3.3. SEPARATION OF VARIABLES 137 3.3.2. Spherical Coordinates In the examples considered so far, Cartesian coordinates were clearly appropriate, since the boundaries were planes. For round objects spherical coordinates are more natural. In the spherical system, Laplace’s equation reads: peel Por I shall assume the problem has azimuthal symmetry, so that V is independent of 37 in that case Eq, 3.53 teduces to 1 v os 3. agi 7° (3.53) (3.54) As before, we look for solutions that are products: V(r, 8) = R(P)O@). (3.55) Putting this into Eq. 3.54, and dividing by V, sas (P%) + Rar \' ar Since the first term depends only on r, and the second only on 8, it follows that each must be a constant: CLAN lod dO) _ Ran (- +) =ll+), Osind do (sino a) 1b+1). (3.57) Here [(/ + 1) is just a fancy way of writing the separation constant—you'll see in a minute why this is convenient. 3, a (3.56) (snes The general case, for g-dependent potentials, is treated in all the graduate texts. See, for instance, J. D. Jackson’s Classical Electrodynamics, 3rd ed., Chapter 3 (New York: John Wiley, 1999)138 CHAPTER 3, SPECIAL TECHNIQUES ‘As always, separation of variables has converted a partial differential equation (3.54) into ordinary differential equations (3.57). The radial equation, d a (°) =1l +R, (3.58) dr dr has the general solution B RO) = Ar +5, (3.59) r as you can easily check; A and B are the two arbitrary constants to be expected in the solution of a second-order differential equatign. But the angular equation, £ (sno) =I +1)sin6 @, (3.60) is not so simple. The solutions are Legendre polynomials in the variable cos: (6) = Pi(cos8). (3.61) Pi(x) is most conveniently defined by the Rodrigues formula: 1 (a\' PO) = 5 (4) «1, (3.62) ‘The first few Legendre polynomials are listed in Table 3.1. Pot) 1 Pia) x P(x) = Gx? =1)/2 Px(x) (3x3 — 3x)/2 Pa(x) (35x4 — 30x? + 3)/8 Ps(x) (63x5 — 70x3 + 15x)/8 Table 3.1 Legendre Polynomials Notice that P)(.x) is (as the name suggests) an /th-order polynomial in x; it contains only even powers, if is even, and odd powers, if is odd. The factor in front (1/2!) was chosen in order that Py =1. (3.63) The Rodrigues formula obviously works only for nonnegative integer values of 1. More- over, it provides us with only one solution. But Eq. 3.60 is second-order, and it should pos- sess two independent solutions, for every value of /. It turns out that these “other solutions”SEPARATION OF VARIABLES 139 O and/or @ = x, and are therefore unacceptable on physical grounds.* For is blow up at @ instance, the second solution for ! 0) =In (0 5) 7 3.64) ‘You might want to check for yourself that this satisfies Eq. 3.60. Inthe case of azimuthal symmetry, then, the most general separable solution to Laplace's equation, consistent with minimal physical requirements, is B Vine) = (" + a) Pi(cos 0). (There was no need to include an overall constant in Eq, 3.61 because it can be absorbed into A and B at this stage.) As before, separation of variables yields an infinite set of solutions, one for each /. The general solution is the linear combination of separable solutions: oO B 4] vor.0=>° (a0! +a) Pi(cos9). | (3.65) Les ‘The following example: lustrate the power of this important result. Example 3.6 The potential Vo(@) is specified on the surface of a hollow sphere, of radius R. Find the potential inside the sphere. Solution: In this case B; = 0 for all —otherwise the potential would blow up at the origin, Thus, Vir.8 Y Arr! Py(c0s 9). (3.66) 0 Atr = R this must match the specified function Vo(@): V(R, 8) = Y>ApR'P)(cos#) = Vol). (3.67) = Can this equation be satisfied, for an appropriate choice of coefficients Ay? Yes: The Legendre polynomials (like the sines) constitute a complete set of functions, on the interval —1 iar P1005) = Vol). i= Multiplying by P) (cos @) sin @ and integrating —exploiting, again, the orthogonality relation 3.68-—we have oe f worn, (cos @) sin 6 dé. RFT 4 fy VOOR en or 21 ya ha Vo(0)Pi(cos@) sin 8 dé. Equation 3.72, with the coefficients given by Eq. 3.73, is the solution to our problem. Example 3.8 Anuncharged metal sphere of radius 2 is placed in an otherwise uniform electric field E = Eo [The field will push positive charge to the “northern” surface of the sphere, leaving a negative charge on the “southern” surface (Fig. 3.24). This induced charge, in turn, distorts the field in the neighborhood of the sphere.] Find the potential in the region outside the sphere. Solution: The sphere is an equipotential—we may as well set it to zero. Then by symmetry the entire xy plane is at potential zero. This time, however, V does nor go to zeto at large z. In fact, far from the sphere the field is Eo@, and hence Vi -Eg +C142 CHAPTER 3. SPECIAL TECHNIQUES Since V = 0 in the equatorial plane, the constant C must be zero. Accordingly, the boundary conditions for this problem are @ V=0 whenr = (i) V > —Egrcosé for > RB. ay We must ft these boundary conditions with a function ofthe form 3.65, “The first condition yields B, es AiR + oe 2 By = ~A, R241, (3.75) so EI 241 vir,0) =o Ar (" ao ) rico = Forr >> R, the second term in parentheses is negligible, and therefore condition (ii) requires that Y Air! P(cos a Evidently, only one term is present: ] = 1. In fact, since Pj(cos@) = cos@, we can read off immediately —Eor cosé. Aj =—Eo, all other Ay’s zero. Conclusion: RB V(r,0) = —Eo (r — = } cose. (3.76) The first term (—Eor cos) is due to the external field: the contribution attributable to the induced charge is evidently 3 Foy 0s If you want to know the induced charge density, it can be calculated in the usual way: av aW)=~ ote] = 3eq Fy cos 8. B77 As expected, it is positive in the “northern” hemisphere (0 < @ < 7/2) and negative in the “southern” (1/2 <6 < 7). Example 3.9 A specified charge density 09 8) is glued over the surface of a spherical shell of radius R. Find the resulting potential inside and outside the sphere. Solution You could, of course, do this by direct integration:a3) SEPARATION OF VARIABLES 143 but separation of variables is often easier. For the interior region we have VOr.8) => Ayr! Py(cos8) (r= R) (3.78) i= (no By terms—they blow up at the origin); in the exterior region SB Vir.) = Pcos8) (r= R) 3.79) 10" (no A; terms—they don’t go to zero at infinity). These two functions must be joined together by the appropriate boundary conditions at the surface itself, First, the potential at = R (Eq. 2.34) s continuous Dark! Acos8) = > (3.80) io = It follows that the coefficients of like Legendre polynomials are equal: By = ApRa+h 81) (To prove that formally, multiply both sides of Eq, 3.80 by Py (cos) sin@ and integrate from 0 to x, using the orthogonality relation 3.68.) Second, the radial derivative of V suffers a discontinuity at the surface (Eq, 2.36) (Ge- F") ar © a 2 -ya+ D grag Piteos) — DAR! Pycos) i= i = -+ a6) (3.82) © Thus or, using Eq. 3.81 < 1 Deal + parr’! Py(cos6) = —o0(6). (3.83) 9 0 From here, the coefficients can be determined using Fourier's trick: A= f a9(8)P)(cos6) sind dé. (3.84) RI“! Jo Equations 3.78 and 3.79 constitute the solution to our problem, with the coefficients given by Eqs. 3.81 and 3.84. For instance, if (8) = kcosé = kP|(cos8), (3.85) for some constant k, then all the Aj’s are zero except for ! = 1, and k 7 2 k A= [[Acosey? sina = 1 Peal (Cas6)P sino dd =144 CHAPTER 3. SPECIAL TECHNIQUES ‘The potential inside the sphere is therefore (rs R), 3.86) whereas outside the sphere vo. =F cose > &). G87) In particular, if 9(9) is the induced charge on a metal sphere in an external field Ep, so that k = 3eqEq (Eq. 3.77), then the potential inside is Egr cos@ = Eoz, and the field is — Egi—exactly right to cancel off the external field, as of course it should be. Outside the sphere the potential due to this surface charge is 3 E> c0s8, our conclusion in Ex. 3.8. consistent Problem 3.16 Derive P;(x) from the Rodrigues formula, and check that P3(cos 9) satisfies the angular equation (3.60) for! = 3. Check that P3 and P, are orthogonal by explicit integration. Problem 3.17 (a) Suppose the potential is a constant Vo over the surface of the sphere. Use the results of Ex. 3.6 and Ex. 3.7 to find the potential inside and outside the sphere. (Of course, you know the answers in advance—this is just a consistency check on the method.) (b) Find the potential inside and outside a spherical shell that carries a uniform surface charge ap. using the results of Ex. 3.9. Problem 3.18 The potential at the surface of a sphere (radius R) is given by Vo = kos 30, where & is a constant. Find the potential inside and outside the sphere, as well as the surface charge density o (0) on the sphere. (Assume there's no charge inside or outside the sphere.) Problem 3.19 Suppose the potential Vp(@) at the surface of a sphere is specified, and there is no charge inside or outside the sphere. Show that the charge density on the sphere is given by os 2, o(@)= aR Let 1)°C; Pi(cos#), (3.88) where cr = [vote (cos) sind (3.89) f3.3. SEPARATION OF VARIABLES 145, Problem 3.20 Find the potential outside a charged metal sphere (charge Q, radius R) placed in an otherwise uniform electric field Eo. Explain clearly where you are setting the zero of potential Problem 3.21 In Prob, 2.25 you found the potential on the axis of a uniformly charged disk: Vor,0) WP ER =r) 0 (a) Use this, together with the fact that P)(1) = 1, to evaluate the first three terms in the expansion (3.72) for the potential of the disk at points offthe axis, assuming r > R. (b) Find the potential for r < R by the same method, using (3.66). (Note: You must break the interior region up into two hemispheres, above and below the disk. Do not assume the coefficients A; are the same in both hemispheres.) Problem 3.22 A spherical shell of radius R carries a uniform surface charge ap on the “north- em” hemisphere and a uniform surface charge —op on the “southern” hemisphere. Find the Potential inside and outside the sphere, calculating the coefficients explicitly up to Ag and Bg, Problem 3.23 Solve Laplace's equation by separation of variables in cylindrical coordinates, assuming there is no dependence on z cylindrical symmetry). [Make sure you find all solutions to the radial equation; in particular, your result must accommodate the case of an infinite line charge, for which (of course) we already know the answer.] Problem 3.24 Find the potential outside an infinitely tong metal pipe, of radius R, placed at right angles to an otherwise uniform electric field Eo. Find the surface charge induced on the pipe. (Use your result from Prob. 3.23.) Problem 3.25 Charge density o@)= (where a is a constant) is glued over the surface of an infinite cylinder of radius R (Fig, 3.25). Find the potential inside and outside the cylinder. [Use your result from Prob. 3.23. sin Sp Figure 3.25146 CHAPTER 3. SPECIAL TECHNIQUES 3.4 Multipole Expansion 3.4.1 Approximate Potentials at Large Distances Ifyou are very far away from a localized charge distribution, it “looks” like a point charge. and the potential is—to good approximation—(1/42r¢9)Q/r, where Q is the total charge. We have often used this as a check on formulas for V. But what if Q is zero? You might reply that the potential is then approximately zero, and of course, you're right, in a sense (indeed, the potential at large r is preity small even if Q is not zero). But we're looking for something a bit more informative than that. Example 3.10 A (physical) electric dipole consists of two equal and opposite charges (iq) separated by a distance d. Find the approximate potential at points far from the dipole. Solu : Let 2 be the distance from —q and 2. the distance from +g (Fig. 3.26). Then and (from the law of cosines) a Frd cos8 “). We're interested in the régime r >> d, so the third term is negligible, and the binomial expansion yields Leif_a. yl? D(a 2 cond) Het (1+ feos) Thus and hence Figure 3.263.4. MULTIPOLE EXPANSION 147 . ao =F s . — ¥ Monopole Dipole Quadrupole Octopole (V~ Ur) (V~ lr (V~ lr?) (V~ Ur’) Figure 3.27 Evidently the potential of a dipole goes like I/r? at large r; as we might have anticipated, it falls off more rapidly than the potential of a point charge. Incidentally, if we put together a pair of equal and opposite dipoles to make a quadrupole, the potential goes like 1/r?; for back-to-back quadrupoles (an octopole) it goes like 1/r*; and so on. Figure 3.27 summarizes this hierarchy; for completeness I have included the electric monopole (point charge), whose Potential, of course, goes like I/r. Example 3.10 pertained to a very special charge configuration. I propose now to develop a systematic expansion for the potential of an arbitrary localized charge distribution, in powers of 1/r. Figure 3.28 defines the appropriate variables; the potential at r is given by tft V(r) = sag f foeeae 91) axe J 2 Using the law of cosines, 24 (r')? — 2rr' cos’ or =rvite (3.92) c=(2) (6-0) For points well outside the charge distribution, ¢ is much less than 1, and this invites a binomial expansion: where (3.93) Figure 3.28