Vous êtes sur la page 1sur 3

c F.

Walee, Math 321, 2008/03/14

2.2+ Gradient extras


Geometric denition of gradient: Given a (suciently nice) scalar eld f (r), e.g. temperature as a function of position, its gradient f at point r is a vector pointing in the direction of greatest increase of f . The magnitude of f is the rate of change of f with distance in that direction. It follows that the gradient at a point is perpendicular to the isosurface of f that passes through that point.

Fundamental examples: (1) If f = f (r) where r = |r|, the scalar eld depends only on distance to the origin, then f is constant if r is constant, so the isosurfaces are spheres, f (r) is in the radial direction r and its magnitude is the rate of change in the radial direction df /dr so f (r) = df r. dr

(2) If f = f (|r r c |) so the scalar eld depends only on distance to point r c , then f is constant if |r r c | is constant, so the isosurfaces are spheres centered at r c and, letting r r c = s s with s = |r r c | f (s) = df df r r c s= . ds ds |r r c |

(3) If f = f (r c) where c is a xed vector, then f is constant if r c is constant, so the isosurfaces are planes perpendicular to c. Let s = r c = |c| where is distance in the c direction so df df c = c. f (r c) = d |c| ds In particular if c = x then r x = x and f (x) = x df /dx.

It also follows from the geometric denition that the dierential change df in the value of f when r changes by an arbitrary dierential dr is f dr
d
f f+

df = dr
df

(1)

That is because, locally, the isosurfaces are planes perpendicular to the gradient, so a displacement dr leads to a change in f in proportion to the component of the displacement in the direction of the gradient d = dr( f )/| f |. The change in f is df = d | f | = dr f . That general relationship (1) between df and dr allows us to obtain the expression for the f in various sets of coordinates, including non-cartesian coordinates.

c F. Walee, Math 321, 2008/03/14 Gradient in cartesian coordinates

The position vector reads r = x + y y + z z . Picking dr = x dx, dr = y dy and dr = z dz x in (1), respectively, gives r f = x x r f = y y f r = z z These three relationships imply f =x Gradient in spherical coordinates Here x = r sin cos , y = r sin sin , z = r cos , so r = r = r ( sin cos + y sin sin + z cos ), r x (6) f f f +y +z . x y z (5) f =x f =y f =z f, f, f. (2) (3) (4)

where r is the distance to the origin, is the polar angle (co-latitude) and is the azimuthal angle (longitude). In earlier sections on spherical coordinates and volume parametrizations, we discussed/derived r = r, r r = r , r = r sin . (7)

These results can be obtained via the hybrid cartesian/spherical expression (6) for r or directly from the geometry and understanding of partial derivatives (see (14, 15) below). Then, from (1) and (7), f r = r r f r = f r = f =r f = r f, f, f. (8) (9) (10)

f = r sin

Since r , and are orthonormal, these expressions imply that f =r f 1 f + 1 f . + r r r sin (11)

Note that the gradient in spherical coordinates is not That expression is not even dimensionally correct.

f = r f /r + f / + f /!

c F. Walee, Math 321, 2008/03/14

To derive the spherical coordinates expression for other operators such as divergence v, curl v and Laplacian 2 = , one needs to know the rate of change of the unit vectors r , and with the coordinates (r, , ). These vectors change with (r, , ) unlike the cartesian direction vectors x, y , z which are the same at every point. Since a partial with respect to r means the rate of change in a xed radial direction (, xed), it should be clear geometrically that r , and do not change with r r = = = 0. r r r (12)

To deduce the rates of change with respect to and , we could start from the hybrid expression (6) for r(r, , ) and crank it out, but a much faster, geometric approach is to use our knowledge of rotation: if a vector v rotates about then its governing equation is dv/dt = v where t is time. In dierential form, this is dv = dt v or dv = d v (13)

where d = || dt is the dierential angle of rotation during the time interval dt. The / derivatives correspond to innitesimal rotation in meridional planes since r and are xed. This is rotation by d about hence from (13), we obtain (14) r = r = , = = , r = = 0. (14)

Likewise / corresponds to innitesimal rotation about z by angle d, hence from (13), z r Exercises: 1. What is v in spherical coordinates where v = r u + v + w ? 2. Derive the gradient in cylindrical coordinates and the derivatives of the cylindrical direction vectors. z r = r = sin , z = = cos , z = = sin r cos . z (15) (16) (17)