1

PHY481 - Lecture 12: Solutions to Laplaces equation

Griths: Chapter 3 Before going to the general formulation of solutions to Laplaces equations we will go through one more very important problem that can be solved with what we know, namely a conducting sphere (or cylinder) in a uniform eld. We will solve this problem using a superposition of solutions including the dipole potential we found above. Conducting sphere or cylinder in a constant electric eld We consider a grounded conducting sphere placed in a uniform electric eld. In this case it is not clear how to use the method of images as there are no charges dened in the problem. The potential due to a uniform electric eld E = E0 z is E0 z or in polar co-ordinates E0 rcos. If we combine this with the point charge and dipole potentials, with arbitrary constants, we have, V (r, ) = A cos + B + C 2 + Drcos r r r>a (1)

where a is the radius of the sphere. We know that each term in this expression satises Laplaces equation, so the sum does too - that is superposition again. Note that this is not the general form of the solution in spherical co-ordinates, but it will turn out to be sucient for the problem of a conducting sphere in an electric eld. We dont expect the rst two terms to contribute as the rst corresponds to a net charge on the sphere and the second to a nite potential. By symmetry we expect the rst term to be zero while the second term is zero because the sphere is grounded. To ensure that the solution is correct as r , we need D = E0 . Since the conducting sphere is grounded, we must have the boundary conditions that the potential at the surface of the sphere be equal to zero V (r = a, ) = 0. If we choose C = a3 E0 , A = B = 0 then we satisfy the boundary conditions on the potential so we suspect the solution is, V (r, ) = E0 cos[r a3 pinduced r ] = E0 z + k 2 2 r r (2)

where the induced dipole moment is given by, pinduced = E0 a3 /k. The induced dipole moment is then proportional to the electric eld. The electric eld outside the sphere is given by, Er = V 2a3 = E0 (1 + 3 )cos; r r E = 1 V a3 = E0 (1 3 )sin r r (3)

Notice that E (a, ) = 0 as it must be to ensure that the electric eld at the surface of the conductor is perpendicular to the surface. At this point we have satised all the boundary conditions so we know the solution is correct and unique. The induced charge density at the surface of the sphere is calculated from () = 0 Er (r = a) = 3 0 E0 cos. Integrating over the sphere surface gives the net charge on the sphere,

Q=
0

2a2 sin(3 0 E0 cos)d = 0

(4)

which shows that no charge is transferred from ground to the sphere, so this result also applies when the sphere is isolated. Moreover if we integrate over the top hemisphere,
/2

Q=
0

2a2 sin(3 0 E0 cos)d = 3a2 0 E0

(5)

we see that a positive charge is induced in the top hemisphere. There is an equal amount of negative charge on the lower hemisphere. Note that the net force on the conducting sphere is zero, as the net force on a dipole in a uniform eld is zero. A conducting or dielectric sphere is attracted into an electric eld, in order to screen out the eld and hence reduce the energy. This is the principle that allows optical tweezers (lasers) to manipulate quantum dots and biomolecules tethered to latex spheres. If the sphere is charged, with an additional charge Q0 , this additional charge is distributed uniformly on the sphere surface and its eect on the potential is found by superposition. Exactly the same procedure works for conducting cylindrical in a constant applied electric eld, where the solution is, V (r, ) = E0 cos(s + a2 ), s (6)

2 by following the same reasoning as for the spherical case above. The electric eld components are Er = E0 cos(1 + a2 /r2 ); E = E0 sin(1 a2 /r2 ) and the surface charge density is 2 0 E0 cos. Integration shows that there is no net charge transfer to the conducting cylinder so this solution also applies for an isolated conducting cylinder in a constant applied eld. Addition of a charge to an isolated cylinder leads to a uniform charge density on the surface of the cylinder and an additional term Alnr in the potential. NOTE: this calculation does not allow the cylinder to move. What if we allow the cylinder to rotate? What orientation do you think it will adopt? Laplaces equation in the three co-ordinate systems Now we are ready to look at more general procedures for solving Laplaces equation, any othogonal co-ordinate system is (see Lecture 4),
2 2

V = 0. The Laplacian in

V =

1 h2 h3 V h1 h3 V h1 h2 V [ ( )+ ( )+ ( )] h1 h2 h3 u1 h1 u1 u2 h2 u2 u3 h3 u3

(7)

For Cartesian co-ordinates we have u1 = x, u2 = y, u3 = z, and h1 = 1, h2 = 1, h3 = 1, so that,

2

V =

2V 2V 2V + + . x2 y 2 z 2

(8)

In cylindrical co-ordinates u1 = s, u2 = , u3 = z, and h1 = 1, h2 = s, h3 = 1, so that,

2

V =

1 V 2V 1 2V + . (s )+ 2 s s s s 2 z 2

(9)

In spherical polar co-ordinates u1 = r, u2 = , u3 = , and h1 = 1, h2 = r, h3 = rsin, so that,

2

V =

1 2 V 1 V 1 2V (r )+ 2 (sin )+ 2 2 . 2 r r r r sin r sin 2

(10)

Problems with dependence on only one co-ordinate (i.e., z or s or r) In Cartesian co-ordinates, this corresponds to the potential outside a sheet of charge, or equivalently the potential of a constant electric eld E0 z. In fact the general solution is quite trivial, namely, 2V =0 z 2 so V (z) = a + bz (11)

where a and b are two constants that can be used to t the boundary conditions. The solution is thus always linear. In cylindrical co-ordinates, this corresponds to the potential outside a line charge, so we expect a logarithmic solution to be possible (i.e. ln(s)) Laplaces equation in this case is,
2

V =

1 V V (s ) = 0; so s =b s s s s

(12)

where b is a constant. Integrating we nd, V (s) = a + bln(s), which is the general solution in the cylindrical case. In spherical polar co-ordinates,
2

V =

1 2 V V (r ) = 0; so r2 =b 2 r r r r

(13)

Integrating, we nd the general solution V (r) = a + b/r, which is the potential outside charge distributions with spherical symmetry. The one variable solutions to Laplaces equation are monotonic i.e. decreasing or increasing with no minima or maxima on their interior. Nevertheless electrostatic potential can be non-monotonic if charges are added as a change in slope can occur at charges. However if a charge is included, the one dimensional solutions to Laplaces equation only apply to the regions of space that do not have charge. They do NOT apply in regions where there is charge, for example inside a sphere with a constant charge density. In that case we have to solve Poissons equation, for example in the case of a sphere with constant charge density we solve,
2

V =

1 2 V (r )= ; 2 r r r 0

(14)

3 We can also solve this. First move the r2 to the other side and integrate, to nd, r2 V r3 = + b Solving yields r 3 0 V (r) = r2 b +d 6 0 r (15)

Recall that we found the solution in Problem 2:21, kQ/R + (R2 r2 )/(6 0 ), which is of course consistent with the solution found from Poissons equation. Problems with dependence on two co-ordinates - separation of variables General properties First we can easily generalize the remark about monotonicity of solutions to Laplaces equation to higher dimensional cases. The generalization is that if we solve Laplaces equation on a nite domain, there are no maxima or minima of the potential in the interior of the domain. Therefore the minima and maxima only occur on the boundaries. To see why this is true, consider Lapalaces equation in Cartesian co-ordinates. It is clear that at least one of the terms must be negative and at least one must be positive, implying that in at least one direction the curvature of the solution must be negative and in at least one direction the curvature of the solution must be positive. This implies that there are no minima or maxima which require all positive and all negative curvature respectively. Instead all we can have on the interior of the domain are saddle points. This is very dierent than solutions in quantum mechanics where the presence of the energy in the equation leads to oscillatory eigenfunctions, where the number of minima and maxima on the domain is related to the energy of the quantum particle. Cartesian co-ordinates in two dimensions Laplaces equation is then, 2V 2V + =0 2 x y 2 (16)

This can no longer be directly integrated, so we try to reduce it to a simpler form by assuming a solution of the form V (x, y) = X(x)Y (y). Subsituting this expression into Eq. (16), we nd that, d2 X d2 Y Y +X 2 =0 dx2 dy or 1 d2 X 1 d2 Y = = constant = k 2 X dx2 Y dy 2 (17)

The requirement that the last expressions have to equal a constant occurs as if we have f (x) = g(y), then both f and g must be constant. This is proven by taking a derivative with respect to x, so that df /dx = dg/dy = 0. Then integrate with respect to x to show f (x) = constant. It is convenient to call the constant k 2 . The sign dierence between the X equation and the Y equation, and whether k is real or imaginary, is very important and determines whether the solutions to X or Y are oscillatory or decaying functions. From this expression we have, X(x) = A (k)eikx + B (k)eikx ; Y (y) = C (k)eky + D (k)eky (18)

Often it is convenient form to write the solutions as combinations of odd and even functions, namely, X(x) = A(k)cos(kx) + B(k)sin(kx); Y (y) = C(k)cosh(ky) + D(k)sinh(ky) (19)

The constants A, B, C, D can in general depend on the separation constant k. We can also add the one dimensional solutions ie ax + b and cy + d which correspond to the separation constant k = 0. The general solution is then, V (x, y) = (a + bx)(c + dy) +
k

[A(k)cos(kx) + B(k)sin(kx)][C(k)cosh(ky) + D(k)sinh(ky)]

(20)

The values of k that are used depends on the problem and often an innite series of values is required. That is, we use superposition to add solutions with dierent values of the separation constant k. This gives us a lot of freedom in choosing which terms to try, and a lot of constants that we can use to nd a solution. Usually many of these are removed using symmetry arguments. A similar procedure may be used to nd solutions in three dimensions. Circular co-ordinates Circular co-ordinates are the (s, ) part of cylindrical polar co-ordinates. The Laplacian in cylindrical polar coordinates is given by,
2

V =

1 V 1 2V (s )+ 2 =0 s s s s 2

(21)

4 Assuming that V = S(s)(), we nd, s S 1 2 =0 (s ) + S s s 2 We now separate the equation using, s S 1 2 = k2 (s ) = S s r 2 (23) (22)

The equation has solutions cos(k), sin(k) when k is nite. When k = 0, it has solution C + D. Noting that () = ( + 2) is required to ensure that is single valued, we set k = n, where n = 0, 1.... We could also use n negative, but that would just change the sign of the constants Dn in a general solution of the form Cn cos(n) + Dn sin(n). The s equation has the form, s2 d2 S dS +s n2 S = 0 2 ds ds (24)

which for n = 0 has solutions S = An sn +Bn sn , while for n = 0, S = A+Blns. With these observations, we nd that the general solution to Laplaces equation in circular co-ordinates (ie. cylindrical co-ordinates with no dependence on z) is,

V (s, ) = (A + Bln(s))(C + D) +
n=1

(An sn +

Bn )(Cn cos(n) + Dn sin(n)) sn

(25)