · ∫ R+ n∏ j=1 [ p (β) + ( ~t n , zj+1, zj ) e−tV (zj)/~n ] ψ(z1)dz1 · · · dzn. 6.2 Real-time Feynman–Kač formula The Trotter product formula for real time is generally applicable to negative as well as nonnega- tive operators. However as noted above, when the operator Ṽ is negative (i.e., when \|J+\| < 1 2), the intersection of the domains D(T̃β) ∩ D(Ṽ ) is too small to determine a unique self-adjoint extension of H̃, and the standard Trotter product formula does not apply. When both T̃β and Ṽ are nonnegative (\|J+\| ≥ 1 2), a result due to Exner et al. (see [5, Theorem 2.2]) allows us to conclude that a version of the Trotter product formula in real time holds with respect to the L2 norm over time. That is, for nonnegative self-adjoint T̃β and Ṽ with densely defined form sum T̃β+̇Ṽ and for any time T > 0 and any state ψ ∈ L2(R+), lim n→∞ ∫ T −T ∥∥∥(e−itT̃β/~ne−itṼ /~n)nψ − e−it(T̃β+̇Ṽ )/~ψ ∥∥∥2 2 dt = 0. (6.10) As discussed in [5], this result is weaker than the standard Trotter product formula; however it implies the existence of a subsequence {nk} along which convergence in the strong operator topology holds. This is sufficient to define a time-sliced path integral. The conditions of the extended Trotter product formula (6.10) are identical to those of the strong Kato–Trotter product formula. Hence an identical analysis determines the self-adjoint extension of H̃ admitting a time-sliced path integral expression. As before, for \|J+\| ≥ 1, T̃0+̇Ṽ coincides with the unique self-adjoint extension of H̃, while for 1 2 < \|J+\| < 1, the restriction (6.3) picks out the θ = 0 extension (Dirichet boundary condition) as that to which (6.10) applies. For \|J+\| = 1 2 , the correspondence (6.8) equates D(T̃β+̇Ṽ ) to the self-adjoint domain for H̃ with 16 J. Haga and R.L. Maitra asymptotics given by (6.7), and each extension admits a Feynman–Kač formula for the real-time evolution operator. For orderings satisfying \|J+\| ≥ 1 or \|J+\| = 1 2 , we have that D(T̃β) ∩ D(Ṽ ) is large enough to determine a unique self-adjoint extension of H̃, implying that the standard Trotter product formula applies. 6.2.1 Factor orderings with \|J+\| ≥ 1 For factor orderings with \|J+\| ≥ 1, the self-adjoint form sum T̃0+̇Ṽ must coincide with the unique self-adjoint extension of H̃. At the same time, D(T̃0)∩D(Ṽ ) ⊆ D(T̃0+̇Ṽ ) and D(T̃0)∩D(Ṽ ) is dense (since it contains C∞0 (R+)), so that H̃ = T̃0+Ṽ is essentially self-adjoint on D(T̃0)∩D(Ṽ ). Thus the standard Trotter product formula applies, and together with the well-known integral kernel [3, 6, 9] gD+ (t, y, z) = 1√ 4πit [ ei(y−z) 2/4t − ei(y+z)2/4t ] for eit∆ permits a real-time Feynman–Kač formula for e−itH̃/~: e−itH̃/~ψ = lim n→∞ ( e−itT̃0/~ne−itV (y)/~n)nψ = lim n→∞ ( ei~t∆/ne−itV (y)/~n)nψ = lim n→∞ N ∫ R+ · · · ∫ R+ n∏ j=1 [( e in(zj+1−zj)2 4~t − e in(zj+1+zj)2 4~t ) e −itV (zj) ~n ] ψ(z1)dz1 · · · dzn, N = ( n 4πi~t )n/2 . (6.11) As in (6.6), the n-fold measure appearing in (6.11) corresponds to the avoiding measure sup- ported on paths which do not intersect the singularity. 6.2.2 Factor orderings with 1 2 < \|J+\| < 1 As discussed above, in this range of orderings the ordinary Trotter product formula does not apply, but the result (6.10) of [5] permits a Feynman–Kač formula. As in the imaginary-time case, the identification of D(T̃β+̇Ṽ ) with the Dirichlet self-adjoint domain of H̃ (θ = 0), inde- pendent of β, implies that without loss of generality, T̃0 may be taken as the kinetic term. The Feynman–Kač formula will thus be as in (6.11), with the limit understood to be taken over the subsequence {nk} for which strong operator convergence of the product formula is guaranteed. 6.2.3 Factor orderings with \|J+\| = 1 2 As discussed in the imaginary-time case, for \|J+\| = 1 2 the intersection D(T̃β)∩D(Ṽ ) is contained in a unique self-adjoint extension of H̃, with the correspondence between β and the angle θ of the self-adjoint extension given by (6.8). Since C∞0 (R+) ⊂ D(T̃β) ∩D(Ṽ ) is dense, H̃ is essentially self-adjoint on D(T̃β) ∩D(Ṽ ) and the standard Trotter product formula again applies. The integral kernel for eit∆ with boundary condition ψ(0) = βψ′(0) is given by [3, 6, 9] g (β) + (t, y, z) = 1√ 4πit [ ei(y−z) 2/4t + ei(y+z)2/4t ] − β−1 exp { it β2 + (y + z) β } erfc [( it β2 )1/2 + ( y + z 2β )( it β2 )−1/2 ] . (6.12) Factor Ordering and Path Integral Measure for Quantum Gravity in (1+1) Dimensions 17 Using (6.12), the time-sliced path integral expression for the real-time evolution operator be- comes e−itH̃/~ψ = lim n→∞ ∫ R+ · · · ∫ R+ n∏ j=1 [ g (β) + ( ~t n , zj+1, zj ) e−itV (zj)/~n ] ψ(z1)dz1 · · · dzn. As with imaginary time, the special cases β = 0 and β = ∞ correspond to the Dirichlet and Neumann boundary conditions, with g (β) + (t, y, z) =  1√ 4πit [ ei(y−z) 2/4t − ei(y+z)2/4t ] , β = 0, 1√ 4πit [ ei(y−z) 2/4t + ei(y+z)2/4t ] , β =∞. The interpretation of the path integral measures as supported on avoiding and reflecting paths for the Dirichlet and Neumann cases, respectively, and of the erfc(·) term in the integral ker- nel (6.12) as interpolating between these two cases, remains as in imaginary time. 7 Conclusion This investigation highlights the effects of factor ordering in circumscribing the range of al- lowed behaviors for wave functions near the Big Bang/Big Crunch singularity. By generating a sufficiently steep positive quantum potential, orderings with \|J+\| ≥ 1 force the correspon- ding path integral measure to be supported only on paths which avoid the singularity, while orderings with \|J+\| = 1 2 (including the Laplace–Beltrami/conformal ordering), generating no quantum potential, require an additional boundary condition on wave functions to determine a path integral measure. This boundary condition dictates whether the path integral measure is supported solely on singularity-avoiding paths or whether it has support on paths which reflect off of the singularity. These results affirm the observation of Kontoleon and Wiltshire [19] regarding the sensitivity of boundary behavior in quantum cosmology to factor ordering. Moreover, the importance of boundary behavior in addressing the question of quantum resolution of the classical Big Bang/Big Crunch singularity has been discussed by Kiefer in [17]; the results in Section 6 reflect the close connection between the support of the path integral measure and the behavior of wave functions near the singularity. In the sequel we aim to investigate the construction of Hartle–Hawking-type wave functions of the universe using the path integral measures above, and solving the Wheeler–DeWitt equation with corresponding factor ordering. To address the inherent time-reparametrization invariance of (1+1) gravity, it would be of interest to construct time-independent propagators between initial and final states as by Bunster in [33, 34]; in [15], Halliwell implements such methods to yield solutions and Green’s functions for the Wheeler–DeWitt equation in quantum cosmology. Additionally, one could implement the promising methods of Gerhardt in [10] to characterize the space of physical states. Eventually we expect to broaden the class of factor orderings considered, and ultimately to extend analysis to cosmological models such as Friedmann–Lemâıtre–Robertson–Walker and Bianchi spacetimes. Acknowledgements The authors are grateful to Renate Loll for many fruitful conversations which informed this project, to Jan Ambjørn for the reference to Nakayama’s work on 2D quantum gravity, and to 18 J. Haga and R.L. Maitra Vincent Moncrief, Antonella Marini, and Maria Gordina for ongoing useful discussions about quantization and mathematical physics. Additionally, the authors would like to thank the editor and the anonymous reviewers for their thoughtful and insightful comments and suggestions. References [1] Ambjørn J., Glaser L., Sato Y., Watabiki Y., 2d CDT is 2D Hořava–Lifshitz quantum gravity, Phys. Lett. B 722 (2013), 172–175, arXiv:1302.6359. [2] Anderson E., Relational motivation for conformal operator ordering in quantum cosmology, Classical Quan- tum Gravity 27 (2010), 045002, 18 pages, arXiv:0905.3357. [3] Clark T.E., Menikoff R., Sharp D.H., Quantum mechanics on the half-line using path integrals, Phys. Rev. D 22 (1980), 3012–3016. [4] DeWitt B.S., Dynamical theory in curved spaces. I. A review of the classical and quantum action principles, Rev. Modern Phys. 29 (1957), 377–397. [5] Exner P., Neidhardt H., Zagrebnov V.A., Remarks on the Trotter–Kato product formula for unitary group, Integral Equations Operator Theory 69 (2011), 451–478. [6] Farhi E., Gutmann S., The functional integral on the half-line, Internat. J. Modern Phys. A 5 (1990), 3029–3051. [7] Faris W.G., Self-adjoint operators, Lecture Notes in Math., Vol. 433, Springer-Verlag, Berlin – New York, 1975. [8] Fülöp T., Singular potentials in quantum mechanics and ambiguity in the self-adjoint Hamiltonian, SIGMA 3 (2007), 107, 12 pages, arXiv:0708.0866. [9] Gaveau B., Schulman L.S., Explicit time-dependent Schrödinger propagators, J. Phys. A: Math. Gen. 19 (1986), 1833–1846. [10] Gerhardt C., Quantum cosmological Friedmann models with an initial singularity, Classical Quantum Gra- vity 26 (2009), 015001, 29 pages, arXiv:0806.1769. [11] Gitman D.M., Tyutin I.V., Voronov B.L., Self-adjoint extensions and spectral analysis in the Calogero problem, J. Phys. A: Math. Theor. 43 (2010), 145205, 34 pages, arXiv:0903.5277. [12] Gitman D.M., Tyutin I.V., Voronov B.L., Self-adjoint extensions in quantum mechanics. General theory and applications to Schrödinger and Dirac equations with singular potentials, Progress in Mathematical Physics, Vol. 62, Birkhäuser/Springer, New York, 2012. [13] Gorbachuk V.I., Gorbachuk M.L., Boundary value problems for operator differential equations, Mathematics and its Applications (Soviet Series), Vol. 48, Kluwer Academic Publishers Group, Dordrecht, 1991. [14] Hall B.C., Quantum theory for mathematicians, Graduate Texts in Mathematics, Vol. 267, Springer, New York, 2013. [15] Halliwell J.J., Derivation of the Wheeler–DeWitt equation from a path integral for minisuperspace models, Phys. Rev. D 38 (1988), 2468–2481. [16] Hartle J.B., Hawking S.W., Wave function of the universe, Phys. Rev. D 28 (1983), 2960–2975. [17] Kiefer C., Can singularities be avoided in quantum cosmology?, Ann. Phys. 19 (2010), 211–218. [18] Kleinert H., Path integrals in quantum mechanics, statistics, polymer physics, and financial markets, 5th ed., World Sci. Publ., Hackensack, NJ, 2009. [19] Kontoleon N., Wiltshire D.L., Operator ordering and consistency of the wave function of the Universe, Phys. Rev. D 59 (1999), 063513, 8 pages, gr-qc/9807075. [20] Lax P.D., Functional analysis, Pure and Applied Mathematics (New York), Wiley-Interscience, New York, 2002. [21] Lebedev N.N., Special functions and their applications, Dover Publications, Inc., New York, 1972. [22] Maitra R.L., Can causal dynamical triangulations probe factor-ordering issues?, Acta Phys. Polon. B Proc. Suppl. 2 (2009), 563–574, arXiv:0910.2117. [23] Moss I., Quantum cosmology and the self observing universe, Ann. Inst. H. Poincaré Phys. Théor. 49 (1988), 341–349. https://doi.org/10.1016/j.physletb.2013.04.006 http://arxiv.org/abs/1302.6359 https://doi.org/10.1088/0264-9381/27/4/045002 https://doi.org/10.1088/0264-9381/27/4/045002 http://arxiv.org/abs/0905.3357 https://doi.org/10.1103/PhysRevD.22.3012 https://doi.org/10.1103/RevModPhys.29.377 https://doi.org/10.1007/s00020-011-1867-2 https://doi.org/10.1142/S0217751X90001422 https://doi.org/10.1007/BFb0068567 https://doi.org/10.3842/SIGMA.2007.107 http://arxiv.org/abs/0708.0866 https://doi.org/10.1088/0305-4470/19/10/024 https://doi.org/10.1088/0264-9381/26/1/015001 https://doi.org/10.1088/0264-9381/26/1/015001 http://arxiv.org/abs/0806.1769 https://doi.org/10.1088/1751-8113/43/14/145205 http://arxiv.org/abs/0903.5277 https://doi.org/10.1007/978-0-8176-4662-2 https://doi.org/10.1007/978-94-011-3714-0 https://doi.org/10.1007/978-94-011-3714-0 https://doi.org/10.1007/978-1-4614-7116-5 https://doi.org/10.1103/PhysRevD.38.2468 https://doi.org/10.1103/PhysRevD.28.2960 https://doi.org/10.1002/andp.201010417 https://doi.org/10.1142/9789814273572 https://doi.org/10.1103/PhysRevD.59.063513 https://doi.org/10.1103/PhysRevD.59.063513 http://arxiv.org/abs/gr-qc/9807075 http://arxiv.org/abs/0910.2117 Factor Ordering and Path Integral Measure for Quantum Gravity in (1+1) Dimensions 19 [24] Nakayama R., 2D quantum gravity in the proper-time gauge, Phys. Lett. B 325 (1994), 347–353, hep-th/9312158. [25] Nelson E., Feynman integrals and the Schrödinger equation, J. Math. Phys. 5 (1964), 332–343. [26] NIST digital library of mathematical functions, available at http://dlmf.nist.gov/. [27] Patel D., Rivera J., Maitra R.L., Propagators for 2d quantum gravity in proper-time gauge with varied factor ordering of the Hamiltonian operator, in preparation. [28] Polyakov A.M., Quantum geometry of bosonic strings, Phys. Lett. B 103 (1981), 207–210. [29] Reed M., Simon B., Methods of modern mathematical physics. I. Functional analysis, Academic Press, New York – London, 1972. [30] Reed M., Simon B., Methods of modern mathematical physics. II. Fourier analysis, self-adjointness, Aca- demic Press, New York – London, 1975. [31] Saloff-Coste L., The heat kernel and its estimates, in Probabilistic Approach to Geometry, Adv. Stud. Pure Math., Vol. 57, Math. Soc. Japan, Tokyo, 2010, 405–436. [32] Šteigl R., Hinterleitner F., Factor ordering in standard quantum cosmology, Classical Quantum Gravity 23 (2006), 3879–3893, gr-qc/0511149. [33] Teitelboim C., Quantum mechanics of the gravitational field, Phys. Rev. D 25 (1982), 3159–3179. [34] Teitelboim C., Proper-time gauge in the quantum theory of gravitation, Phys. Rev. D 28 (1983), 297–309. [35] Verlinde H., Conformal field theory, two-dimensional quantum gravity and quantization of Teichmüller space, Nuclear Phys. B 337 (1990), 652–680. https://doi.org/10.1016/0370-2693(94)90023-X http://arxiv.org/abs/hep-th/9312158 https://doi.org/10.1063/1.1704124 http://dlmf.nist.gov/ https://doi.org/10.1016/0370-2693(81)90743-7 https://doi.org/10.1088/0264-9381/23/11/013 http://arxiv.org/abs/gr-qc/0511149 https://doi.org/10.1103/PhysRevD.25.3159 https://doi.org/10.1103/PhysRevD.28.297 https://doi.org/10.1016/0550-3213(90)90510-K 1 Introduction 2 Mathematical background 3 Action for (1+1)-dimensional gravity 4 Schrödinger quantization 5 Self-adjoint extensions of 5.1 Deficiency subspaces 5.2 Self-adjoint extensions in terms of boundary spaces 5.3 Self-adjoint extensions in terms of asymptotics 6 Formulation of the propagator and path-integral representation of the fundamental solution 6.1 Imaginary-time Feynman–Kac formula 6.1.1 Factor orderings with \|J+\| > 12 6.1.2 Factor orderings with \|J+\|=12 6.2 Real-time Feynman–Kac formula 6.2.1 Factor orderings with \|J+\| 1 6.2.2 Factor orderings with 12<\|J+\|<1 6.2.3 Factor orderings with \|J+\| = 12 7 Conclusion References

Factor Ordering and Path Integral Measure for Quantum Gravity in (1+1) Dimensions

Institution

Ähnliche Einträge