Physics 322: Example of multipole expansion

Carl Adams, St. FX Physics November 25, 2009

x (4d,0,3d)

curlyr q 00 11
11 00 11 00

d x

All distances in this problem are scaled by d. The source charge q is oset by distance d along the z-axis. 1. Lets start by calculating the exact potential at the eld point r = 4d + 3d. This is easy to x z do since there is only one source charge. V (4d, 0, 3d) = With V0 =
q 40 d

q 1 q 1 0.223607 V0 = 40 40 d 20

(1)

as a convenient unit of voltage.

2. As I said in class the beauty of the multipole expansion is that we attribute QT , p, and Q2 to properties of the charge distribution in the same way we would give our own height, weight, and shoe size; they dont depend on the eld point. Also when they are complete they contain information about the source coordinates but they dont contain the source coordinates as direct variables. Now with just one charge in the distribution I would not recommend using a multipole expansion to approximate V (r) (i.e. you can easily write the closed, exact expression) so the steps I am showing here are merely meant to illustrate how you would do this for general charge distributions. (a) The total charge Q is simply q. 1

(b) The dipole moment p=

qk rk = qd z

(2)

The sum k is over all of the charges in the array. There are analogous expressions for line charges, surface charges, and general (r ) that involve integrals rather than sums. (c) The quadrupole moment tensor Q2 = qk qd2 2 3rk rk rk I = 3z I z 2 2 (3)

We can also represent the quadrupole moment tensor as a matrix where we multiply column vector z by row vector z rather than the opposite order (which would give a dot product). 0 Q2 = 3 0 2 1 qd2

0 0 1

1 0 0 1 0 0 qd2 0 1 0 = 0 1 0 2 0 0 2 0 0 1

(4)

So which one is correct expression 3 or 4? Both are equally correct since we are all ||||, trying to do is represent a geometrical, mathematical object. (i.e. ve, V, 25, a hand held up in the air with ngers spread apart all represent 5 the integer that is one bigger than four and one smaller than six but one of them is good for Superbowls and movie sequels while another is great for keeping count). 3. Now lets calculate the separate monopole, dipole, and quadrupole contributions to the potential. The multipole expansion is V (r) 1 40 r p r Q2 r QT + 2 + 3 r r r (5)

Notice that this expression contains r only. There is neither nor any source coordinates. (a) First we need r r= r = r 4d + 3d x z = 0.8 + 0.6 x z 2 + (3d)2 (4d) 1 = 0.200000 V0 + (3d)2 (6)

(b) The monopole contribution is Vmono (r) = 1 QT q = 40 r 40 (4d)2 (7)

Wow, we are already pretty close. The quality of the monopole approximation should go as roughly the size of the charge distribution divided by the distance to the eld point. In this case this is d/5d or 20%. This looks reasonable. (c) The dipole contribution is x z z z 1 rp 1 (0.8 + 0.6) qd q d 0.6 == 0.024000 V0 (8) = == 2 2 2 40 r 40 25d 40 25d A bit improvement. The percentage dierence between the multipole approx with two terms and the actual answer is 0.18 %. This is actually way better than we would expect! (and should have been my rst clue to the problems I would enounter later). We would expect it to be as least as good as 4% (square of the size/distance ratio). Vdip (r) = 2

(d) The quadrupole contribution is Vquad (r) = 1 r Q2 r 3 40 r (9)

Lets use the matrix multiplication technique rst. The rst r is represented as a row matrix and the second as a column matrix. The idea is that Vquad must be a scalar Vquad = 1 3 d 40 2 125 1 250
q d2

0.8 0 0.6

= =

0.8 0 0.6

0.64 + 0.72 V0 = 0.00032 V0 (10) 250 As long as we realize that r I r = 1 and remember to do the dot products from the front and back we can also use expression 3 Vquad (r) = = 1 qd2 r (3z I) r z 40 2 (5d)3 (0.8 + 0.6) (3z I) (0.8 + 0.6) x z z x z

0.8 0 V0 1.2

1 0 0 0.8 0 1 0 0 0 0 2 0.6

V0 2 125 3 0.6 0.6 1 = V0 = 0.00032 V0 (11) 250 So the 3-term multipole expansion now gives us V 0.200000 + 0.024000 + 0.00032 = 0.22432V0 . Hmm, but this is slightly worse than a 2-term expansion (but still pretty good 0.33%). More confusing is that if we subtract the quadrupole term we get V 0.2236800 which is almost exactly the exact answer given in equation 1 (within 0.033%). Arrgh! 4. The problem is that I have choosen a location in space where the magnitude of the quadrupole correction is very small (I did know that going in) and where the octupole correction is roughly twice as big in magnitude and opposite in sign (that was a surprise)! You probably dont want to work with third-rank tensors or go back and work out the other terms in the 1/ expansion so I will use a shortcut to gure out the octupole moment that works when the charges are distributed along the z-axis. (a) Recognize that r z = cos if we give r in (r, , ). This allows us to replace the dot products in Vdip and Vquad to give V (r, , )mono = V (r, , )dip = V (r, , )quad = 1 q (1) 40 r 1 qd (cos ) 40 r 2 1 qd2 1 (3 cos2 1) 40 r 3 2 (12) (13) (14)

Hey, those terms in brackets are P (u = cos ). This is not an accident. We have tried to rank the contributions to the potential in terms of powers of r so each term should be a separable solution to Laplaces equation. 3

(b) So with this in mind V (r, , )oct = 1 qd3 1 (5 cos3 cos ) 40 r 4 2 (15)

since P=3 (u) = 1 (5u3 u). (This isnt true for all charge distributions; just this one). 2 In this case cos = 0.6 V (r, , )oct = 1 qd3 1 (5(0.6)3 0.6) = 0.000576 V0 40 (5d)4 2 (16)

(c) Now V 0.200000 + 0.024000 + 0.00032 0.000576 = 0.223744, a percentage dierence of 0.06% from the exact answer. We expect something on the order of (d/r)4 = 0.16% so we are still doing better than expected. (d) The lesson here. There are times that because of the charge distribution or because of the eld point that the poles can get out of order. If you look at the expression in terms of you can see where the nodes are. Even a clock that doesnt work is exactly right twice a day.