Upper bounds on second order operators, acting on metric function

We prove upper bounds on the general second order operator acting on metric function. The suggested approach does not use traditional formulas for deviations of geodesics and Jacobi fields construction and leads to the manifolds generalization of the classical coercitivity and dissipativity conditio...

Повний опис

Збережено в:
Бібліографічні деталі
Опубліковано в: :Український математичний вісник
Дата:2007
Автор: Antoniouk, A.V.
Формат: Стаття
Мова:Англійська
Опубліковано: Інститут прикладної математики і механіки НАН України 2007
Онлайн доступ:https://nasplib.isofts.kiev.ua/handle/123456789/124513
Теги: Додати тег
Немає тегів, Будьте першим, хто поставить тег для цього запису!
Назва журналу:Digital Library of Periodicals of National Academy of Sciences of Ukraine
Цитувати:Upper bounds on second order operators, acting on metric function / A.V. Antoniouk // Український математичний вісник. — 2007. — Т. 4, № 2. — С. 163-172. — Бібліогр.: 12 назв. — англ.

Репозитарії

Digital Library of Periodicals of National Academy of Sciences of Ukraine
_version_ 1859618845933699072
author Antoniouk, A.V.
author_facet Antoniouk, A.V.
citation_txt Upper bounds on second order operators, acting on metric function / A.V. Antoniouk // Український математичний вісник. — 2007. — Т. 4, № 2. — С. 163-172. — Бібліогр.: 12 назв. — англ.
collection DSpace DC
container_title Український математичний вісник
description We prove upper bounds on the general second order operator acting on metric function. The suggested approach does not use traditional formulas for deviations of geodesics and Jacobi fields construction and leads to the manifolds generalization of the classical coercitivity and dissipativity conditions for diffusion equations.
first_indexed 2025-11-29T00:10:58Z
format Article
fulltext Український математичний вiсник Том 4 (2007), № 2, 163 – 172 Upper bounds on second order operators, acting on metric function Alexander Val. Antoniouk (Presented by S. Ya. Makhno) Abstract. We prove upper bounds on the general second order op- erator acting on metric function. The suggested approach does not use traditional formulas for deviations of geodesics and Jacobi fields con- struction and leads to the manifolds generalization of the classical co- ercitivity and dissipativity conditions for diffusion equations. 2000 MSC. 35A15, 53C21, 58E35. Key words and phrases. Second order operators, metric distance function, geodesic deviations, upper bounds. In this paper we turn to the geometric problem of upper estimates on the general second order operators, acting on metric function. We consider operators of form Lf = 1 2 d∑ σ=1 Aσ(Aσf) + A0f, (1) where A0, Aα represent C∞ smooth globally defined vector fields on the oriented smooth complete connected Riemannian manifold M without boundary. The known approaches of differential geometry were mostly invented for the case of Laplace-Beltrami ∆ or similar operators [5, 6], because it was hard to find the implicit representations for arbitrary differential operators on metric function, defined as a minimum of length functional ρ2(x, y) = inf { 1∫ 0 |γ̇(ℓ)|2dℓ, γ(0) = x, γ(1) = y } . (2) Received 20.04.2004 The research was supported by A.von Humboldt Foundation (Germany). ISSN 1810 – 3200. c© Iнститут математики НАН України 164 Upper bounds on second order operators... Corresponding techniques were related with the use of geodesic deviations formulas and related Jacobi fields, with the study of the global geometry of manifold, e.g. [5, 6], survey [7] and references therein. However, for upper bounds one does not need the precise represen- tations for differential operators on metric (!). Below we develop such estimates and demonstrate, that the traditional approach of geodesic de- viations is a little advanced for such simple problem. The found conditions on coefficients of operator L generalize the clas- sical dissipativity and coercitivity conditions from the linear base space to manifold. They relate the coefficients of operator with the geometric properties of manifold, without traditional separation of geometry: • coercitivity: ∃ o ∈ M such that ∀C ∈ R+ ∃KC ∈ R 1 such that ∀x ∈ M 〈Ã0(x),∇xρ2(x, o)〉 + C d∑ σ=1 ‖Aσ(x)‖2 ≤ KC(1 + ρ2(x, o)); (3) • dissipativity: ∀C, C ′ ∈ R+ ∃KC ∈ R 1 such that ∀x ∈ M , ∀h ∈ TxM 〈∇Ã0(x)[h], h〉 + C d∑ σ=1 ‖∇Aσ(x)[h]‖2 − C ′ d∑ σ=1 〈Rx(Aσ(x), h)Aσ(x), h〉 ≤ KC‖h‖ 2, (4) where Ã0 = A0 + 1 2 ∑d σ=1 ∇Aσ Aσ and [R(A, h)A]m = R m p ℓqA pAℓhq denotes the curvature operator, related with (1,3)-curvature tensor with components R 2 1 34 = ∂Γ 2 1 3 ∂x4 − ∂Γ 2 1 4 ∂x3 + Γ j 1 3Γ 2 j 4 − Γ j 1 4Γ 2 j 3. (5) For simplicity of further calculations we only point the positions of corresponding indexes. Notation ∇H[h] means the directional covariant derivative, defined by (∇H(x)[h])i = ∇jH i(x) · hj . (6) Main result of article provides A. V. Antoniouk 165 Theorem 1. Suppose that conditions (3)–(4) hold. Then there is constant K such that at the points of C2-regularity of metric distance { AI 0 + AII 0 + 1 2 d∑ σ=1 (AI σ + AII σ )2 } ρ2(x, y) ≤ Kρ2(x, y). (7) Notations AI , AII mean vector fields, acting on the first and second vari- ables x and y of function ρ2(x.y) correspondingly, for example AIIρ2(x, y) = 〈A(y),∇y〉ρ 2(x, y). Similarly ∀C ∃KC such that LIρ2(x, o) + C d∑ σ=1 (AI σρ2(x, o))2 ρ2(x, o) ≤ K(1 + ρ2(x, o)). (8) Proof. Step 1. First note, that for smooth vector field X in a vicinity of some point z of manifold N and smooth function f on N there are following representations Xf(z) = lim ε→0 1 ε ε∫ 0 Xf(zs) ds = lim ε→0 1 ε ε∫ 0 d ds f(zs) ds = lim ε→0 f(zε) − f(z) ε , X(Xf)(z) = lim ε→0 1 ε2 ε∫ 0 ds s∫ −s X(Xf)(zℓ) dℓ = lim ε→0 1 ε2 ε∫ 0 ds s∫ −s d dℓ (Xf)(zℓ) dℓ = lim ε→0 1 ε2 ε∫ 0 {(Xf)(zs)−(Xf)(z−s)} ds = lim ε→0 1 ε2 ε∫ 0 d ds {f(zs)+f(z−s)} ds = lim ε→0 f(zε) + f(z−ε) − 2f(z) ε2 . (9) Here we used notation zε for the differential flow along field X: zε = z + ∫ ε 0 X(zs) ds. Step 2. In the vicinity of geodesic γ(ℓ), ℓ ∈ [0, 1] from γ(0) = x to γ(1) = y that minimizes (2) consider smooth vector field H. Introduce a family of paths [0, 1] × (−δ, δ) ∋ (ℓ, s) → γ(ℓ, s) ∈ M 166 Upper bounds on second order operators... such that at s = 0 path γ(ℓ, s) ∣∣ s=0 = γ(ℓ) gives geodesic γ above and parameter s appears as a result of evolution along H: ∂ ∂s γ(ℓ, s) = H(γ(ℓ, s)). (10) Note that for s 6= 0 each path γ(ℓ, s)s-fixed must not be geodesic, unlike in formulas for geodesic deviations. H ]1,0[Îthe only geodesic is here field H Figure 1: Field H (white vectors) in a vicinity of geodesic from x to y deter- mines a set of paths, parameterized by s. The resulting surface is parameterized by (ℓ, s) ∈ [0, 1]× (−δ, δ). Note that for s 6= 0 each path γ(ℓ, s)s-fixed should not be geodesic Step 3. Now let’s apply (9) with N = M × M , X = HI ⊗ HII and function f(z) = ρ(x, y) for z = (x, y). Using the minimal property of geodesic, i.e. that the path γ(ℓ, s) is longer than geodesic from γ(0, s) to γ(1, s), we can estimate terms with ±ε in (9) from above and obtain (point at which we get rid of implicit representations, see also (16)) (HI + HII)ρ2(x, y) = lim ε→0 ρ2(γ(1, ε), γ(0, ε)) − ρ2(x, y) ε ≤ lim ε→0 ∫ 1 0 | ∂ ∂ℓ γ(ℓ, ε)|2dℓ − ∫ 1 0 | ∂ ∂ℓ γ(ℓ, 0)|2dℓ ε = 1∫ 0 ∂ ∂s ∣∣∣ s=0 ∣∣∣ ∂ ∂ℓ γ(ℓ, s) ∣∣∣ 2 dℓ. (11) A. V. Antoniouk 167 To find derivative ∂ ∂s ∣∣ s=0 in the above expression let us use that by continuity arguments, for any ℓ and sufficiently small δ(ℓ) the path {γ(ℓ, z)}z∈(−δ(ℓ),δ(ℓ)) completely lies in some coordinate vicinity (xi). In this coordinate system relation (10) has integral form γi(ℓ, s) = γi(ℓ) + s∫ 0 H i(γ(ℓ, z)) dz (12) with point γ(ℓ) on initial geodesic. Therefore γ̇i(ℓ, s) = γ̇i(ℓ) + s∫ 0 ∂kH i(γ(ℓ, z))γ̇k(ℓ, z) dz and ∂ ∂s γ̇i(ℓ, s) = ∂kH i(γ(ℓ, s))γ̇k(ℓ, s) = (∇kH i − Γ i k h)γ̇k(ℓ, s), (13) where we changed to the covariant derivatives and introduced notation γ̇ = ∂ ∂ℓ γ. In particular, from above formula and (10) we conclude com- mutation ∂ ∂s ∂ ∂ℓ γi(ℓ, s) = ∂ ∂ℓ ∂ ∂s γi(ℓ, s). Relation (13) and autoparallel property of Riemannian connection ∂kgmn(x) = ghnΓ h k m + gmhΓ h k n (14) lead to ∂ ∂s |γ̇(ℓ, s)|2 = ∂ ∂s [gij(γ(ℓ, s))γ̇i(ℓ, s)γ̇j(ℓ, s)] = ∂kgij ∂ ∂s γk · γ̇iγ̇j + 2gij γ̇ i ∂ ∂s γ̇j = 2gij γ̇ i(∇kH j)γ̇k = 2〈γ̇,∇H[γ̇]〉. (15) Therefore estimate (11) transforms to (HI + HII)ρ2(x, y) ≤ 2 1∫ 0 〈∇H[γ̇], γ̇〉 dℓ. 168 Upper bounds on second order operators... Step 4. In a similar to (11) way, (HI + HII)(HI + HII)ρ2(x, y) = lim ε→0 ρ2(γ(1, ε), γ(0, ε)) + ρ2(γ(1,−ε), γ(0, ε)) − 2ρ2(x, y) ε2 ≤ lim ε→0 ∫ 1 0 | ∂ ∂ℓ γ(ℓ, ε)|2dℓ + ∫ 1 0 | ∂ ∂ℓ γ(ℓ,−ε)|2dℓ − 2 ∫ 1 0 | ∂ ∂ℓ γ(ℓ, 0)|2dℓ ε2 = 1∫ 0 ∂2 ∂s2 ∣∣∣ s=0 ∣∣∣ ∂ ∂ℓ γ(ℓ, s) ∣∣∣ 2 dℓ. (16) Using relation (15) we find 1 2 ∂2 ∂s2 ∣∣∣ ∂ ∂ℓ γ(ℓ, ε) ∣∣∣ 2 = ∂ ∂s 〈γ̇(ℓ, s),∇H[γ̇(ℓ, s)]〉 = ∂ ∂s { gij(γ)γ̇i[∇kH j(γ)]γ̇k } = ∂mgij(γ)Hmγ̇i[∇kH j(γ)]γ̇k + gij{ (∇mH i − Γ i m h)γ̇m }[∇kH j(γ)]γ̇k + gij γ̇ i[∂m∇kH j(γ) · Hm(γ)]γ̇k + gij(γ)γ̇i[∇kH j(γ)]{ (∇mHk − Γ k m h)γ̇m }, where, after the differentiation of product, we substituted relations (10) and (13). Using property (14), transforming partial derivative ∂m∇kH j to co- variant ∇m∇kH j and contracting the terms with connection Γ we have 1 2 ∂2 ∂s2 |γ̇(ℓ, ε)|2 = gij(∇mH i)γ̇m(∇kH j)γ̇k + gij γ̇ i(∇m∇kH j)Hmγ̇k + gij(γ)γ̇i(∇kH j)(∇mHk)γ̇m. Now we commute the covariant derivatives in the second term ∇m∇kH j = ∇k∇mHj + R j h kmHh to obtain 1 2 ∂2 ∂s2 |γ̇(ℓ, ε)|2 = |∇H[γ̇] |2 + gij γ̇ i(∇k∇mHj + R j h kmHh)Hmγ̇k + gij γ̇ i(∇kH j)(∇mHk)γ̇m = |∇H[γ̇] |2 − 〈γ̇, R(H, γ̇)H〉 + gij γ̇ i(∇k∇mHj)Hmγ̇k + gij γ̇ i(∇kH j)(∇mHk)γ̇m with curvature operator. A. V. Antoniouk 169 Redenoting indexes m ↔ k in the third term we have 3rd + 4th terms = gij γ̇ i(∇m∇kH j)Hkγ̇m + gij γ̇ i(∇kH j)(∇mHk)γ̇m = gij γ̇ i(∇m{Hk∇kH j})γ̇m which leads to the final representation 1 2 ∂2 ∂s2 |γ̇(ℓ, ε)|2 = |∇H[γ̇] |2 − 〈γ̇, R(H, γ̇)H〉 + 〈γ̇,∇(∇HH)[γ̇]〉. Taking now H = A0 for the first order estimate and H = Aα for the second order estimate we find { AI 0 + AII 0 + 1 2 d∑ α=1 (AI α + AII α )2 } ρ2(x, y) ≤ 1∫ 0 ( 2〈∇Ã0[γ̇], γ̇〉 + d∑ α=1 {|∇Aα[γ̇] |2 − 〈R(Aα, γ̇)Aα, γ̇〉} ) dℓ. (17) Dissipativity condition (4), in view of (2) leads to the statement (7). Step 5. To get estimate (8), one proceeds like above with a choice y = o and ∂ ∂s γ(ℓ, s) = c(ℓ, s)H(γ(ℓ, s)) instead of (10), taking c(ℓ, s0) = a(ℓ) and c(ℓ, sα) = b(ℓ) for the first and second order operators A0 and (Aα)2 correspondingly. One has from (17) { AI 0 + 1 2 d∑ α=1 (AI α)2 } ρ2(x, o) ≤ 1∫ 0 ( 2〈∇(aA0 + 1 2 d∑ α=1 ∇bAα [bAα] )[γ̇], γ̇〉 + d∑ α=1 {|∇(bAα)[γ̇] |2 − 〈R(bAα, γ̇)bAα, γ̇〉} ) dℓ. Using that ∇c(ℓ)[γ̇] = ∂c(ℓ) ∂ℓ we can further rewrite the last inequality { AI 0 + 1 2 d∑ α=1 (AI α)2 } ρ2(x, o) ≤ 1∫ 0 (2a〈∇A0[γ̇], γ̇〉 + 2 ∂a ∂ℓ 〈A0, γ̇〉 + d∑ α=1 {b2|∇Aα[γ̇] |2+ ∂b2 ∂ℓ 〈Aα,∇Aα[γ̇]〉+ (∂b ∂ℓ )2 |Aα| 2−b2〈R(Aα, γ̇)Aα, γ̇〉 + b2〈∇(∇Aα Aα)[γ̇], γ̇〉 + ∂b2 ∂ℓ 〈∇Aα Aα, γ̇〉}) dℓ. (18) 170 Upper bounds on second order operators... To get the last line we also applied ∇Aα b(ℓ) = ∂b(ℓ) ∂s = 0, leading to calculation ∇(∇bAα [bAα])[γ̇] = ∇γ̇(b2∇Aα Aα) = b2∇(∇Aα Aα)[γ̇] + ∂b2 ∂l ∇Aα Aα. Taking further a(ℓ) = b2(ℓ), b(ℓ) = 1 − ℓ and using estimate |〈∇Aα[γ̇], Aα〉| ≤ (1 − ℓ) 2 |∇Aα[γ̇] |2 + 1 2(1 − ℓ) |Aα| 2 we find LIρ2(x, o) ≤ 1∫ 0 { (1 − ℓ)2(2〈∇Ã0[γ̇], γ̇〉 +2 d∑ α=1 |∇Aα[γ̇] |2 − d∑ α=1 〈R(Aα, γ̇)Aα, γ̇〉) (19) + 4(ℓ − 1)〈Ã0(γ), γ̇〉 + 2 d∑ α=1 |Aα| 2 } dℓ. (20) Using that ∇γ(ℓ)ρ2(γ(ℓ), o) = 2ρ(γ(ℓ), o)∇γ(ℓ)ρ(γ(ℓ), o) = 2(ℓ − 1)ρ(x, o) γ̇(ℓ) ρ(x, o) = 2(ℓ − 1)γ̇, the first term in (20) gives 2(ℓ − 1)〈Ã0(γ), γ̇〉 = 〈Ã0(γ),∇γ(ℓ)ρ2(γ(ℓ), o)〉. Finally, using the coercitivity and dissipativity assumptions (3)–(4) for lines (19) and (20) correspondingly, we conclude LIρ2(x, o) ≤ 1∫ 0 { 2KC(1 − ℓ)2|γ̇|2 + KC′(1 + ρ2(γ(ℓ), o)) } dℓ ≤ K(1 + ρ2(x, o)) where we applied (1 − ℓ) ≤ 1 for ℓ ∈ [0, 1] and that path γ(ℓ, 0) = γ(ℓ) realizes the geodesic distance. A. V. Antoniouk 171 Term (AI αρ2(x,o))2 ρ2(x,o) in (8) is treated like the first order term in (11) with choice of coefficient c(ℓ) = b(ℓ) = 1 − ℓ. We get AI αρ2(x, o) ≤ 2 1∫ 0 〈∇(bAα)[γ̇], γ̇〉 dℓ = 1∫ 0 {∂b ∂ℓ 〈Aα(γ), γ̇〉 + b〈∇Aα[γ̇], γ̇〉 } dℓ. Therefore (AI αρ2(x, o))2 ≤ 1∫ 0 {(∂b ∂ℓ )2 ‖Aα‖ 2 + b2‖∇Aα[γ̇] ‖2 } dℓ + 2 1∫ 0 |γ̇|2 dℓ. The last integral gives ρ2(x, o) by (2). Moreover, we can add the first and second terms to line (18) to apply, like before, the coercitivity and dissipativity conditions and finish estimate (8). Remark 1. Let us note, that the upper bounds in (11) and (16) gave us more freedom in a choice of paths {γ(ℓ, s)}ℓ∈[0,1] for s 6= 0. Otherwise they all should be geodesics (compare with picture 1) and we would have to work with curvature, arising in Jacobi equation on geodesic deviations ∇γ̇∇γ̇ ∂γ(ℓ, s) ∂s + R ( γ̇, ∂γ ∂s ) γ̇ = 0. (21) The work with Jacobi equation, as a second order Sturm-Liuville equa- tions which depends on curvature, would require the knowledge of global geometry instead of pointwise conditions (3)–(4), i.e., for example, more precise information about the structure of harmonic tensors and Betti numbers, etc, e.g. [5, 6]. Remark 2. The applications of upper bounds (7)–(8) to the smooth properties of parabolic equations on manifolds are discussed in [1–4]. Here we show that under conditions (3)–(4) plus some additional assumption on the behaviour of coefficients on the infinity, one has C∞– regularity of process yx t with respect to the initial data x and regularity properties of corresponding diffusion semigroups. In particular, dissipa- tivity condition (4) actually represents the coercitivity condition for the high order variational systems for process yx t . Finally, in [4] we apply the technique of upper bounds on opera- tors, acting on metric function, to state the existence and uniqueness of solutions to the Stratonovich diffusions on noncompact manifolds with globally non-Lipschitz coefficients. 172 Upper bounds on second order operators... Author is grateful to referees for their comments about previous ver- sion of the article. References [1] A. Val. Antoniouk, Nonlinear Symmetries of Variational Calculus and Regularity Properties of Differential Flows on Non-Compact Manifolds, In Proceedings of 5th Intern. Conf. “Symmetry in Nonlinear Mathematical Physics”, 2003, 1228–1235. [2] A. Val. Antoniouk, A. Vict. Antoniouk, Regularity of Nonlinear Flows on Non- compact Riemannian Manifolds: Differential vs. Stochastic Geometry or What kind of Variations are Natural? // Ukrainian Math. Journal, 58 (2006), N 8, 1011–1034. [3] A. Val. Antoniouk, A. Vict. Antoniouk, Nonlinear Calculus of variations for dif- ferential flows on manifolds: geomentrically correct generalization of covariant and stochastic variations // Ukrainean Math. Bulletin, 1 (2004), N 4, 449–484. [4] A. Val. Antoniouk, A. Vict. Antoniouk, Non-explosion and solvability of nonlin- ear diffusion equations on noncompact manifolds, to appear in Ukrainian Math. Journal, 16 pp. [5] A. L. Besse, Manifolds all of whose geodesics are closed, Springer–Verlag, 1978. [6] J. Cheeger, D. G. Ebin, Comparison theorems in Riemannian geometry, North- Holland Publ.Co., 1975. [7] A. Grigor’yan, Analytic and geometric background of recurrence and non- explosion of the Brownian motion on Riemannian manifolds // Bull. Amer. Math. Soc., 36 (1999), N 2, 135–249. [8] E. P. Hsu, Stochastic Analysis on Manifolds, Graduate studies in Mathematics, 38, Providence, Rhode Island: American Math. Soc., 2002. [9] N. Ikeda, S. Watanabe, Stochastic differential equations and diffusion processes, Dordrecht: North-Holland publishing, 1981. [10] N. V. Krylov, B. L. Rozovskii, On the evolutionary stochastic equations, Ser. “Contemporary problems of Mathematics”, VINITI, Moscow, 14 (1979), 71–146. [11] H. Kunita, Stochastic flows and stochastic differential equations, Cambridge Uni. Pess, 1990. [12] E. Pardoux, Stochastic partial differential equations and filtering of diffusion pro- cesses // Stochastics, 3 (1979), 127–167. Contact information Alexander Val. Antoniouk Department of Nonlinear Analysis, Institute of Mathematics NAS Ukraine, Tereschenkivska 3, 01601 MSP Kiev-4, Ukraine E-Mail: antoniouk@imath.kiev.ua
id nasplib_isofts_kiev_ua-123456789-124513
institution Digital Library of Periodicals of National Academy of Sciences of Ukraine
issn 1810-3200
language English
last_indexed 2025-11-29T00:10:58Z
publishDate 2007
publisher Інститут прикладної математики і механіки НАН України
record_format dspace
spelling Antoniouk, A.V.
2017-09-28T13:36:56Z
2017-09-28T13:36:56Z
2007
Upper bounds on second order operators, acting on metric function / A.V. Antoniouk // Український математичний вісник. — 2007. — Т. 4, № 2. — С. 163-172. — Бібліогр.: 12 назв. — англ.
1810-3200
2000 MSC. 35A15, 53C21, 58E35.
https://nasplib.isofts.kiev.ua/handle/123456789/124513
We prove upper bounds on the general second order operator acting on metric function. The suggested approach does not use traditional formulas for deviations of geodesics and Jacobi fields construction and leads to the manifolds generalization of the classical coercitivity and dissipativity conditions for diffusion equations.
Author is grateful to referees for their comments about previous version of the article.
en
Інститут прикладної математики і механіки НАН України
Український математичний вісник
Upper bounds on second order operators, acting on metric function
Article
published earlier
spellingShingle Upper bounds on second order operators, acting on metric function
Antoniouk, A.V.
title Upper bounds on second order operators, acting on metric function
title_full Upper bounds on second order operators, acting on metric function
title_fullStr Upper bounds on second order operators, acting on metric function
title_full_unstemmed Upper bounds on second order operators, acting on metric function
title_short Upper bounds on second order operators, acting on metric function
title_sort upper bounds on second order operators, acting on metric function
url https://nasplib.isofts.kiev.ua/handle/123456789/124513
work_keys_str_mv AT antonioukav upperboundsonsecondorderoperatorsactingonmetricfunction