Green’s Function For Radial Heat Conduction in Two-Region Composite Cylinders With Perfect Boundary Contact by Donald E. Amos Albuquerque, NM Abstract This paper presents the derivation of the Green’s function for composite cylinders 0<r<a and r>a in perfect contact on the surface r=a. Because the source function can be in either region, there are two pairs of functions which define the Green’s function. Each pair is the solution to a two-region conduction problem with zero initial temperatures and continuity of temperature and flux on the cylinder r=a. These pairs are used in conjunction with a general formula to get the solution to other problems where the cylinders are in perfect contact, but may have non-zero initial conditions and/or possibly a distribution of internal heat sources. The Green’s function approach and a direct approach agree when applied to three problems with known solutions. A fourth problem illustrates a complication where the Laplace transform of the general solution is more useful. Keywords Two Regions Heat Conduction Unsteady State Laplace Transform 1.0 Introduction The main thrust of this paper is to derive the radial Green’s function G(r,r',t) for adjacent cylinders in perfect contact on the common surface r=a. The inner cylinder, 0 r a , is denoted as Region 1 and the outer cylinder, a r , is denoted as Region 2 In the notation of [7, Chapter 2], we are solving the heat conduction problem labeled (R0C0). In order to apply the concept of a Green’s function to a problem, one has to use these functions in conjunction with the general formula [6]. t j t j Tj (P', t) G j (P, P', t )q j (P, )dVd G 3j j (P, P', t )q 3 j(P, )dVd k j 0 Vj 0 V3 j G jj (P, P', t 0)f j (P)dV k 3- j j G (P, P', t 0)f 3 j (P)dV j 1,2 k j 3 j V3 j t G jj (P, P', t ) T (P, ) j Tj (P, ) G jj (P, P', t ) j dS jd n i n i 0 S j SC k t G j (P, P', t ) T (P, ) 3- j j T3 j (P, ) 3 j G 3j j (P, P', t ) 3 j dS3 jd k j 0 S3 j SC n i n i The indices j and 3-j with j=1,2 can only have the values of 1 or 2. Here V1 and V2 are the volumes of Regions 1 and 2 with surfaces S1 and S 2 ; Sc is the contact boundary and S j Sc , j 1,2 are external (1.1) j 3 j Vj boundaries. q1, q2 and f1, f 2 are heat source and initial temperature distributions respectively. The operators / ni are internal normal derivatives. Since the Green’s function source can be in either region, the construction requires the solution of two problems each of which satisfy the conditions on the common boundary and each yielding two functions. The complete Green’s function therefore consists of two pairs (G11( r, r ', t ), G21 ( r, r ', t )) and (G12 ( r, r ', t ), G22 ( r, r ', t )) where the superscript designates the region where the source occurs and the subscript designates the region where the function applies. 1 For this application, Region 1 has no external boundary and the external boundary for Region 2 is at infinity where we select zero values for all temperatures. Consequently, (1.1) reduces to only the terms involving the source distributions q1, q2 and the initial temperatures f1, f 2 . Laplace transforms of the Green’s functions are used in most of the manipulations because the transforms contain well documented handbook functions. 2. Conduction Equations for the Green’s Function We solve the two-region problem of heat conduction in concentric cylinders with a heat source and perfect boundary contact. The inner cylinder we designate as Region 1, 0<r<a, and the outer cylinder as Region 2, a r . The conduction equations for the Green’s functions are 1 G1 1 G2 2G1 , 0<r<a , 2G2 , a<r< , G2 ( r, t ) 0 for r (2.1) 1 t 2 t with boundary conditions, expressing continuity of temperature and flux at r=a, as G G G1( a, t ) G2 ( a, t ) , k1 1 ( a, t ) k 2 2 ( a, t ) . (2.2) r r The initial conditions are (2.3) G j ( r,0) 0 , j 1,2 except possibly at the source point where r r ' . The energy comes from the point source solution which can appear in either Region 1 or Region 2. 3. Green’s Function for the Source in Region 1 (G11( r, r ', t ), G21 ( r, r ', t )) For the sake of simplicity, we suppress the superscripts for the analysis of this section, and set G1( r, r ', t ) G11( r, r ', t ) and G2 ( r, r ', t ) G21 ( r, r ', t ) . With the source in Region 1 along the cylindrical surface r r ' , the source solution of the heat equation is [ 8, p.533], [9, Chapter XIII] 2 2 B rr ' 1 S1( r, r ', t ) 1 e ( r r ' ) /(41t ) I 0 ( ) , B1 (3.1) . 2t 21t 21 We use the Laplace transform with parameter p and write the solution in Region 1 as the sum of the source solution and regular solutions of the conduction equation (2.1) which are finite for r=0. In Region 2 we use only regular solutions of the conduction equation which vanish for r . Then, the (Laplace) transformed equations from Section 2 lead to modified Bessel functions and we select the appropriate solutions to form the equations G1( r, r ', p ) S1( r, r ', p ) AI 0 (q1r ) , 0 r a q j p / j (3.2) G2 ( r, r ', p ) BK 0 (q2 r ) , ar where the transform of (3.1) is [9, Chapter XIII] I (q r ')K0 (q1r ) r r ' (3.3) S1( r, r ', p ) B1 0 1 I 0 (q1r )K0 (q1r ') r r ' and I 0 and K0 are modified Bessel functions of zero order. The next step is to apply the boundary conditions for continuity of temperature and flux, G G G1(a, p ) G2 ( a, p ) , k1 1 ( a, p) k2 2 ( a, p) (3.4) r r 2 and solve for A and B. Since a> r and a r ' in Region 1, we use the first form of S1 for S1( a, r ', p ) with (3.2) in (3.4) B1I 0 (q1r ')K 0 (q1a ) AI 0 (q1a ) BK 0 (q2a ) (3.5) q1k1B1I 0 (q1r ')K1(q1a ) q1k1 AI1(q1a ) q2k2 BK1(q2a ) or AI 0 ( q1a ) BK 0 (q2a ) B1I 0 (q1r ')K 0 (q1a ) (3.6) . q1k1 AI1(q1a ) q2k2 BK1(q2a ) q1k1B1I 0 (q1r ')K1(q1a ) Then K 0 ( q1a ) K 0 ( q2a ) I 0 ( q1a ) K 0 ( q1a ) (3.7) A B1I 0 ( q1r ') q1k1K1( q1a ) q2k2 K1(q2a ) , I 0 ( q1a ) K 0 ( q2a ) q1k1I1( q1a ) q2k2 K1(q2a ) B B1I 0 ( q1r ') q1k1I1(q1a ) q1k1K1(q1a ) I 0 ( q1a ) K 0 ( q2a ) q1k1I1(q1a ) q2k 2 K1(q2a ) or A B1I 0 (q1r ') q2k2 K1(q2a )K 0 (q1a ) q1k1K1(q1a )K 0 (q2a ) / D (3.8) B B1I 0 (q1r ') q1k1K1(q1a ) I 0 (q1a ) q1k1I1(q1a )K 0 (q1a ) / D k1 B1I 0 (q1r ') / D a D q2k2 K1( q2a ) I 0 (q1a ) q1k1I1(q1a )K 0 (q2a ) where we have used the Wronskian of the I and K functions I1( z )K0 ( z ) I 0 ( z )K1( z ) 1/ z (3.9) in the computation of B. The solution (3.2) becomes r' r a r r' a I ( q r ') K 0 ( q1r ) G1( r, r ', p ) B1 0 1 B1I 0 ( q1r ) I 0 ( q1r ')C , I 0 ( q1r ) K 0 ( q1r ') (3.10) k G2 ( r, r ', p ) 1 B1K 0 ( q2 r ) I 0 ( q1r ') / D , a ar C q2k 2 K1( q2a ) K 0 ( q1a ) q1k1K1( q1a ) K 0 ( q2a ) / D D q2k 2 K1( q2a ) I 0 ( q1a ) q1k1I1( q1a ) K 0 ( q2a ) or (3.11) F ( r, r ', p ) G1( r, r ', p ) 1 F1( r ', r, p ) G2 ( r, r ', p ) r' r a r r' a k1 B1I 0 ( q1r ')K 0 (q2r ) / D , a where 3 ar qj p / j p / j , j 1,2 D q2k2 K1(q2a ) I 0 (q1a ) q1k1I1(q1a )K 0 (q2a ) , q j (3.12) . q k K (q a ) I 0 (q1a )K 0 (q1r ) I 0 (q1r )K 0 (q1a ) F1( r, r ', p ) B1I 0 ( q1r ') 2 2 1 2 / D q1k1K 0 ( q2a ) I1( q1a )K 0 (q1r ) I 0 (q1r )K1(q1a ) Now we follow the procedure of [9, Chapter XII, Section 128] and invert the Laplace transform by the complex inversion integral. Then with G as G1 or G2 , i 1 G( r, r ' t ) G( r, r ', p )e pt dp , 2 i i (3.13) 0. The indication from [9] is that the integrand has no poles, but it is clear from (3.11) that G1 and G2 have branch points at p=0. Therefore we select the branch cut along the negative real axis and use Cauchy’s theorem to close the contour about the cut with a classical keyhole contour. Then with G as G1 or G2 we have 0 G ( r, r ' t ) (3.14) 2 G ( r, r ', p )e pt 2 i pw e 2 i 1 0 2 i wei dw lim 2 G( r, r ', p )e pt 2 i 0 G( r, r ', p)e pt pw e 2 i we i p e iei d i . dw 0 Now, the integrand is real on the positive real axis. Therefore, by the reflection principle, the values on the top and bottom of the cut where p w2ei and p w2ei , 0 w , are complex conjugates of one another and the difference is 2i times the imaginary part on p w2ei . Then, (3.14) becomes (3.15) G( r, r ' t ) Im G( r, r ', p ) pw e 2 2 i we w2 t 0 1 dw lim 0 2 G( r, r ', p)e pt p e i ei d . The task now is to explicitly compute the real and imaginary parts of G1 and G2 on p w2ei and evaluate the limit 0 . The details of the algebra are relegated to APPENDIX A and we jump to the final result. The Solution for G11( r, r ', t ) and G21 ( r, r ', t ) with the Source in Region 1 With the essentials presented in APPENDIX A, the final solution, with the proper superscript attached, is r' r J 0 ( 1w ) J 0 ( 1w ) 4 k k a a e w2 t dw , r a G11( r, r ' t ) 21 22 B1 (3.16) r' a a ( D12 D22 )w 0 G21 ( r, r ', t ) (3.17) 2 k1 B1 a 0 r' r r J 0 ( 1w ) D1J 0 ( 2 w ) D2Y0 ( 2 w ) ra a a a w2 t e dw , 2 2 r' a D1 D2 D1( w) 2 J 0 ( 1w)Y1( 2 w) 1J1( 1w)Y0 ( 2 w), D2 ( w) 2 J 0 ( 1w) J1( 2 w) 1J 0 ( 2 w) J1( 1w) B1 1 21 The asymptotic expressions , j k j / j j a / j , j 1,2 4 2 for z 0 z show that 1/ ( D12 D22 )w O( w), D1 / ( D12 D22 ) O( w) and D2 / ( D12 D22 ) =O(w3 ) for w 0 making each integrand integrable at w=0. J 0 ( z ) 1, J1( z ) z / 2, Y0 ( z ) 2 ln( z / 2), Y1( z ) 4. Green’s Function for the Source in Region 2 (G12 ( r, r ', t ), G22 ( r, r ', t )) We take the source in Region 2, a r ' , on the cylindrical surface r r ' , 2 2 B rr ' 1 S2 ( r, r ', t ) 2 e ( r r ' ) /(4 2 t ) I 0 ( ) , B2 (4.1) 2t 2 2t 2 2 and solve the heat conduction problem in the manner of Section 3. We use the Laplace transform and write the solution in Region 2 as the sum of the source solution and regular solutions of the conduction equation which vanish for r . In Region 1 we use only regular solutions of the conduction equation which are finite at r=0. Then, proceeding as before with the superscript suppressed, we have G1( r, r ', p ) AI 0 ( q1r ) , 0ra , (4.2) G2 ( r, r ', p ) S2 ( r, r ', p ) BK 0 ( q2r ) , a r ', r qj p / j , j 1,2 where [9, Chapter XIII ] I (q r ')K0 (q2r ) r r ' S2 ( r, r ', p ) B2 0 2 I 0 (q2r )K0 (q2r ') r r ' and I 0 and K0 are modified Bessel functions of zero order. The next step is to apply the boundary conditions G G G1(a, p ) G2 ( a, p ) , k1 1 ( a, p) k2 2 ( a, p) (4.4) r r and solve for A and B. Since a r and a r ' in Region 2, we use the second form of S 2 for (4.3) S2 ( a, r ', p ) with (4.2) in (4.4), (4.5) B2 I 0 (q2a )K 0 (q2 r ') BK 0 (q2a ) AI 0 (q1a ) q2k2 B2 I1(q2a )K 0 (q2 r ') q2k2 BK1(q2a ) q1k1AI1(q1a ) Solving for A and B with the help of the Wronskian (4.6) gives I 0 ( z )K1( z ) I1( z )K0 ( z ) 1/ z G1( r, r ', p ) (4.7) k2 B2 K 0 ( q2 r ') I 0 ( q1r ) / D , a F ( r, r ', p ) G2 ( r, r ', p ) 2 F2 ( r ', r, p ) r a r' r r' a r' r a where 5 . p / j , j 1,2 D q2k2 K1(q2a ) I 0 (q1a ) q1k1I1(q1a )K 0 (q2a ) , q j (4.8) q k I ( q a ) I 0 ( q2 r ')K1(q2a ) I1(q2a )K 0 (q2r ') F2 ( r, r ', p ) B2 K 0 ( q2 r ) 2 2 0 1 / D q1k1I1( q1a ) I 0 ( q2 r ')K 0 (q2a ) I 0 (q2a )K 0 (q2r ') . Again, we follow Section 3 and invert the Laplace transform by the complex inversion integral using the keyhole contour. The solution with the proper designation (G12 ( r, r ', t ), G22 ( r, r ', t )) is (4.9) G 2j ( r, r ' t ) 2 pw e Im G j ( r, r ', p ) 0 2 i 1 0 2 we w t dw lim 2 G j (r, r ', p )e pt p e i ei d 2 i for j=1 or 2. The details for F2 ( r, r ', p ) on p w e , 0 w and p ei , - are contained in APPENDIX B. The Solution for G12 ( r, r ', t ) and G22 ( r, r ', t ) with the Source in Region 2 J 0 ( 1wr / a ) D1J 0 ( 2 wr '/ a ) D2Y0 ( 2 wr '/ a ) e w t 2k 2 B2 dw, a 0 D12 D22 2 G12 ( r, r ' t ) (4.10) ra D1( w) 2 J 0 ( 1w)Y1( 2 w) 1J1( 1w)Y0 ( 2 w) D2 ( w) 2 J 0 ( 1w) J1( 2 w) 1J 0 ( 2 w) J1( 1w) B2 0 G22 ( r, r ', t ) B2 0 (4.11) 2 J 0 ( 1w) ( r ', w, 2 ) 1J1( 2 w) ( r ', w, 2 ) we w t dw 2 2 • D1J 0 ( 2 wr / a ) D2Y0 ( 2 wr / a ) D1 D2 r r' 2 J 0 ( 1w) ( r, w, 2 ) 1J1( 2 w) ( r, w, 2 ) we w t dw 2 2 • D1J 0 ( 2 wr '/ a ) D2Y0 ( 2 wr '/ a ) D1 D2 r r' 2 , ra 2 r r ( r, w, 2 ) J 0 ( 2 w )Y0 ( 2 w) J 0 ( 2 w)Y0 ( 2 w ) a a r a r a ( r, w, 2 ) J1( 2 w)Y0 ( 2 w ) J 0 ( 2 w )Y1( 2 w) 1 , j k j / j j a / j , j 1,2 2 2 Notice that (a, w, 2 ) 0 , and because (a, w, 2 ) is a Wronskian, (a, w, 2 ) 2 /( 2w) . B2 5. Applications of the Two Region Radial Green’s Functions In Applications I and II, we have a common geometry with perfect boundary contact. In application I, we consider the problem of the cooling of Region 1 by Region 2, which we take initially to be at zero temperature. There are no heat generation terms, q1 q2 0 . In Application II, we take the initial temperatures to be zero, f1 f 2 0 , and heat the media with a uniform source over Region 1, q1 Q1 >0 , q2 0 . In Application III, we take the material properties of both regions to be the same, 6 which gives a one region infinite medium. It is verified that the solution for the Green’s function is just the source term. In application IV, we take an initial temperature distribution which is uniform over both Region 1 and Region 2 with no heat generation. Since there is no gradient, the temperature does not change and the solution should reflect this. The analytics, however, become cumbersome and this is a case where the Laplace transform is better suited to solving the problem. Application I We apply the radial Green’s function for cylinders 0 r a and a r in perfect contact on the surface r=a to the cooling of the inner cylinder 0 r a 2T1 1 T1 , 0ra 1 t T1 (a, t ) T2 (a, t ) (5.1) T1 ( r ,0) V 2T2 k1 1 T2 , ar 2 t T1 T (a, t ) k1 1 (a, t ) r r T2 ( r ,0) 0 lim T2 ( r , t ) 0 r This problem can be solved by a direct application of the Laplace transform and the solution is given in Carslaw and Jaeger [9, Chapter XII, Section 128]. The general solution of a two-region problem is expressed in terms of the Green’s function by (1.1). We designate Region 1 as 0 r a and Region 2 as a r . In the context of this problem, G11( P, P ', t 0) G11( r, r ', t ) and G21 ( P, P ', t 0) G21 ( r, r ', t ) from Section 3. Region 1 has no S1 boundary, the S 2 boundary values at infinity are zero and there is no heat generation. We are left with only the contributions from the initial temperature distributions. Then with f1( P ) V and f 2 ( P ) 0 (5.2) and j=1 in (1.1) we have (5.3) T1 (P', t) V G11 (P,P', t 0)dV+ V1 k 21 G 2 (P,P', t 0)f 2 (P)dV V G11(r,r ', t 0)dV . k1 2 V2 V1 To get T1 (P', t) T1(r ', t) , we need to integrate the Green’s function over the interval 0 r a . The integration, applied to (3.16), is ( dV 2 rdr ) a (5.4) r 2 a 2 2 a 2 2 J 0 ( 1w )rdr wJ ( w ) J1( 1w) 2 1 1 1 a w w 1 0 1 and from (3.15) we have the solution for Region 1: T1( r ', t ) B1V (5.5) 4k1k2 2 a 2a 2 1 2 0 r' J 0 ( 1w ) J1( 1w) w2 t a e dw ( D12 D22 )w2 D1( w) 2 J 0 ( 1w)Y1( 2 w) 1J1( 1w)Y0 ( 2 w) D2 ( w) 2 J 0 ( 1w) J1( 2 w) 1J 0 ( 2 w) J1( 1w) j k j / j j a / j , j 1,2 7 Our goal now is to bring this formula into the form shown in Carslaw and Jaeger [9, Chapter XII, Section 128]. We multiply the numerator and denominator by 1 2 and change the variable with w v 1 . Then with D1 and D2 redefined, we have T1( r ', t ) (5.6) V 4k1k2 2 a 2 1 2 21 2a 2 1 1 J 0 ( vr ') J1( va ) ( D 2 D 2 )v 2 1 0 e1v t dv, 2 r' a 2 D1( v ) k2 1 J 0 ( va )Y1( va ) k1 2 J1( va )Y0 ( va ) = 1/ 2 D2 ( v ) k2 1 J 0 ( va ) J1( va ) k1 2 J 0 ( va ) J1( va ) The final form is (5.7) T1( r ', t ) 4Vk1k2 2 a 2 J (vr ') J (va ) ( D0 2 D12 )v 2 e 0 1 1v 2 t dv, r' a 2 which agrees with the solution presented in Carslaw and Jaeger [9, Chapter XII]. In order to solve for the temperatures in Region 2, we need the Green’s function with the source in Region 2. To this end we take j=2 in (1.1). Again, since there is no heat source and the solution vanishes at infinity, we get only the contribution from the initial temperature in Region 1. Then, with f1( P ) V , and f 2 ( P ) 0 (1.1) becomes k1 2 k1 2 V 2 2 T (P', t) G (P,P', t 0)f (P)dV+ G (P,P', t 0)f (P)dV G12 (P,P', t)dV . 2 2 2 1 1 (5.8) k 21 V1 k 21 V1 V2 The Green’s function for Region 1 with the source in Region 2 is taken from Section 4, equation (4.10). Again, to get T2 ( r ', t ) we need to integrate over 0<r<a. The integral is given in (5.4). The result is 2k 2 k1 2V 2 a 2 T2 ( r ', t ) B2 a k 21 1 J1( 1w) D1J 0 ( 2 wr '/ a ) D2Y0 ( 2 wr '/ a ) e w 2 t dw, ( D12 D22 )w 0 r' a D1 2 J 0 ( 1w)Y1( 2 w) 1J1( 1w)Y0 ( 2 w ) (5.9) . D2 2 J 0 ( 1w) J1( 2 w) 1J 0 ( 2 w) J1 ( 1w) j k j / j j a / j , j 1,2 In order to get this to look like the formula in [9, Chapter XII], we need to change the variable w v 1 and multiply the denominator by 1 2 and the numerator by 1 2 . This manipulation redefines D1 and D2 and puts a compensatory factor 1 2 in the coefficient to give J ( va ) D1J 0 ( vr ') D2Y0 ( vr ') e 1v t 2k k V 2 a 2 T2 ( r ', t ) 2 B2 1 2 1 2 1 dv, 2 2 a k 21 1 ( D D ) v 1 2 0 2 (5.10) D1( v ) k2 1 J 0 ( va )Y1( va ) k1 2 J1( va )Y0 ( va ) r' a D2 ( v ) k2 1 J 0 ( va ) J1( va ) k1 2 J 0 ( va ) J1( va ) = 1/ 2 The coefficient reduces to (5.11) 2k V 2 2k2 k12V 2 a 2 B2 12 1 a k 21 1 8 and this formula agrees with the [9, Chapter XII]. Application II In this application, also taken from Carslaw & Jaeger [9, Chapter XII, Section 128], we apply the general formula (1.1) to a problem with the same geometry and Green’s function. In this problem, we have perfect contact between cylinders with zero initial conditions, but there is constant heat production q1 Q1 of heat units per unit area per unit length of cylinder in Region 1 and no production in Region 2, q2 0 . Then (1.1) becomes T1 (P', t) (5.12) 1 t 1 G (P, P', t )q1 (P, )dVd + G 2 (P, P', t )q 2 (P, )dVd k1 0 V1 0 V2 1Q1 k1 t 1 1 t . G (r,r ', t )dVd 1 1 0 V1 The integration on r and the change of variables w v 1 are the same as that in Part I. Therefore (replace V with 1Q1 / k1 in (5.7)) t 4k k Q T1( r ', t ) 1 22 2 1 1 k1 0 a (5.13) J 0 ( vr ') J1( va ) ( D 2 D 2 )v 2 1 0 e1v 2 ( t ) dvd , r' a 2 D1( v ) k 2 1 J 0 ( va )Y1( va ) k1 2 J1( va )Y0 ( va ) = 1/ 2 D2 ( v ) k2 1 J 0 ( va ) J1( va ) k1 2 J 0 ( va ) J1( va ) The integration on produces T1( r ', t ) (5.14) 4Q1k2 2 (1 e1v t ) J 0 (vr ') J1(va ) a 2 2 ( D12 D22 )v 4 0 dv, r' a and this agrees with [9, Chapter XII]. To get T2 ( r ', t ) for Region 2, we take j=2 in (1.1), T2 (P', t) (5.15) 2 t t 2 2 G (P, P', t )q (P, )dVd + G (P, P', t )q (P, )dVd 2 2 1 1 0 V k 2 0 V2 1 2Q1 k2 t G (P,P', t )dVd 2 1 0 V1 Since the integration on r is the same as Part I, we use the form (5.10) from Part I ( by 2Q1 ): k2 2k 2 2 a 2 2Q1 T2 (r ', t) B2 1 2 a 1 k2 0 t (5.16) 0 k1 2V is replaced k 21 J1 (va) D1J 0 ( vr ') D2Y0 ( vr ') e 1v (D12 D22 )v D1 (v) k 2 1 J 0 (va)Y1 ( va) k1 2 J1 (va)Y0 ( va) D2 (v) k 2 1 J 0 (va)J1 ( va) k1 2 J 0 ( va)J1 (va) 9 2 ( t ) r' a = 1/ 2 dvd , or (1 e1v t )J1 (va) D1J 0 ( vr ') D2Y0 ( vr ') 2k 2 2 a 2 2Q1 1 2 (5.17) T2 (r ', t) B2 dv a 1 k2 1 0 (D12 D22 )v3 and this agrees with [9, Chapter XII] since the coefficients are the same: 2k 2 2 a 2 2Q1 1 2 2Q1 2 (5.18) . B2 a 1 k2 1 By using two different applications of the Green’s functions, we have provided a partial check on the j k 3- j j coefficients in the general formula (1.1) and a partial check on the formulas for T1( P ', t ) and kj k j3 j 2 and T2 ( P ', t ) . Application III In this application, we take the physical properties of both regions to be the same, 1 2 , k1 k2 k to confirm that the reduction yields the Green’s function for the whole plane, 0 r . The solution is given by the source solution, r 2 r '2 rr ' 1 S ( r, r ', t ) exp (5.19) I 0 . 4 t 4 t 2 t In this case, the parameters become 1 2 k / and 1 2 a / . The Wronskian J1( z )Y0 ( z ) J 0 ( z )Y1( z ) 2 /( z ) helps in the reduction process. In all of the formulas in (3.16), (3.17), (4.10) and (4. 11) the reduction comes down to 2 D1( w) , D2 ( w) 0 (5.20) w and the integral (5.21) 1 2 0 2 r r' J 0 ( w ) J 0 ( w )we w t dw , a a which can be evaluated from a table of Hankel transforms to yield (5.19). Application IV This example is instructive because, in using both terms of the initial distribution, we give credence to the formula and at the same time show, as we have observed in previous papers [3],[4],[5],[6], that working with the Laplace transform may be a better approach than a direct evaluation. We consider the problem with (5.22) f1( P ) V , 0 r a , f 2 ( P ) V , a r< , q1 q2 0 and try to apply (1.1). Since there is no gradient in the initial distribution, the temperature does not change and the solution to the conduction problem is 10 T1( r ', t ) V , 0 r' a (5.23) T2 ( r ', t ) V , a r'< . , To see what the difficulties are, we apply (1.1) using (3.16) and (3.17) with the temperature for r>a limited to a finite interval a r R in order to keep the integration finite, a T1( r ', t ) 2V (5.24) R G11( r, r ', t )rdr 0 Using the relations (5.25) J 0 ( z) zdz zJ1( z ) k 2V 2 1 G21 ( r, r ', t )rdr . k1 2 a Y0 ( z ) zdz zY1( z ) , the integrals are a (5.26) 0 R r a2 J 0 ( 1w )rdr J1( 1w) a 1w r r D1J 0 ( 2 w a ) D2Y0 ( 2 w a ) rdr (5.27) a a R a R D1 RJ 0 ( 2 w ) aJ 0 ( 2 w) D2 RY0 ( 2 w ) aY0 ( 2 w) 2w a a 2w and with (3.17) we have an indeterminate form r' R R J 0 ( 1w ) D1J 0 ( 2 w ) D2Y0 ( 2 w ) 2 ak1 a a a w2 t B1 lim R e dw (5.28) a 2 R 0 ( D12 D22 )w to evaluate. Basically, f 2 ( r ) does not have the proper behavior at infinity to give a convergent integral. However, there is a way around this complication. If we work with the Laplace transform, a (5.29) T1( r ', p ) 2V G11( r, r ', p )rdr 2V 0 (5.30) k21 G21 ( r, r ', t )rdr k1 2 a F ( r, r ', p ) G11( r, r ', p ) 1 F1( r ', r, p ) r' r a r r' a k1 B1I 0 ( q1r ')K 0 (q2r ) / D , ar a D q2k2 K1(q2a ) I 0 (q1a ) q1k1I1(q1a )K 0 (q2a ) , q j p / j , j 1,2 G21 ( r, r ', p ) (5.31) , q k K (q a ) I 0 (q1a )K 0 (q1r ) I 0 (q1r )K 0 (q1a ) F1( r, r ', p ) B1I 0 ( q1r ') 2 2 1 2 / D q1k1K 0 ( q2a ) I1( q1a )K 0 (q1r ) I 0 (q1r )K1(q1a ) there is no problem with the integration because of the exponential decay of the K functions, but one has to go through the inversion process again. The integration on G11 takes the form a (5.32) G11( r, r ', p )rdr while the integration on a 0 r' F1( r ', r, p )rdr F1( r, r ', p )rdr 0 G21 r' is simple 11 (5.33) G21 ( r, r ', p )rdr a k B I (q r ') 1 1 0 1 K 0 ( q2r )rdr . a D a In each of these cases, we use the integrals (5.34) I0 ( z) zdz zI1( z) K0 ( z) zdz zK1( z ) . The exponential decay of the K functions guarantees the convergence. We need the integrals a (5.35) r' r' a K 0 ( q1r )rdr K1( q1r ') K1(q1a ) , q1 q1 r' (5.36) I 0 ( q1r )rdr 0 r' I1(q1r ') , q1 a I 0 (q1r )rdr r' K 0 ( q2 r )rdr a a r' I1(q1a ) I1(q1r ') q1 q1 a K1(q2a ) . q2 Then, for r r ' substitution and reduction using the Wronskian (3.9 ) gives 1 q2k2 K1( q2a ) I 0 ( q1a ) K1( q1r ') q r ' a I (q r ')K (q a ) 1 r' 0 1 1 1 / D . (5.37) F1(r, r ', p)rdr B1 q1 I 0 (q1r ') I1( q1a ) K1( q1r ') r' q1k1K 0 ( q2a ) I1( q1r ') K1( q1a ) For the integral where r ' r we have r' r' q2k 2 K1( q2a ) I 0 ( q1a ) K 0 ( q1r ') I 0 ( q1r ') K 0 ( q1a ) F1( r ', r, p )rdr B1 I 0 (q1r )rdr q1k1K0 (q2a ) I1(q1a )K 0 (q1r ') I 0 (q1r ')K1(q1a ) / D 0 0 (5.38) q k K ( q a ) I 0 ( q1a ) K 0 ( q1r ') I 0 ( q1r ') K 0 ( q1a ) r' B1 I1( q1r ') 2 2 1 2 /D q1 q1k1K 0 ( q2a ) I1( q1a ) K 0 ( q1r ') I 0 ( q1r ') K1( q1a ) and, with the Wronskian again, (5.37) and (5.38) combine to give the sum in (5.32) I ( q a ) I ( q r ') q2k2 K1( q2a ) 0 1 0 1 a q1r ' q1r ' r' 1 G ( r , r ', p ) rdr B (5.39) 1 /D 1 q 1 I (q a ) 0 q k K (q a ) 1 1 1 1 0 2 q r' 1 After noting q2k2 / q12 k21 /( 2 p ) k21 /( 2 q2 ) , (5.40) we combine (5.39) with (5.41) k21 k k B I (q r ') a k G21 ( r, r ', p )rdr 2 1 1 1 0 1 K1(q2a ) 2 1 B1I 0 (q1r ')K1(q2a ) / D k1 2 a k1 2 a D q2 2q2 to get the main result for (5.29), a T1( r ', p ) k G11( r ', r, p )rdr 2 1 G21 ( r, r ', p )rdr 2V k1 2 a (5.42) 0 2 B1 ( q2k2 / q1 ) K1( q2a ) I 0 ( q1a ) ( q1k1 / q12 ) K 0 ( q2a ) I1( q1a ) / D B1 / q12 where we have used the definition of D in (5.31) in the reduction. Then the transform becomes T1( r ', p) 2VB1 / q12 V / p (5.43) 12 since B1 1/(21) and q1 p / 1 . The inversion gives what one expects T1( r ', t ) V (5.44) , 0 r' a . The result for T2 ( r ', t ) should be similar. References [1] Abramowitz S, Stegun IA (1965) Handbook of Mathematical Functions, AMS 55, Dover Publications Inc., New York, 1046pp [2] Amos DE (2006) Handbook of Integrals Related to Heat Conduction and Diffusion, http://nanohub.org/resources/13874 [3] Amos DE, Beck JV, de Monte F (2011) Transient Heat Conduction in Adjacent Quadrants Separated by a Thermal Resistance, http://nanohub.org/resources/12465 [4] Amos DE (2011) Transient Heat Conduction in Adjacent Materials Heated on Part of the Common Boundary, http://nanohub.org/resources/12390 [5] Amos, DE (2012), Green’s Functions For Heat Conduction in Adjacent Materials, http://nanohub.org/resources/12856 [6] Amos, DE (2012), Theory of Heat Conduction for Two-region Problems Using Green’s Functions, http://nanohub.org/resources/13671 [7] Amos, DE (2012), 1-D Green’s Functions For Heat Conduction Between Semi-infinite Slabs With Perfect and Imperfect Boundary Contact, http://nanohub.org/resources/15237. [8] Cole DC, Beck JV, Haji-SheikhA, Litkouhi B (2010) Heat Conduction Using Green’s Functions, 2nd Ed., CRC Press, 643p. [9] Carslaw HS, Jaeger JC (1948) Conduction of Heat in Solids, Oxford Univ Press, London, 386pp APPENDIX A Calculation of the integrands for (3.15) Evaluation of F1( r, r ', p ) and G2 ( r, r ', p ) on the contour p w2ei , 0 w Because of the symmetry in the cases for r r ' and r r ' we consider only the case for r r ' . We take F1( r, r ', p ) , q k K ( q a ) I 0 ( q1a ) K 0 ( q1r ) I 0 ( q1r ) K 0 ( q1a ) F1( r, r ', p ) B1I 0 ( q1r ') 2 2 1 2 / D , q1k1K 0 ( q2a ) I1( q1a ) K 0 ( q1r ) I 0 ( q1r ) K1( q1a ) and start by applying the analytic continuation relations (A.1) (A.2) K0 ( zei / 2 ) K1( zei / 2 ) 2 2 i J 0 ( z ) iY0 ( z ) = J1( z ) iY1( z ) , iJ 0 ( z ) Y0 ( z ) 2 I 0 ( zei / 2 ) J 0 ( z ) , 13 I1( zei / 2 ) iJ1( z ) and the Wronskian (A.3) J1( z )Y0 ( z ) J 0 ( z )Y1( z ) 2 /( z ) to F1( r, r ', p ) . Then with the definitions, j kj / j , (A.4) j a/ j we have D q2k2 K1(q2a ) I 0 (q1a ) q1k1I1(q1a )K 0 (q2a ) w 2 D1 iD2 (A.5) , , j 1,2 iw 2 J 0 ( 1w) J1( 2 w) iY1( 2 w) 2 iw1J1( 1w) J 0 ( 2 w) iY0 ( 2 w) 1 2 D1 iD2 D w D12 D22 . D1 2 J 0 ( 1w)Y1( 2 w) 1J1( 1w)Y0 ( 2 w) D2 2 J 0 ( 1w) J1( 2 w) 1J 0 ( 2 w) J1( 1w) The first term of F1( r, r ', p ) in braces is q2k2 K1(q2a ) p w e 2iwK1(i 2w) 2iw J1( 2w) iY1( 2w) 2 i 2 I 0 (q1a)K0 (q1r ) I 0 (q1r )K0 (q1a ) p w e 2 i (A.6) r r J 0 ( 1w )Y0 ( 1w) J 0 ( 1w)Y0 ( 1w ) 2 a a 2 ( r, w, 1 ) (note is real and ( a, w, 1) 0) q2k2 K1(q2a ) I 0 (q1a )K0 (q1r ) I 0 (q1r )K0 (q1a ) p w e 2 i and the second term is q1k1K0 (q2a ) p w e 2 i i1wK 0 (i 2 w) 1w I1(q1a )K0 (q1r ) I 0 (q1r )K1(q1a ) p w e 2 i (A.7) i 2 2 w 2 4 Y1( 2 w ) i 2w 2 4 J1( 2w ) J 0 ( 2w) iY0 ( 2w) r r J1( 1w)Y0 ( 1w ) J 0 ( 1w )Y1( 1w) 2 a a = i ( r, w, 1 ) 2 q1k1K0 (q2a ) I1(q1a )K0 (q1r ) I 0 (q1r )K1(q1a ) p w2 ei . (note is real and ( a, w, 1 ) 2 /( 1w) 1w 2 4 Y0 ( 2 w ) i1w 2 4 J 0 ( 2 w ) Then, with (A.6) and (A.7), F1( r, r ', p ) becomes F1(r, r ', p) p w e 2 i B1J 0 ( 1wr '/ a ) (A.8) B1J 0 ( 1wr '/ a ) 2 2 wY1( 2 w) ( r, w, 1 ) i 2 wJ1( 2 w) ( r, w, 1 ) / D( w) 4 1wY0 ( 2 w) ( r, w, 1 ) i1wJ 0 ( 2 w) ( r, w, 1 ) 2 w 2Y1( 2 w) ( r, w, 1 ) 1Y0 ( 2 w) ( r, w, 1 ) 2 D iD 1 2 4 i 2 J1( 2 w) ( r, w, 1 ) 1J 0 ( 2 w) ( r, w, 1 ) w D12 D22 or 14 D1 2Y1( 2 w) ( r, w, 1 ) 1Y0 ( 2 w) ( r, w, 1 ) F1(r, r ', w2ei ) B1 J 0D(21wrD'/2a) , 2 D J ( w) (r, w, ) J ( w) (r, w, ) , 1 (A.9) +iB1 2 2 2 1 2 1 1 0 2 1 J 0 ( 1wr '/ a ) D1 2 J1( 2 w) ( r, w, 1 ) 1J 0 ( 2 w) ( r, w, 1 ) , D12 D22 2 D2 2Y1( 2 w) ( r, w, 1 ) 1Y0 ( 2 w) ( r, w, 1 ) and finally with r a r a ( r, w, 1 ) J 0 ( 1w )Y0 ( 1w) J 0 ( 1w)Y0 ( 1w ) (A.10) r a r a ( r, w, 1 ) J1( 1w)Y0 ( 1w ) J 0 ( 1w )Y1( 1w) we get r' Im F1( r, r ', p ) p w 2 ei B1 J 0 ( 1w a ) D1 2 J1( 2 w) ( r, w, 1 ) 1J 0 ( 2 w) ( r, w,) 1 2 D12 D22 D2 2Y1( 2 w) ( r, w, 1 ) 1Y0 ( 2 w) ( r, w, 1 ) r' (A.11) B1 B1 where (A.12) J 0 ( 1w a ) 2 D1J1( 2 w) 2 D2Y1( 2 w) ( r, w, 1 ) 2 D12 D22 1D1J 0 ( 2 w) 1D2Y0 ( 2 w) ( r, w, 1 ) r ' A( w) ( r, w, 1 ) B( w) ( r, w, 1 ) J 0 ( 1w ) 2 a D12 D22 A( w) 2 D1J1( 2 w) D2Y1( 2 w) , D1( w) 2 J 0 ( 1w)Y1( 2w) 1J1( 1w)Y0 ( 2w) B( w) 1 D1J 0 ( 2 w) D2Y0 ( 2 w) , D2 ( w) 2 J 0 ( 1w) J1( 2 w) 1J 0 ( 2w) J1( 1w) Further reduction with (A.10) and the Wronskian (A.3) gives 2 J ( w) A( w) 2 D1J1( 2 w) D2Y1( 2 w) = 1 2 1 1 2 w (A.13) B( w) 1 D1J 0 ( 2 w) D2Y0 ( 2 w) = A( w) ( r, w, 1 ) B( w) ( r, w, 1 ) 21 2 J 0 ( 1w) 2 w 41 2 r J 0 ( 1w ) a 1 2 w 2 2 and with j k j / j j a / j , j 1,2 we have (A.14) Im F1( r, r ', p ) p w2 ei r' r J 0 ( 1w ) J 0 ( 1w ) 2 k1k2 a a B1 a 2 w2 D12 D22 For G2 ( r, r ', p ) on p w2ei we have (A.15) G2 ( r, r ', p ) 1 B1I 0 ( q1r ') K 0 ( q2 r ) / D , 1 Then, 15 a r . . r' r r J 0 ( 1w ) i J 0 ( 2 w ) iY0 ( 2 w ) a 2 a a and the denominator from (A.5) gives the product I 0 (q1r ')K 0 (q2r ) p w e (A.16) (A.17) 2 i 1 r' r r r r J 0 ( 1w ) D1Y0 ( 2 w ) D2 J 0 ( 2w ) i D1J 0 ( 2w ) D2Y0 ( 2w ) /( D12 D22 ) . w a a a a a Then (A.18) Im G2 ( r, r ', p ) 2 i pw e k1 1 r' r r B1 J 0 ( 1w ) D1J 0 ( 2 w ) D2Y0 ( 2 w ) /( D12 D22 ) a w a a a Evaluation of F1( r, r ', p ) and G2 ( r, r ', p ) on the contour p ei , for 0 We have completed the combination of the contours along the top and bottom of the branch, but we still need to evaluate the line integral around the origin with the radius and take to zero, 1 lim 0 2 (A.19) F1(r, r ', p)e pt p e i ei d . This requires the evaluation of q k K (q a ) I 0 (q1a )K 0 (q1r ) I 0 (q1r )K 0 (q1a ) F1( r, r ', p ) B1I 0 (q1r ') 2 2 1 2 / D q k K ( q a ) I ( q a ) K ( q r ) I ( q r ) K ( q a ) 1 1 0 1 0 1 1 1 11 0 2 (A.20) D q2k2 K1( q2a ) I 0 (q1a ) q1k1I1(q1a )K 0 (q2a ) with p ei for small . The power series representations for the I and K functions are needed to get the proper limits: K0 ( z ) ln( z / 2) O( z 2 ln z ) , I 0 ( z ) 1 O( z 2 ) , I1( z ) z / 2 O( z 3 ) (A.21) . 1 K1( z ) ( z / 2)ln( z / 2) ( z / 2)( 1/ 2) O( z 3 ln z) z The limit lim ln b 0 in the form lim ln b 0 helps in the evaluation. Then, the only terms 0 0 which give a non-zero contribution for 0 are D q2k2 K1(q2a ) I 0 (q1a ) q1k1I1(q1a )K 0 (q2a ) (A.22) F1( r, r ', p ) k2 / a B1I 0 ( q1r ')q2k2 K1(q2a ) I 0 (q1a )K 0 (q1r ) / D O B1( k2 / a) K 0 ( q1r ) /( k2 / a) O ln( ei / 2 1r / a) / 2 One can see clearly now that the limit in (A.15) is (A.23) 1 lim 0 2 F1( r, r ', p)e pt p e i ei d 0 . Similarly for G2 ( r, r ', p ) , it is clear that in the integrand dominates the logarithmic term in K0 (q2r ) , (A.24) G2 ( r, r ', p ) 1 B1I 0 ( q1r ') K 0 ( q2 r ) / D , 1 and 16 ar 1 lim 0 2 (A.25) G2 ( r, r ', p)e pt p e i ei d 0 . APPENDIX B Calculation of the integrands for (4.9) Evaluation of F2 ( r, r ', p ) on the contour p w2ei , 0 w Because of the symmetry of F2 ( r, r ', p ) in the cases for r r ' and r r ' we consider only the case for r r ' and note that the computation of D is the same as in APPENDIX A: w D = D1 iD2 2 1 2 D1 iD2 D1 2 J 0 ( 1w)Y1( 2 w) 1J1( 1w)Y0 ( 2 w) (B.1) D w D12 D22 D2 2 J 0 ( 1w) J1( 2 w) 1J 0 ( 2 w) J1( 1w) We need the Bessel function relations for the values on the top of the cut: (B.2) K0 ( zei / 2 ) K1( zei / 2 ) 2 2 i J 0 ( z ) iY0 ( z ) = J1( z ) iY1( z ) iJ 0 ( z ) Y0 ( z ) 2 . , I 0 ( zei / 2 ) J 0 ( z ) , I1( zei / 2 ) iJ1( z ) and the Wronskian J1( z )Y0 ( z ) J 0 ( z )Y1( z ) 2 /( z ) is helpful in some reductions. Then, q k I ( q a ) I 0 ( q2 r ')K1(q2a ) I1(q2a )K 0 (q2 r ') F2 ( r, r ', p ) B2 K 0 ( q2 r ) 2 2 0 1 / D q k I ( q a ) I ( q r ') K ( q a ) I ( q a ) K ( q r ') 1 1 1 1 0 2 0 2 0 2 0 2 The first term in braces is (B.3) q1k1I1(q1a ) p w e 1iwI1(i 1w) 1wJ1( 1w) 2 i I 0 (q2r ')K0 (q2a) I 0 (q2a)K0 (q2r ') p w e 2 i (B.4) r' r' J 0 ( 2 w )Y0 ( 2 w) J 0 ( 2 w)Y0 ( 2 w ) 2 a a ( r ', w, 2 ) (note is real and ( a, w, 2 ) 0) 2 q1k1I1(q1a ) I 0 (q2r ')K0 (q2a ) I 0 (q2a )K0 (q2r ') p w2ei 2 1wJ1( 1w) (r ', w, 2 ) and the second term is q2k2 I 0 (q1a ) p w2 ei i 2wI 0 (i 1w) i 2wJ 0 ( 1w) I 0 (q2 r ')K1(q2a ) I1(q2a )K 0 (q2r ') p w e 2 i (B.5) i r' r' J ( w ) Y ( w ) J ( w )Y1( 2 w) 1 2 0 2 0 2 2 a a = i ( r ', w, 2 ) (note is real and ( a, w, 2 ) 1/( 2 w)) 2 q2k2 I 0 (q1a ) I 0 (q2 r ')K1(q2a ) I1(q2a )K 0 (q2r ') p w e 2 i 17 2 2 wJ 0 ( 1w) ( r ', w, 2 ) Then, F2 ( r, r ', p ) becomes F2 (r, r ', p ) p w e 2 i (B.6) = = or 2 wJ 0 ( 1w) ( r ', w, 2 ) i J 0 ( 2 wr / a ) iY0 ( 2 wr / a ) / D( w) 2 2 1wJ1( 2 w) ( r ', w, 2 ) iJ 0 ( 2 wr / a ) 2 D1 iD2 B2 2 J 0 ( 1w) ( r ', w, 2 ) 1J1( 2 w) ( r ', w, 2 ) 2 2 Y0 ( 2 wr / a ) w D12 D22 w 2 B2 2 J 0 ( 1w) ( r ', w, 2 ) 1J1( 2w) ( r ', w, 2 ) D12 D22 Im F2 ( r, r ', p ) = (B.7) B2 2 B2 p w2 ei D1Y0 ( 2 wr / a ) D2 J 0 ( 2 wr / a ) i D1J 0 ( 2 wr / a ) D2Y0 ( 2 wr / a ) 2 J 0 ( 1w) ( r ', w, 2 ) 1J1( 2w) ( r ', w, 2 ) D1J 0 ( 2wr / a ) D2Y0 ( 2wr / a ) D12 D22 r' r' ( r ', w, 2 ) J 0 ( 2 w )Y0 ( 2 w) J 0 ( 2 w)Y0 ( 2 w ) a a r' a r' a ( r ', w, 2 ) J1( 2 w)Y0 ( 2 w ) J 0 ( 2 w )Y1( 2 w) For G1 we have G1( r, r ', p ) k2 B2 K 0 ( q2 r ') I 0 ( q1r ) / D a G1(r, r ', p) p w e 2 i (B.8) k2 B2 J 0 ( 1wr / a ) i J 0 ( 2 wr '/ a ) iY0 ( 2 wr '/ a ) / D a 2 = k2 B2 J 0 ( 1wr / a) Y0 ( 2 wr '/ a ) iJ 0 ( 2 wr '/ a ) 2 D1 iD2 w D12 D22 2a k2 B2 J 0 ( 1wr / a ) D1Y0 ( 2 wr '/ a ) D2 J 0 ( 2 wr '/ a ) /( D12 D22 ) aw i k2 B2 J 0 ( 1wr / a ) D1J 0 ( 2 wr '/ a ) D2Y0 ( 2 wr '/ a ) /( D12 D22 ) aw and (B.9) ImG1( r, r ', p ) 2 i pw e k2 J ( wr / a ) B2 0 2 1 2 D1J 0 ( 2wr '/ a ) D2Y0 ( 2wr '/ a) aw D1 D2 Evaluation of F2 ( r, r ', p ) and G1( r, r ', p ) on the contour p ei , - An inspection of (B.3) with the relations in (A.21) shows that the only terms which will produce a nonzero contribution for 0 are 18 k 2 / a for 0 D q2k2 K1( q2a ) I 0 (q1a ) q1k1I1(q1a )K 0 (q2a ) (B.10) F2 ( r, r ', p ) B2 K 0 ( q2 r )q2k 2 I 0 (q1a ) I 0 (q2r ')K1(q2a ) / D O K 0 (q2r ) O ln(q2r ) / 2 O ln( ei / 2 2 r / a ) / 2 Therefore, we have (B.11) 1 lim 0 2 F2 (r, r ', p)e pt p e i ei d 0 Similarly for G1( r, r ', p ) , it is clear that in the integrand dominates the logarithmic term in K0 (q2r ') , k G1( r, r ', p ) 2 B2 K 0 ( q2 r ') I 0 ( q1r ) / D , (B.12) a and (B.13) 1 lim 0 2 G1(r, r ', p)e pt p e 19 i ei d 0 .