This action might not be possible to undo. Are you sure you want to continue?
Vector differentiation, the ∇ operator, grad, div and curl.
Syllabus topics covered: 1. Vector ﬁelds 2. Grad, div and curl operators in Cartesian coordinates. Grad, div, and curl of products etc. Here we cover differentiation of vectors. Note that this differs from the gradient introduced in Chapter 1, which we obtained by differentiating a scalar ﬁeld.
3.1 Vector functions of one or more variables
(See Thomas 13.1) In many physical contexts one is interested in vectors that vary with position or time. For example, the position of a point can be described by a vector r. Thus, if we consider a moving particle, its position can be described as a function of time t by the vector r(t ), and its rate of change with respect to t will be the velocity (which has magnitude and direction, i.e. is a vector: its magnitude is the speed). The position vector is then a function of one variable. Another context is where we have a vector deﬁned at each point, say F(r) = F(x, y, z) and a curve with a parameter u, say, so its points are (x(u), y(u), z(u)). Then we can deﬁne a vector function of u, F(u) = F((x(u), y(u), z(u)). We can deal with this and the moving particle case as follows. A vector function of a scalar u, F(u), can be deﬁned by specifying its components as functions of u: F(u) = ( f1 (u), f2 (u), f3 (u)) . The derivative dF/du of F with respect to u is then: dF = du d f1 d f2 d f3 , , du du du .
This simply goes back to the fundamental deﬁnition of a derivative: dF F(u + δ u) − F(u) = lim . du δ u→0 δu 28
that if F and G are vector functions of u. Example 3.G. by writing out the components and collecting terms. by differentiating the components of F the required number of times.Clearly one can compute higher derivatives. = (cos t )i + j.G. + . as a function of time t . Also. Example 3. The continuous parameter t can take all real values. du du du Proof: d(F.1.G) dG dF = F. 2π π t=π t = π/2 -1 −π − 2π t=0 1 Figure 3. + .1: Sketch of the curve deﬁned parametrically by r = (sin t )i + t j It is easy to prove. du du = Exercise 3. Sketch the curves whose parametric equations are 29 . The ﬁrst and second derivatives are dr dt d2 r dt 2 The sketch is shown in Fig. such as d2 F/du2 . 3. then d(F. = (− sin t )i. Write down the derivatives dr/dt and d2 r/dt 2 for the vector r = (sin t )i + t j.1.E.D. then dr/dt is the velocity v of the particle. Also dv/dt ≡ d2 r/dt 2 is the particle’s acceleration. sketch the curve whose parametric equation is r = r(t ).1. If r(t ) is the position vector of a particle. Q.G) du d ( f 1 g1 + f 2 g2 + f 3 g3 ) du d f2 d f3 dg1 dg2 dg3 d f1 = f1 + f2 + f3 + g1 + g2 + g3 du du du du du du dG dF = F.2.
or we may only be interested in their values on a particular path r(s). v). Here we will be assuming that F is a smoothly-varying function of position. y. v). ∂u ∂u ∂u . y. in terms of partial derivatives of its components. z. if F = ( f1 (u. and this vector (in general) varies with position so it is a vector ﬁeld. 3. We can add vector ﬁelds and multiply them by a constant in the obvious way.(a) r = (3 sin π t )i + (2 cos π t )j (b) r = (cos π t )j (c) r = t i + t 2k (−∞ ≤ t ≤ ∞). moving because of currents and tides. v).2 Vector Fields (See Thomas 16. Note: the set of all 9 derivatives of a component by a coordinate forms a quantity of a new kind. called a tensor. f3 (u. solid mechanics and relativity. y. One case where this arises is if the vector is deﬁned at each point on a surface (x(u. If we attach a velocity vector to each point of the ﬂowing ﬂuid. so at each point in space we have a vector. then ∂F = ∂u ∂ f1 ∂ f2 ∂ f3 . and if λ is a constant then λ F is also a vector ﬁeld. z(u. then it is straightforward to deﬁne its partial derivatives with respect to u or v. A physical example of a vector ﬁeld is the velocity in a ﬂowing ﬂuid (e. i. this gives us a total of 9 quantities for each of the x. we shall be concerned mostly with vectors and scalars which depend on position in three-dimensional space. moving because of winds). v)). Thus. which are functions of three variables x. so if F and G are two vector ﬁelds then F + G is also a vector ﬁeld. 30 . for example. F3 . v)). say F = F(u.2) For the rest of this course. These are used in ﬂuid mechanics. for example. the water in the oceans. y. y. z). z derivatives of F1 . Sometimes they may depend also on a fourth variable.g. now things are not necessarily moving with time. v). which again depend on position: F = (F1 (x. y. The velocity at any point in the ﬂuid is a vector quantity – it has magnitude and direction. f2 (u.2. We can also write F in terms of its components. y(u. F2 . Given a vector ﬁeld. we could of course now differentiate the vector ﬁeld with respect to each of the coordinates (x. z)) . We write a vector F that varies with position as F = F(x. or the air in the atmosphere.e. z) in turn. in the manner described in the previous section. A vector depending on position is said to constitute a vector ﬁeld. so all these derivatives exist at all points of interest r. and write down the derivatives dr/dt and d2 r/dt 2 where they are deﬁned. F2 (x. . 2 If F is a vector function of more than one variable. we have a vector ﬁeld deﬁned in the region occupied by the ﬂuid. y. (except possibly for one or two singular points). but the magnetic ﬁeld has a direction and a strength at each point in space. z). F3 (x. Another physical example is a magnetic ﬁeld. such as time t . z) ≡ F(r) An example is shown in Figure 3. v).
it will turn out that we have to take certain combinations which are “well behaved” if we rotate the x. so it is a scalar ﬁeld.0 0. We can also get the above result if we write out ∇ and F in components. in this course we will not deal with tensors.0 Figure 3.8 1.6 0. To do this.1.6 y 0. where ∂ ∂ ∂ ∇ = i +j +k ∂x ∂y ∂z is the operator called “del” which we met previously in forming the gradient of a scalar. z) = F1 i + F2j + F3k is a vector ﬁeld. but it is a vector differential operator. y.2: Example of a ﬂow.2 0.0 0. In this case the speed and direction at each point is a function of the position (x. we will restrict ourselves to forming scalar and vector quantities from these derivatives. these will turn out to be forming the dot and cross products of ∇ with F . z axes. y.8 0.8) Suppose F(x.0 0.4 x 0. Note again that ∇ is not a true vector (because on its own we can’t deﬁne its length or direction). 3. y) However.3 The Divergence of a vector ﬁeld (See Thomas 16.2 0.(F1 i + F2j + F3k) ∂x ∂y ∂z 31 . The divergence of F is deﬁned to be ∇·F = ∂ F1 ∂ F2 ∂ F3 + + ∂x ∂y ∂z Here div F is a scalar and depends on position. ∇ · F = (i ∂ ∂ ∂ + j + k ).4 0.
that if F. G are any two vector ﬁelds.2. ∂x ∂y ∂z Exercise 3. If F = (y − x)i + (z − y)j + (x − z)k. we have produced a scalar ﬁeld ∇ · F. that div behaves as expected for addition and multiplication by a constant. − − − ∂y ∂z ∂z ∂x ∂x ∂y Note that curl F is another vector ﬁeld. We can write ∇ × F as curl F – again the two notations are completely interchangeable. F1 F2 F3 It is easy to verify. ∇ · (F + G) = (∇ · F) + (∇ · G) . Note that.e. and a magnetic ﬁeld B is an example of a ﬁeld that has ∇ · B = 0 everywhere (this is an observational fact. then on balance the vector ﬁeld is pointing away from the point and out of the surface. calculate ∇ · F. 3. It is convenient to remember ∇ × F in terms of a determinant like the one for v × w: i j k ∇ × F = ∂ /∂ x ∂ /∂ y ∂ /∂ z .3. ∇·F = ∂ (3xy2 ) ∂ ez ∂ (xy sin z) + + = 3y2 + xy cos z. by writing out the components. by writing out the determinant in full. [Answer: -3] 2 3. If the divergence is negative. by direct calculation. calculate ∇ · F. We can also write ∇ · F as div F.i = 1. etc. and one considers a small closed surface surrounding that point. if at some point in space the divergence is positive. Loosely speaking. i. i. (See Fig.4 The Curl of a vector ﬁeld (See Thomas 16. then on balance the vector ﬁeld is pointing towards the point and into the surface.and use the properties i. The reason for the name solenoidal is historical: that a solenoid is a coiled wire that produces a magnetic ﬁeld. we found a vector ﬁeld ∇V . Example 3. that this is equivalent to the original deﬁnition. ∇ · (λ F) = λ (∇ · F) The geometrical meaning of the divergence is as follows. ∇ × (F + G) = (∇ × F) + (∇ × G) 32 .7) The curl of a vector ﬁeld F is deﬁned to be ∇×F = ∂ F3 ∂ F2 ∂ F1 ∂ F3 ∂ F2 ∂ F1 i+ j+ k. Here. These notations are completely interchangeable.3.) This idea will be made more precise when we come to the Divergence Theorem in the next Chapter. If F = 3xy2 i + ez j + xy sin zk. given a vector ﬁeld F. It is also easy to show. A vector ﬁeld F for which ∇ · F = 0 everywhere is called divergence-free or solenoidal. It is easy to show. given a scalar ﬁeld V . and arises because magnetic monopoles have never been found in many searches).j = 0.
see the next section for multiplying a scalar and vector ﬁeld. (See Fig. Loosely speaking. ﬁnd ∇ × F.2 1 y 0 -1 -2 -2 -1 0 x 1 2 Figure 3.3: Example of a vector ﬁeld with positive divergence (everywhere): F = xi + yj. Example 3. it would mean that the vector ﬁeld tends to go round in a clockwise direction. If the component of the curl were negative. and if λ is any constant then ∇ × (λ F) = λ (∇ × F) Note in the above that λ must be independent of position. it means that in the vicinity of the point and in a plane normal to n. Exercise 3.) This idea will be made more precise when we come to Stokes’s Theorem. i ∇ × v = ∂ /∂ x y j k ∂ /∂ y ∂ /∂ z = i(0 − 0) + j(0 − 0) + k(−1 − 1) = −2k. Find the divergence (∇ · F) and curl (∇ × F) of the following vector ﬁelds: F = x2 i + xzj − 3zk F = x2 i − 2xyj + 3xzk F = ∇(1/r) where r = (x2 + y2 + z2 )1/2 = 0. The velocity in a ﬂuid is v = yi − xj + 0k.4. If F = (x2 + y2 + z2 )i + (x4 − y2 z2 )j + xyzk. Find ∇ × v. 3.4. −x 0 Exercise 3.4. the vector ﬁeld tends to go round in an anticlockwise direction if one looks along vector n.3. The geometrical meaning of the curl is as follows. if at some point in space the component of the curl in the n direction is positive. A vector ﬁeld F for which ∇ × F = 0 everywhere is called curl-free or irrotational. 33 2 .
However. if we have two scalar ﬁelds U (r).G(r) and cross product F × G(r) in the obvious way.2 1 y 0 -1 -2 -2 -1 0 x 1 2 Figure 3. (3. 2 3. where U is a scalar ﬁeld. we can also multiply scalar and vector ﬁelds together: e. so the possible products we could have are U F. by taking the dot or cross products of each ﬁeld at the same point r. likewise for a scalar ﬁeld U and a vector ﬁeld F(r) multiplying them gives U F(r).G.1) dx dx dx Some of the vector cases are just like that. but some are more complicated: we next give the results. if we have two vector ﬁelds F. We saw above that grad. the derivative would just give the well-known product rule for derivatives. Also. Div and curl can only be applied to vector ﬁelds. F.7 and the exercises to 16. or a scalar product (dot product) of two vector ﬁelds. or the cross product F × G of two vector ﬁelds. and discuss the details afterwards.4: Example of a vector ﬁeld with positive curl (in the z direction): F = xj − yi. say UV . div and curl to these products. we have to have a product which is itself a scalar ﬁeld: that can be an ordinary product of two scalar ﬁelds. Div and Curl of products (See Thomas 16. 34 . dg d f d( f g) =f + g. G we can deﬁne their dot product F. V (r) we can multiply them (at each point r) to get a new scalar ﬁeld UV (r). div and curl behave in the “obvious” way for addition and multiplication by a constant.5 Grad. div and curl to products. but only for the following allowed combinations: to apply grad.g.8) We can now consider the application of grad. If we were dealing with functions of a single variable. We can now apply grad.
G1 + G2 + G3 .∇): Here the notation (G. (Warning: the form of this deﬁnition will not persist in curvilinear coordinates.∇)G (3. it is |G| times the derivative d F/ds along the direction of the unit vector parallel to G.4. and involve the new operator (G.3 and 3. (G.3. we have (G. G.7 are antisymmetric.6 look quite similar to 3.e.7.6) (3.7) (3. Note also that Eqs. and r = |r| as usual.∇F2 . except for the minus sign in 3. taking F = (F1 . The other two 3. but the directional derivative will remain the same). Note: you are not expected to memorise Eqs.5) (3.∇)G + (G.∇)F above.F ∇ · (F × G) = G.∇)F = G1 ∂ F1 ∂ F1 ∂ F1 ∂ F2 ∂ F2 ∂ F2 ∂ F3 ∂ F3 ∂ F3 + G2 + G3 . G2 .3 are symmetrical in the two variables. using Eq 3.∇)V = G1 ∂ ∂ ∂ ∂V ∂V ∂V V = G1 + G2 + G3 + G2 + G3 .∇)F − G(∇ · F) − (F.5 and 3.5. ∂x ∂y ∂z ∂x ∂y ∂z Thus writing out the whole thing. where for a scalar ﬁeld V . 3. if G = (G1 .2 and 3. G are swapped due to the antisymmetry of the cross product.6.∇)F is to be interpreted as (G. 3. G. Let a be a constant vector. F3 ).For grad of products we have: ∇(UV ) = U (∇V ) + V (∇U ) grad (UV ) = U grad V + V grad U ∇(F.1. G1 + G2 + G3 ∂x ∂y ∂z ∂x ∂y ∂z ∂x ∂y ∂z This is essentially the directional derivative of vector F in the direction of G. F2 .2. You should know the deﬁnition of (G.∇)F For div of products we have: ∇ · (U F) = U (∇ · F) + (∇U ). they must change sign if F.∇F3 ).∇F1 .2) (3. we have: ∇ × (U F) = U (∇ × F) + (∇U ) × F = U (∇ × F) − F × (∇U ) ∇ × (F × G) = F(∇ · G) + (G. but you may be given those formulae in an exam question. Then. i.5 and 3. 3.e. 3. G3 ).(∇ × G) and for curl of products. while 3.(∇ × F) − F.6. 3. Example 3.7 are more complicated. i.G) = F × (∇ × G) + G × (∇ × F) + (F.3) We see above that equations 3.4) (3. ∇ × (ra) = r(∇ × a) − a × ∇r a×r = 0− r 35 .5 and the possible minus sign in 3.
the two middle terms differentiate the constant a so are both zero. and then we can apply grad to that. ∇ × (a × r) = a(∇ · r) + (r. For a couple of examples: ﬁrstly for Eq. this can be fairly long.2 it is simple.6 Vector second derivatives: applying ∇ twice We also have a second set of identities arising from applying two of grad. Much shorter proofs can be given using index notation. Then. using Equation 3. div or curl in succession. Let a be a constant vector. This gives a total of ﬁve allowed cases.3. expanding out using the ordinary product rule and doing some rearrangement. ∇ × (U F) = i ∂ ∂ ∂ ∂ ∂ ∂ (UF3 ) − (UF2 ) + j (UF1 ) − (UF3 ) + k (UF2 ) − (UF1 ) ∂y ∂z ∂z ∂x ∂x ∂y ∂ F3 ∂U ∂ F2 ∂U ∂ F1 ∂U ∂ F3 ∂U = i U +j U + F3 −U − F2 + F1 −U − F3 ∂y ∂y ∂z ∂z ∂z ∂z ∂x ∂x ∂ F2 ∂U ∂ F1 ∂U + F2 −U − F1 +k U ∂x ∂x ∂y ∂y Now we just re-order the 12 terms so that the six with a U ∂ Fi come ﬁrst. Here grad U and curl F produce vector ﬁelds. but this is no longer on the syllabus. and it is simple to check from the deﬁnitions that ∇ · r = 3 and (a.2 to 3. but is not difﬁcult. 3.∇)a − r(∇ · a) − (a. to which either div or curl can be applied. though it gets quite long for Eqs. The others can be proved in a similar way.7.∇)r = a.6: from the deﬁnition of curl.3 and 3.7 can be proved directly from the deﬁnitions by inserting components.7. then the six with an Fi ∂ U come next. 3. and from the deﬁnitions. while div F produces a scalar ﬁeld.∇)r = 3a + 0 − 0 − a = 2a (On the top line. it becomes clear that the result is ∇ × (U F) = U (∇ × F) + (∇U ) × F QED.) Proofs: All of the equations 3.Example 3. we have ∇(UV ) = i ∂ ∂ ∂ (UV ) + j (UV ) + k (UV ) ∂x ∂y ∂z ∂V ∂U ∂V ∂U ∂V ∂U = i(U +V ) + j(U +V ) + k(U +V ) ∂x ∂x ∂y ∂y ∂z ∂z ∂V ∂V ∂V ∂U ∂U ∂U +j +k +j +k +V i = U i ∂x ∂y ∂z ∂x ∂y ∂z = U (∇V ) + V (∇U ) QED. Next we’ll prove Eq. which are as follows: 36 . 3.6.
grad div F and ∇2 F.] The relation curl grad U = 0 is particularly useful. and we’ll also show how to construct the desired U with a suitable integral. we assume that the function U is sufﬁciently well-behaved for its partial second derivatives to commute. 37 . − ∂ y∂ z ∂ z∂ y ∂ z∂ x ∂ x∂ z ∂ x∂ y ∂ y∂ x = 0. giving a relationship between curl curl F. − . the vector ﬁeld deﬁned by H ≡ ∇U will have curl H = 0.div(grad U) = ∇ · (∇U ) ≡ ∇2U curl(grad U ) = ∇×(∇U ) = 0 div(curl F) = ∇·(∇× F) = 0 curl(curl F) = ∇ × (∇ × F) = ∇(∇ · F) − ∇ F grad(div F) = ∇(∇ · F) = ∇ × (∇ × F) + ∇2F 2 (3.13) This operator can be applied to either a scalar ﬁeld or a vector ﬁeld. this is true for any ﬁelds.11 and 3.9: curl(∇U ) = = i j k ∂ /∂ x ∂ /∂ y ∂ /∂ z ∂ U /∂ x ∂ U /∂ y ∂ U /∂ z ∂ 2U ∂ 2U ∂ 2U ∂ 2U ∂ 2U ∂ 2U − .10) (3. [Note. These two zero cases can be helpfully memorised by the fact that they would also give zero if ∇ was replaced by an ordinary vector a . Now we can see immediately that if curl F = 0.: Proof of 3.g. see below. producing a ﬁeld of the same type. it is impossible to ﬁnd such a scalar ﬁeld U : because for any U . this is very important in a wide range of physical problems. it is simply ∇ 2U ≡ ∂2 ∂2 ∂2 U + 2U + 2U 2 ∂x ∂y ∂z (3.9) (3. since it is often interesting to ask. as expected from the deﬁnition of ∇. given some vector ﬁeld F. In components. All of the relations above can be proved by direct substitution. but beware. therefore H = F. However. Note that the last two of the above equations 3. this simpliﬁes things from 3 components to 1. and div curl F) are identically zero .8) (3.12 are just a rearrangement of each other. as long as they are sufﬁciently well behaved that the partial derivatives commute.11) (3. e. so ∇2 F means apply ∇2 to each component of F separately.8 introduces a new operator ∇ 2 called the Laplacian. we will show in the next chapter that if curl F = 0 everywhere in a given domain then we can ﬁnd such a scalar ﬁeld U . and we will meet it extensively in Chapter 8. giving ∇2 F = i∇2 F1 + j∇2 F2 + k∇2 F3 so ∇2 F is another vector ﬁeld. The ﬁrst equation above Eq. can we ﬁnd a scalar ﬁeld U such that ∇U = F ? If we can. this sort of rule is not applicable to every equation containing ∇.12) We see here that two of these cases (curl grad U. 3.
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue reading from where you left off, or restart the preview.