Supplementary Material to: Restricted Diffusion in Annular Geometrical Pores Bahman Ghadirian, Allan M. Torres, Nirbhay N. Yadav and William S. Price This supplementary material contains the derivations of the important equations that constitute the results of this paper. It also contains the cross checks with the limiting cases that can be found in the literature. Contents Concentric Cylinder 2 1. Diffusion Propagator 2 2. Pulse Gradient Spin-Echo Attenuation in a Concentric Cylinder 8 3. Mean Square Displacement in a Concentric Cylinder 17 Concentric Sphere 21 1. Diffusion Propagator 21 2. Pulse Gradient Spin-Echo Attenuation in a Concentric Sphere 28 3. Mean Square Displacement in a Concentric Sphere 33 The limiting cases 36 1. Concentric Cylinder 36 2. Concentric Sphere 39 REFERENCES 40 A. B. C. D. 1 A. Concentric Cylinder 1. Diffusion Propagator (a) Zero eigenvalue solution For the case of λ = 0 (i.e., zero eigenvalue), 2 0 0 . (S1) Consequently, the general form of the spatial part of the propagator is also given by a constant (arbitrarily set to 1) 0 BRZ C 1, (S2) where B and C are constants and that cannot simultaneously be equal to zero (i.e., a trivial solution). Thus, P(r , , z , t ) 0 0T 0 A0 . (S3) The normalisation constant is determined by N 0 0 0 dr dr L a 2 b 2 . V V Therefore P 0 r0 , 0 , z0 , r1 ,1 , z1 , t 0 r0 0 r1 N 0 N 0 2 1 . L a 2 b2 (S4) (b) Positive eigenvalue solution When 2 2 0, where the spatial eigenvalues (separation constants) , and n (a non-negative integer) are defined by 2 2 n2 1 2 Z 1 2 R R 1 1 2 n 2 , , and r 2 . Z z 2 r 2 2 Rr r 2 r r r The solution for the Z component is Z A1 cos( z ) A2 sin( z ). (S5) The solution for the Φ component is B1 cos(n ) B2 sin(n ) (S6) where B1 and B2 are constants. According to the periodic boundary condition B2 = 0, therefore B1 cos(n ). From 2 R 1 R 2 n2 2 r 2 r r r (S7) R 0, the solution is given by R r En,m Cn ( nm r ), 3 (S8) where En,m are constants corresponding to each eigenvalue. Cn ( nm r ) F1 J n ( nm r ) F2Yn ( nm r ), where the index m is a non-negative integer, F1 and F2 are constants and Jn and Yn are Bessel functions of the first and second kind respectively. The boundary condition on the inner surface of the top of the cylinder is D M L /2 Z z 0 (S9) z L /2 where ML/2 is the relaxivity on the inner surface of the cylinder, when applied to the cos function in Eq. (S5) gives L M s tan s L / 2 . D 2 (S10) This root function also holds for the boundary condition at z = -L/2 (but changing M-L/2). The boundary condition on the inner surface of the bottom of the cylinder is D M L /2 Z z 0 (S11) z L /2 where M-L/2 is the relaxivity. Applied to the sin function in Eq. (S5) gives the root function M u L L/2 . D 2 u cot 4 (S12) This root function also holds for the boundary condition at z = L/2 (but changing ML/2). Thus the z-component eigenfunction is s 0 u 0 Z As cos( s z ) Au sin( u z ). (S13) The normalisation constants are L /2 z sin(2 s z ) L sin( s L) N s A cos( s z )cos(v z )dz 1 . L /2 4 s L /2 2 s L 2 2 s L /2 and L /2 z sin(2 u z ) L sin( u L) Nu A sin( u z )sin( w z )dz 1 . L /2 4 u L /2 2 uL 2 2 u L /2 The inner radial boundary condition is given by 0 D Ma R r r a (S14) where Ma is the relaxivity on the outer annular surface at r = a, when applied to the J solution the following root function is obtained nl J n ( nl a) M a. J n ( nl a) D From the Bessel function relation1 5 (S15) Yn x J n x cos n J n x , sin n (S16) Yn x J n x cos n J n x , sin n (S17) and thus also the following property J n x 1 J n x , this boundary condition for the Y solution give the n same eigenvalues as for the J solution. The inner radial boundary condition is given by D M b R 0 r r b (S18) where Mb is the relaxivity on the outer annular surface at r = b. Again applying this operator to the Y solution gives the following root function for computing the corresponding eigenvalues nk Yn nk b Yn nk b Mb . D (S19) Regardless of whether the J or Y solution is used on this boundary the eigenvalues will be the same. Therefore the radial solution is given by 6 R Fl J n nl r Fk Yn nk r . n 0 l 0 (S20) n 0 k 0 Thus the diffusion propagator for the annular cylinder will be of the form P(r , , z, t ) H nls J n nl r cos n cos s z exp nl2 s2 Dt n 0 l 0 s 0 H nlu J n nl r cos n sin u z exp nl2 u2 Dt n 0 l 0 u 0 (S21) H nks Yn nk r cos n cos s z exp nk2 s2 Dt n 0 k 0 s 0 H nku Yn nk r cos n sin u z exp nk2 u2 Dt . n 0 k 0 u 0 The above propagator is not complete yet as it should be normalised. For this purpose we use the definition of normalisation for each term in the above expression and boundary conditions of the radial eigenfunctions at r = a and r = b, thus N1 J n nl r cos n cos s z J m mu r cos m cos v z dv 2 n2 L sin s L 2 2 1 a J a a 2 n nl 4 s L nl 2 n2 2 2 2 J a b J b n nl b 2 n nl nl (S22) 2 J b . n nl Note since we put F2 = 0 corresponding to the boundary condition on r = a, the second part of the radial eigenfunction has no contribution to the normalisation factor. Similarly, the following normalisation factors are obtained for the second, third and last part of the propagator, N2 N3 L sin u L 2 2 n2 2 a J a a 2 u L n nl nl 2 n2 2 2 2 J a b J b n nl b 2 n nl nl 2 J n nl b , (S23) L sin s L 2 2 n2 2 a Y a a 2 s L n nk nk 2 n2 2 2 2 Y a b Y b n nk b 2 n nk nk 2 Yn nk b , (S24) 1 4 1 4 and 7 N4 L 1 4 sin u L 2 2 n2 2 a Y a a 2 u L n nk nk 2 n2 2 2 2 Y a b Y b n nk b 2 n nk nk 2 Yn nk b . (S25) Finally the normalised diffusion propagator in an annular cylinder with inner and outer radius equal to b and a, respectively and length L is given by P r0 , 0 , z0 , r1 ,1 , z1 , t 4 J n nl r1 cos n1 cos s z1 J n nl r0 cos n 0 cos s z0 exp nl2 s2 Dt n0 l 0 s 0 L 1 4 J n nl r1 cos n1 cos u z1 J n nl r0 cos n 0 cos u z0 exp nl2 u2 Dt n 0 l 0 u 0 sin s L 2 2 n2 2 n2 2 2 2 2 2 a J n nl a a 2 J n nl a b J n nl b b 2 J n nl b s L nl nl L 1 n 0 k 0 s 0 L 1 n 0 k 0 u 0 2 n2 2 2 2 J a b J b n nl b 2 n nl nl sin s L 2 2 n2 2 a Yn nk a a 2 s L nk 2 n2 2 2 2 Yn nk a b Yn nk b b 2 nk 2 J n nl b 4Yn nk r1 cos n1 cos s z1 Yn nk r0 cos n 0 cos s z0 exp nk2 s2 Dt sin u L 2 2 n2 2 a J a a 2 u L n nl nl 2 Yn nk b 4Yn nk r1 cos n1 cos u z1 Yn nk r0 cos n 0 cos u z0 exp nk2 u2 Dt sin u L 2 2 n2 2 n2 2 2 2 2 2 L 1 a Yn nk a a 2 Yn nk a b Yn nk b b 2 Yn nk b uL nk nk . (S26) 2. Pulse Gradient Spin-Echo Attenuation in a Concentric Cylinder (a) Zero eigenvalue contribution Substituting the propagator for = 0 for the concentric cylinder (Eq. (11)) in to the SGP equation (Eq. (1)), where ρ(r0) is the equilibrium spin density (equal to 1/(volume) of the restricting pore) at the point r0 and initial time zero gives, 8 E q, 0 r0 P 0 r0 , r1 , e 4 a b L/2 a 0 L/2 b 0 i 2 q r1 r0 dr0 dr1 i 2 q r r 1 1 L / 2 L a 2 b2 L a 2 b2 e 1 0 r0 r1dr0 d0 dz0 dr1d1dz1 L/2 4 I I , 2 L2 a 2 b 2 2 r0 r1 where I r0 a b L/2 0 L/2 exp i 2 r0 q sin cos 0 z0 q cos r0 dr0 d 0 dz0 . I r1 is the same as I r0 with r1, θ1 and z1 instead of r0, θ0 and z0 and negative exponential term. Thus, I r0 a b 0 r0 exp i 2 qr0 sin cos 0 L/2 L/2 a b 0 exp i 2 z0 q cos dz0 d 0 dr0 1 i 2 q cos exp 2i qL cos 1 a 1 r0 exp i 2 qr sin cos 0 d 0 dr0 i 2 q cos exp i qL cos b 0 1 i 2 q sin 2 2 r0 exp i 2 qr0 sin cos 0 exp i qL cos exp i qL cos d 0 dr0 exp 2i qL cos 1 aJ1 2 qa sin bJ1 2 qb sin . exp i qL cos Similarly for I r the following expression is obtained 1 I r1 exp 2i qL cos 1 1 aJ1 2 qa sin bJ1 2 qb sin . 2 2 i 2 q sin exp i qL cos Then the product of these two integrals will be 1 cos 2 qL cos 2 I r0 I r1 aJ1 2 qa sin bJ1 2 qb sin 2 4 2 2 q sin 2 9 (S27) The echo attenuation corresponding to the zero-eigenvalue becomes E q, 0 2 1 cos 2 qL cos aJ1 2 qa sin bJ1 2 qb sin L q sin 2 a b 4 2 4 2 2 2 2 2 . (S28) (b) Positive eigenvalue contribution Substituting the propagator for > 0 for the concentric cylinder (Eq. (12)) in to the SGP equation (Eq. (1)) gives, E q, 4 a b L/2 a L/2 0 L/2 b 0 L/2 r0 P r0 , r1 , ei 2 qr r r0 r1dr1d1dz1dr0 d0 dz0 . 1 0 (S29) This multiple integrals are treated as follows, E q, n0 l 0 s 0 n 0 l 0 u 0 n 0 k 0 s 0 n 0 k 0 u 0 16 exp nl2 s2 D I r0 I r1 a b 2 L2 2 2 sin s L 2 2 n2 2 n2 2 2 2 2 1 a J a a J a b J b b 2 n nl s L n nl nl2 n nl nl 16 exp nl2 u2 D I r0 I r1 2 2 2 2 a b L 2 J n nl b sin u L 2 2 n2 2 n2 2 2 2 2 1 a J a a J a b J b b 2 n nl u L n nl nl2 n nl nl 16 exp nk2 s2 D I r0 I r1 2 2 a b 2 L2 J n nl b sin s L 2 2 n2 2 n2 2 2 2 2 1 a Y a a Y a b Y b b 2 n nk s L n nk nk2 n nk nk 16 exp nk2 u2 D I r0 I r1 2 2 2 2 a b L 2 Yn nk b sin u L 2 2 n2 2 1 a Yn nk a a 2 uL nk 2 Yn nk b 2 n2 2 2 2 Yn nk a b Yn nk b b 2 nk (S30) 10 , where L/2 0 L/ 2 0 a L/2 b 0 L/ 2 0 I r0 a I r0 b I a r0 L/2 r J n nl r0 cos n 0 cos s z0 exp i 2 r0 q sin cos 0 zq cos dr0 d 0 dz0 , r J n nl r0 cos n 0 cos u z0 exp i 2 r0 q sin cos 0 zq cos dr0 d 0 dz0 , r0Yn nl r0 cos n 0 cos s z0 exp i 2 r0 q sin cos 0 zq cos dr0 d 0 dz0 , b 0 L/ 2 a L/2 0 L/ 2 0 n I r0 b (S31) r Y nl r0 cos n 0 cos u z0 exp i 2 r0 q sin cos 0 zq cos dr0 d 0 dz0 . The corresponding integrals with r1 are the same as Eq. (S31) but with r1, θ1 and z1 instead of r0, θ0 and z0 and negative exponential terms. Next calculating each integral by parts, the first integral will be subdivided as ei 2 q cos z i 2 q cos cos s z0 s sin s z0 cos z exp i 2 q cos z dz s 0 0 L / 2 s2 4 2 q 2 cos 2 L / 2 L/2 L/2 L L L L ei q cos L i 2 q cos cos s s sin s e i q cos L i 2 q cos cos s s sin s 2 2 2 2 2 2 2 2 s 4 q cos L L 2 s cos qL cos sin s 4 q cos sin qL cos cos s 2 2 . 2 2 2 2 s 4 q cos Similarly the same expression is obtained for the integral corresponding to r0, L/2 L/2 cos s z0 exp i 2 q cos z0 dz0 L/2 L/2 cos s z0 exp i 2 q cos z0 dz0 2 s L L 4 q cos sin qL cos cos s 2 s cos qL cos sin 2 2 . 2 2 2 2 s 4 q cos Also the second and third parts of the integral can be evaluated together as 11 a b 0 i n r0 J n nl r0 cos n 0 exp i 2 qr sin cos 0 d 0 dr0 2 qa sin J n nl a J n 2 qa sin nl a J n nl a J n 2 qa sin 2 qb sin J n nl b J n 2 qb sin nl b J n nl b J n 2 qb sin nl2 4 2 q 2 sin 2 . The integral corresponding to r0, which contains a negative exponential instead, is given by a b 0 r0 J n nl r0 cos n 0 exp i 2 qr0 sin cos 0 d 0 dr0 2 qa sin J n nl a J n 2 qa sin nl a J n nl a J n 2 qa sin 2 qb sin J n nl b J n 2 qb sin nl b J n nl b J n 2 qb sin n i . nl2 4 2 q 2 sin 2 Then I r0 I r1 i 2 1 2 n n 2 qa sin J n nl a J n 2 qa sin nl a J n nl a J n 2 qa sin 2 qb sin J n nl b J n 2 qb sin nl b J n nl b J n 2 qb sin nl2 2 q sin 2 2 2 2 s L s L 2 s cos qL cos sin 2 4 q cos sin qL cos cos 2 , 2 2 2 s 2 q cos which after simplification gives 2 qa sin J n nl a J n 2 qa sin nl a J n nl a J n 2 qa sin 2 qb sin J n nl b J n 2 qb sin nl b J n nl b J n 2 qb sin 2 cos qL cos sin s L 4 q cos sin qL cos cos s L s 2 2 I r0 I r1 nl2 2 q sin 2 s2 2 q cos 2 12 2 (S32) The second integral I r , is again subdivided to be determined, viz. 0 ei 2 q cos z i 2 q cos sin u z0 u cos u z0 L / 2 sin u z0 exp i 2 q cos z0 dz0 u2 4 2 q 2 cos 2 L / 2 L/2 L/2 L L L L ei q cos L i 2 q cos sin u u sin u e i q cos L i 2 q cos sin u u cos u 2 2 2 2 2 2 2 2 u 4 q cos uL L i 2 u sin qL cos cos u 2 2 . 2 2 2 2 u 4 q cos i 4 q cos cos qL cos sin Similarly the above expression for the integral corresponding to r0 is given by L /2 L /2 sin u z0 exp i 2 q cos z0 dz0 L L i 4 q cos cos qL cos sin u i 2 u sin qL cos cos u 2 2 . 2 2 2 2 u 4 q cos Then L/2 L/2 sin u z0 exp i 2 q cos z0 dz L/2 L/2 sin u z0 exp i 2 q cos z0 dz0 2 uL L 4 q cos cos qL cos sin u 2 u sin qL cos cos 2 2 . 2 2 2 2 u 4 q cos Again the following product of two multiple integrals after simplification gives 13 (S33) 2 qa sin J n nl a J n 2 qa sin nl a J n nl a J n 2 qa sin 2 qb sin J n nl b J n 2 qb sin nl b J n nl b J n 2 qb sin 2 sin qL cos cos u L 4 q cos cos qL cos sin u L u 2 2 I r0 I r1 2 2 2 2 nl 2 q sin u 2 q cos 2 (S34) The third integral I r , is determined similarly, 0 L /2 L /2 cos s z0 exp i 2 q cos z0 dz0 L L 2 s cos qL cos sin s 4 q cos sin qL cos cos s 2 2 . 2 2 2 2 s 4 q cos Similarly for the integral corresponding to r0 we get the same expression as above, thus the product of these two integrals is given by L/2 L/2 cos s z exp i 2 q cos z dz L/2 L/2 cos s z0 exp i 2 q cos z0 dz0 2 s L L 4 q cos sin qL cos cos s 2 s cos qL cos sin 2 2 . 2 2 2 2 s 4 q cos Also a b 0 r0 Yn nk r0 cos n 0 exp i 2 qr0 sin cos 0 d 0 dr0 2 qa sin Yn nk a J n 2 qa sin nk a Yn nk a J n 2 qa sin 2 qb sin Yn nk b J n 2 qb sin nk b Yn nk b J n 2 qb sin n i , nk2 4 2 q 2 sin 2 14 and a b 0 r1Yn nk r1 cos n1 exp i 2 qr1 sin cos 1 d1 dr1 2 qa sin Yn nk a J n 2 qa sin nk a Yn nk a J n 2 qa sin 2 qb sin Yn nk b J n 2 qb sin nk b Yn nk b J n 2 qb sin i n nk2 4 2 q 2 sin 2 . Therefore the product of two integrals is 2 qa sin Yn nk a J n 2 qa sin nk a Yn nk a J n 2 qa sin 2 qb sin Yn nk b J n 2 qb sin nk b Yn nk b J n 2 qb sin 2 cos qL cos sin s L 4 q cos sin qL cos cos s L s 2 2 I r0 I r1 2 2 2 2 nk 2 q sin s 2 q cos Similarly the fourth integral in Eq. (S31) can be calculated as L/2 L/2 sin u z0 exp i 2 q cos z0 dz0 L/2 L/2 sin u z1 exp i 2 q cos z1 dz1 2 uL L 4 q cos cos qL cos sin u 2 u sin qL cos cos 2 2 . 2 2 2 2 u 4 q cos Also we obtain the following integral as a b 0 r0 Yn nk r0 cos n 0 exp i 2 qr0 sin cos 0 d 0 dr0 2 qa sin Yn nk a J n 2 qa sin nk a Yn nk a J n 2 qa sin 2 qb sin Yn nk b J n 2 qb sin nk b Yn nk b J n 2 qb sin n i , nk2 4 2 q 2 sin 2 15 2 (S35) and also a b 0 i n r1Yn nk r1 cos n1 exp i 2 qr1 sin cos 1 d1 dr1 2 qa sin Y a J 2 qa sin a Y a J 2 qa sin n nk n nk n nk n 2 qb sin Y b J 2 qb sin b Y b J 2 qb sin n nk n nk n nk n nk2 4 2 q 2 sin 2 . Thus the product of the two integrals becomes, 2 qa sin Yn nk a J n 2 qa sin nk a Yn nk a J n 2 qa sin 2 qb sin Yn nk b J n 2 qb sin nk b Yn nk b J n 2 qb sin 2 sin qL cos cos u L 4 q cos cos qL cos sin u L u 2 2 I r0 I r1 nk2 2 q sin 2 u2 2 q cos 2 (S36) 2 Substituting the integral products (i.e., Eqs. (S32), (S34), (S35) and (S36)) into the expression for the signal attenuation (Eq. (1)), gives the final result for the spin-echo attenuation for diffusion in a concentric cylinder of finite length with arbitrary angle between the longitudinal axis and the direction of the applied magnetic field, viz. E q, 0 16 2 a b2 L2 n0 2 O q, a O q , b 2 L M q, l 0 H q, a H q, b s 0 2 2 L L L A q, C q, N q, B q, F q, 2 2 2 u 0 2 P q, a P q, b L M q, k 0 K q, a K q, b s 0 2 2 L L L A q, D q, N q, B q, G q, , 2 2 2 u 0 2 16 (S37) where 1 sin 2 s z A q, z 1 , 2 s z 1 sin 2 u z B q, z 1 , 2 u z exp nl2 s2 Dt C q, t , 2 2 nl2 2 q sin 2 s2 2 q cos 2 F q, t exp nl2 u3 Dt , 2 nl2 2 q sin 2 u2 2 q cos 2 2 2 exp nk s Dt D q, t , 2 2 nk2 2 q sin 2 s2 2 q cos 2 2 2 exp nk u Dt G q, t , 2 2 nk2 2 q sin 2 u2 2 q cos 2 n2 2 H q, r r 2 J n nl r r 2 2 J n2 nl r , nl n2 2 K q, r r 2Yn nk r r 2 2 Yn2 nk r , nk M q, z 2 s cos 2 qz cos sin s z 4 q cos sin 2 qz cos cos s z , (S38) N q, z 2 u sin 2 qz cos cos u z 4 q cos cos 2 qz cos sin u z , O q, r 2 qr sin J n nl r J n 2 qr sin nl r J n nl r J n 2 qr sin , P q, r 2 qr sin Yn nk r J nk 2 qr sin nk r Yn nk r J n 2 qr sin . 3. Mean Square Displacement in a Concentric Cylinder The MSD in a concentric cylinder is calculated from Eq. (20) and the expressions for the diffusion propagator (Eqs. (11) and (12) Eqs. (S4) and (S26)). It is possible to redefine the expression for the MSD in terms of the mean propagator2,3 P R, t r0 P r0 , r0 R, t dr0 , (S39) where R is the vector of displacement during t, that is R = r1 – r0. An equivalent expression for the MSD to Eq. (20) but using the mean propagator is3 17 r1 r0 2 R t 2 P R, t R 2 dR . (S40) Using Eqs. (S39) and (S40), the MSD component corresponding to the zero-eigenvalue component (Eq. (11)) of the propagator will be r1 r0 2 0 2 1 2 L2 a 2 b2 . 3 (S41) Using the cylindrical coordinate system the MSD component for the non-zero eigenvalues is calculated by substituting Eq. (12) into Eq. (20). The integral over r0 is over the three elements of the coordinate system (i.e., z0, 0, and r0). Integration over the z0 coordinate gives r1 r0 2 1 0 2 z1 z0 2 cos s z0 dz0 2 2 L L 4 L s cos s 8 L2 4 r0 r1 0 1 z12 s2 sin s 2 2 . 2 s2 L/2 L/2 (S42) Also, the integral over 0 is 0 L 4 L s cos s 2 2 2 s L 2 2 2 8 L 4 r0 r1 0 1 z1 s sin 2 cos n d 0 0 2 s2 (S43) L 8n1 s2 8n 1 s2 cos n sin s . 2 Note that in the above integration the terms containing sin(nπ) are eliminated as they are zero for different values of n. The integration over r0 gives 18 L 8n1 s2 8n 1 s2 cos n sin s J n nl r0 r0 dr0 2 n 2i 1 n L 1 8n1 s2 8n 1 s2 cos n sin s aJ n 2i 1 nl a bJ n 2i 1 nl b . 2 nl i 0 2 1 n i 1 1 n i 2 2 a b (S44) By using the above integration results (i.e., Eqs. (S42)–(S44)) in the above equation and repeating the integration for r1 (i.e., z1, θ1 and r1) the final form of the MSD expression the for non-zero eigenvalues is obtained, 2 L 16 s 1 cos2 n sin 2 s n 2 i 1 n 2 aJ n 2i 1 nl a bJ n 2i 1 nl b . n nl i 0 2 1 n i 1 1 n i 2 2 2 (S45) This calculation is then repeated for the second term in the propagator. The third and last terms that are related to Bessel function of the second kind are calculated in a similar way and the final form of the MSD for diffusion in a concentric cylinder (i.e., including zero and non-zero eigenvalues components) is 19 r1 r0 2 2 1 1 2 L2 a 2 b 2 3 L a2 b2 2 2 n 2i 1 n 2 2 s L 64 s 1 cos n sin aJ n 2i 1 nl a bJ n 2i 1 nl b 2 i 0 2 1 n i 1 1 n i 2 2 n0 l 0 s 0 sin s L 2 2 n2 2 n2 2 2 2 2 2 n nl Ln 1 a J n nl a a 2 J n nl a b J n nl b b 2 J n nl b s L nl nl 2 2 exp nl s Dt 2 2 n 2i 1 n L 64 u 1 cos 2 n sin 2 u aJ n 2i 1 nl a bJ n 2i 1 nl b 2 i 0 2 1 n i 1 1 n i 2 2 sin u L 2 2 n2 2 n2 n 0 l 0 u 0 2 2 2 2 2 n nl Ln 1 a J n nl a a 2 J n nl a b J n nl b b 2 J n nl b L nl nl u 2 2 exp nl u Dt 2 2 n 2 i 1 n L 64 s 1 cos 2 n sin 2 s aYn 2i 1 nk a bYn 2i 1 nk b 2 i 0 2 1 n i 1 1 n i 2 2 sin s L 2 2 n2 2 n2 2 2 2 2 n 0 k 0 s 0 2 n nk Ln 1 a Y a a 2 Yn nk a b Yn nk b b 2 Yn nk b s L n nk nk nk exp 2 2 Dt s nk 2 2 n 2 i 1 n L 2 2 u 64 1 cos n sin aY a bY b u n 2 i 1 nk 2 1 n 2i 1 nk i 0 2 n i 1 1 n i 2 2 2 2 n 0 k 0 s 0 n Ln 1 sin u L a 2Y a 2 a 2 n Y a 2 b 2Y b 2 b 2 n Y b 2 nk n nk u L n nk nk2 n nk nk2 n nk exp nk2 u2 Dt (S46) This can be simplified to r1 r0 2 2 2 1 1 2 L2 a 2 b 2 1 cos n 2 2 3 L a b n0 V 1 a V 1 b 2 L L L L S 1 A q , R 1 t U 1 B q , R 2 t l 0 H q, a H q, b s 0 2 2 2 2 u 0 2 W 1 a W 1 b L L L L S 1 A q , R 3 t U 1 2 B q , 2 R 4 t , k 0 u 0 H q, a H q, b s 0 2 2 20 (S47) where L L S1 64 s sin 2 s , 2 2 L L U 1 64 u sin 2 u , 2 2 n 1 2i n r J 2i n 1 nl r , n n i 0 2 i 1 i 2 2 n 1 2i n W 1 r rY2i n 1 nk r , n n i 0 2 i 1 i 2 2 V 1 r R1 t R2 t R3 t R4 t exp nl2 s2 Dt nL nl , exp nl2 u2 Dt nL nl , exp nk2 s2 Dt nLnk , exp nk2 u2 Dt nLnk . (S48) B. Concentric Sphere 1. Diffusion Propagator (a)Zero eigenvalue solution For the case of λ = 0 (i.e., zero eigenvalue), the time variable is equal to a constant T 0 A0 (S49) and similarly for the spatial eigenfunction 2 0 0 . 21 (S50) Consequently, the general form of the spatial part of the propagator is also given by a constant (arbitrarily set to 1) 0 BR C 1, (S51) where B and C can be any constants that do not simultaneously. Therefore the diffusion propagator corresponding to the zero-eigenvalue will be given by P r , , t 0 0T 0 A0 . (S52) The normalised propagator of zero-eigenvalue (i.e., Nλ=0) is obtained as P 0 r0 , 0 , r1 ,1 , t 0 r0 0 r1 N 0 N 0 3 . 4 a 3 b3 (S53) (b) Positive eigenvalue solution For non-zero eigenvalues, it is convenient to set λ = -α2, the Eq. (23) becomes 1 T 2 1 2 R 2 R 1 1 2 1 2 2 , 2 DT t R r r r r (S54) n n 1 1 1 1 2 , 2 2 r r (S55) setting 22 and rearranging gives a Legendre’s differential equation of the form n n 1 0. 1 2 (S56) The solutions of this equation are given in terms of Legendre’s polynomials of the first Pn and second Qn kind, C1Pn C2Qn (S57) where C1 and C1 are constants. Since Qn(μ) have poles at μ = ±1, this solution has been excluded. The spatial part of Eq. (S54) can be written as 2 R 2 R 2 n n 1 R 0. r 2 r r r2 (S58) The solution for this equation is R r En,m nmr 1/2 Cn1/2 (nmr ) (S59) where Cn 1/ 2 ( nm r ) F1 J n 1/ 2 ( nm r ) F2Yn 1/ 2 ( nm r ) . The temporal part of the diffusion equation (Eq. (S54)) is 1 T 2 , DT t 23 (S60) with solution T t exp D 2t . (S61) Thus, combining Eqs. (S57), (S59), and (S61) the solution for this propagator is given by P r , , t r , T t Anm nm r n 0 m0 1/ 2 2 Cn 1/ 2 nm r Pn exp D nm t . (S62) From boundary condition on the outer surface (r = a) given by D M a R 0, r r a (S63) where Ma is the relaxivity on the outer surface, gives the root-function, nl J n1/2 nl a M a 1 . J n1/2 nl a D 2a (S64) Similar to the case of the concentric cylinder, by using the definition of Y function in terms of J, it can be shown that both solutions on this surface results in the same eigenvalues. Similarly for the boundary condition on r = b the boundary condition D M b R 0, r r b gives the equation for the eigenvalues to be 24 (S65) nk Yn1/2 nk b Yn 1/2 nk b Mb 1 . D 2b (S66) This boundary condition applied on either Y or J solutions gives the same eigenvalues. Thus, R Bnl nl r 1/2 n 0 l 0 J n1/2 nl r Bnk nk r n 0 k 0 1/2 Yn1/2 nk r . (S67) Therefore the non-normalised diffusion propagator for diffusion within a concentric sphere is P r , , t Bnl nl r 1/ 2 n 0 m 0 Bnk nk r n 0 k 0 1/ 2 J n 1/ 2 nl r Pn exp D nl2 t (S68) Yn 1/ 2 nk r Pn exp D t . 2 nk The above propagator must be normalised by determining the normalisation factor for each part in keeping with the boundary conditions of the radial eigenfunctions at r = a and r = b. The first normalisation factor is, N1 nl r 1/ 2 J n 1/ 2 nl r Pn mu r J m 1/ 2 mu r Pm dV a2 n 1 / 2 2 2 J n1/ 2 nl a 2 1 J n 1/ 2 nl a 2 2 nl a 4 . 2n 1 nl b2 n 1 / 2 2 2 2 J n 1/ 2 nl b J n1/ 2 nl b 1 2 b 2 nl 1/ 2 25 (S69) Since F2 = 0 corresponds to the boundary condition on r = a, the second part of the radial eigenfunction has no contribution to the normalisation factor. The normalisation factor for the second part is, N 2 nk r 1/ 2 Yn 1/ 2 nk r Pn mv r Ym 1/ 2 mv r Pm dV 2 a2 n 1 / 2 2 Yn1/ 2 nk a 2 1 Y a 2 n 1/ 2 nk 2 a nk 4 . 2 2 n 1 nk b2 n 1 / 2 Y 2 b 2 Yn1/ 2 nk b 1 n 1/ 2 nk 2 2 nk b 1/ 2 (S70) Thus, the normalised diffusion propagator for diffusion in a concentric sphere is P r0 , 0 , r1 , 1 ; t n0 l 0 2n 1 nl nl r1 2 1/ 2 J n 1/ 2 nl r1 Pn 1 nl r0 1/ 2 J n 1/ 2 nl r0 Pn 0 n 1 / 2 2 a 2 J n1/ 2 nl a 2 1 J a n 1/ 2 nl 2 2 a nl 2 2 n 1 / 2 2 J n21/ 2 nl b b J n1/ 2 nl b 1 2 2 b nl 2 exp nl2 Dt 1/ 2 1/ 2 2n 1nk nk r1 Yn1/ 2 nk r1 Pn 1 nk r0 Yn1/ 2 nk r0 Pn 0 exp 2 Dt . nk 2 n0 k 0 n 1 / 2 2 2 2 2 a Yn1/ 2 nk a 1 Y a (S71) 2 a 2 n 1/ 2 nk nk 2 n 1 / 2 2 2 2 Y b b Yn1/ 2 nk b 1 nk2 b 2 n 1/ 2 nk The half-integer order Bessel functions of the first kind can be converted to spherical Bessel using the following definitions1 26 J n1/2 z 2z J n 1/2 z 1 2z jn z j z , n 2 z Yn1/2 z 2z Yn1/2 z 1 2z yn z y z . n 2 z zJ n1/2 z z 1 jn z zjn z , 2 2 jn z , (S72) yn z , Therefore z zYn1/2 z 2 1 2 yn z zyn z . (S73) And thus Eq. (S71) by using Equations (S72) and (S73) becomes P r0 , 0 , r1 , 1 , t 2n 1 nl exp nl2 Dt n0 l 0 jn nl r1 Pn 1 jn nl r0 Pn 0 5 1 2 2 4 nl a nl a 2 n jn nl a 2 nl a jn nl a jn nl a nl a jn nl a 2 2 2 5 1 2 2 4 nl b nl b 2 n jn nl b 2 nl b jn nl b jn nl b nl b jn nl b 2 2 2 n 0 k 0 2n 1nk exp nk2 Dt yn nk r1 Pn 1 yn nk r0 Pn 0 2 5 1 2 2 4 nk a nk a 2 n yn nk a 2 nk a yn nk a yn nk a nk a yn nk a 2 2 2 5 1 2 2 4 nk b nk b 2 n yn nk b 2 nk b yn nk b yn nk b nk b yn nk b 2 2 (S74) 27 . 2. Pulse Gradient Spin-Echo Attenuation in a Concentric Sphere (a) Zero eigenvalue contribution Substituting the propagator for = 0 for the concentric sphere (Eq. (24)) in to the SGP equation (Eq. (1)) gives, the signal attenuation corresponding to this part of the propagator to be E q, 0 4 2 a b r P r , r , e 1 a 1 b 1 0 1 9 a 3 b 0 0 i 2 q r1 r0 1 d 1r12 dr1d 0 r0 2 dr0 sin 2 qa 2 qa cos 2 qa sin 2 qb 2 ab cos 2 qb . 2 2 q 3 2 6 (S75) (b) Positive eigenvalue contribution Substituting the propagator for > 0 for the concentric sphere (Eq. (S71)) and (r0) that is the equilibrium spin density (equal to 1/(volume) of the restricting pore) at the point r0 and initial time zero in to the SGP equation (Eq. (1)) gives, E q, 0 4 2 a b r P r , r , e 1 a 1 b 1 1 0 0 i 2 q r1 r0 1 d 1r12 dr1d 0 r0 2 dr0 2n 1 nl exp nl2 D I r0 I r1 3 2 a 3 b3 n 0 l 0 n 1 / 2 2 2 a 2 J n1/ 2 nl a 2 1 J a n 1/ 2 nl 2 2 a nl 2 2 n 1 / 2 2 J n21/ 2 nl b b J n1/ 2 nl b 1 2 2 nl b 2n 1nk exp nk2 D I r0 I r1 3 , 2 2 a 3 b3 n 0 k 0 n 1 / 2 2 a 2 Yn1/ 2 nk a 1 Yn21/ 2 nk a 2 2 a nk n 1 / 2 2 2 2 2 Yn 1/ 2 nk b b Yn1/ 2 nk b 1 2 2 b nk where μ = cos θ, (θ is inclination angle and 0 ) and also 28 (S76) r r a 1 b 1 I r0 I r1 a 1 b 1 1/ 2 J n 1/ 2 nl r0 Pn 0 exp i 2 qr0 0 d 0 r0 2 dr0 1/ 2 J n 1/ 2 nl r1 Pn 1 exp i 2 qr1 1 d 1 r12 dr1 , nl 0 nl 1 and r I r a I r0 r1 1 b 1 a 1 b 1 1/ 2 Yn 1/ 2 nk r0 Pn 0 exp i 2 qr0 0 d 0 r0 2 dr0 1/ 2 Yn 1/ 2 nk r1 Pn 1 exp i 2 qr1 1 d 1 r12 dr1 . nk 0 nk 1 The first integral becomes I r0 nl r0 a 1/2 nl r0 J n1/2 nl r0 i n qr0 1/2 b i n nl q 1/2 nl2 2 q 1 1 b a J n1/2 nl r0 Pn 0 exp i 2 qr0 0 d 0 r0 2 dr0 2 1/2 J n1/2 2 qr0 r0 2 dr0 2 qa J n1/2 nl a J n1/2 2 qa nl a J n1/2 2 qa J n1/2 nl a (S77) 2 qb J n1/2 nl b J n1/2 2 qb nl b J n1/2 2 qb J n1/2 nl b . The second integral becomes I r1 nl r1 a 1/2 b nl r1 a b 1/2 J n1/2 nl r1 Pn 1 exp i 2 qr11 d 1r12 dr1 1 1 J n1/2 nl r1 i n qr1 1/2 J n1/2 2 qr1 r12 dr1 i nl q 2 qa J 2 n1/2 nl a J n1/2 2 qa nl a J n1/2 2 qa J n1/2 nl a 2 nl 2 q 2 qb J n1/2 nl b J n1/2 2 qb nl b J n1/2 2 qb J n1/2 nl b . n 1/2 The third integral becomes 29 (S78) I r0 nk r0 a 1/2 i nk q 1/2 nk2 2 q 1 1 b n Yn1/2 nk r0 Pn 0 exp i 2 qr0 0 d 0 r0 2 dr0 2 2 qa Yn1/2 nk a J n1/2 2 qa nk a J n1/2 2 qa Yn1/2 nk a (S79) 2 qb Yn1/2 nk b J n1/2 2 qb nk b J n1/2 2 qb Yn1/2 nk b . And the fourth integral becomes i nk q 2 qa Y a J 2 qa a J 2 n1/2 nk n1/2 nk n1/2 2 qa Yn1/2 nk a 2 nk 2 q 2 qb Yn1/2 nk b J n1/2 2 qb nk b J n1/2 2 qb Yn1/2 nk b . n I r1 1/2 Combining Eqs. (S76)-(S80), the spin-echo attenuation can be written as 30 (S80) E q, 0 n 0 l 0 3 2n 1 nl exp nl D 2 a b3 3 n 1 / 2 2 2 n 1 / 2 2 2 2 2 2 2 J a b J b 1 J b a J n 1/ 2 nl a 1 n 1/ 2 nl n 1/ 2 nl n 1/ 2 nl nl2 a 2 nl2 b 2 nl q 1/ 2 2 qa J a J 2 qa a J 2 qa J a 2 n 1/ 2 nl n 1/ 2 nl n 1/ 2 n 1/ 2 nl 2 nl 2 q 2 qb J n 1/ 2 nl b J n1/ 2 2 qb nl b J n 1/ 2 2 qb J n1/ 2 nl b 1 3 2n 1nk exp nk D 2 a 3 b3 n 0 k 0 2 n 1 / 2 2 2 n 1 / 2 2 2 2 2 2 2 Yn 1/ 2 nk a b Yn 1/ 2 nk b 1 Yn 1/ 2 nk b a Yn 1/ 2 nk a 1 2 2 2 2 nk a nk b 2 nk q 1/ 2 2 qa Y a J 2 qa a J 2 qa Y a 2 n 1/ 2 nk n 1/ 2 nk n 1/ 2 n 1/ 2 nk 2 nk 2 q 2 qb Yn 1/ 2 nk b J n1/ 2 2 qb nk b J n 1/ 2 2 qb Yn1/ 2 nk b . 1 (S81) Transforming from half-integer order Bessel functions of the first kind to spherical Bessel functions using Eqs. (S72) and (S73), Eq. (S81) simplifies to 31 E q, 0 2 a3 2 qa jn nl a jn 2 qa nl a jn 2 qa jn nl a 2 2 nl a 2 qa 3 b 2 qb jn nl b jn 2 qb nl b jn 2 qb jn nl b b 2 2 qb 2 nl 2 5 n0 l 0 1 2 2 4 nl a nl a 2 n jn nl a 2 nl a jn nl a jn nl a nl a jn nl a 2 2 2 5 1 2 2 4 nl b 2 nl b 2 n 2 jn nl b 2 nl b jn nl b jn nl b nl b jn nl b 6 2n 1 nl3 exp nl2 D a 3 b3 2 a3 2 qa yn nk a jn 2 qa nk a jn 2 qa yn nk a 2 2 nk a 2 qa 3 b 2 qb y b j 2 qb b j 2 qb y b n nk n nk n n nk 2 2 nk b 2 qb 2 n0 k 0 5 1 2 2 4 nk a nk a 2 n yn nk a 2 nk a yn nk a yn nk a nk a yn nk a 2 2 2 5 1 2 2 4 nk b 2 nk b 2 n 2 yn nk b 2 nk b yn nk b yn nk b nk b yn nk b 6 2n 1 nk3 exp nk2 D a 3 b3 (S82) In compact form we have for the annular sphere attenuation function E q, 0 2 G q, a G q , b A 6 3 C q, a C q, b n 0 l 0 a b3 2 H q, a H q, b B , D q, a D q, b k 0 where 32 (S83) A t 2n 1 nl2 exp nl2 Dt , B t 2n 1nk2 exp nk2 Dt , 2 5 1 2 2 4 C q, r nl r nl r 2 n jn nl r 2 nl r jn nl r jn nl r nl r jn nl r , 2 2 2 5 1 2 2 4 D q, r nk r nk r 2 n yn nk r 2 nk r yn nk r yn nk r nk r yn nk r , 2 2 r G q, r 2 2 qr jn nl r jn 2 qr nl r jn 2 qr jn nl r , 2 nl 2 q H q, r r 2 q 2 nk 2 qr yn nk r jn 2 qr nk r jn 2 qr yn nk r . 2 (S84) 3. Mean Square Displacement in a Concentric Sphere The MSD for diffusion in a concentric sphere is determined similarly to the MSD for diffusion within a concentric cylinder (see Section A.3). For the case of zero-eigenvalue the propagator (i.e., Eq. (24) Eq. (S53)), the MSD component is r1 r0 2 4 3 3 a b . 3 3 2 2 (S85) Using the spherical coordinate system the MSD component corresponding to non-zero eigenvalues (Eq. (25) Eq. (S74)) can be obtained using Eq. (20). For this we start with the following integral, r r arccos 1 2 1 1 0 2 1 0 1 2 arccos0 Pn 0 d 0 r1 r0 2 1 0 2 12 Pn 0 d 0 2arccos1 0 Pn cos 0 sin 0d 0 1 1 02 Pn cos0 sin 0d0 . 33 (S86) It can be shown from standard tables of integrals (page 50, Eq. 1.14.1 Eq(1))4 that the first integral in the above will be zero. Also the integrand in the third integral is an odd function and thus the contribution will vanish when integrated from –π to π. Thus, the only remaining integral to be evaluated is the second integral. Making the change of variable θ π/2-θ, changing the limits of integration to - π/2 to π/2 (and consequently multiplying by -2), the integral is transformed into the standard integral (page 435, Eq. 2.17.7 Eq(15) in ref.4), 2 n n 2 !! 1 1 2arccos 1 0 Pn cos 0 sin 0 d 0 4 arccos1 . 2n 2 n 1 ! 2 (S87) Also the integration over r0 which involves Bessel functions of the first kind gives 2 2 n n a 1 1 1 1 n 2 !! r 1/2 J n 2 !! 2 4 arccos r r dr 4 arccos 1 nl 0 n 1/2 nl 0 0 0 1 b 2n2 2n 2 n 1 ! n 1 ! 2 2 3 3 1 1 n n 2i n i 1 3 2 2 2 3/2 a 3/2 J J n 2i 3/2 nl b n 2 i 3/2 nl a b 5 nl3/2 1 i 0 1 n n i 2 2 2 2 n n 2 !! 1 1 arccos1 2n 2 n 1 ! 2 3 3 1 1 n n 2i n i 8 3 2 2 2 2 a 2 j n 2i 1 nl a b jn 2i 1 nl b . 5 1 nl 1 n i 0 n i 2 2 2 (S88) Doing the similar calculations for θ1 and r1 also gives the same results and the final MSD with contribution from both zero and non-zero eigenvalues is 34 r1 r0 2 4 3 a 3 b3 3 3 4 a 3 b3 2 2 2 1 3 3 1 n n 2i n i 2n 1 3 2 2 2 2 a2 j n 2i 1 nl a b jn 2i 1 nl b 5 2 nl2 1 i 0 1 n n i 2 2 2 2 n0 l 0 5 1 2 2 4 nl a nl a 2 n jn nl a 2 nl a jn nl a jn nl a nl a jn nl a 2 2 2 2 n 1 1 n 2 !! 2 8 exp nl Dt 2n 2 n 1 ! 2 2 5 nl b nl b 2 2 n 1 jn nl b 2 2 nl b 4 jn nl b jn nl b nl b jn nl b 2 2 2 1 3 3 1 n n 2i n i 2n 1 3 2 2 2 2 a2 y a b y b n 2 i 1 nl n 2 i 1 nl 2 2 5 nk 1 i 0 1 n n i 2 2 2 2 5 1 n0 k 0 2 2 4 a a 2 n nk nk yn nk a 2 nk a yn nk a yn nk a nk a yn nk a 2 2 2 2 n 1 1 n 2 !! exp 2 Dt 8 nk n2 2 n 1 ! 2 2 5 1 2 2 4 nk b nk b 2 n yn nk b 2 nk b yn nk b yn nk b nk b yn nk b 2 2 (S89) This can be simplified to r1 r0 2 2 2 1 1 n n 2 !! 2 2 4 3 a 3 b3 8 2n 1 3 3 2n 2 n 1 / 2 ! 4 a 3 b3 n 0 2 V 2 a V 2 b 2 W 2 a W 2 b Q1 t Q2 t , l 0 C q, a C q, b k 0 D q, a D q, b (S90) where 35 3 3 1 1 n n 2i n i 2 2 2 2 r2 j V 2r n 2 i 1 nl r , 5 1 i 0 1 n n i 2 2 2 3 3 1 1 n n 2i n i 2 2 2 2 r2 y W 2r n 2 i 1 nl r , 1 1 5 i 0 n n i 2 2 2 Q1 t exp nl2 Dt Q2 t 2 nl2 , exp nk2 Dt 2nk2 (S91) . C. The limiting cases 1. Concentric Cylinder The new model for diffusive spin-echo attenuation in a concentric cylinder (i.e., Eqs. (7), (17) and (18) was successfully tested against known solutions for diffusion between parallel planes5 and in a normal cylinder6,7 (although the literature solution is for an infinite cylinder) for the reflecting boundary condition (i.e., M = 0 for all surfaces) by setting the inner radius to zero (i.e. b0) in the concentric model and simulating the attenuation profiles for q oriented at = 0 (i.e., parallel planes – see Figure 1), = 90 (i.e., perpendicular), and tilted at = 45° to the long axis of the cylinder as illustrated in Figure S1. The corresponding simulations and comparisons against known solutions for diffusion between parallel planes7-9 and in a normal cylinder7 under the relaxing boundary condition (i.e., M 0 for all surfaces) are given in Figure S2. 36 Figure S1. Simulated spin-echo attenuation profiles for diffusion in a reflecting concentric cylindrical pore in the limiting case b = 0 m, where the symmetrical axis of the cylinder is oriented at (A) = 0° (equivalent to parallel planes; ) and = 90° (equivalent to cylinder; ) and (B) = 45° to the direction of the gradient q. The simulations were performed with a = 20 m, b = 0 m, L = 40 m, Δ = 0.87 s and D = 2.3 10-9 m2s-1 such that D / ( a b ) 4 D / L 2 2 5 . In (A) the existing literature solutions for attenuation in a simple infinite cylinder () and between parallel plane () is included for comparison and are observed to be in excellent agreement. 37 Figure S2. Simulated spin-echo attenuation profiles for diffusion in a concentric cylindrical pore with relaxing boundary conditions in the limiting case b = 0 m, whose symmetrical axis is oriented at (A) = 0° (equivalent to parallel planes; ) and = 90° (equivalent to cylinder; ) and (B) = 45° to the direction of q. In these simulations, a = 20 m, b = 0 m, L = 40 m, Δ = 0.87 s and D = 2.3 10-9 m2s-1 such that D / ( a b ) 4 D / L determining the eigenvalues were M z 2 2 5 . The values of the relaxivity used in 2 D / L, M r D / a, o 38 and M ri D / b . In (A) the existing literature solution for attenuation in simple infinite cylinder () and parallel plane () is included for comparison and are observed to be in excellent agreement. 2. Concentric Sphere The new model for diffusive spin-echo attenuation in a concentric sphere (i.e., Eqs. (7), (28) and (29)) was successfully tested against known solutions for a sphere under reflecting (i.e., M = 0 for all surfaces)10-12 and relaxing (i.e., M 0 for all surfaces)7 boundary conditions by setting the inner radius to zero (i.e. b0) in the concentric model as illustrated in Figure S3. Figure S3. A comparison of the PGSE attenuation for the reflecting (M = 0; ) and relaxing ( M ro D / a ; ) limiting cases for the concentric sphere, when the inner radius is shifted to zero (i.e., b = 0) with Δ = 0.87 s such that D / ( a b) 2 5 . The analytical solution for attenuation in a simple reflecting () and relaxing () sphere is included for comparison and are observed to be in excellent agreement. 39 D. REFERENCES 1. Handbook of Mathematical Functions, 1st ed. edited by M. Abramowitz and I. A. Stegun (Dover, New York, 1970). 2. J. Kärger and W. Heink, J. Magn. Reson. 51, 1 (1983). 3. W. S. Price, NMR Studies of Translational Motion: Principles and Applications, 1st ed. (Cambridge University Press, Cambridge, 2009). 4. A. P. Prudnikov, Yu. A. Brychkov, and O. I. Marichev, Integrals and Series, 1st ed. (Gordon and Breach, The Netherlands, 1992), Vol. 2. 5. J. E. Tanner and E. O. Stejskal, J. Chem. Phys. 49, 1768 (1968). 6. O. Söderman and B. Jönsson, J. Magn. Reson. A 117, 94 (1995). 7. P. T. Callaghan, J. Magn. Reson. A 113, 53 (1995). 8. A. Coy and P. T. Callaghan, J. Chem. Phys. 101, 4599 (1994). 9. S. Frey, J. Kärger, H. Pfeifer, and P. Walther, J. Magn. Reson. 79, 336 (1988). 10. B. Balinov, B. Jönsson, P. Linse, and O. Söderman, J. Magn. Reson. A 104, 17 (1993). 11. B. Balinov, B. Jönsson, P. Linse, and O. Söderman, J. Magn. Reson. A 108, 130 (1994). 12. W. S. Veeman, in Diffusion in a Closed Sphere, edited by Graham. A. Webb (Elsevier, London, 2003), Vol. 50, pp.201-216. 40