J ni o r Elect o u a rn l o f P c r ob abili ty Vol. 3 (1998) Paper no. 4, pages 1–36. Journal URL http://www.math.washington.edu/˜ejpecp/ Paper URL http://www.math.washington.edu/˜ejpecp/EjpVol3/paper4.abs.html GEOMETRIC EVOLUTION UNDER ISOTROPIC STOCHASTIC FLOW M. Cranston and Y. LeJan University of Rochester Rochester, New York, USA cran@math.rochester.edu Université de Paris, Sud Orsay 91405, France Yves.LeJan@math.u-psud.fr Abstract. Consider an embedded hypersurface M in R3 . For Φt a stochastic flow of differomorphisms on R3 and x ∈ M, set xt = Φt (x) and Mt = Φt (M). In this paper we will assume Φt is an isotropic (to be defined below) measure preserving flow and give an explicit descripton by SDE’s of the evolution of the Gauss and mean curvatures, of Mt at xt . If λ1 (t) and λ2 (t) are the principal curvatures of Mt at xt then the vector of mean curvature and Gauss curvature, (λ1 (t) + λ2 (t), λ1 (t)λ2 (t)), is a recurrent diffusion. Neither curvature by itself is a diffusion. In a separate addendum we treat the case of M an embedded codimension one submanifold of Rn . In this case, there are n−1 principal curvatures λ1 (t), . . . , λn−1 (t). If Pk , k = 1, n − 1 are the elementary symmetric polynomials in λ1 , . . . , λn−1 , then the vector (P1 (λ1 (t), . . . , λn−1 (t)), . . . , Pn−1 (λ1 (t), . . . , λn−1 (t)) is a diffusion and we compute the generator explicitly. Again no projection of this diffusion onto lower dimensions is a diffusion. Our geometric study of isotropic stochastic flows is a natural offshoot of earlier works by Baxendale and Harris (1986), LeJan (1985, 1991) and Harris (1981). 1991 Mathematics Subject Classification. 60H10, 60J60. Key words and phrases. Stochastic flows, Lyapunov exponents, principal curvatures. Research supported by NSA and Army Office of Research through MSI at Cornell. Submitted to EJP on February 19, 1996. Final version accepted on February 12, 1998. Typeset by AMS-TEX 1 2 M. CRANSTON AND Y. LEJAN §1 Isotropic Flows and Geometric Setting. We begin with a nonnegative measure F (dρ) on [0, +∞) with finite moments of all orders. The pth moment of F will be denoted by µp . To F we can associate a covariance Z ∞Z (1.1) Cij (z) = eiρhz,ti (δji − ti tj )σn−1 (dt)F (dρ), S n−1 0 where σn−1 is normalized Lebesgue measure on S n−1 . By a result of Kolmogorov, to Cij is associated a smooth (in x) vector-field valued Brownian motion Ut (x) such that (1.2) EUti (x)Usj (y) = (t ∧ s)C ij (x − y), 1 ≤ i, j ≤ n. A Brownian flow on Rn is constructed by solving the equation Z (1.3) t Φt (x) = x + ∂Us (Φs (x)) 0 (∂ denotes Stratonovich differential.) Existence and uniqueness of such flows was proved in Baxendale (1984), LeJan and Watanabe (1984). Then if Ta is translation by a ∈ Rn , Ta Φt T−a and Φt are identical in law so Φ is stationary. If R is unitary, RΦt R−1 and Φt are identical in law. When these two properties hold we say Φ is isotropic. These properties both follow from the nature of Cij so our Φ is isotropic. The field Ut (x) is smooth so we differentiate it writing Wji = (1.4) i Bjk = (1.5) ∂ i U ∂xj ∂2 U i. ∂xj ∂xk Then we have as direct consequence of (1.1) and (1.2) (using h, i to denote both Euclidean inner product as well as quadratic variation for martingales, d will denote the Itô differential.) (1.6) hdWji (t, y), dW`k (t, y)i = (1.7) µ2 [(n + 1)δki δ`j − δji δ`k − δ`i δjk ]dt n(n + 2) i hdBjk (t, y), dWqp (t, y)i = 0 and for vectors u, v ∈ Rn , (1.8) hhdB(u, u), vi, hdB(u, u), vii = 3µ4 [(n + 3)||u||4||v||2 − 4hu, vi2 ||u||2 ]dt n(n + 2)(n + 4) GEOMETRIC EVOLUTION UNDER ISOTROPIC STOCHASTIC FLOW 3 By isotropy, (1.6), (1.7) and (1.8) remain valid under any unitary change of coordinates. We define, ik Cj` dt = hdWji , dW`k i = µ2 [(n + 1)δki δ`j − δji δ`k − δ`i δjk ]dt n(n + 2) where both dWji , dW`k are evaluated at (t, y), and ij Ck`pq dt = hdB i (ek , e` ), dB j (ep , eq )i again both dB terms are evaluated at (t, y). A quick calculation shows Pn i 2 E = 0 so div U = 0 and therefore Φ is a measure-preserving flow. i=1 dWi (n+3)µ4 nn nn nn Also, it’s rather easy to check that C1111 = 3C1122 and C1122 = n(n+2)(n+4) . ik dt for isotropic flows is The most general form of hdWji , (t, y), dW`k (t, y)i = Cj` ik i,k i k i k Cj` = aδ δj,` + bδj δ` + cδ`δj where a + c, a − c, a + c + db are nonnegative (see LeJan (1985) and the references therein). We retrict our attention to the present µ2 µ2 µ2 covariance structure; a = (n+1) n(n+2) , b = − n(n+2) , c = − n(n+2) , as computations are massively simplified by the fact that certain Ito correction terms vanish. The works of Baxendale and Harris (1986) and LeJan (1985) focused on first order properties of the general isotropic flows in Euclidean space. That is, on properties of the derivative flow DΦt (x). Perhaps the most fundamental information about the derivative flow is its Lyapunov spectrum. To describe the Lyapunov spectrum, first extend DΦt to p-forms α = v1 ∧ · · · ∧ vp (we’re in Euclidean space so we can identify forms and vectors) by defining DΦt (x)α = DΦt (x)v1 ∧ · · · ∧ DΦt (x)vp . Then if v1 , . . . , vp are linearly independent, a.s. 1 log kDΦt (x)αk = γ1 + · · · + γp t→∞ t lim is the sum of the first p Lyapunov exponents. In the case of the isotropic, measure2 preserving flow on Rn , γ1 + · · · + γp = np(n−p)µ (see LeJan 1985). The problem 2(n+2) dealt with in this work involves second order behavior of the flow. The evolving curvature of a submanifold moving under the flow would depend on the properties of the second order flow, (x, u, v) → (Φt (x), DΦt (x)u, D2 Φt (x)(DΦt (x)u, DΦt (x)u) + DΦt (x)v) . In particular, our results rely on the properties given by (1.6), (1.7) and (1.8). Initially, we will consider M an embedded hypersurface in Rn , later specializing to the case n = 3. Take Πt : Txt Rn → Txt Mt to be the orthogonal projection and ∇ to be the canonical connection on Euclidean space. Then St , the second fundamental form of Mt at xt , is defined for u, v ∈ Txt Mt by St (u, v) = (I −Πt )∇u v where v is extended in an arbitrary but smooth manner in a neighborhood of xt . Now St induces a linear map from the exterior algebra Λk (Txt Mt ) → Λk (Txt Mt ) 4 M. CRANSTON AND Y. LEJAN for k = 1, 2, . . . , n − 1. This arises as follows. First observe that since we are in Euclidean space we can identify vectors and forms. Next take νt ∈ Tx⊥t Mt to be a unit vector (there are two choices and we may select the process νt to be continuous in t.) Now in the case k = 1, u → hSt (u, ·), νt i is a map from Λ1 (Txt Mt ) to Λ1 (Txt Mt ). When k > 1 we need to introduce the index set * Ik = {m ∈ {1, . . . , n − 1}k : m1 < m2 < · · · < mk } and for σ ∈ Sk , the set of permutations on k letters, define (−1)σ to be the sign of σ. Then for u1, . . . , uk , v1 , . . . , vk ∈ Txt Mt , set S (k)(u1 ∧ · · · ∧ uk , v1 ∧ · · · ∧ vk ) = X (−1)σ k Y hSt (uσ(j) , vj ), νt i j=1 σ∈Sk This gives rise to the linear map u1 ∧ · · · ∧ uk → S (k)(u1 ∧ · · · ∧ uk , ·) from Λk (Txt Mt ) → Λk (Txt Mt ). The object of our study will be the trace of S (k), T rS (k) , for k = 1, 2, . . . , n − 1. This is obtained via contraction on indices (see Bishop and Goldberg (1980)). Take {u1 , . . . , un−1} to be any basis for Txt Mt . For k ∈ {1, . . . , n − 1}, ~` ∈ Ik , set * | ` | = `1 + · · · + `k , α* = u`1 ∧ u`2 ∧ · · · ∧ u`k ` * * α ` = (−1)| ` |+k u1 ∧ · · · ∧ û`1 ∧ · · · ∧ û`k ∧ · · · ∧ un−1 ( ˆ indicates the indicated vector is omitted from the wedge product) α = u1 ∧ · · · ∧ un−1 . Define an inner product * i = dethu` , umj i. hα* , αm i ` Then we may express T rS (k) as * (1.9) * ` m * )hα , α T rS (k) = S (k) (α* , αm i.||α||−2 , ` * * where the sum (Einstein’s convention) is over ` , m ∈ Ik . Now this is invariant under change of basis so we may select {u1 , . . . , un−1 } to be the unit principal directions of curvature (see Spivak (1975).) Namely, the eigenvectors of S (1) : Λ1 (Txt Mt ) → GEOMETRIC EVOLUTION UNDER ISOTROPIC STOCHASTIC FLOW 5 Λ1 (Txt Mt ). Let λ1 , . . . , λn−1 be the corresponding eigenvalues, otherwise known as the principal curvatures. Now the ui are orthonormal so * * * * hα ` , αm i||α||−2 = δ ` ,m and as hSt (uj , v), νt i = λj huj , vi X σ∈Sk (−1)σ k Y hSt (u`σ(j) , u`j ), νt i = λ`1 . . . λ`k ≡ λ* Thus (1.10) ` j=1 T rS (k) = X * λ* ≡ Pk (λ1 , . . . , λn−1 ). ` ` ∈Ik Pk is the elementary symmetric polynomial of degree k. The process ξt = (xt , Πt , νt , P1 (λ1 , . . . , λn−1 ), · · · Pn−1 (λ1 , . . . , λn−1 )) will, in our case be a diffusion. By the translation invariance part of isotropy, it is clear that xt is not needed, i.e., (Πt , νt , P1 (λ1 , . . . , λn−1 ), · · · , Pn−1 (λ1 , . . . , λn−1 )) has the Markov property. By the invariance of the law of the flow under unitary actions, it follows that Πt and νt are also superfluous. Thus, for isotropic flows the vector (P1 (λ1 , . . . , λn−1 ), . . . , Pn−1 (λ1 , . . . , λn−1 )) is a diffusion. The point of the present article is to describe this diffusion explicitly in the measure-preserving isotropic case. We can derive consequences about the behavior of this diffusion from the explicit description. These properties suggest what sort of behavior one might have with more general flows. What we expect for more general flows is that the process of curvatures for a k-dimensional submanifold evolving under a general flow will be recurrent (though not necessarily a diffusion) if γk > 0. That is, generically the k dimensional tangent space Txt Mt will ‘see’ only stretching directions and this will tend to smooth out the curvatures. This condition may even be too strong. §2 Preliminary Results. Recall the correlations of dW given by (1.6). The flow Φt : Rn → Rn is a C ∞ diffeomorphism whose first derivative satisfies the Stratonovich stochastic differential equation, where x ∈ M, (2.1) dDΦt (x) = ∂W (xt )DΦt (x) Given u ∈ Tx M, DΦt (x)u = u(t) ∈ Txt Mt satisfies (2.2) du = ∂W u where we have suppressed the variable (xt ) in ∂W . In fact, due to the choice of C in (1.1) (2.3) du = dW u, 6 M. CRANSTON AND Y. LEJAN that is, the Itô corrections vanish. Furthermore, if we select a basis {u1 , . . . , un−1} for Tx M, then {u1(t), . . . , un−1 (t)}, where uj (t) = DΦt (x)uj , forms a basis for Txt Mt . We will write uj for uj (t) for simplicity of notation. We now proceed to derive a Stratonovich equation for the second fundamental form St of Mt at xt . Given vectors u, v ∈ Tx M move them via the flow, u(t) = DΦt (x)u, v(t) = DΦt (x)v. Extend v to a smooth vector field V in a neighborhood −1 of x. Set Vt (y) = DΦt (Φ−1 t (y))V (Φt (y)) so that Vt is a vector field near xt such that Vt (xt ) = v(t). Then if Zt ≡ ∇u(t)Vt (xt ) we claim (2.4) dZ(u, v) = ∂W Z(u, v) + ∂B(u, v). For proof of (2.4), take γ to be a curve on M with γ(0) = x, γ 0 (0) = u and define γt = Φt (γ) so that γt0 (0) = u(t) = DΦt u. Then, ∇ut Vt = lim s−1 (Vt (γt (s)) − Vt (γt (0)) s→0 = lim s−1 (DΦt (γ(s))V (γ(s)) − DΦt (x)v(x)) s→0 = lim s−1 [(DΦt (γ(s)) − DΦt (x))V (γ(s)) s→0 + DΦt (x)(V (γ(s)) − v(x))] = D2 Φt (x)(u, v) + DΦt (x)∇u v Since Z t Φt (x) = x + ∂U(Φs (x)) Z 0 t DΦt (x) = I + ∂W (Φs (x))DΦs (x) 0 Z t 2 D Φt (x)(·, ·) = ∂B(Φs (x))(DΦs (x)·, DΦs (x)·) 0 Z t ∂W (Φs (x))D2 Φs (x)(·, ·) + 0 (2.4) follows. A word about notation: ∂Wt Z is an abbreviation for ∂Wji (Φt (x))(Ztj ) i and ∂B(u, v) is short for (∂Bjk (Φt (x))uj (t)vk (t). Next we need an equation for Πt : Txt Rn → Txt Mt . For any ξ ∈ Txt Mt , let {u1 , . . . , un−1} be the moving frame mentioned above, set α = u1 ∧ · · · ∧ un−1 and defining (2.5) αi = (−1)i+1 u1 ∧ · · · ∧ ûi ∧ · · · ∧ un−1 we have (2.6) Πt ξ = n−1 X hξ ∧ αi , αi||α||−2 ui . i=1 The next two lemmas are elementary but crucial. GEOMETRIC EVOLUTION UNDER ISOTROPIC STOCHASTIC FLOW 7 Lemma 2.1. If ξt is a Txt Mt -valued continuous semimartingale, then (I − Πt ) · ∂W ξt = ∂Πt ξt . Proof. Suppose Πt ξt = ξt is as in the statement of the lemma with ξt = xit ui . Then (I − Πt )∂W ξt − ∂Πt ξt = ∂W ξt − Πt ∂W ξt h∂αt , αt i ξt − ∂W ξt ||αt ||2 hξt ∧ αit , ∂αt i hξt ∧ ∂αit , αt i − u − ui i ||αt ||2 ||αt ||2 h∂αt , αt i = −Πt ∂W ξt + 2 ξt ||αt ||2 hαt , ∂αt i i hui ∧ ∂αi , αt i − xit u − x ui i t ||αt ||2 ||αt ||2 +2 Since dαt = −ui ∧ ∂αi + ∂W ui ∧ αi the last term becomes hui ∧ ∂αi , αi ui ||αt ||2 h∂αt , αt i h∂W ui ∧ αit , αt i i =− ξt + xt ui ||αt ||2 ||αt ||2 h∂αt , αt i =− ξt + Πt ∂W ξt ||αt ||2 − xit From this we get the conclusion of the lemma. Lemma 2.2. If ηt ∈ Txt Mt⊥ is a continuous semimartingale, then ∂Πt ηt = Πt ∂ W̃ ηt where W̃ is the transpose of W . Proof. Recalling that h∂αt , αt i Πt v + ∂W Πt v ||αt ||2 hv ∧ αit , ∂αt i hv ∧ ∂αit , αt i + ui + ui ||αt ||2 ||αt ||2 ∂Πt (v) = −2 8 M. CRANSTON AND Y. LEJAN we see that since Πt ηt = 0, ∂Π(ηt ) = hηt ∧ αit , ∂αt i ui . ||αt ||2 But ∂αt = ∂W uk ∧ αkt so ∂Πt (ηt ) = hηt ∧ αit , ∂W uk ∧ αkt i ui ||αt ||2 = h∂ W̃ ηt , uk (t)ihαkt (t), αkt (t)iui (t) ||α(t)||2 = h∂ W̃ ηt ∧ αit , αt i i ut ||αt ||2 = Πt ∂ W̃ ηt and the lemma is proved. Using S = (I − Π)Z together with Lemmas 2.1 and 2.2 we obtain the following. Proposition 2.3. For u, v ∈ Tx M, u(t) = DΦt (x)u, v(t) = DΦt (x)v, (2.7) dS(u, v) = (I − Π)dB(u, v) + ∂W S(u, v) − Π(∂W + ∂ W̃ )S(u, v) where W̃ denotes the transpose of W . Proof. We shall abbreviate by suppressing the dependence on u and v. So, dS = (I − Π)∂Z − ∂ΠZ = (I − Π)dB + (I − Π)∂W Z − ∂ΠZ, by (2.4), = (I − Π)dB + (I − Π)∂W S + (I − Π)∂W R − ∂ΠR − ∂ΠS , ΠZ ≡ R = (I − Π)dB + (I − Π)∂W S − ∂ΠS, by Lemma 2.1 = (I − Π)dB + ∂W S − Π(∂W + ∂ W̃ )S, by Lemma 2.2. The equation (2.7) can be simplified. Set dP = (I − Π)dW (2.8) dQ = ΠdW̃ and define λ and µ so that (I − Π)∂W = dP + dλ (2.9) Π∂ W̃ = dQ + dµ. GEOMETRIC EVOLUTION UNDER ISOTROPIC STOCHASTIC FLOW Lemma 2.4. dΠ = dP Π + dλΠ + dQ(I − Π) + dµ(I − Π). (2.10) 1 dS(u, v) = (I − Π)dB(u, v) + (dP − dQ)S(u, v) + (dP − dQ)2 S(u, v) 2 (2.11) + d(λ − µ)S(u, v) Proof. Let ζt be an Rn -valued semimartingale. Then dΠ = (I − Π)∂W Π + Π∂ W̃ (I − Π) (2.12) holds since ∂Πt ζt = ∂Πt (Πt ζt ) + ∂Πt ((I − Πt )ζt ) = (I − Πt )∂Wt Πt ζt + Πt ∂ W̃t (I − Πt )ζt , by Lemmas 2.1 and 2.2. Thus, we have the following quadratic variations, 1 (−(I − Π)dW ΠdW − ΠdW̃ (I − Π)dW ) 2 1 = (−dP ΠdW − dQ(I − Π)dW ) , 2 1 dµ = ((I − Π)dW ΠdW̃ + ΠdW̃ (I − Π)dW̃ ) 2 1 = (dP ΠdW̃ + dQ(I − Π)dW̃ ). 2 dλ = Consequently, performing Itô corrections on (2.12), we get dΠ = dP Π + dλΠ + dQ(I − Π) + dµ(I − Π) which proves (2.10). In order to derive (2.11), rewrite (2.7) as dS(u, v) = (I − Π)dB(u, v) + (I − Π)∂W S(u, v) − Π∂ W̃ S(u, v). Then (2.11) follows from the definitions of dP, dQ, dλ and dµ in (2.9). Lemma 2.5. If ζt is a Txt Mt⊥ -valued semi-martingale, then (2.12) dλζ = (n − 1)µ2 ζdt 2n(n + 2) 9 10 M. CRANSTON AND Y. LEJAN (2.13) dµζ = (n − 1)(n + 1)µ2 ζdt 2n(n + 2) (n − 1)nµ2 1 (dP − dQ)2 ζ = − ζdt 2 2n(n + 2) (2.14) Proof. Select an orthonormal basis {e1 , . . . , en } for Txt Mt such that en ∈ Txt Mt⊥ . Then, 2dλζ = −dΠdW ζ − dP ΠdW ζ − dQ(I − Π)dW ζ = −(I − Π)dW ΠdW ζ − ΠdW̃ (I − Π)dW ζ n−1 X = −( in Cni )ζdt, since hdWin , dWnn i = 0, 1 ≤ i ≤ n − 1, i=1 (n − 1)µ2 ζdt, by (1.6) = n(n + 2) 2dµζ = dP ΠdW̃ ζ + dQ(I − Π)dW̃ ζ n−1 X =( ii Cnn )ζdt i=1 = (n − 1)(n + 1)µ2 ζdt, by (1.6). n(n + 2) Thus, 1 1 (dP − dQ)2 ζ = (dP − dQ)((I − Π)dW ζ − ΠdW̃ ζ) 2 2 n−1 X 1 n = (dP − dQ)(dWn ζ − dW̃in ei hζ, en i) 2 i=1 X 1 (dWnn dWnn ζ − dWni dW̃in ζ) 2 i=1 n−1 = n−1 X 1 nn ii = (Cnn ζ − ( Cnn )ζ)dt 2 i=1 µ2 = ((n − 1) − (n − 1)(n + 1))ζdt , by (1.6), 2n(n + 2) (n − 1)µ2 =− ζdt. 2(n + 2) Recall that we have selected an initial basis for Tx M with un ∈ Tx M ⊥ and ui (t) = DΦt (x)ui gives a basis for Txt Mt when i runs through {1, . . . , n − 1}. GEOMETRIC EVOLUTION UNDER ISOTROPIC STOCHASTIC FLOW 11 (I−Π)un Theorem 2.6. If ν(t) denotes the unit vector ν = ||(I−Π)u perpendicular to Mt n|| at xt , then (n2 − 1)µ2 dν = −dQν − νdt. 2n(n + 2) Proof. Set vn = (I − Π)un . Then by Lemmas 2.2 and 2.3, dvn = −Π∂ W̃ vn + (I − Π)∂W vn 1 = (dP − dQ)vn + (dP − dQ)2 vn + (dλ − dµ)vn 2 (n − 1)µ2 vn dt, by Lemma 2.5 = (dP − dQ)vn − (n + 2) Then, applying d||vn ||2 = 2hvn , dvn i + h(dP − dQ)vn i 2(n − 1)µ2 ||vn ||2 dt + h(dP − dQ)vn i (n + 2) 2(n − 1)µ2 (n − 1)µ2 ||vn ||2 dt + ||vn ||2 dt = 2||vn ||2 dWnn − (n + 2) n (n − 1)(n − 2)µ2 ||vn ||2 dt = 2||vn ||2 dWnn − n(n + 2) = 2hvn , (dP − dQ)vn i − and so d||vn||−1 = −||vn ||−1 dWnn + Consequently, as ν = (n − 1)(n + 1)µ2 ||vn ||−1 dt. 2n(n + 2) vn ||vn || , (n − 1)µ2 (n − 1)(n + 1)µ2 νdt − (I − Π)dW ν + νdt (n + 2) 2n(n + 2) − h(dP − dQ)ν, dWnn i dν = (dP − dQ)ν − (n − 1)2 µ2 (n − 1)µ2 νdt − νdt 2n(n + 2) n(n + 2) (n + 1)(n − 1)µ2 = dQν − νdt , as desired. 2n(n + 2) = −dQν − We can now derive the equation for hS(u, v), νi. 12 M. CRANSTON AND Y. LEJAN Theorem 2.7. Suppose u(t) = DΦt u, v(t) = DΦt v and ν is as in Theorem 2.6. Then dhS(u, v), νi = hdB(u, v), νi + hS(u, v), νihdW ν, νi − (n − 1)2 µ2 hS(u, v), νidt. 2n(n + 2) Moreover, if ui (t) = DΦt ui , then 33 hdhS(ui , uj ), νi, dhS(uk , u` ), νii = C1122 [hui , uj ihuk , u` i + hui , uk ihuj , u` i + hui , u`ihuj , uk i]dt + µ2 (n − 1) hS(ui , uj ), νihS(uk , u` i, νidt. n(n + 2) Proof. From Lemmas 2.4 and 2.5 we have dS(u, v) = (I − Π)dB(u, v) + (dP − dQ)S(u, v) − (n − 1)µ2 S(u, v)dt. (n + 2) Also −h(dP − dQ)S(u, v), dQνi = hdQS(u, v), dQνi = hS(u, v), νi n−1 X ii Cnn dt i=1 (n − 1)µ2 hS(u, v), νidt. n(n + 2) 2 = So, dhS(u, v), νi = hdS(u, v), νi + hS(u, v), dνi + hdS(u, v), dνi = hdB(u, v), νi + hS(u, v), νidWnn − (n − 1)µ2 hS(u, v), νidt (n + 2) (n2 − 1)µ2 hS(u, v), νidt − h(dP − dQ)S(u, v), dQνi 2n(n + 2) (n − 1)(n − 1)2 µ2 = hdB(u, v), νi + hS(u, v), νidWnn − hS(u, v), νidt 2n(n + 2) − nn For the quadratic variation term, since B and W are independent and C111 = nn 3C1122, hdS(ui , uj ), νihdS(uk , u` ), νi = hhdB(ui , uj ), νihdB(uk , u` ), νii nn + hS(ui , uj ), νihS(uk , u`), νiCnn dt nn = C1122 [hui , uj ihuk , u`i + hui , uk ihuj , u` i + hui , u` ihuj , uk i]dt + (n − 1)µ2 hS(ui , uj ), νihS(uk , u` ), νidt. n(n + 2) GEOMETRIC EVOLUTION UNDER ISOTROPIC STOCHASTIC FLOW 13 We now split the exposition. The case of a hypersurface in R3 will be developed below. The case of a hypersurface in Rn , n > 3, will be treated in an addendum. The latter involves lengthy computations together with combinatorial arguments, whereas the former is more readable. Theorem 2.8. Let u, v ∈ Tx M with u and v linearly independent and define u(t) = DΦt u, v(t) = DΦt v, (however in our computations we shall suppress the t dependence). The process (T r S (1), T r S (2)) satisfies the equations d T r S (1) = [hdB(u, u), νikvk2 + hdB(v, v), νikuk2 − 2hdB(u, v), νihu, vi]kαk−2 + T r S (1) (dW33 − 2(dW11 + dW22 )) + 2[S(u, u)hdW v, vi + S(v, v)hdW u, ui − S(u, v)(hdW u, vi + hu, dW vi)]kαk−2 4µ2 + T r S (1) dt 15 d T r S (2) = [S(v, v)hdB(u, u), νi + S(u, u)hdB(v, v), νi − 2S(u, v)hdB(u, v), νi]kαk−2 + 2 T r S (2) (dW33 − (dW11 + dW22 )) 4µ2 T r S (2) dt . + 15 16µ4 µ2 (1) 2 (2) + [14(T r S ) − 24 T r S ] dt hd T r S i = 35 15 2µ4 32µ2 (2) (1) 2 (2) (2) 2 hd T r S i = (3(T r S ) − 4 T r S ) + (T r S ) dt 35 15 8µ4 16µ2 (1) (2) (1) (1) (2) hd T r S , dT r S i = Tr S + Tr S Tr S dt . 35 15 (1) Proof. Recall we have selected an orthomal basis {e1 , e2 , e3 } for Txt Mt such that ν = e3 . Then with α = u ∧ v, from LeJan (1986) we have (2.15) dkαk−2 2µ2 = −2(dW11 + dW22 ) − dt . −2 kαk 15 Using (1.6), it follows that (2.16) dhu, vi = hdW u, vi + hu, dW vi + 10µ2 hu, vidt . 15 Now by (1.9), the mean curvature is (2.17) T r S (1) = [S(u, u)kvk2 + S(v, v)kuk2 − 2S(u, v)hu, vi]kαk−2 14 M. CRANSTON AND Y. LEJAN so by Itô’s formula, d T r S (1) = [kvk2 dS(u, u) + kuk2 dS(v, v) − 2hu, vidS(u, v) + S(u, u)dkvk2 + S(v, v)dkuk2 − 2S(u, v)dhu, vi + hdS(u, u), dkvk2 i + hdS(v, v), dkuk2 i − 2hdS(u, v), dh(u, v)i]kαk−2 + T r S (1) dkαk−2 + hd(T r S (1) kαk2 ), dkαk−2 i kαk−2 = [hdB(u, u), νikvk2 + hdB(v, v), νikuk2 − 2hdB(u, v), νihu, vi]kαk−2 2µ2 T r S (1) dt + T r S (1) dW33 − 15 + [S(u, u)hdW v, vi + S(v, v)hdW u, ui − S(u, v)(hdW u, vi + hu, dW vi)]kαk−2 10µ2 2µ2 + T r S (1) dt − T r S (1) dt 15 15 2µ2 − 2 T r S (1) (dW11 + dW22 ) − T r S (1) dt , 15 using Theorem 2.7, (2.15), (2.16) and the fact that hd(T r S (1) kαk2 ), dkαk−2 i = −2kαk−2 [hkαk2 T r S (1) dW33 , dW11 + dW22 i + 2hS(u, u)hdW v, vi + S(v, v)hdW u, ui − S(u, v)(hdW u, vi + hu, dW vi), dW11 + dW22 i] h −2µ i 2µ2 2 = −2 + T r S (1) dt 15 15 =0. Thus, (2.18) d T r S (1) = [hdB(u, u), νikvk2 + hdB(v, v), νikuk2 − 2hdB(u, v), νi hu, vi]kαk−2 + T r S (1)(dW33 − 2(dW11 + dW22 )) + 2[S(u, u)hdW v, vi + S(v, v)hdW u, ui − S(u, v)(hdW u, vi + hu, dW vi)]kαk−2 4µ2 + T r S (1)dt . 15 Consequently, the quadratic variation of T r S (1) is given by (2.19) hd T r S (1) i = [ hdB(u, u), νikvk2 + hdB(v, v), νikuk2 − 2hdB(u, v), νi hu, vi ]kαk−4 + (T r S (1) )2 hdW33 − 2(dW11 + dW22 )i + 4[S(u, u)hdW v, vi + S(v, v)hdW u, ui − S(u, v)(hdW u, vi + hu, dW vi)]kαk−4 + 4T r S (1) hdW33 − 2(dW11 + dW22 ), S(u, u)hdW v, vi + S(v, v)hdW u, ui − S(u, v)(hdW u, vi + hu, dW vi)ikαk−2 Setting qn = (n + 3)µ4 nn = C1122 n(n + 2) (n + 4) GEOMETRIC EVOLUTION UNDER ISOTROPIC STOCHASTIC FLOW and if ui , uj , uk , u` are all orthogonal to ν, then (2.20) hdB(ui , uj ), νi hdB(uk , u` ), νi = qn [hui , uj i huk , u` i + hui , uk i huj , u`i + hui , u` i huj , uk i]dt . Then (2.21) h hdB(u, u), νikvk2 + hdB(v, v), νikuk2 − 2hdB(u, v), νi hu, viikαk−4 = q3 [3kuk4kvk4 + 3kvk4 kuk4 + 4(2hu, vi2 + kuk2 kvk2 )hu, vi2 + 2(kuk2 kvk2 + 2hu, vi2 )kuk2 kvk2 − 4(3kuk2 hu, vi)kvk2 hu, vi − 4(3kvk2 hu, vi)kuk2 hu, vk]kαk−4 dt = q3 [8kuk4 − 16kuk2kvk2 hu, vi2 + 8hu, vi4 ]kαk−4 dt = 8q3 dt . Next, observe hdW33 − 2(dW11 + dW22 )i = hdW33 i − 4hW33 , dW11 + dW22 i + 4hdW11 + dW22 i 18µ2 = dt 15 and hhdW v, vii = hhdW v, vi hdW u, uii = hhdW u, vi, hu, dW vii = hhdW u, vi, hdW u, vii = hhdW u, vi, hdW v, vii = hhdW v, ui, hdW u, uii = hdW v, ui hv, vi = 2µ2 kvk4 dt 15 µ2 (3hu, vi2 − kuk2 kvk2 )dt 15 µ2 (3hu, vi2 − kvk2 )dt 15 µ2 (4kuk2 kvk2 − 2hu, vi2 )dt 15 2µ2 hu, vikvk2 dt 15 2µ2 hu, vikuk2 dt 15 2µ2 hu, vikvk2 dt . 15 15 16 M. CRANSTON AND Y. LEJAN Thus, [S(u, u)hdW v, vi + S(v, v)hdW u, ui − S(u, v)(hdW u, vi + hu, dW vi)] µ2 = [2S(u, u)2 kvk4 + 2S(v, v)2 kuk4 15 + 2S(u, v)2 (3kuk2 kvk2 + hu, vi2 ) + 2S(u, u)S(v, v)(3hu, vi2 − kuk2 kvk2 ) − 2S(u, u)S(u, v)(1hu, vikuk2 + 2hu, vikuk2 )kvk2 − 2S(v, v)S(u, v)(2hu, vikuk2 + 2hu, vikuk2 )kuk2 ]dt µ2 = [2{S(u, u)2 kvk4 + S(v, v)2 kuk4 + 2S(u, u)S(v, v)kuk2 kvk2 15 − 4S(u, u)S(u, v)kuk2 kvk2 hu, vi − 4S(v, v)S(u, v)kuk2 kvk2 hu, vi 4S(u, v)2 hu, vi2 } − 6(S(u, u)S(v, v) − S(u, v)2 )kαk2 ]dt µ2 = [2(T r S (1) )2 − 6T r S (2) ]kαk4 dt 15 It’s easy to verify that µ2 kvk2 dt 15 µ2 hdW33 , hdW u, vii = hdW33 , hu, dW vii = − hu, vidt 15 µ2 1 2 hdW1 + dW2 , hdW u, vii = hu, vidt 15 hdW33 , hdW v, vii = − so hdW33 , S(u, u)hdW v, vi + S(v, v)hdW u, ui − S(u, v)(hdW u, vi + hu, dW vi)i µ2 = − [S(u, u)kvk2 + S(v, v)kuk2 − 2S(u, v)hu, vi]dt 15 µ2 = − T r S (1)kαk2 dt , 15 1 hdW1 + dW22 , S(u, u)hdW v, vi + S(v, v)hdW u, ui − S(u, v)(hdW u, vi + hu, dW vi)i µ2 [S(u, u)kvk2 + S(v, v)kuk2 − 2S(u, v)hu, vi]dt = 15 µ2 = T r S (1)kαk2 dt . 15 Therefore 4 T r S (1)kαk−2 hdW33 − 2(dW11 + dW22 ), S(u, u)hdW v, vi + S(v, v)hdW u, ui (2.23) − S(u, v)(hdW u, vi + hu, dW vi)i µ2 2µ2 (1) 2 = 4(T r S ) − − dt 15 15 12µ2 =− (T r S (1))2 dt . 15 GEOMETRIC EVOLUTION UNDER ISOTROPIC STOCHASTIC FLOW 17 Combining (2.20), (2.21) and (2.22) we arrive at (2.24) hd T rS (1) i = µ2 16µ4 dt + [14(T r S (1))2 − 24 T r S (2)]dt . 35 15 Turning now to the Gauss curvature, by (1.9) we have, (2.25) T r S (2) = [S(u, u)S(v, v) − S(u, v)2 ]kαk−2 . So, by Itô’s formula, d T r S (2) = (S(v, v)dS(u, u) + S(u, u)dS(v, v) − 2S(u, v)dS(u, v) + hdS(u, u), dS(v, v)i − hdS(u, v), dS(u, v)i]kαk−2 + [S(u, u)S(v, v) − S(u, v)2 ]dkαk−2 + hS(v, v)dS(u, u) + S(u, u)dS(v, v) − 2S(u, v)dS(u, v), dkαk−2 i = [S(v, v)hdB(u, u), νi + S(u, u)hdB(v, v), νi − 2S(u, v)hdB(u, v), νi]kαk−2 4µ2 + 2T r S (2)dW33 − T r S (2)dt 15 + [hhdB(u, u), νi, hdB(v, v), νii − hdB(u, v), νi, hdB(u, v), νii]kαk−2 dt 2µ2 + T r S (2)dt 15 2µ2 T r S (2)dt − 2 T r S (2)(dW11 + dW22 ) − 15 8µ2 + T r S (2)dt . 15 Thus, dT r S (2) = [S(v, v)hdB(u, u), νi + S(u, u)hdB(v, v), νi (2.26) − 2S(u, v)hdB(u, v), νi]kαk−2 4µ2 T r S (2)dt , 15 since hhdB(u, u), νi, hdB(v, v), νii = h(dB(u, v), νi, hdB(u, v), νii . + 2T r S (2)(dW33 − (dW11 + dW22 )) + 18 M. CRANSTON AND Y. LEJAN Using (2.20), (2.26) and our remarks above on W correlations, we get (2.27) hd T r S (2)i = hS(v, v)hdB(u, u), νi + S(u, u)hdB(v, v), νi − 2S(u, v)hdB(u, v), νiikαk−4 + 4(T r S (2))2 hdW 3 − (dW11 + dW22 )i = q3 [3S(v, v)2 kuk4 + 3S(u, u)2kvk4 + 4S(u, v)2 (kuk2 kvk2 + 2hu, vi2 ) + 2S(u, u)S(v, v)(kuk2kvk2 + 2hu, vi2 ) − 12S(v, v)S(u, v)kuk2hu, vi − 12S(u, u)S(u, v)kvk2hu, vi]kαk−2 dt 32µ2 + (T r S (2))2 dt 15 32µ2 = 3q3 (T r S (1))2 dt − 4q3 T r S (2)dt + (T r S (2))2 dt 15 32µ2 2µ4 (3(T r S (1))2 − 4 T r S (2))dt + (T r S (2))2 dt . = 35 15 Finally we need (2.28) hdT r S (1) , dT r S (2) i = 2T r S (1) T r S (2) hdW33 − 2(dW11 + dW22 ), dW33 − (dW11 + dW22 )i + 4T r S (2) hS(u, u)hdW v, v) + S(v, v)hW u, ui − S(u, v)(hdW u, vi + hu, dW vi), dW33 − (dW11 + dW22 )ikαk−2 + kαk−4 hhdB(u, u), νikvk2 + hdB(v, v), νikuk2 − 2hdB(u, v), νihu, vi , S(v, v)hdB(u, u), νi + S(u, u)hdB(v, v), νi − 2S(u, v)hdB(u, v), νii 24µ2 8µ2 = T r S (1) T r S (2) dt − T r S (1) T r S (2) dt + 4q3 T r S (1) dt 15 15 16µ2 8µ4 = T r S (1) T r S (2) dt + T r S (1) dt . 15 35 This completes the proof. Theorem 2.9. The process (T r St(1), T r St(2)) is a recurrent diffusion with generator 1 µ2 16µ4 ∂ 2 f 2 Lf(κ, m) = (14κ − 24m) + (κ, m) 2 15 35 ∂κ2 2 16µ2 8µ4 ∂ f κm + κ (κ, m) + 15 35 ∂κ ∂m 2 1 32µ2 2 2µ4 ∂ f 2 + m + (3κ − 4m) (κ, m) 2 15 35 ∂m2 4µ2 ∂f 4µ2 ∂f + κ (κ, m) + m (κ, m) . 15 ∂κ 15 ∂m This diffusion never enters the region κ2 ≤ 4m at strictly positive times a.s. . GEOMETRIC EVOLUTION UNDER ISOTROPIC STOCHASTIC FLOW 19 Remarks. (1) In n = 3, the three Lyapunov exponents are (see LeJan 1986) γ1 = 3, γ2 = 0, γ3 = −3. Thus there is always a stretching direction, a neutral direction and a compressing direction. A curve will always “see” the stretching direction. This roughly explains the positive recurrence of the curvature of a curve proved in LeJan 1986. In the case of a surface, the tangent plane will “see” the stretching and the neutral direction. This also explains recurrence. The authors plan a complete account of the situation in higher dimensions in a forthcoming paper. (2) Since X = T r S (1) = λ1 +λ2 and Y = T r S (2) = λ1 λ2 , it follows that (X, Y ) stays in the region κ2 ≥ 4m. The boundary of this region corresponds to λ1 = λ2 . That is κ2 = 4m when the surface in question has an umbilic (i.e. a point where λ1 = λ2 ). Thus, x may be an umbilic for M but xt , t > 0, will never be an umbilic for Mt . One easily checks that the matrix a appearing in the generator for this diffusion (2nd order derivative coefficients) degenerates on the curve κ2 − 4m = 0. Moreover, the two eigenvectors of this matrix are, respectively, perpendicular to and tangent to the curve κ2 − 4m = 0. The eigenvector perpendicular to this curve has eigenvalue zero. Notice the drift vector 4µ2 4µ2 2 15 κ , 15 m points away from the region κ − 4m < 0. Thus, when the point x on the initial manifold M is an umbilic point, the diffusion (Xt , Yt ) has diffusion component only tangential the curve κ2 = 4m and drifts away from this curve never to return. (3) Neither T r S (1) nor T r S (2) is a diffusion by itself. (4) In the case n = 3 one can assert that (Xt , Yt ) does not spend a lot of time in the region λ = {(κ, m) : κ > M1 , m > M2 } with M1 and M2 large positive constants. Recall the following results of LeJan (1986): draw a curve γ on the initial manifold M with γ(0) = x. Set γt = Φt (γ). Then the curvature τt of γt at γt (0) is a positive recurrent diffusion. Thus, the average amount of time that τt spends above a large positive constant is small. Since it is impossible to draw a curve on Mt with zero curvature (or even small curvature) when (Xt , Yt ) ∈ Λ, (Xt , Yt ) can not stay for long in Λ. More precise statements can be made by referring to the article of LeJan where the exact form of the invariant measure for τt is given. Proof (of Theorem 2.9). The claim follows immediately √ that (Xt , Yt ) is a diffusion µ2 2 from Theorem 2.8. Define h = κ − 4m. Then with a = 15 , b = µ354 , in (m, h)coordinates ∂2f ∂2f + 14aκh ∂κ2 ∂κ∂h 2 ∂ f + (3aκ2 + 4ah2 + 4b) 2 ∂h ∂f ∂f + 4aκ + h−1 (3aκ2 + ah2 + 4b) . ∂κ ∂h Lf(κ, h) = (4aκ2 + 3ah2 + 8b) 20 M. CRANSTON AND Y. LEJAN We prove the claim regarding lack of umbilics first as it eases a technical point which might arise in the proof of the recurrence. For this, it suffices to show that Ht does not hit zero a.s. where (Xt , Ht ) is diffusion given in (κ, h) coordinates. Rt First define ηt = 3aXt2 + 4aHt2 + 4b and then τt = 0 dx ηs . Then Z τt Hτt = h + bt + Z Hs−1 (3aXs2 + aHs2 + 4b)ds 0 t = h + bt + 0 Hτ−1 (3aXτ2s + aHτ2s + 4b)ητ−1 ds s s with bt a one-dimensional Brownian motion. Fix now an ∈ (0, 1/6) and suppose 0 < h < 1/6 as well. Set ρt = h + bt + q If σ = inf{t > 0 : Hτt > 2 2 ( 12 −)b (1+4)a }, Z t 1 ds + . 2 0 ρs then since for h ≤ q ( 12 −)b (1+4)a , one has 2 3aκ +ah +4b ( 12 + ) < 3aκ 2 +4ah2 +4b for all κ. By an elementary comparison theorem (see IkedaWatanabe, 1989), it follows that ρt ≤ Hτt for t < σ . Since ρt can not hit zero (it is a Bessel process above the critical dimension), Ht can’t hit zero either. For recurrence, we shall show (Xt , Ht ) is recurrent which will suffice. f(κ, h) = `n(κ6 + h6 ). Then 6κ5 6h5 , f = h κ6 + h6 κ6 + h6 30κ4 h6 − 6κ10 36κ5 h5 = , f = − ; κh (κ6 + h6 )2 (κ6 + h6 )2 fκ = fκκ fhh = 30κ6 h4 − 6h10 (κ6 + h6 )2 and Lf = 6(κ6 + h6 )−2 [(4aκ2 + 3ah2 + 8b)(5κ4 h6 − κ10 ) − 84aκ6 h6 + (3aκ2 + 4ah2 + 4b)(5κ6 h4 − h10) + 4aκ12 + 4aκ6 h6 + h4 (3aκ2 + ah2 + 4b)(κ6 + h6 )] = 6(κ6 + h6 )−2 [−a(3κ10 h2 − 18κ8 h4 + 39κ6 h6 − 15κ4h8 + 3h12 ) − b(8κ10 − 24κ6 h4 − 40κ4 h6 )] κ = uh = −6(κ6 + h6 )−2 h10 [3ah2 (u10 − 6u8 + 13u6 − 5u4 + 1) + 8bu4 (u6 − 3u2 − 5)] The polynomial P (u) = u10 − 6u8 + 13u6 − 5u4 + 1 > 0 for all u. Set GEOMETRIC EVOLUTION UNDER ISOTROPIC STOCHASTIC FLOW 21 Also, there is a u0 > 0 such that Q(u) = u6 − 3u2 − 5 ≥ 0 whenever |u| ≥ u0. Thus, Lf = −6(κ6 + h6 )−2 h10 [3ah2 P (u) + 8bu4 Q(u)] has {Lf ≤ 0} ⊇ {(κ, h) : |κ| ≥ u0 h} . Also, denote M = supu Q(u) P (u) , one sees ( {Lf ≤ 0} ⊇ r (κ, h) : h ≥ 86M 3a ) Consequently, off of the compact triangle ( T = {(κ, h) : |κ| ≤ u0 h} ∩ (κ, h) : h ≤ r 86M 3a ) we have Lf ≤ 0. Now, define C(r) = {(κ, h) : h > 0, κ6 + h6 < r} and select r0 large enough so that C(r0 ) ⊃ T . Put for R > r0 `n(κ6 + h6 ) − `nR . `nr0 − `nR Then on C(R) \ C(r0 ), LuR ≥ 0 and uR ∂C(r0 )∩{h>0} = 1, uR ∂C(R)∩{h>0} = 0. Thus, if τr = inf{t > 0 : (Xt , Ht ) ∈ ∂C(r)} (recall, Ht never hits 0), then as uR is L-subharmonic and has the same boundary values as P(κ,w) (τr0 < TR ) uR (κ, h) = uR (κ, h) ≤ P(κ,h) (τr0 < τR ) . But limR↑∞ uR (κ, h) = 1 which implies the recurrence of (Xt , Ht ) and therefore the recurrence of (Xt , Yt ). 22 M. CRANSTON AND Y. LEJAN References P. Baxendale, (1984), Brownian motions in the diffeomorphism Group. I, Compositio Math. 53, 19–50. P. Baxendale and T.Harris, (1986), Isotropic stochastic flows, Annals of Prob., Vol 14, 1155–1179. R. Bishop and S. Goldberg, (1980), Tensor Analysis on Manifolds, Dover, Toronto. Hardy, Littlewood, Polya, (1973), Inequalities, Cambridge University Press, Cambridge. T. Harris, (1981), Brownian motions on the homeomorphisms of the plane, Annals of Prob. Vol. 9, 232–254. N. Ikeda, S. Watanabe, (1989), Stochastic Differential Equations and Diffusion Processes, North-Holland, New York. Y. LeJan, (1985), On isotropic Brownian motions, Zeit. für Wahr., 70, 609–620. Y. LeJan, (1991), Asymototic properties of isotropic Brownian flows, in Spatial Stochastic Processes, ed. K. Alexander, J. Watkins, Birkhäuser, Boston, 219–232. Y. LeJan, S. Watanabe (1984), Stochastic flows of diffeomorphisms, Proc. of the Taniguchi Symposium, 307–332, North Holland, New York. M. Spivak, (1979), A comprehensive introduction to differential geometry, Publish or Perish, Berkeley. GEOMETRIC EVOLUTION UNDER ISOTROPIC STOCHASTIC FLOW 23 Addendum. Higher dimensional hypersurfaces. ~ ~ ∈ Ik . Define Recall α~` = u`1 ∧ · · · ∧ u`k , αm ~ = um1 ∧ · · · ∧ umk , for `, m α~` = (−1)p+1 u`1 ∧ · · · ∧ û`p ∧ · · · ∧ u`k , i+1 αm um1 ∧ · · · ∧ ûmi ∧ · · · ∧ umk . ~ i = (−1) p (p) (i) Then α~`p = (−1)p+1 u`1 ∧ · · · ∧ up` (i) (i) (i) and um , and set S (0)(·, ·) ≡ 1. Theorem A.1. dS (k) k X (α~`, αm ~)= S (k−1)(α~`p , αm ~ i )hdB(u`p , umi ), νi i,p=1 n + kS (k)(α~`, αm ~ )dWn − k(n − k)(n − 1)µ2 (k) S (α~`, αm ~ )dt , 2n(n + 2) or, in a more synthetic form (k) dS(α) = S (k−1) ∧ dB ν (α) + kS (k)(α)hdhW ν, νi − −k(n − k)(n − 1) (k) S (α) , 2n(n + 2) with hB ν u, vi = hB(u, v), νi. Proof. By Theorem 2.7 and Itô’s formula, dS (k) (α~`, αm ~) = X (−1)σ k k Y X [ hS(u`σ(j) , umj ), νi]hdB(u`σ(i) , umi ), νi i=1 j=1 j6=i σ∈Sk n + kS (k) (α~`, αm ~ )dWn − + 1 2 X X (−1)σ σ∈Sk 1≤i6=p≤k [ k Y k(n − 1)(n − 1)µ2 (k) S (α~`, αm ~ )dt 2n(n + 2) hS(u`σ(j) , umj ), νi][dB n (u`σ(i) , umi ), dB n (u`σ(p) , ump )i j=1 j6={i,p} +hdB n (u`σ(i) , ump ),dB n (u`σ(p) , umi )i + hdB n (u`σ(i) , u`σ(p) ), dB n (umi , ump )i] + k(k − 1)(n − 1)µ2 (k) S (α~`, αm ~ )dt. 2n(n + 2) The first term may be written k X X (−1)σ [ = [ X p+i (−1) k Y hS(u`σ(j) , umj ), νi]hdB(u`p , umi ), νi j=1 j6=i i,p=1 σ∈Sk σ(i)=p k X σ (−1) [ k Y (p) hS(u` σ(j) (i) , umj ), νi]]hdB(u`p , umi ), νi j=1 j6=i i,p=1 σ=Sk−1 = k X S (k−1) (α~` , αm ~ i )hdB(u`p , umi ), νi. p i,p=1 (p) p+1 αm um1 ∧ · · · ∧ umk−1 defines u` ~ i = (−1) k−1 24 M. CRANSTON AND Y. LEJAN The third term combines with the fifth term to give − k(n − k)(n − 1)µ2 (k) S (α~`, αm ~ )dt. 2n(n + 2) Finally, in the fourth term, notice that given σ ∈ Sk and i, p ∈ {1, . . . , k} there is exactly one other η ∈ Sk such that both {σ(i), σ(p)} = {η(i), η(p)} and σ(j) = η(j) for j ∈ / {i, p}. Furthermore, (−1)σ = −(−1)η , and [ k Y hS(u`σ(j) , umj ), νi][hB n (u`σ(i) , umi ), dB n (u`σ(p) , ump )i j=1 j ∈{i,p} / +hdB n (u`σ(i) , ump ), dB n (u`σ(p) , umi )i + hdB n (u`σ(i) , u`σ(p) ), dB n (umi , ump )i] =[ k Y hS(u`η(j) , umj ), νi][hdB n (u`η(i) , umi ), dB n (u`η(p) , ump )i j=1 j ∈{i,p} / +hdB n (u`η(i) , ump ), dB n (u`η(p) , umi )i + hdB n (u`η(i) , u`η(p) ), dB n (umi , ump )i] Thus exchanging the sums in the fourth term reveals the cancellation of terms from these σ − η pairs and so the whole term vanishes. This finishes the proof. From LeJan (1985) we recall the notation τ`j β = ej ∧ i(e` )β where i(e` )β = Σ(−1)k+1 hβk , e` iβ1 ∧ · · · ∧ β̂k ∧ · · · ∧ βm . The situation for general n is given by the following Theorem which we prove in the Addendum. Theorem A.2. For 1 ≤ k ≤ n − 1, dT rS (k) k X =[ ~ m ~ −2 ` S (k−1) (α~` , αm ~ i )hdB(u`p , umi ), νi]hα , α i||α|| p i,p=1 + T rS (k) [kdWnn −2 n−1 X dWii ] i=1 ~ j m ~ ` + S (k) (α~`, αm ~ )(hτi α , α i − ~ ~ + hα` , τij αm i)dWji ||α||−2 (n + 1)k(n − k)µ2 T rS (k) dt , or 2n(n + 2) d T rS (k) = dhS (k)α, αi kαk2 hdW α, αi = hS ∧ dB (α) + hdW ν, νi − 2 S (k)(α) kαk2 (n + 1)k(n − k)µ2 (k) + S (k) ∧ dW (α), αi kαk2 − hS (α), αi kαk2 . 2n(n + 2) (k−1) ν GEOMETRIC EVOLUTION UNDER ISOTROPIC STOCHASTIC FLOW Proof. For ~` ∈ Ik , ~ ~ + (k + 1)(n − 1 − k)µ2 ~` m hα , α ~ idt n n−1 X d||α||−2 (n − 2)(n − 1)µ2 = −2 dWii − dt −2 ||α|| n(n + 2) i=1 (A.2) (A.3) ~ ~ ~ ~ dhα` , αm i = (hτij α` , αm i + hα` , τij αm i)dWji (A.1) d||α||−2 = −2||α||−2dWii + At , Since At has bounded variation ~ ` m ~ −2 Tr S (k) = S (k)(α~`, αm ~ )hα , α i||α|| by Itô’s formula, ~ ` m ~ −2 dT rS (k) = (dS (k)(α~`, αm ~ ))hα , α i||α|| ~ ` m ~ −2 + S (k)(α~`, αm ~ )(dhα , α i)||α|| + T rS (k) d||α||−2 ||α||−2 ~ ` m ~ −2 + hdS (k) (α~`, αm ~ ), dhα , α ii||α|| ~ −2 ~ + hdS (k) (α~`, αm ihα` , αm i ~ ), d||α|| ~ ` m ~ −2 + S (k)(α~`, αm i ~ )hdhα , α i, d||α|| Proceeding in order through the terms, we have using Theorem A.1 (A.4) ~ ` m ~ −2 dS (k)(α~`, αm ~ )hα , α i||α|| =[ k X ~ ` m ~ −2 S (k−1) (α~`p , αm ~ i )hdB(u`p , umi ), νi]hα , α i||α|| i,p=1 + kT rS (k) dWnn − k(2n − k)(n − 1)µ2 T rS (k) dt. 2n(n + 2) From (A.1) follows, (A.5) ~ ` m ~ −2 S (k)(α~`, αm ~ )dhα , α i||α|| ~ ~ j ` j m ` m ~ ~ i −2 = S (k)(α~`, αm ~ )(hτi α , α i + hα , τi α i)dWj ||α|| + (k + 1)(n − k − 1)µ2 T rS (k) dt. n 25 26 M. CRANSTON AND Y. LEJAN By (A.2) (A.6) d||α||−2 T rS (k) ||α||−2 = −2T rS (k) n−1 X dWii − i=1 (n − 2)(n − 1)µ2 T rS (k) dt. n(n + 2) Combining Theorem A.1 and (A.1), ~ m ~ −2 ` hdS (k) (α~`, αm ~ ), dhα , α ii||α|| ~ ~ j ` j m m ~ ~ i n ` = kS (k) (α~`, αm ~ )(hτi α , α i + hα , τi α i)hdWj , dWn i kµ2 ~ j ~ j m m ~ ~ i j i n i j ` ` S (k) (α~`, αm ~ )(hτi α , α i + hα , τi α i)[(n + 1)δn δn − δj δn − δn δn ]dt n(n + 2) kµ2 ~ i ~ m ~ i m ~ ` ` =− S (k) (α~`, αm ~ )(hτi α , α i + hα , τi α i)dt n(n + 2) 2k(n − k − 1)µ2 T rS (k) dt. =− n(n + 2) = By Theorem A.1 and (A.3) ~ −2 ~ hdS (k) (α~`, αm ihα` , αm i ~ ), d||α|| (A.7) ` n i m ~ −2 = −2kS (k) (α~`, αm ~ )hdWn , dWi ihα , α i||α|| ~ = 2k(n − 1)µ2 T rS (k) dt. n(n + 2) From (A.1) and (A.3) we get (A.8) ~ m ~ −2 ` S (k) (α~`, αm i ~ )hdhα , α i, d||α|| ~ ~ j ` j m p m ~ ~ i −2 ` = −2S (k) (α~`, αm ~ )(hτi α , α i + hα , τi α i)hdWp , dWj i||α|| 2µ2 ~ i ~ m ~ i m ~ −2 ` ` = S (k) (α~`, αm dt ~ )(hτi α , α i + hα , τi α i)||α|| n(n + 2) 4(n − k − 1)µ2 =− T rS (k) dt n(n + 2) Summing (A.4)–(A.8) shows dTr S (k) =[ k X ~ i (−1)i+p+1 S (k−1) (α~p , αm )hdB(u`p , umi ), νi]hα` , αm i||α||−2 ~ ` ~ m ` i,p=1 + T rS (k) [kdWnn − 2dWii ] ~ ~ m ~ m ~ i −2 ` ` + S (k) (α~`, αm ~ )(hτi α , α i + hα , τi α i)dWj · kαk j + k(n − k)(n + 1)µ2 T rS (k) dt 2n(n + 2) and the proof is complete. j GEOMETRIC EVOLUTION UNDER ISOTROPIC STOCHASTIC FLOW 27 Theorem A.3. For n − 1 ≥ k ≥ ` ≥ 1, hdT rS (k) , dT rS (`) i = µ2 [(k` + 2(k + `) + 4)(n − 1) n(n + 2) − 2(k + 2)(n − ` − 1) − 2(` + 2)(n − k − 1) − 4(n − ` − 1)(n − k − 1)]T rS (k) T rS (`) dt X (n−k−1)∧` k + (1 − δn−1 ) 4µ2 (n + 2) d(n, (k, `); j)T rS (k+j) T rS (`−j) dt j=0 X (n−k)∧(`−1) nn + C1122 a(n, (k, `); j)T rS (k+j−1) T rS (`−j−1) dt j=0 where d(n, (k, `); 0) = (n − 1 − k) k − ` + 2j X k − ` + 2j d(n, (k, `); j) = (n − 1 − k − j) − d(n, (k, `); r) ,j ≥ 1 j j −r j−1 r=0 b(n, (k, `); 0) = (n − k)(n − ` + 2) k − ` + 2j k − ` + 2j − 2 b(n, (k, `); j) = 3(n − k − j) + (k − ` + 2j)(k − ` + 2j − 1) j j−1 k − ` + 2j − 1 k − ` + 2j − 1 + (k − ` + 2j)(n − k − j)[ + ] j j −1 k − ` + 2j + (n − k − j)(n − k − j − 1) ,j ≥ 1 j a(n, (k, `); 0) = b(n, (k, `); 0) a(n, (k, `); j) = b(n, (k, `); j) − j−1 X k − ` + 2j j−m a(n, (k, `); m) m=0 nn C1122 (n2 − 1) = (n + 3)µ4 . n(n + 2)(n + 4) Proof. We compute the quadratic variation terms in hdT rS (k) , dT rS (`)i, k ≥ `, which do not arise from dB. These are hT rS (k) [kdWnn − 2 n−1 X ~ ~ m ~ m ~ −2 ` ` dWii ] + S (k) (α~`, αm dWjh , ~ )(hτh α , α i + hα , τh α i)||α|| j j i=1 X n−1 T rS (`) [`dWnn − 2 m dWm ] + S (`) (α~r , α~s )(hτq α~r , α~s i + hα~r , τq α~s i)||α||−2 dWp i p p q m=1 unless k = n − 1 or both ` = k = n − 1 in which case, the term S (k)(α~`, αm ~) ~ j ~ ` ` j m m ~ ~ −2 i (hτi α , α i + hα , τi α i)||α|| dWj = 0 or both this and the corresponding S (`) term vanishes. Thus we get T rS (k) T rS (`) hkdWnn − 2 n−1 X i=1 X n−1 dWii , `dWnn − 2 j=1 dWjj i 28 M. CRANSTON AND Y. LEJAN ` +T rS (k) S (`) (α~r , α~s )(hτqp α~r , α~s i + hα~r , τqp α~s i)||α||−2 (1 − δn−1 ) hkdWnn − 2 n−1 X dWii , dWpq i i=1 +T rS (`) S (k) h`dWnn ~ ~ m ~ m ~ −2 k ` ` (α~`, αm (1 − δn−1 ) ~ )(hτh α , α i + hα , τh α i)||α|| −2 j n−1 X j dWii , dWjh i i=1 +S (k) ~ ~ (`) ~ ~ (α~`, αm (α~r , α~s )(hτhj α` , αm i + hα` , τhj , αm i)(hτqp α~r , α~s i ~ )S ` + hα~r , τqp α~s i)(1 − δn−1 ) k · (1 − δn−1 )||α||−4 hdWjh , dWpq i = µ2 T rS (k) T rS (`) [k · `(n − 1) + 2(k + `)(n − 1) + 4(n − 1)]dt n(n + 2) X ip µ2 np ` + T rS (k) S (`) (α~r , α~s )(hτpp α~r , α~s i + hα~r , τpp α~s i)||α||−2 (1 − δn−1 )[kCnp −2 Cip ]dt n(n + 2) n−1 i=1 X ij µ2 ~ j ~ j m nj m ~ ~ −2 k ` ` + T rS (`) S (k) (α~`, αm (1 − δn−1 )[`Cnj −2 Cij ]dt ~ )(hτj α , α i + hα , τj α i)||α|| n(n + 2) n−1 i=1 µ2 ~ ~ (k) ~ ~ S (k) (α~`, αm (α~r , α~s )(hτhj α` , αm i + hα` , τhj αm i)(hτqp α~r , α~s i + ~ )S n(n + 2) k + hα~r , τqp α~s i)(1 − δn−1 )||α||−4 [(n + 1)δqh δpj − δjh δpq − δph δjq ]dt We now proceed singly through the last three terms. Using the invariance of our expressions under a change of basis we may take the ui to be the unit principal curvature directions. Then (A.9) ~ ` S (k) (α~`, αm ~ ) = λ~ ~ ` δm S (`) (α~r , α~s ) = λ~r δ~s~r hτqp α~r , α~s iδ~s~r = δqp δ~s~r if and only if p ∈ ~r , = 0 ~ ~ ` hτhj α` , αm iδm ~ = j ~ ` δh δm ~ if and only if j ∈ ~ `,= 0 otherwise, otherwise. Thus, (A.10) X ip µ2 np ` T rS (k) S (`) (α~r , α~s )(hτpp α~r , α~s i + hα~r , τpp α~s i)(1 − δn−1 )||α||−2 [kCnp −2 Cip ]dt n(n + 2) n−1 i=1 µ2 p ` n p = T rS (k) λ~r hτp α~r , α~r i(1 − δn−1 )[k(nδpn − δn δp ) − 2]dt n(n + 2) 2(k + 2)(n − ` − 1)µ2 =− T rS (k) T rS (`) dt. n(n + 2) Similarly, (A.11) X ij µ2 ~ j ~ j m nj m ~ ~ −2 k ` ` T rS (`) S (k) (α~`, αm (1 − δn−1 )(`Cnj −2 Cij )dt ~ )(hτj α , α i + hα , τj α i)||α|| n(n + 2) n−1 i=1 −2(` + 2)(n − k − 1)µ2 = T rS (k) T rS (`) dt. n(n + 2) GEOMETRIC EVOLUTION UNDER ISOTROPIC STOCHASTIC FLOW 29 Finally, the last term is (A.12) 4µ2 ~ ~ ` λ~λ~r hτhj α` , α` ihτqp α~r , α~r i(1 − δn−1 )[(n + 1)δqh δpj − δjh δpq − δph δjq ]dt n(n + 2) ` 4µ2 ~ ~ ~ ~ k =(1 − δn−1 ) λ~`λ~r [(n + 1)hτhj α` , α` ihτhj α~r , α~r i − hτjj α` , α` ihτpp α~r , α~r i n(n + 2) k (1 − δn−1 ) ~ ~ −hτhj α` , α` ihτjh α~r , α~r i]dt 4µ2 ~ ~ ~ ~ k λ~λ~r [nhτjj α` , α` ihτjj α~r , α~r i − hτjj α` , α` ihτpp α~r , α~r i]dt ) =(1 − δn−1 n(n + 2) ` =(1 k − δn−1 ) 4µ2 λ~λ~r [nδj6∈~`δj6∈~r − δj6∈~`δp6∈~r ]dt n(n + 2) ` (δj6∈~` = 1, when j 6∈ ~ `, ) 0, otherwise 4nµ2 4(n − k − 1)(n − ` − 1µ2 ) k =(1 − δn−1 )( λ~λ~r · δj6∈~`δj6∈~r dt − T rS (k) T rS (`) dt). n(n + 2) ` n(n + 2) However, by summing on j we see that (A.13) λ~`λ~r δj6∈~`δj6∈~r = n−`−1 X mλ~`λ~r m=0 rc |=m |~ `c ∩~ which is a symmetric homogeneous polynomial in λ1 , . . . , λn−1 of degree k + ` so we seek an expansion for (A.13) of the form X (n−k−1)∧` (A.14) d(n, (k, `); j)T rS (k+j) T rS (`−j) . j=0 The principal term in T rS (k+j) T rS (`−j) (which doesn’t appear in T rS (k+m) T rS (`−m) for m > j) is In λ21 . . . λ2`−j λ`−j+1 . . . λk+j . Pn−`−1 r m=0 mλ~ ` λ~ |~ `∩~ r |=m this term appears (n − 1 − k − j) In T rS (k+r) T rS (`−r) for r < j this term appears P expansion for n−`−1 mλ~`λ~r m=0 k−`+2j j k−`+2j j−r times. times. Thus in the rc |=m |~ `c ∩~ d(n, (k, `); 0) = (n − 1 − k) k − ` + 2j X k − ` + 2j − d(n, (k, `); r) ,j ≥ 1. j j−r j−1 d(n, (k, `); j) = (n − 1 − k − j) r=0 Putting (A.10), (A.11) and (A.12) together results in (A.15) µ2 T rS (k) T rS (`) [(k` + 2(k + `) + 4)(n − 1) − 2(k + 2)(n − ` − 1) n(n + 2) − 2(` + 2)(n − k − 1) − 4(n − ` − 1)(n − k − 1)] [ X (n−k−1)∧` + (1 k − δn−1 ) 4µ2 n(n + 2) j=0 d(n, (k, `); j)T rS (k+j) T rS (`−j) ]dt. 30 M. CRANSTON AND Y. LEJAN The quadratic variation term in hdT rS (k) , dT rS (`)i, k ≥ `, arising from dB is, Q(n, (k, `))dt =h k X ~ m ~ ` S (k−1) (α~`p , αm ~ i )hα , α ihdB(u`p , umi ), νi, i,p=1 X̀ S (`−1) (α~rq , α~sj )hα~r , α~s ihdB(urq , usj ), νii||α||−4 j,q=1 nn = C1122 k X X̀ (`−1) S (k−1) (α~`p , αm (α~rq , α~sj )[hu`p , umi ihurq , usj i ~ i )S i,p=1 j,q=1 ~ ~ + hu`p , urq ihumi , usj i + hu`p , usj ihurq , umi i]hα` , αm ihα~r , α~s i||α||−4 dt and taking advantage of the invariance of this expression under a change of basis, we take ui to be the ith unit principal direction of curvature. Then ~ ~ m ~ ` ` i S (k−1) (α~`p , αm ~ i )hα , α i = λ~ ~ δp `p δm S `−1 (α~rq , α~sj )hα~r , α~s i = λ~rq δ~s~r δqj ~ r ` ~ [hu`p , umi ihurq , usj i + hu`p , urq ihumi , usj i + hu`p , usj ihurq , umi i]δm s ~ δ~ ` r ` ` r ~ r ` ~ = (δmpi δsjq + δrpq δsmj i + δspj δmqi )δm s. ~ δ~ Thus, Q(n, (k, `)) = nn C1122 k X̀ X ` λ~` λ~rq (1 + 2δrpq )dt p p=1 q=1 nn = C1122 (3 k X X̀ λ~` λ~rq + k X X̀ p p=1 q=1 `p =rq λ~` λ~rq ). p p=1 q=1 `p 6=rq Since Q(n, (k, `)) is a symmetric homogeneous polynomial of degree k + ` − 2 in λ1 , . . . , λn−1 , it may be written X (n−k)∧(`−1) Q(n, (k, `)) = nn C1122 a(n, (k, `); j)T rS (k+j−1) T rS (`−j−1) . j=1 We now employ some elementary counting arguments to compute the coefficients a(n, (k, `); j). In T rS (k+j−1) T rS (`−j−1) the term (A.16) λ21 . . . λ2`−j−1 λ`−j . . . λk+j−1 appears GEOMETRIC EVOLUTION UNDER ISOTROPIC STOCHASTIC FLOW 31 but no such term (with ` − j − 1 squares) appears in T rS (k+m−1)T rS (`−m−1) for m > j. However, if r < j, λ21 . . . λ2`−j−1 λ`−j . . . λk+j−1 (A.17) k−`+2j j−r appears times in T rS (k+r−1) T rS (`−r−1) . We now count the number of appearances of (A.16) in Q(n, (k, `)). X̀ 3 λ~`p λ~rq we must fill in the empty spaces in λ~r and λ~` From k X p=1 q=1 `p =rq λ~r = λ1 . . . λ`−j−1 − · · · − (A.18) λ~` = λ1 . . . λ`−j−1 − · · · −` · · · −k with the terms λ`−j , . . . , λk+j−1 , λx , λx , where x = `p = rq . Notice λt appears before λu only if t < u. One λx must go in λ~r the other in λ~`. If the order is not to be violated, this forces x > k − 1 + j and λx must end each of the terms λ~r and λ~`. There are (n − 1) − (k − 1 + j) = n − k − j such choices for x. This leaves k − 1 + j − (` − j − 1) j k − ` + 2j = j remaining possibilities for filling in the empty spaces in (A.18) in λ~r and λ~`. Note that if we interpret 00 = 1 this is still correct for k = ` and j = 0. Thus (A.19) 3 k X̀ X λ~`p λ~rq p=1 q=1 `p =rq k − ` + 2j has 3(n − k − j) j terms of the form λ21 . . . λ2`−j−1 λ`−j . . . λk+j−1 . In k X̀ X λ~`p λ~rq we must fill in the blanks of p=1 q=1 `p 6=rq λ~r = λ1 . . . λ`−j−1 − · · · − λ~` = λ1 . . . λ`−j−1 − · · · −` · · · −k with λ`−j , . . . , λk+j−1 , λx , λy with x 6= y, λx going into λ~r , λy going into λ~`. If ` − j ≤ x, y ≤ k + j − 1, then λx and λy appear on the list λ`−j , . . . , λk+j−1 . Since identical terms can’t appear twice in λ~r or λ~` the twin of λx must go in λ~`, the twin of λy must go in λ~r . There are thus k−`+2j−2 ways to fill in the spaces j−1 once λx and λy are given with ` − j ≤ x, y ≤ k + j − 1. There are (k + j − 1 − (` − 32 M. CRANSTON AND Y. LEJAN j − 1))(k + j − 1 − (` − j − 1) − 1) choices for the ordered pair (x, y). Notice that when j = 0 there are no such terms. Thus (A.20) ` − j ≤ x, y ≤ k + j − 1 accounts for k − ` + 2j − 2 (k − ` + 2j)(k − ` + 2j − 1) j−1 terms of the form λ21 . . . λ2`−j−1 λ`−j . . . λk+j−1 in Q(n, (k, `)) (interpreting 0.) z −1 = If ` − j ≤ x ≤ k + j − 1 < y ≤ n − 1, then λy must terminate thesequence λ~` k − ` + 2j − 1 and the twin of λx must go into λ~`. There are then ways to fill j in the spaces in (A.18) once such a pair (x, y) is given. We interpret this expression as 0 when k = ` and j = 0. There are (k − ` + 2j)(n − k − j) such pairs so we have (A.21) corresponding to ` − j ≤ x ≤ k +j − 1 < x ≤ n − 1 k − ` + 2j − 1 there are (k − ` + 2j)(n − k − j) j terms of the form λ21 · · · λ2`−j−1 λ`−j · · · λk+j−1 in Q(n, (k, `)) . If ` − j ≤ y ≤ k + j − 1 < x ≤ n − 1, then the twin of λy must go in λ~r . When k − ` + 2j − 1 j = 0, this is impossible. There are thus ways to fill the spaces j −1 in (A.18) once (x, y) is given and (k − ` + 2j)(n − k − j) such pairs so (A.22) corresponding to ` − j ≤ y ≤ k +j − 1 < x ≤ n− 1, k − ` + 2j − 1 there are (k − ` + 2j)(n − k − j) j−1 2 2 terms of the form λ1 · · · λ`−j−1 λ`−j · · · λk+j−1 in z Q(n, (k, `)), for j ≥ 0, (interpreting −1 = 0.) k − ` + 2j Finally, when k + j − 1 < x, y ≤ n − 1, there are ways to fill in the j spaces in (A.18) once (x, y) is given and (n − k − j)(n − k − j − 1) such pairs (x, y). Thus (A.23) corresponding to k + j −1 < x, y ≤ n− 1, there are k − ` + 2j (n−k−j)(n−k−j −1) terms of the j form λ21 · · · λ2`−j−1 λ`−j · · · λk+j−1 in Q(n, (k, `)), interpreting 00 = 1. GEOMETRIC EVOLUTION UNDER ISOTROPIC STOCHASTIC FLOW 33 Totalling (A.18)–(A.23), we see (A.24) Q(n, k) has b(n, (k, `); j) = 3(n − k − j) k − ` + 2j j k − ` + 2j − 2 + (k − ` + 2j)(k − ` + 2j − 1) j−1 k − ` + 2j − 1 + (k − ` + 2j)(n − k − j) j k − ` + 2j − 1 + (k − ` + 2j)(n − k − j) j−1 k − ` + 2j + (n − k − j)(n − k − j − 1) j z = 0, −1 terms with exactly ` − j − 1 squares. Notice that we interpret −1 0 = 0, 0 = 1. Thus, b(n, (k, `); 0) = (n − k)(n − k − 2). Upon decomposing Q(n, (k, `)) 0 into sums of expressions with exactly ` − j − 1 squares summed over j, we are lead to, using (A.16) and (A.23), X (n−k)∧(`−1) Q(n, k) = nn C1122 a(n, (k, `); j) Tr S (k+j−1) Tr S (`−j−1) j=0 where a(n, (k, `); j) = b(n, (k, `); j) − j−1 X k − ` + 2j m=0 j−m a(n, (k, `); m) , a(n, (k, `); 0) = b(n, (k, `); 0) = (n − k)(n − ` + 2) . From Theorems A.2 and A.3, we get immediately Theorem A.4. Suppose Φt is the isotropic, measure preserving flow on Rn given by (1.3). Let M be a smooth hypersurface in Rn with x ∈ M. If (λ1 (t), . . . , λn−1 (t)) are the principal curvatures of Φt (M) at Φt (x), then Tr S (k) = X λ~` ~ `∈Ik and (Tr S (1), Tr S (2), . . . , Tr S (n−1)) is an (n − 1)-dimensional diffusion. The 34 M. CRANSTON AND Y. LEJAN generator is given by X Lf(x1 , . . . , xn−1 ) = 1≤`≤k≤n−1 1 1 k + (1 − δ` ) 2 2 µ2 [((k` + 2(k + `) + 4)(n − 1) n(n + 2) − 2(k + 2)(n − ` − 1) − 2(` + 2)(n − k − 1) − 4(n − ` − 1)(n − k − 1))xk x` + 1− k δn−1 X (n−k−1)∧` 4n d(n, (k, `); j)xk+j x`−j ] j=0 X (n−k)∧(`−1) nn +C1122 j=0 + n−1 X k=1 ∂2f a(n, (k, `); j)xk+j−1 x`−j−1 ∂xk ∂x` (n + 1)k(n − k))µ2 ∂f xk 2n(n + 2) ∂xk where, for k ≥ ` d(n, (k, `); 0) = (n − 1 − k) d(n, (k, `); j) = (n − 1 − k − j) k − ` + 2j j b(n, (k, `); j) = 3(n − k − j) k − ` + 2j j − d(n, (k, `); r) k − ` + 2j − 1 j k − ` + 2j j + a(n, (k, `); j) = b(n, (k, `); j) − r=0 nn C1122 = (n + 3)µ4 , n(n + 2)(n + 4) and x−1 = x0 = 1. a(n, (k, `); r) k − ` + 2j j −r ,j > 0 ,j ≥ 1 k − ` + 2j − 2 j −1 k − ` + 2j − 1 j −1 ; j ≥ 0 , with a(n, (k, `); 0) = b(n, (k, `); 0) = (n − k)(n − ` + 2), j−1 X k − ` + 2j j−r r=0 + (n − k − j)(n − k − j − 1) j−1 X + (k − ` + 2j)(k − ` + 2j − 1) + (k − ` + 2j)(n − k − j) z −1 0 = , = 1, −1 0 0 GEOMETRIC EVOLUTION UNDER ISOTROPIC STOCHASTIC FLOW 35 Remarks. (1) Straightforward computations show that for n = 3 k = ` = 1; d(3, (1, 1); 0) = 1 d(3, (1, 1); 1) = −2 a(3, (1, 1); 0) = 8 k = 2, ` = 1; a(3, (2, 1); 0) = 4 k = 2, ` = 2; a(3, (2, 2); 0) = 3 a(3, (2, 2); 1) = −4 Thus, when M is a hypersurface in R3 , the pair ( Tr S (1), Tr S (2)) is a diffusion with generator 16µ4 ∂ 2 f 1 µ2 2 Lf(κ, m) = (14κ − 24m) + (κ, m) 2 15 35 ∂κ2 16µ2 8µ4 ∂2f + κm + κ (κ, m) 15 35 ∂κ∂m 2 1 32µ2 2 2µ4 ∂ f 2 + m + (3κ − 4m) (κ, m) 2 15 35 ∂m2 4µ2 ∂f 4µ2 ∂f κ (κ, m) + mm (κ, m) + 15 ∂κ 15 ∂m which agrees with Theorem 2.9. 44 4 (2) For n = 4, a = µ242 , b = C1122 = 7µ 192 , 1 ∂2f (a(19x2 − 32y) + 15b) 2 2 ∂x 1 ∂2 f + (a(44y2 − 32xz) + b(8x2 − 4y)) 2 2 ∂y 2 1 ∂ f + (a75z 2 + b(3y2 − 4xz)) 2 2 ∂z ∂2f ∂2f + (a(22xy − 48z) + b10x) + (a25xz + b5y) ∂x∂y ∂x∂z ∂2f + (a50yz + b(4xy − 6z)) ∂y∂z 15 ∂f ∂f 15 ∂f +a x + a10y +a z . 2 ∂x ∂y 2 ∂z Lf(x, y, z) = (3) For any n, no proper subset of {Tr S (1), . . . , Tr S (n−1)} can be a diffusion. 36 M. CRANSTON AND Y. LEJAN (4) The authors intend to resolve such issues as transience or recurrence for higher dimensional cases. (5) Since x = λ1 + λ2 + λ3 , y = λ1 λ2 + λ1 λ3 + λ2 λ3 , z = λ1 λ2 λ3 , Newton’s inequalities, Hardy-Littlewood- Polya (1973), imply x2 ≥ 3y, y2 ≥ 3xz with equality if and only if λ1 = λ2 = λ3 . As in the case n = 3, the diffusion matrix degenerates on λ1 = λ2 = λ3 and one expects that the process will not reach this set at positive times.