JOURNAL OF APPLIED PHYSICS VOLUME 21. NUMBER 8 AUGUST. 1956 Charged Right Circular Cylinder W. R. SMYTHE Norman Bridge Laboratory, California Institute of Technology, Pasadena, California (Received March 5, 1956) A new method permits the calculation of the electric field surrounding a charged conducting surface of revolution without the use of orthogonal functions. Detailed formulas show how to find the charge density on a right circular cylinder with any desired precision. The numerical examples worked out give the maximum deviation of the actual surface from that of a true cylinder of diameter d to be -0.OO15d, -0.OOO37d, -0.OO017d, +0.OO16d, and +0.014d for length to diameter ratios 1, I, 1, 2, and 4, respectively. The capacitance calculation gives an accuracy of one part in 30000 for the ratios I, 1, and 2. A capacitance formula is worked out which is accurate to one part in 1000 over the ratio range 0 to 4. Additional formulas indicate the method of solution for the bodies in longitudinal and transverse electric fields and the extension of twobody problems such as the thick plate parallel plate capacitor. A way to calculate the flow about bodies of revolution is indicated. Digital computers are well suited to this method as no function tables are needed. 1. INTRODUCTION of the most common figures occuring in apO NEparatus is the right circular cylinder. Thus, its capacitance and the fields surrounding it when it is charged are of considerable practical importance. The reason the literature is blank on this subject is evident when one sees the hopeless complications the usual orthogonal function techniques involve when applied to this problem. A new approach is needed. The method to be described gives a complete solution, as accurate as desired, of this and similar problems. The charge density CT. on the end must be a function of p2 to have a minimum (dCT./dp=O) at the pole and that on the side (CT.) must be a function of Z2 to have a minimum (dCT./dz=O) at the equator. Furthermore, very near the edge the charge densities become equal and approach that on the charged rectangular wedge which is proportional to o-t where ~ is the distance from the edge. Suitable forms are therefore co CT.=L: An(b2-z12)n- i (4) n=O co 2. METHOD OF SOLUTION CT.=L: Bn(1-PI2)n-i, An axially symmetrical solution of Laplace's equation in the cylindrical coordinate system P, 11', z is (5) n=O where b, ZI, and PI are dimensionless numbers given by (6) (1) Differentiation 2p times with respect to z introduces a factor (lP into the integrand. Thus, when the Bessel function is expanded in the usual way in even powers of tp, the result may be written This expression is widely used in electrostatic electron microscope lens calculations and gives, in terms of the potential V(z,O) on the axis, the potential at any point off the axis which can be reached without crossing a charge sheet. The charge distribution on a closed conducting surface that encloses the origin gives a constant potential Vo inside so that P d2 V(Z,O)] [ dz2 p .-0 = {VO when p=O (3) 0 when p,.r:O. The important n=O terms will be called the "fundamental" terms and the remaining ones the "correction" terms. The latter are zero at the edge and in order that the fundamental terms match there it is necessary that Ao=blBo. (7) The number of correction terms required for a given accuracy is much smaller for the forms given in Eqs. (4) and (5) than when integral powers of p and z are used. Also, as will appear shortly, numerical checking is simpler. When the potential on the axis due to the charge densities of Eqs. (4) and (5) is calculated, differentiated with respect to ZI and ZI is equated to zero the result is d2 p V 00 MBn -=L: --F[p+t, n-p+i; n+1+i; (1+b2)-IJ dZ 12p n=O n+i +MNGA nF[P+t, n- p+i; n+ 1+i; b2(1 +b2)-IJ, (8) where Thus our problem is to find a charge distribution on the surface of the right circular cylinder bounded by the planes z=±c and the cylinder p=a which satisfies (3). b2n+i(n-i) !(p-t)! ta(2p)! M E(l +b2)P+t' N (-1)p2 f (n+i) !P! 917 Downloaded 31 Mar 2011 to 134.208.24.193. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions W. 918 R. and G= [r (l)J2[rmjl =0.684463408, where F(a,b j C j z) is a hypergeometric function. For simplicity, negative factorials are taken unity throughout this paper so that (n-i)! means 1·i· 5/3· .. (n-i). The units are mks so a is in meters, V in volts, and E is 8.85434X 10-12 f/m. SMYTHE viation. From symmetry one of these (of zero radius) is at the pole and one at the equator and the deviations at these points should have the same sign when p is odd and opposite signs when p is even. From Eq. (2) the amplitude of these deviations should increase with increasing distance from the origin. Except at the TABLE 3. DETERMINATION OF COEFFICIENTS b=l The determination of Ao( =b1Bo), AI, .. " Ar and B 1, B 2, •• " B. in Eqs. (4) and (5) requires writing the equations in (3) for p values 0, 1, 2, .. ·r+s and solving these equations simultaneously. The chief labor is the calculation of the hypergeometric functions F(a,b j C j z). This is greatly facilitated by use of the following recursion formulas. c(c-1)F(a, b-1 j c-1 j z)+c[ (a-b)z-c+ 1J XF(a,b j C j z)+b(c- a)zF(a, b+ 1 j c+ 1 j z) =0 cF(a,bj Cj z)-cF(a+1, b-1 j Cj z) - (a-b+1)zF(a+1, c(a-b)F(a,bj Cj (9) A 0=0.39326697 A 1 = 10.911218 A2= -181.16229 Bo=0.62427242 B 1 = -0.09128092 B. = 0.02058206 EV Potentials EV Pole 1.00000002 V EV Equa. 0.9923 V EV Edge 1.00079692 V EV Capacitance 10.420264 Ea EV A o=0.40304719 Al = 6.4134459 A 2 = -97.322125 B o=0.63979755 B 1 = -0.13360489 B. = 0.06215840 B3= -0.01398595 Potentials EV EV Pole 1.00000000 V EV Equa. 0.9968 V Edge 0.99972774 V EV EV Capacitance EV 10.4155203 Ea .V A 0=0.46928422 Al =0.89919472 A.= -2.4818068 B o=0.59126109 B 1 = -0.09703091 B.=0.02927739 .V Potentials .V Pole 1.00000395 V EV Equa. 0.99937427 V .V Edge 0.99725336 V .V Capacitance EV 12.107972 .a A 0= 0.55905946 A 1=0.14867157 A.= -0.05617693 Bo = 0.55905946 B 1 = -0.14082266 B 2 =0.07155254 EV Potentials EV Pole 1.00010243 V EV Equa. 0.99997755 V EV EV Capacitance .V 14.973964 fa A 0=0.67570982 A 1=0.03921759 A.= -0.022289015 Bo=0.53631121 B 1 = -0.38793856 B 2 =0.35626582 .V Potentials .V Pole 1.00440238 V .V Equa. 0.99999954 V EV .V Capacitance .V 19.776506 Ea A 0=0.66050233 A 1=0.05455850 A 2 = -0.027136699 A 3 = 0.024926044 B o=0.52424102 B 1 = -0.23638748 B.=0.18391OO1 EV Potentials EV Pole 1.00149239 V EV Equa. 1.00000001 V EV EV Capacitance EV 19.775418 Ea .V A o=0.83974760 Al =0.012420258 A.= -0.021371704 B o=0.52900782 B 1 = -2.1617990 B.=2.8472879 EV Potentials .V Pole 1.0893612 V • V Equa. 0.99999935 V EV .V Capacitance EV 27.745204 Ea A o=0.79451208 A 1=0.020672127 A.= -0.()374287843 A 3 =0.()419885966 A.=-0.0623777893 B o=0.50051123 B 1 = -0.65430090 B 2 =0.73249240 .V EV EV EV EV EV EV EV Deviations +O.OOOOOOa -0.0083a +0.OOOO25a Deviations +O.OOOOOOa -0.003Oa -0.000005a b=i bj c+1 j z)=O z)-a(c-b)F(a+1, bj c+1; z) +b(c-a)F(a, b+1 j c+1 j z)=O. (10) (11) Only four functions need be calculated by series, say p=O for AoAl, B o, B 1• The remainder are then found by repeated application of Eqs. (10) and (11). The maximum p value (r+s) terms are checked by Eq. (9). If these check to eight places, all intermediate p values are good to eight places and the calculation need not be repeated. If these fail to check the error can be located by application of Eq. (9) to intermediate values. These operations are simple and fast on a digital computer. Little labor is involved in checking the factors that multiply F(a,bj Cj z) in Eq. (8) by repetition of the calculation. The first function in Eq. (8) converges fast for large values of b and the second for small values. For hand calculation of other values analytic continuation formulas 1 save time. Sets of coefficients for b values t, !, 1, 2,4 are given in Table I. One must use the complete set for omission of any term may cause large errors. Sets of from four to six give adequate accuracy over this range. Eight places were carried in all calculations so the last digit is uncertain due to cumulative rounding off errors. 4. ACCURACY OF RESULTS To evaluate the results some estimate of accuracy is needed. If there are m-correction terms then there are m+ 1 constants which are adjusted to some optimum value by Eq. (3). The actual Vo surface can intersect the cylinder on only m+ 1 circles between pole and equator. These separate m+ 2 circles of maximum deErdelyi, Magnus, Oberhettinger, and Tricomi, Higher Transcendental Functi9ns (McGraw-Hill Book Company, Inc., New York, 1953), Vol. 1, Eq. (1) p. 108, and Eq. (12) p. 110. 1 I." Deviations +0.0000075a -0.000732a -0.000100 b=l Deviations +0.000209a -0.OOOO344a b=2 Deviations +0.00872a -0.000001 a Deviations +0.00316a -0.000000a b=4 Potentials Pole 1.01596239 V Equa. 0.99999942 V Deviations +0.1496a -0.000002a Deviations +0.02700 -0.000001 a Capacitance 27.844690 fa • Note that 0.0'3 means 0.0003. Downloaded 31 Mar 2011 to 134.208.24.193. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions CHARGED RIGHT CIRCULAR edge, the distance between the actual surface and the cylinder is found by multiplication of the potential difference at that point by eju. These deviations are shown in Table I. It is simpler to calculate the coefficients than the potentials. The potential at the pole due to the A .. side term is alrn+l(n-t) !GA" 00 (s-!) !lr' 2E21(1 +lr)l s=O (n+s+t) !(1 +b )8 CYLINDER wedge existed there which is strictly true only for positive deviations. 5. POTENTIAL OF CHARGED CYLINDER The potential outside the charged cylinder is V(p,\O,z)= -----L -----2 XP2,[b(1 +lr)-lJ, (12) where P 2 ,(x) is a Legendre polynomial. For hand calculation for large b values a much more complicated but rapidly convergent formula was used. The pole potential due to the Bn end terms is given by aBn{ (n-t)!G +FCt,n+ i ;n+(5/3); (1+4b2)-IJ}. (n+i) (1 +4b2)1 4E 2f(n+t)! ab2n+lGA .. 00 21r21E s=O X [In(b-18) +!¢t(s+n+ ll)-!¢t(s+!)-N ,J n (14) where No=O, 1 N,=2 -1 +-+ ... +1 ]. [ 1·2 3·4 2s(2s-1) For large values of b this was replaced by ab2n+l(n-t) !GAn 00 2E(1 +b )t2t(n+t)! ,91 [(s-!) !J2 - - -2 - - - L - - - 2 (s !)2(1 +b )' XF[s+!, n-s+t; n+l+t; b2(1+b 2)-I]. (15) The potential at the equator due to the two ends is (16) This converges slowly for small b values for which no formula suitable for hand calculation to eight places was found. Note that only four places are given for the equatorial potential when b=i. The potential at the edge is of no value in error estimation as, except by accident, it falls on neither a maximum nor a minimum. Its value is of some interest, however, and is given for small b values where a good formula was available. The actual distance between the Vo surface and the cylinder at the edge is small because of the large value of u there. The edge deviation given in the table is calculated as if the field of a right angle (17) The capacitance is found by integration of Eqs. (4) and (5) over the surface to get the total charge and division by V o which give C=1raE L [ (s!)2(s+n+t)W f 6. CAPACITANCE [(s-!)!J3(n-t)!(-b 2)' --- L -------- udS --, s 41rEr where r is the distance from the field point to the surface element dS and u is u. on the ends (5) and u, on the sides (4). An expression for the potential of the ends is easily found as a sum of spherical harmonics by taking the origin for each end at its center and integrating over ring elements. For the sides, however, all integrals found were double sums and none was valid over the whole external region. It is probable one must resort to numerical integration. (13) For small values of b an analytic continuation formulal was used. For small values of b the potential at the equator due to the sides is 919 2Bn (2b)ib 2n (n-t) !GAn] n+t (n+t)! --+ . (18) The error in this value should be much smaller than the surface deviations because positive and negative deviations tend to cancel. Thus at b= t a change in the equatorial potential deviation from 0.77% to 0.32% produces a change in C of only 0.0047%. Possibly even and odd numbers of correction terms give different sign errors in the capacitance because of a reversal in the dominance of positive and negative deviations. Note that successive correction terms alternate in sign. In any case more terms give greater accuracy. The following capacitance values are believed to be significant in the last digit. The thin disk for which the capacitance is exactly 8Ea is inserted for completeness. We write cia for b. cia o t t 1 2 4 C in farads 0.708347 X 10-10a 0.9222 X 10-loa 1.07208 X 10-10a 1.32585 X 10-10a 1.75098 XIo- l oa 2.465 X 10-10a. It would be useful to have some capacitance formula valid for intermediate values in this range. An approximate capacitance formula may be found from Eqs. (3), (4), (5), and (7) using only the fundamental terms. It gives values which deviate smoothly from the more exact values and so can be corrected by a simple addi- Downloaded 31 Mar 2011 to 134.208.24.193. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions w. 920 R. tive term. Thus, we obtain _ 1 C=af: - - - - - - - - - - - - - - - - F(!, Ii l+Ii x)+1.7248b i F(!, Ii l+*i Y) +0.714 exp( -1.55b l )-0.232b4 / 3 ] farads, (19) where x is (1+b2)-t, y is b2(1+b2)-1, b is the ratio of length to diameter and a is the radius in meters. The accuracy is one part in 1000 or better over the range 0<: b<: 4. In the ranges of x and y where the hypergeometric series converge slowly analytic continuation formulas1 will give rapid convergence. 7. OTHER APPLICATIONS Evidently the foregoing calculations can be carried out for other charged solids of revolution provided suitable expansions such as (4) and (5) can be found for the surface charge density. It is by no means certain that those used are best, even for a cylinder. Taylor has suggested certain orthogonal polynomials which may be better. The problem of the solid in an axial uniform field is easily solved by differentiation of Eq. (2) with respect to z so that -Ez= aV(z,p) az =I: (-1)P(~)2Pd2P.HV(Z'0). 0 (p 1)2 00 V(p,cp,z)-coscp L 12.566(1 +b2)l(1 + 1. 7248b l ) [ SMYTHE 2 dZ (20) 2 P+l Instead of Eq. (3) one now requires that E. cancel the applied field inside the cylinder and that higher derivatives vanish. In the cylinder case, this leads to almost identical hypergeometric functions but odd instead of even z dependence. For a traverse field one differentiates Eqs. (1) and (2) with respect to p, integrates with respect to z, and multiplies by coscp so that (p)2P-l d2p-l V (z,O) . pl(p-l)1 2 dZ 2p-l (-l)p (21) Instead of Eq. (3) one now requires that V=Ep cOScp cancel the applied potential of the uniform field Ell: inside and that higher derivatives vanish. Potential flow around a body may be found in exactly the same way by using the vector potential expansion2 analogous to Eq. (2) and requiring that A", be zero inside the body which makes the boundary a line of flow. The same method applies to two identical mutually external coaxial bodies of rotation with equal charges of the same or opposite sign. For example, the parallel plate capacitor problem where each plate is a short cylinder can be solved by Eq. (3) by taking the origin at the center of one of the cylinders and adding to d2 p V jdz2p the contribution from the image charge distribution on the other cylinder. Two equations like Eq. (5) will be needed because Bn will be different on the two cylinder faces. Lack of symmetry on the sides will require two factors (b-z1)n-l and (b+z1)n-l instead of the single factor W-zl)n-l in Eq. (5). Thus for the same accuracy twice as many coefficients will probably be needed as for the single cylinder. There is also another parameter, the distance between cylinders. Digital computers make such calculations feasible. There have been experimental indications that the old Kirchoff formula 3 is badly off in some cases. Other problems such as a charged right circular cylinder inside a coaxial conducting tube, potential flow about a right circular cylinder inside a coaxial pipe, etc., can be set up in similar fashion. Useful discussions on this problem have been held with Taylor who is working on the application of a similar method to electromagnetic wave antenna problems. I W. R. Smythe, Static and Dynamic Electricity (McGraw-Hill Book Company, Inc., New York, 1950), Eq. (5), p. 268. 8 H. Geiger and Karl Scheel, Handbuch der Physik (Verlag Julius Springer, Berlin, 1927), Vol. XII, p. 485. Downloaded 31 Mar 2011 to 134.208.24.193. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
© Copyright 2024 Paperzz