STABILITY OF HALF-DEGREE POINT DEFECT PROFILES FOR 2-D NEMATIC LIQUID CRYSTAL ZHIYUANGENG, WEI WANG, PINGWEN ZHANG,ANDZHIFEI ZHANG 6 1 0 Abstract. In this paper, we prove the stability of half-degree point defect profiles in R2 for the 2 nematic liquid crystal within Landau-deGennes model. n a J 2 1 1. Introduction Defects in liquid crystal are known as the places where the degree of symmetry of the nematic ] P order increases so that the molecular direction cannot be well defined. The most striking feature A of liquid crystal is a variety of visual defect patterns. Predicting the profiles of defect as well as . h stability is thus of great practical importance and theoretical interest. We mention some works t [2, 15, 22, 25] on the defects based on the topological properties of the order parameter manifolds. a m There exist three commonly used continuum theories describing the nematic liquid crystal: [ Oseen-Frank model, Ericksen model and Landau-de Gennes model. In the Oseen-Frank model, the state of nematic liquid crystals is described by a unit-vector filed which represents the mean 1 local orientation of molecules, and defects are interpreted as all singularities of this vector field v 5 [9, 10, 6, 18]. However, the core structure of defects in nematic liquid crystals, such as the disclina- 4 tion lines observed in experiments, cannot be represented by the usual director field and requires 8 description by Landau-de Gennes model [4]. In this model, the state of nematic liquid crystals is 2 0 described by a 3 3 order tensor Q belonging to × . 1 = Q :Q M3 3, Q = QT, trQ = 0 . 0 × Q ∈ 6 n o 1 For Q , one can find s,b R,n,m S2 with n m= 0 such that : ∈ Q ∈ ∈ · v 1 1 i Q = s(n n I)+b(m m I), X ⊗ − 3 ⊗ − 3 r where I is 3 3 identity matrix. The local physical properties of nematic liquid crystals depend on a × the degree of symmetry of order tensor Q. Specifically, there are three different states: (1) s = b = 0, which describes the isotropic distribution; (2) s = 0,b = 0, which corresponds to the uniaxial distribution; 6 (3) s = 0,b = 0, which describes the biaxial distribution. 6 6 Configuration of nematic liquid crystals corresponds to local minimizers of Landau-de Gennes energy functional, whose simplest form is given by L (1.1) [Q]= Q(x)2 +f (Q(x)) dx, LG B F 2|∇ | ZΩn o where L > 0 is a material-dependent elastic constant, and f is the bulk energy density, which can B be taken as follows a2 b2 c2 f (Q) = tr(Q2) tr(Q3)+ tr(Q2)2, B − 2 − 3 4 1 2 ZHIYUANGENG,WEIWANG,PINGWENZHANG,ANDZHIFEIZHANG where a2,b2,c2 are material-dependent and non-zero constants, which may depend on temperature. A well-known fact is that f (Q) attains its minimum on a manifold given by B N 1 = Q :Q = s+(n n I), n R3, n = 1 , N ∈ Q ⊗ − 3 ∈ | | n o where s+ = b2+√b4+4a2c2. It is easy to see that is a smooth submanifold of , homemorphic to 4c2 N Q the real projective plane RP2, and contained in the sphere Q : Q = 2s+ . Critical points ∈ Q | | 3 of Landau-de Gennes functional satisfy the Euler-Lagrangenequation q o 1 (1.2) L∆Q= a2Q b2(Q2 Q 2I)+c2Q Q 2. − − − 3| | | | The Landau-de Gennes energy (1.1) and Euler-Lagrange equation (1.2) are widely used to study the behavior of defects, see [1, 3, 7, 21] and references therein. However, there still exist many challenging problems in understandingthe mechanism which generates defects and predicting their profiles as well as stability, see [11]for many conjectures. Theradial symmetric solution in a ball or in R3, named hedgehog solution, is regarded as a potential candidate profile for the isolated point defect in 3-D region. The property and stability of this solution are well studied and it is shown that the radial symmetric solution are not stable for large a2 and stable for small a2 [13]. We also refer [26, 20, 16, 12] and references therein for related works. In this paper, we are concerned with a class of point defects in R2, which correspond to “radial” solutions of the Euler-Lagrange equation (1.2). Here “radial” means that the eigenvectors of Q don’t change along the radial direction. Precisely speaking, we study the solution with the form (1.3) Q(r,ϕ) = u(r)F +v(r)F , 1 2 where (r,ϕ) is the polar coordinate in R2, and coskϕ sinkϕ 0 1 0 0 − (1.4) F =2nn I = sinkϕ coskϕ 0 , F = 3e e I = 0 1 0 . 1 2 2 3 3 − − ⊗ − − 0 0 0 0 0 2 The boundary condition on these solutions is taken to be 1 k k (1.5) lim Q(r,ϕ) = s (n(ϕ) n(ϕ) I), n(ϕ) = (cos ϕ,sin ϕ,0), + r + ⊗ − 3 2 2 → ∞ which has degree k about origin as an RP2-valued map. Here k Z 0 . Note that if we assume 2 ∈ \{ } theinvariance of Q along thedefect line, then disclination linein 3-D domain can beideally treated as a point defect in 2-D domain. In [14], Ignat, Nguyen, Slastikov and Zarnescu proved the existence of the radial solution for any non-zero integer k. Moreover, the solution is also a local minimizer of the reduced functional (2.3). An important question is whether the radical solution they constructed is a local minimizer of the energy . This problem was also partially answered in [14], where the instability result LG F is proved for k > 1. However, the question of whether the k-radially symmetric solutions (1.3) | | subject to (1.5) for k = 1 are stable remains open. ± The goal of this paper is to give a positive answer to this question. Precise result will be stated in next section. We remark that this problem is somewhat analogous to the stability of radial solutions of the Ginzburg-Landau equation(see [17, 23, 19, 8, 24] for example). STABILITY OF HALF-DEGREE POINT DEFECT PROFILES FOR 2-D NEMATIC LIQUID CRYSTAL 3 2. The stability of radially symmetric solution with k = 1 ± We make the following rescaling c2 2 b2 Q = Q, x = , b2 L c r and let t = ab24c2. Then Landau-deeGennes eneregy functional (1.1) is rescaled into the form(drop the tildes): 1 t 1 1 [Q] = Q(x)2 tr(Q2) tr(Q3)+ tr(Q2)2 dx. LG F R2 2|∇ | − 2 − 3 4 Z n o Therefore, without loss of generality, we may take L = b = c = 1 and a2 = t > 0. In such case, substituting (1.3) into (1.2), (u,v) satisifes the following ODE system(see [5, 11]): u + u′ k2u= u[ t+2v+(6v2 +2u2)], (2.1) ( v′′′′+ vrr′ =−vr2[−t−v−+(6v2 +2u2)]+ 13u2, together with the boundary conditions s+ s+ (2.2) u(0) = 0, v (0) = 0, u( ) = , v(+ ) = , ′ ∞ 2 ∞ − 6 where s+ = 1+√1+24t. The system (2.1) can be also viewed as the Euler-Lagrange equation of the 4 functional 1 ∂u ∂v 2k2u2 t E(u,v) = ∞ 2( )2+6( )2+ (2u2+6v2) 2 ∂r ∂r r2 − 2 Z0 n (cid:2) 1 (cid:3) (2.3) 2v(v2 u2)+ (2u2+6v2)2 rdr, − − 4 o which is a reduced form of (1.1) under the assumption (1.3). Let us recall some basic properties of the solution (u,v) constructed in [14]: (H1) u > 0,v < 0,u+3v < 0, u2+3v2 < s2/3 for r (0, ); (H2) For t < 1/3: v > s+ > 1; + ∈ ∞ − 6 −6 (H3) For t > 1/3: v < s+ < 1; − 6 −6 (H4) For t = 1/3: v = s+ = 1; − 6 −6 (H5) u > 0, v (1+6v) 0. ′ ′ ≤ Remark 2.1. In fact, (H5) was not verified in [14]. In the appendix, we will present a proof by using (H1)-(H4) and the fact that (u,v) is a minimizer of (2.3). The main result of this paper is stated as follows. Theorem 2.1. Let (u,v) be a stable critical point of (2.3) for k = 1 with the properties (H1)- ± (H4). Then the solution Q = u(r)F +v(r)F is a local minimizer of Landau-de Gennes energy 1 2 (1.1). That is, for any perturbation V H1(R2, ), it holds ∈ Q d2 1 1 t (2.4) (V), (Q+εV)2 Q 2 (Q+εV 2 Q 2) I dε2 R2 2|∇ | − 2|∇ | − 2 | | −| | Z n 1 1 tr((Q+εV)3) tr(Q3) + (Q+εV 4 Q 4) dx − 3 − 4 | | −| | ε=0 o (cid:12) = V 2(cid:2) t V 2 2tr(QV2)+ Q(cid:3) 2 V 2+2(tr(QV))2 dx (cid:12)0. (cid:12) R2 |∇ | − | | − | | | | ≥ Z n o 4 ZHIYUANGENG,WEIWANG,PINGWENZHANG,ANDZHIFEIZHANG Moreover, the equality holds if and only if V span V ,V ,V ,V ,V , where 0 1 2 3 4 ∈ { } V = uE , 0 2 ku V = (u(r)E +√3v (r)E )cosϕ E sinϕ, 1 ′ 1 ′ 0 2 − r ku V = (u(r)E +√3v (r)E )sinϕ+ E cosϕ, 2 ′ 1 ′ 0 2 r V = (ucoskϕ 3v)E +usinkϕE , 3 3 4 − V = usinkϕE (ucoskϕ+3v)E . 4 3 4 − Here E ,E , ,E are defined in (3.2). 0 1 4 ··· Remark 2.2. The above null space is generated by the invariance of the energy F under the LG rotation and translation, i.e., (2.5) (RQRT) = (Q), Q(x+x ) = (Q(x)), LG LG LG 0 LG F F F F for any constant R SO(3) and constant x0 R2. (cid:0) (cid:1) ∈ ∈ Remark 2.3. This result implies that the solution (1.3) for k = 1 can be regarded as the profile of point defects in R2 or the local profile of line defects in R3. ± 3. The second variation of Landau-de Gennes energy To prove the stabillity, we need to compute the second variation of at critical point Q = LG u(r)F +v(r)F . For any V H1(R2, ), we have F 1 2 ∈ Q d2 1 1 t (3.1) (V), (Q+εV)2 Q 2 (Q+εV 2 Q 2) I dε2 R2 2|∇ | − 2|∇ | − 2 | | −| | Z n 1 1 (tr((Q+εV)3) tr(Q3))+ (Q+εV 4 Q 4) dx − 3 − 4 | | −| | ε=0 o (cid:12) = V 2 t V 2 2tr(QV2)+c2 Q 2 V 2+2(tr(QV))2(cid:12) dx (cid:12) R2 |∇ | − | | − | | | | Z n o = V 2 t V 2 2(u tr(F V2)+v tr(F V2)) 1 2 R2 |∇ | − | | − · · Z n +(6v2 +2u2)V 2+2(u tr(F V)+v tr(F V))2 dx. 1 2 | | · · o We define 1 0 0 3 −3 1 E = 0 1 0 = F , 0 r2 0 −03 2 √6 2 3 coskϕ sinkϕ 0 1 1 E = sinkϕ coskϕ 0 = F , 1 1 √2 − √2 0 0 0 (3.2) sinkϕ coskϕ 0 1 − E = coskϕ sinkϕ 0 , 2 √2 0 0 0 0 0 1 0 0 0 1 1 E = 0 0 0 , E = 0 0 1 . 3 4 √2 √2 1 0 0 0 1 0 STABILITY OF HALF-DEGREE POINT DEFECT PROFILES FOR 2-D NEMATIC LIQUID CRYSTAL 5 A straightforward calculation shows tr(E E ) = δi for 0 i,j 4, i j j ≤ ≤ which implies that E is an orthonormal basis in . Thus, we can write V H1(R2, ) as i 0 i 4 { } ≤≤ Q ∈ Q a linear combination of this basis in the polar coordinate 4 (3.3) V(r,ϕ) = w (r,ϕ)E (ϕ). i i i=0 X Using (3.3), a direct calculation yields that 4 V 2 = w2, tr(F V) = √2w , tr(F V) = √6w , i 1 1 2 0 | | i=0 X V2 =w2E2+(w2+w2)E2 +w2E2+w2E2+(w w +w w )(E E +E E ) 0 0 1 2 1 3 3 4 4 1 3 2 4 1 3 3 1 √6 √6 +(w w +w w )(E E +E E )+w w (E E +E E ) w w E w w E 1 4 2 3 1 4 4 1 3 4 3 4 4 3 0 1 1 0 2 2 − 3 − 3 √6 √6 + w w E + w w E , 0 3 3 0 4 4 6 6 thus we obtain coskϕ 2√3 tr(F V2)= (w2 w2) w w +sinkϕw w , 1 2 3 − 4 − 3 0 1 3 4 1 1 tr(F V2)= w2 w2 w2+ w2+ w2. 2 0 − 1 − 2 2 3 2 4 From the fact V 2 = (∂ V)2+ 1 (∂ V)2, we have |∇ | r r2 ϕ 4 1 V 2 = w2 + w2 +(kw w )2+(kw +w )2+w2 +w2 . |∇ | ir r2 0ϕ 2− 1ϕ 1 2ϕ 3ϕ 4ϕ i=0 X (cid:0) (cid:1) In summary, we conclude that + 2π 4 1 (V)= ∞ w2 + w2 +(kw w )2+(kw +w )2+w2 +w2 I ir r2 0ϕ 2− 1ϕ 1 2ϕ 3ϕ 4ϕ Z0 Z0 nXi=0 (cid:2) (cid:3) 4 1 2 +(6v2 +2u2 t) w2 2 u( coskϕ(w2 w2) w w +sinkϕw w ) − i − 2 3 − 4 − √3 0 1 3 4 i=0 (cid:0)X (cid:1) (cid:2) 1 1 +v(w2 w2 w2+ w2 + w2) +4(uw +√3vw )2 rdrdϕ. 0 − 1 − 2 2 3 2 4 1 0 o (cid:3) In order to prove (V) 0, it suffices to show that for k = 1, I ≥ | | + 2π 1 IA(w ,w ,w ) , ∞ w2 +w2 +w2 + [w2 +(kw w )2+(kw +w )2] 0 1 2 0r 1r 2r r2 0ϕ 2− 1ϕ 1 2ϕ Z0 Z0 n 4 +(6v2 +2u2 t)(w2+w2+w2) uw w − 0 1 2 − √3 0 1 (3.4) +v(w2 w2 w2)]+4(uw +√3vw )2 rdrdϕ 0, 0 − 1 − 2 1 0 ≥ o 6 ZHIYUANGENG,WEIWANG,PINGWENZHANG,ANDZHIFEIZHANG and + 2π 1 IB(w ,w ) , ∞ w2 +w2 + (w2 +w2 )+(6v2 +2u2 t)(w2 +w2) 3 4 3r 4r r2 3ϕ 4ϕ − 3 4 Z0 Z0 n (3.5) u(coskϕ(w2 w2)+2sinkϕw w ) v(w2 +w2) rdrdϕ 0. − 3 − 4 3 4 − 3 4 ≥ o The following lemma shows that C (R2 0 ) is dense in H1(R2). Thus, we may assume V c∞ C (R2 0 ), hence w C (R2 0 ). \{ } ∈ c∞ i c∞ \{ } ∈ \{ } Lemma 3.1. C (R2 0 ) is dense in H1(R2). c∞ \{ } Proof. Since C (R2) is dense in H1(R2), it suffices to show that any C (R2) function can be c∞ c∞ approximated by C (R2 0 ). For this, we introduce a smooth cut-off function χ(r) defined by c∞ \{ } 0 r 2 χ(r) = ≤ − 1 r 1. (cid:26) ≥ − Foranyu∈ Cc∞(R2),letuN(x)= u(x)χ(lnN|x|)forN ≥ 1. Obviously,uN ∈ Cc∞(R2\{0}). Moreover, ln x ln x ku−uNkH1 ≤ u(1−χ( N| |)) L2 + ∇u(1−χ( N| |)) L2 (cid:13) 1 u ln x(cid:13) (cid:13) (cid:13) (cid:13)+ N x χ′( N| (cid:13)|)) L2(cid:13). (cid:13) | | (cid:13) (cid:13) (cid:13) (cid:13) It is easy to see that the first two terms on the right hand side tend to zero as N + . While, → ∞ the third term is bounded by C u(x)2 1 C 2 N(cid:16)Ze−2N≤|x|≤e−N | |x|2| dx(cid:17) ≤ √NkukL∞, which tends to zero as N + . (cid:3) → ∞ 4. Some important integral identities In this section, let us derive some important integral identities, which will play crucial roles in our proof. In the sequel, we assume that η C ((0,+ ),R). c∞ ∈ ∞ Using (2.1), we deduce that (η) := ∞ (vη)2 +(6v2 +2u2 t v)(vη)2 rdr r A − − Z0 n o v 1 = ∞ (vη)2 +(v + ′ u2)vη2 rdr r ′′ r − 3 Z0 n o 1 =(vv η2r) + ∞(v2η2 vu2η2)rdr ′ |∞0 r − 3 Z0 1 (4.1) = ∞ (vη )2 vu2η2 rdr, r − 3 Z0 n o STABILITY OF HALF-DEGREE POINT DEFECT PROFILES FOR 2-D NEMATIC LIQUID CRYSTAL 7 and k2 (η) := ∞ (uη)2 +(6v2 +2u2 t+2v+ )(uη)2 rdr B r − r2 Z0 n o u = ∞[(η u+uη)2 +uη2(u + ′)]rdr ′ ′ ′′ r Z0 =(uuη2r) + ∞u2η2rdr ′ |∞0 r Z0 (4.2) = ∞(uη )2rdr. r Z0 Taking derivative to (2.1) gives u 2u 2u u + ′′ ′ + = u[ t+2v+6v2 +6u2]+2uv (1+6v), ′′′ ′ ′ r − r2 r3 − v v 2uu v + ′′ ′ = v [ t 2v+18v2 +2u2]+ ′(1+6v). ′′′ ′ r − r2 − − 3 Therefore, we have (η) := ∞ (v η)2 +(v η)2(18v2 +2u2 t 2v) rdr ′ r ′ C − − Z0 n o v v 2uu = ∞ (v η+v η )2+v η2(v + ′′ ′ ′(1+6v)) rdr ′′ ′ ′ ′ ′′′ r − r2 − 3 Z0 n o (v η)2 2uuv (1+6v) = ∞ (v′η′)2 ′ ′ ′ η2 rdr+(rv′′v′η2) ∞ − r2 − 3 0 Z0 n (v η)2 2uuv (1+6v) o (cid:12) (4.3) = ∞ (v η )2 ′ ′ ′ η2 rdr, (cid:12) ′ ′ − r2 − 3 Z0 n o and k2 (η) := ∞ (uη)2 +(uη)2(6v2 +6u2 t+2v+ ) rdr D ′ r ′ − r2 Z0 n o u u 2u = ∞ (u η+uη )2+uη2 u + ′′ ′ + 2uv (1+6v) rdr ′′ ′ ′ ′ ′′′ ′ r − r2 r3 − Z0 n (cid:16) (cid:17)o (uη)2 2uuη2 = ∞ (u′η′)2 ′ + ′ 2uu′v′(1+6v)η2 rdr+(ru′′u′η2) ∞ − r2 r3 − 0 Z0 n (uη)2 2uuη2 o (cid:12) (4.4) = ∞ (uη )2 ′ + ′ 2uuv (1+6v)η2 rdr. (cid:12) ′ ′ ′ ′ − r2 r3 − Z0 n o In addition, we have uη k2 (uη)2 (η) := ∞ ( )2+(6v2+2u2 t+2v+ ) rdr E r r − r2 r2 Z0 n o uη u uη2 u = ∞[( ′ +( )η)2+ (u + ′)]rdr ′ ′′ r r r2 r Z0 u η2 u = ∞ (rηr)2+ r4(2ruu′−u2) rdr+(r)′uη2 ∞0 Z0 (cid:16) u η2 (cid:17) (cid:12) (4.5) = ∞ ( η )2+ (2ruu u2) rdr. (cid:12) r r r4 ′− Z0 (cid:16) (cid:17) 8 ZHIYUANGENG,WEIWANG,PINGWENZHANG,ANDZHIFEIZHANG 5. Proof of Theorem 2.1 This section is devoted to the proof of Theorem 2.1. 5.1. Non-negativity of IB(w ,w ). 3 4 Proposition 5.1. For any w ,w C (R2 0 ) and k = 1, we have 3 4 c∞ ∈ \{ } | | + 2π 1 IB(w ,w ):= ∞ w2 +w2 + (w2 +w2 )+(6v2 +2u2 t)(w2+w2) 3 4 3r 4r r2 3ϕ 4ϕ − 3 4 Z0 Z0 n u(coskϕ(w2 w2)+2sinkϕw w ) v(w2 +w2) rdrdϕ 0. − 3 − 4 3 4 − 3 4 ≥ o Proof. Let z = w +iw , i = √ 1. Then we have 3 4 − ∂ z 2 = (∂ w )2+(∂ w )2, ∂ z 2 = (∂ w )2+(∂ w )2, z 2 = w2+w2, | r | r 3 r 4 | ϕ | ϕ 3 ϕ 4 | | 3 4 coskϕ(w2 w2)+2sinkϕw w = Re(coskϕ isinkϕ)(w +iw )2 = Re(e ikϕz2). 3 − 4 3 4 − 3 4 − Thus, we can rewrite IB(w ,w ) as 3 4 2π 1 (5.1) IB(z) = ∞ ∂ z 2 + ∂ z 2+(6v2 +2u2 t v)z 2 uRe(e ikϕz2)rdrdϕ. | r | r2| ϕ | − − | | − − Z0 Z0 Assume that + ∞ z(r,ϕ) = z (r)eimϕ, m m= X−∞ we have + + ∞ ∞ Re(e ikϕz2) =Re e ikϕ e(m+l)ϕz z = Re ei(m+l k)ϕz z . − − m l − m l l,m= l,m= (cid:0) X−∞ (cid:1) (cid:0) X−∞ (cid:1) Substituting it into (5.1), we get IB(z) = 2π +∞ +∞ ∂ z 2+ m2 z 2+(6v2 +2u2 t v)z 2 | r m| r2 | m| − − | m| Z0 nmX=−∞h i u Re(z z ) rdr. m l − mX+l=k o When k = 1, we can write ∞ IB(z) = 2π M , m m=1 X where 1 M = ∞ ∂ z 2+ ∂ z 2+ (m2 z 2+(1 m)2 z 2) m Z0 n| r m| | r 1−m| r2 | m| − | 1−m| +(6v2 +2u2 t v)(z 2+ z 2) 2uRe(z z ) rdr. m 1 m m 1 m − − | | | − | − − o Noticing that m2 1,(1 m)2 0 for m 1, and using the following simple relations ≥ − ≥ ≥ z z Re(z z ), ∂ z 2 (∂ z )2, ∂ z 2 (∂ z )2, m 1 m m 1 m r m r m r 1 m r 1 m | − |≥ − | | ≥ | | | − | ≥ | − | STABILITY OF HALF-DEGREE POINT DEFECT PROFILES FOR 2-D NEMATIC LIQUID CRYSTAL 9 we conclude that 1 M ∞ (∂ z )2+(∂ z )2+ z 2 m ≥ Z0 (cid:16) r| m| r| 1−m| r2| m| +(6v2 +2u2 t v)(z 2+ z 2) 2uz z rdr. m 1 m m 1 m − − | | | − | − | || − | (cid:17) Thus, we only need to show that for any q ,q C ((0, ),R), 0 1 c∞ ∈ ∞ q2 I˜(q ,q ) , ∞ (∂ q )2+(∂ q )2+ 1 +(6v2 +2u2 t v)(q2+q2) 2uq q rdr > 0. 0 1 r 0 r 1 r2 − − 0 1 − 0 1 Z0 n o Let η = q /u and ζ = q /v. Then η,ζ C ((0, )). From (4.1) and (4.2), it is straightforward to 1 0 c∞ ∈ ∞ obtain I˜(q ,q ) = (ζ)+ (η) ∞ 3v(uη)2 +2vu2ζη rdr 0 1 A B − Z0 n o 1 (5.2) = ∞ (uη )2+(vζ )2 vu2(3η+ζ)2 rdr, ′ ′ − 3 Z0 n o which is non-negative since v < 0. The case of k = 1 can be considered similarly. (cid:3) − 5.2. Non-negativity of IA(w ,w ,w ). First of all, we expand w (r,ϕ) as 0 1 2 i ∞ w (r,ϕ) = µ(i)(r)cosnϕ+ν(i)sinnϕ . i n n n=0 X(cid:0) (cid:1) Since ω C (R2 0 ), we may assume µ(i),ν(i) C ((0, ),R) for all n and i. Furthermore, i c∞ n n c∞ ∈ \{ } ∈ ∞ w C (R2) requires µ(0) = ν(0) = 0 for n 1. 0 c∞ n n ∈ ≥ Direct calculation shows that 2π(∂wi)2dϕ = π 2(∂µ(0i))2+ ∞ ((∂µ(ni))2+(∂νo(i))2) , ∂r ∂r ∂r ∂r Z0 n=1 (cid:2) X (cid:3) 2π(∂wi)2dϕ = π ∞ n2((µ(i))2+(ν(i))2), ∂ϕ n n Z0 n=1 X 2π ∂wiw dϕ = π ∞ n(µ(j)ν(i) ν(j)µ(i)), ∂ϕ j n n − n n Z0 n=1 X 2π w2dϕ = π 2(µ(i))2+ ∞ ((µ(i))2+(ν(i))2) , i 0 n n Z0 n=1 (cid:2) X (cid:3) 2π w w dϕ = π 2µ(i)µ(j)+ ∞ (µ(i)µ(j)+ν(i)ν(j)) . i j 0 0 n n n n Z0 n=1 (cid:2) X (cid:3) Thus, we can decompose IA(w ,w ,w ) as 0 1 2 ∞ (5.3) IA(w ,w ,w )= IA +IA + IA, 0 1 2 0,01 0,2 n n=1 X 10 ZHIYUANGENG,WEIWANG,PINGWENZHANG,ANDZHIFEIZHANG where (0) (1) ∂µ ∂µ IA = ∞ ( 0 )2+(µ(0))2(18v2 +2u2 t 2v)+( 0 )2 0,01 ∂r 0 − − ∂r Z0 n k2 4u +(µ(1))2(6v2 +6u2 t+2v+ )+ (1+6v)µ(0)µ(1)rdr, 0 − r2 √3 0 0 ∂µ(2) k2 IA = ∞( 0 )2+(µ(2))2(6v2 +2u2 t+2v+ ) rdr, 0,2 ∂r 0 − r2 Z0 o and 2 ∂µ(i) ∂ν(i) 4kn 2 n2 IA = ∞ ( n )2+( n )2 + (µ(1)ν(2) µ(2)ν(1))+ ((µ(i))2+(ν(i))2) n ∂r ∂r r2 n n − n n r2 n n Z0 nXi=0(cid:16) (cid:17) Xi=0 k2 +((µ(0))2+(ν(0))2)(18v2 +2u2 t 2v)+((µ(1))2+(ν(1))2)(6v2 +6u2 t+2v+ ) n n − − n n − r2 k2 4u +((µ(2))2+(ν(2))2)(6v2 +2u2 t+2v+ )+ (1+6v)(µ(0)µ(1) +ν(0)ν(1)) rdr. n n − r2 √3 n n n n o Thenon-negativity ofIA follows fromthefact that(u,v) is alocal minimizer ofreducedenergy 0,01 (2.3). Indeed, we have for any η,ξ H1((0, ),rdr), ∈ ∞ d2 (η,ξ) , E(u+εη,v +εξ) E(u,v) J dε2 − ε=0 n o(cid:12) = ∞ (∂ η)2 +(∂ ξ)2+η2(18v2 +(cid:12)2u2 t 2v) r r (cid:12) − − Z0 n k2 4u +ξ2(6v2 +6u2 t+2v+ )+ (1+6v)ηξ rdr 0, − r2 √3 ≥ o which implies that (5.4) IA 0. 0,01 ≥ It follows from (4.2) that 2 (5.5) IA = (µ(2)/u) = ∞ u∂ (µ(2)/u) rdr 0. 0,2 B 0 r 0 ≥ Z0 (cid:16) (cid:17) It remains to prove IA 0 for all n 1, which is a consequence of the following proposition. n ≥ ≥ Proposition 5.2. For any µ ,ν ,µ ,ν ,µ ,ν C ((0, )) and k =1, we have 0 0 1 1 2 2 c∞ ∈ ∞ | | IA(µ ,ν ,µ ,ν ,µ ,ν ) n 0 0 1 1 2 2 2 ∂µ ∂ν 4kn 2 n2 , ∞ ( i)2+( i)2 + (µ ν µ ν )+ (µ2+ν2) ∂r ∂r r2 1 2− 2 1 r2 i i Z0 nXi=0(cid:16) (cid:17) Xi=0 k2 +(µ2+ν2)(18v2 +2u2 t 2v)+(µ2+ν2)(6v2 +6u2 t+2v+ ) 0 0 − − 1 1 − r2 k2 4u +(µ2+ν2)(6v2 +2u2 t+2v+ )+ (1+6v)(µ µ +ν ν ) rdr 0. 2 2 − r2 √3 0 1 0 1 ≥ o Proof. From the fact that 2(µ ν µ ν ) (µ2+ν2+µ2+ν2), 1 2− 2 1 ≥ − 1 1 2 2