Vector fields

How do we describe a quantity that has a size and a direction at every point, such as the velocity of moving water or the pull of gravity, and how do we find the paths along which it carries things?

Where you are 8 in Vector calculus

Why this exists

A function of several variables (Functions of several variables and level sets) gives one number at each point. Many quantities need a vector at each point instead: the velocity of water in a tank, the pull of the Sun on a planet, the gradient of a function (Gradient and directional derivatives). A vector field holds such a quantity, and the integrals of the rest of the course, work along a curve and flux through a surface, integrate one.

A field also moves things. A leaf on moving water follows a curve whose velocity is, at every moment, the water’s velocity where the leaf is: a flow line. Example 3 follows a float across a draining tank.

Definition

Let D \subseteq \mathbb{R}^n. A vector field on D is a function \mathbf{F} : D \to \mathbb{R}^n, assigning to each point \mathbf{x} of D a vector \mathbf{F}(\mathbf{x}) (Vectors in n dimensions), pictured as an arrow with its tail at \mathbf{x}. In the plane and in space the component functions are written

\mathbf{F} = \langle P, Q \rangle, \qquad \mathbf{F} = \langle P, Q, R \rangle,

so that \mathbf{F} = P\,\mathbf{i} + Q\,\mathbf{j} + R\,\mathbf{k}. A function from D to \mathbb{R} is, by contrast, a scalar function.

\mathbf{F} is continuous at \mathbf{a} \in D if for every \varepsilon > 0 there is \delta > 0 with \lVert \mathbf{F}(\mathbf{x}) - \mathbf{F}(\mathbf{a}) \rVert < \varepsilon whenever \mathbf{x} \in D and \lVert \mathbf{x} - \mathbf{a} \rVert < \delta, as in Limits and continuity in several variables. When D is open, \mathbf{F} is of class C^1 on D if every first partial derivative of every component exists and is continuous on D.

Write \mathbf{r} = \langle x, y, z \rangle for the position vector and \rho = \lVert \mathbf{r} \rVert (in the plane, \langle x, y \rangle). A gradient field is \mathbf{F} = \nabla f for a differentiable f : D \to \mathbb{R}, called a potential function for \mathbf{F}; when such an f exists is the subject of Conservative fields and potential functions. A unit vector field has \lVert \mathbf{F} \rVert = 1 at every point; if \mathbf{F} is never \mathbf{0}, then \mathbf{F}/\lVert \mathbf{F} \rVert is one.

Let D be open and \mathbf{F} continuous on D. A flow line (or integral curve, or streamline) of \mathbf{F} is a differentiable vector function \mathbf{r} : I \to D on an open interval I (Vector-valued functions) with

\mathbf{r}'(t) = \mathbf{F}(\mathbf{r}(t)) \quad \text{for every } t \in I.

In components this is a system of differential equations, x' = P(x, y) and y' = Q(x, y) in the plane, whose right sides do not involve t.

Intuition

A field cannot be drawn completely, since there is an arrow at every point. A picture shows arrows at a grid of points, all shrunk by one factor so that they do not collide: directions are exact and lengths are in proportion (Figure 1).

If \mathbf{F} is the velocity of a fluid, a speck carried along has velocity \mathbf{F}(\mathbf{r}(t)) at its position \mathbf{r}(t): flow lines are the paths of the specks, tangent to the arrows. A function whose gradient is perpendicular to the field cannot change along them, and for a field with continuous partial derivatives exactly one passes through each point at a given time.

x y
Figure 1The field \mathbf{F}(x, y) = \langle x - y, x + y \rangle of Example 1, every arrow drawn at one fifth of its length. The dashed curve is the flow line e^t\langle \cos t, \sin t \rangle, through (1, 0) at t = 0 (Exercise 5).

Results

Throughout, D \subseteq \mathbb{R}^n is open.

Result 1 (continuity by components). A vector field \mathbf{F} = \langle F_1, \dots, F_n \rangle on D is continuous at \mathbf{a} if and only if every component F_i is continuous at \mathbf{a}.

Result 2 (quantities along flow lines). Let \mathbf{F} be continuous on D and \mathbf{r} : I \to D a flow line of \mathbf{F}.

(a) If g : D \to \mathbb{R} is differentiable and \nabla g \cdot \mathbf{F} = 0 at every point of D, then g(\mathbf{r}(t)) is constant on I: the flow line lies in one level set of g.

(b) If \mathbf{F} = \nabla f with f differentiable on D, then \displaystyle \frac{d}{dt} f(\mathbf{r}(t)) = \lVert \nabla f(\mathbf{r}(t)) \rVert^2, so f(\mathbf{r}(t)) is nondecreasing on I, and \mathbf{r}'(t) is normal to the level set of f through \mathbf{r}(t).

Result 3 (rescaling by a positive function). Let \mathbf{F} be continuous on D, let g : D \to \mathbb{R} be continuous with g > 0, let \mathbf{r} : I \to D be a flow line of \mathbf{F}, and let s_0 \in I. There are an open interval J containing 0 and a strictly increasing differentiable \sigma from J onto I, with \sigma(0) = s_0, such that \mathbf{r} \circ \sigma is a flow line of g\mathbf{F}. So g\mathbf{F} traces the curves of \mathbf{F}, in the same direction, at other speeds.

Result 4 (existence and uniqueness). Let \mathbf{F} be of class C^1 on D, \mathbf{a} \in D and t_0 \in \mathbb{R}. There is a flow line \mathbf{r} : I \to D with t_0 \in I and \mathbf{r}(t_0) = \mathbf{a}. If flow lines \mathbf{r}_1 : I_1 \to D and \mathbf{r}_2 : I_2 \to D have \mathbf{r}_1(t_0) = \mathbf{r}_2(t_0) for some t_0 \in I_1 \cap I_2, then \mathbf{r}_1 = \mathbf{r}_2 on I_1 \cap I_2.

Result 5 (consequences). Let \mathbf{F} be of class C^1 on D.

(a) If \mathbf{r} : I \to D is a flow line and c is a number, then t \mapsto \mathbf{r}(t + c) is a flow line on the interval of t with t + c \in I.

(b) If \mathbf{F}(\mathbf{a}) = \mathbf{0}, every flow line that passes through \mathbf{a} is constant, equal to \mathbf{a}.

(c) If flow lines \mathbf{r}_1 : I_1 \to D and \mathbf{r}_2 : I_2 \to D have \mathbf{r}_1(t_1) = \mathbf{r}_2(t_2), then \mathbf{r}_2(t) = \mathbf{r}_1(t + t_1 - t_2) for every t \in I_2 with t + t_1 - t_2 \in I_1. Flow lines that meet trace the same curve near the meeting point; they never cross.

Why it is true

Result 1. For each i and every \mathbf{x} \in D,

\begin{aligned} \lvert F_i(\mathbf{x}) - F_i(\mathbf{a}) \rvert &\le \lVert \mathbf{F}(\mathbf{x}) - \mathbf{F}(\mathbf{a}) \rVert \\ &\le \sum_{j=1}^n \lvert F_j(\mathbf{x}) - F_j(\mathbf{a}) \rvert. \end{aligned}

The first inequality holds because the squared length is a sum of squares including (F_i(\mathbf{x}) - F_i(\mathbf{a}))^2; the second because \displaystyle \sqrt{\sum c_j^2} \le \sum \lvert c_j \rvert, as squaring both sides shows. If \mathbf{F} is continuous at \mathbf{a}, a \delta making the middle term less than \varepsilon does the same for the left one. Conversely, if each F_j is continuous at \mathbf{a}, choose \delta_j making \displaystyle \lvert F_j(\mathbf{x}) - F_j(\mathbf{a}) \rvert < \frac{\varepsilon}{n} and let \delta be the least \delta_j; the right side is then less than \varepsilon.

Result 2. (a) By the chain rule in gradient form (Result 1 of Gradient and directional derivatives), g(\mathbf{r}(t)) has derivative \nabla g(\mathbf{r}(t)) \cdot \mathbf{F}(\mathbf{r}(t)) = 0 on I, and a function with zero derivative on an interval is constant, by the mean value theorem (The mean value theorem). (b) The same chain rule gives the derivative \nabla f(\mathbf{r}(t)) \cdot \nabla f(\mathbf{r}(t)) \ge 0, and a function with nonnegative derivative on an interval is nondecreasing, again by the mean value theorem. The velocity is \nabla f(\mathbf{r}(t)), normal to the level set by Result 4 of Gradient and directional derivatives.

Result 3. Since \mathbf{r} is continuous, u \mapsto 1/g(\mathbf{r}(u)) is continuous and positive on I. Let

\psi(s) = \int_{s_0}^{s} \frac{du}{g(\mathbf{r}(u))}, \qquad s \in I.

By the fundamental theorem of calculus (The second fundamental theorem of calculus), \psi'(s) = 1/g(\mathbf{r}(s)) > 0, so \psi is continuous and strictly increasing. It maps I onto an interval J containing \psi(s_0) = 0, open because \psi takes no largest or smallest value on the open interval I. By the rule for the derivative of an inverse function, its inverse \sigma : J \to I is differentiable, with

\sigma'(t) = \frac{1}{\psi'(\sigma(t))} = g(\mathbf{r}(\sigma(t))).

By the chain rule in each component, \mathbf{q} = \mathbf{r} \circ \sigma has \mathbf{q}'(t) = \sigma'(t)\,\mathbf{r}'(\sigma(t)) = g(\mathbf{q}(t))\,\mathbf{F}(\mathbf{q}(t)).

Result 4. This is the existence and uniqueness theorem of Picard and Lindelöf for systems of ordinary differential equations, applied to \mathbf{x}' = \mathbf{F}(\mathbf{x}); a C^1 field satisfies its Lipschitz hypothesis near each point. It is stated here without proof and proved in a course on differential equations or in Real Analysis.

Result 5. (a) By the chain rule, \displaystyle \frac{d}{dt}\mathbf{r}(t + c) = \mathbf{r}'(t + c) = \mathbf{F}(\mathbf{r}(t + c)). (b) The constant \mathbf{c}(t) = \mathbf{a} on \mathbb{R} is a flow line, since \mathbf{c}'(t) = \mathbf{0} = \mathbf{F}(\mathbf{a}). A flow line \mathbf{r} with \mathbf{r}(t_1) = \mathbf{a} agrees with \mathbf{c} at t_1, so by Result 4 it equals \mathbf{c} on its whole interval. (c) By (a), \mathbf{s}(t) = \mathbf{r}_1(t + t_1 - t_2) is a flow line, and \mathbf{s}(t_2) = \mathbf{r}_1(t_1) = \mathbf{r}_2(t_2); by Result 4, \mathbf{s} = \mathbf{r}_2 wherever both are defined.

Worked examples

Example 1

A spiral field

Let \mathbf{F}(x, y) = \langle x - y, x + y \rangle on \mathbb{R}^2. (a) Find \mathbf{F} at (1, 0), (0, 1), (-1, 0), (0, -1) and (2, 1). (b) Show that away from the origin \displaystyle \lVert \mathbf{F} \rVert = \sqrt2\,\rho and \mathbf{F} makes the angle \displaystyle \frac{\pi}{4} with the position vector. (c) Describe the picture.

Solution

(a) \mathbf{F}(1, 0) = \langle 1, 1 \rangle, \mathbf{F}(0, 1) = \langle -1, 1 \rangle, \mathbf{F}(-1, 0) = \langle -1, -1 \rangle, \mathbf{F}(0, -1) = \langle 1, -1 \rangle and \mathbf{F}(2, 1) = \langle 1, 3 \rangle.

(b) \lVert \mathbf{F} \rVert^2 = (x - y)^2 + (x + y)^2 = 2(x^2 + y^2) = 2\rho^2, the cross terms \mp 2xy cancelling. And \mathbf{F} \cdot \mathbf{r} = x^2 - xy + xy + y^2 = \rho^2, so for \rho > 0 the angle \theta between them (The dot product and the angle between vectors) has

\cos\theta = \frac{\rho^2}{\sqrt2\,\rho \cdot \rho} = \frac{1}{\sqrt2}, \qquad \theta = \frac{\pi}{4}.

(c) \mathbf{F} = \mathbf{r} + \langle -y, x \rangle, the position vector plus the same vector turned a quarter turn counterclockwise. Every arrow leans \displaystyle \frac{\pi}{4} counterclockwise from straight out, and the arrows grow in proportion to the distance from the origin, where \mathbf{F} = \mathbf{0} (Figure 1). The components are polynomials, so \mathbf{F} is C^1, and by Result 5(b) the only flow line through the origin stays there.

Example 2

The inverse-square field

Let \mathbf{F} = -\dfrac{4}{\rho^3}\,\mathbf{r} on \mathbb{R}^3 without the origin, a field of the form of Newton’s law of gravity: it points toward the origin with length \displaystyle \frac{4}{\rho^2}. (a) Find \mathbf{F}(1, 2, 2) and its length. (b) Show that \mathbf{F} = \nabla f for \displaystyle f = \frac{4}{\rho}. (c) Find the flow line through (1, 2, 2) at t = 0, and its domain.

Solution

(a) At (1, 2, 2), \rho = 3, so \displaystyle \mathbf{F}(1, 2, 2) = -\frac{4}{27}\langle 1, 2, 2 \rangle, of length \displaystyle \frac{4}{27} \cdot 3 = \frac49.

(b) Exercise 16 of Gradient and directional derivatives with n = -1 gives \nabla(\rho^{-1}) = -\rho^{-3}\langle x, y, z \rangle away from the origin; multiply by 4. The components are quotients of C^1 functions with nonzero denominators, so \mathbf{F} is C^1.

(c) The field points along -\mathbf{r}, so look for a flow line on the ray through (1, 2, 2): \mathbf{r}(t) = \ell(t)\,\mathbf{u} with \displaystyle \mathbf{u} = \frac13\langle 1, 2, 2 \rangle and \ell(t) > 0. Then \rho = \ell, \mathbf{r}' = \ell'\,\mathbf{u} and \displaystyle \mathbf{F}(\mathbf{r}) = -\frac{4}{\ell^2}\,\mathbf{u}, so the condition is \displaystyle \ell' = -\frac{4}{\ell^2} with \ell(0) = 3. Then (\ell^3)' = 3\ell^2\ell' = -12, so \ell^3 = 27 - 12t and

\mathbf{r}(t) = \frac{(27 - 12t)^{1/3}}{3}\,\langle 1, 2, 2 \rangle, \qquad t < \frac94.

By Result 4 it is the only flow line through (1, 2, 2) at t = 0. As \displaystyle t \to \frac94 it reaches the origin, outside the domain, with speed \displaystyle \frac{4}{\ell^2} \to \infty. Along it \displaystyle f = \frac{4}{\ell} increases, as Result 2(b) requires.

Example 3

Following a float in a tank

Priya and Omar model the surface water of a round tank at a water treatment plant, of radius 4 meters, draining through an outlet of radius 0.25 meters at its center. With the origin at the center and time in seconds, the water’s velocity in meters per second is

\mathbf{F}(x, y) = \left\langle -\frac{x}{40} - \frac{y}{8},\ \frac{x}{8} - \frac{y}{40} \right\rangle.

Omar sets a float on the water at the rim point (4, 0) at t = 0. (a) How fast does it start? (b) Find its path. (c) When does it reach the outlet, and after how many turns? (d) How far has it traveled?

Solution

(a) \displaystyle \mathbf{F}(4, 0) = \left\langle -\frac{1}{10}, \frac12 \right\rangle, of length \displaystyle \frac{\sqrt{26}}{10} \approx 0.51 meters per second. In general \displaystyle \mathbf{F} = -\frac{1}{40}\mathbf{r} + \frac18\langle -y, x \rangle, a sum of perpendicular vectors of lengths \displaystyle \frac{\rho}{40} and \displaystyle \frac{\rho}{8}, so \displaystyle \lVert \mathbf{F} \rVert = \frac{\sqrt{26}}{40}\,\rho.

(b) That sum is a pull toward the center in proportion to the distance, plus a turning at \displaystyle \frac18 radian per second. So try a distance \ell with \displaystyle \ell' = -\frac{\ell}{40} and an angle growing at rate \displaystyle \frac18:

\mathbf{r}(t) = 4e^{-t/40}\left\langle \cos\tfrac{t}{8}, \sin\tfrac{t}{8} \right\rangle.

By the product rule,

\begin{aligned} \mathbf{r}'(t) &= -\tfrac{1}{40}\mathbf{r}(t) \\ &\quad + \tfrac{1}{8} \cdot 4e^{-t/40}\left\langle -\sin\tfrac{t}{8}, \cos\tfrac{t}{8} \right\rangle, \end{aligned}

and the second vector is \displaystyle \frac18\langle -y, x \rangle at \mathbf{r}(t) = \langle x, y \rangle, so \mathbf{r}' = \mathbf{F}(\mathbf{r}). The components of \mathbf{F} are linear, so \mathbf{F} is C^1, and by Result 4 this is the float’s path (Figure 2).

(c) The distance to the center, 4e^{-t/40}, is 0.25 when e^{t/40} = 16, at t = 40\ln 16 \approx 110.9 seconds. The float has then turned through \displaystyle \frac{t}{8} = 20\ln 2 \approx 13.86 radians, about 2.21 turns.

(d) The distance is the arc length (Arc length and curvature), with speed \displaystyle \lVert \mathbf{F}(\mathbf{r}(t)) \rVert = \frac{\sqrt{26}}{40} \cdot 4e^{-t/40}:

\begin{aligned} \int_0^{40\ln 16} \frac{\sqrt{26}}{10}\,e^{-t/40}\,dt &= 4\sqrt{26}\left(1 - \tfrac{1}{16}\right) \\ &= \tfrac{15}{4}\sqrt{26} \approx 19.12 \end{aligned}

meters.

(4, 0)
Figure 2The tank of Example 3, radius 4 meters, with the outlet of radius 0.25 at its center. The spiral is the float’s path \displaystyle 4e^{-t/40}\langle \cos\frac{t}{8}, \sin\frac{t}{8} \rangle; the arrows show the water’s velocity at eight points, at three times their length.
Example 4

A pendulum and a conserved quantity

A pendulum at angle x from the vertical, with angular velocity y, obeys x' = y and y' = -\sin x in suitable units, so its states move along the flow lines of \mathbf{F}(x, y) = \langle y, -\sin x \rangle. Show that the flow line through (0, 1) stays on the closed curve y^2 = 2\cos x - 1, \displaystyle \lvert x \rvert \le \frac{\pi}{3}, and find which way it moves.

Solution

Let \displaystyle H(x, y) = \frac12 y^2 - \cos x, the pendulum’s energy. Then \nabla H = \langle \sin x, y \rangle and

\nabla H \cdot \mathbf{F} = y\sin x - y\sin x = 0,

so by Result 2(a), H is constant along every flow line. At (0, 1), \displaystyle H = \frac12 - 1 = -\frac12, so the flow line lies in the level set y^2 = 2\cos x - 1. Its points need \displaystyle \cos x \ge \frac12, so x lies in \displaystyle \left[-\frac{\pi}{3}, \frac{\pi}{3}\right] or a translate of it by a multiple of 2\pi, with gaps of x-values between these pieces. The x-coordinate of the flow line is continuous and starts at 0, so by the intermediate value theorem it never reaches another piece: the flow line stays on the oval through (0, 1) (Figure 3). Since \mathbf{F}(0, 1) = \langle 1, 0 \rangle, it moves right along the top of the oval, clockwise, and the pendulum never swings past \displaystyle \frac{\pi}{3}. That it goes all the way round the oval is proved in a course on differential equations.

x y
Figure 3The pendulum field \mathbf{F}(x, y) = \langle y, -\sin x \rangle of Example 4, arrows at about one fifth of their length, with level curves \displaystyle \frac12 y^2 - \cos x = c for \displaystyle c = -\frac12 (heavy, the flow line through (0, 1)), 0 and \displaystyle \frac12, and the dashed curve c = 1 through the zeros (\pm\pi, 0).
Example 5

Same curve, different speeds

Let \mathbf{F}(x, y) = \langle 1, 2x \rangle. (a) Find the flow line of \mathbf{F} through (0, 1) at t = 0, and describe the flow line of \mathbf{G}(x, y) = e^{xy}\langle 1, 2x \rangle through the same point. (b) The unit vector field \mathbf{U} = \mathbf{F}/\lVert \mathbf{F} \rVert has a flow line \mathbf{q} with \mathbf{q}(0) = (0, 1). When does it reach (1, 2)?

Solution

(a) x' = 1 and y' = 2x with x(0) = 0, y(0) = 1 give \mathbf{r}(t) = \langle t, 1 + t^2 \rangle, the parabola y = 1 + x^2 traced left to right. Since e^{xy} is continuous and positive, Result 3 gives a flow line of \mathbf{G} through (0, 1) along the same parabola with x increasing, though the system x' = e^{xy}, y' = 2xe^{xy} is hard to solve directly. \mathbf{G} is C^1, so by Result 4 it is the only one.

(b) Here g = (1 + 4x^2)^{-1/2} > 0 is continuous and \mathbf{U} is C^1, so by Result 4, \mathbf{q} is the \mathbf{r} \circ \sigma of Result 3’s proof, where \sigma inverts

\psi(s) = \int_0^s \sqrt{1 + 4u^2}\,du,

the arc length of the parabola from (0, 1) to \mathbf{r}(s). Since \psi' \ge 1, \psi maps \mathbb{R} onto \mathbb{R}, and \mathbf{q} reaches (1, 2) = \mathbf{r}(1) at t = \psi(1). With v = 2u and the antiderivative

\int \sqrt{1 + v^2}\,dv = \tfrac12 v\sqrt{1 + v^2} + \tfrac12\ln\big(v + \sqrt{1 + v^2}\big),
\psi(1) = \frac{\sqrt5}{2} + \frac{\ln(2 + \sqrt5)}{4} \approx 1.479.

A unit vector field moves at speed 1, so elapsed time is distance traveled.

Example 6

Proof: a gradient field has no closed flow lines

Let f be differentiable on an open set D with \nabla f continuous, and let \mathbf{r} : I \to D be a flow line of \nabla f with \mathbf{r}(t_1) = \mathbf{r}(t_2) for some t_1 < t_2 in I. Prove that \mathbf{r} is constant on [t_1, t_2]. Deduce that \mathbf{F}(x, y) = (1 + x^2 + y^2)\langle -y, x \rangle is not the gradient of a differentiable function on any open set containing the unit circle.

Solution

Let h(t) = f(\mathbf{r}(t)). By Result 2(b), h is nondecreasing on I with h'(t) = \lVert \nabla f(\mathbf{r}(t)) \rVert^2. For t_1 \le t \le t_2,

h(t_1) \le h(t) \le h(t_2) = h(t_1),

so h is constant on [t_1, t_2] and h' = 0 on (t_1, t_2). Hence \mathbf{r}'(t) = \nabla f(\mathbf{r}(t)) = \mathbf{0} on (t_1, t_2). Each component of \mathbf{r} is continuous on [t_1, t_2] with zero derivative inside, so it is constant by the mean value theorem.

For the field, let \mathbf{r}(t) = \langle \cos 2t, \sin 2t \rangle. On the unit circle 1 + x^2 + y^2 = 2, so \mathbf{F}(\mathbf{r}(t)) = 2\langle -\sin 2t, \cos 2t \rangle = \mathbf{r}'(t): a flow line with \mathbf{r}(0) = \mathbf{r}(\pi), not constant on [0, \pi]. Since \mathbf{F} is continuous, the first part rules out \mathbf{F} = \nabla f on any open set containing the circle.

Boundaries

Without continuous partial derivatives, flow lines can split. The field \mathbf{F}(x, y) = \langle 3x^{2/3}, 1 \rangle, with \displaystyle x^{2/3} = (\sqrt[3]{x})^2, is continuous on \mathbb{R}^2, and both \langle 0, t \rangle and \langle t^3, t \rangle are flow lines through the origin at t = 0, since 3(t^3)^{2/3} = 3t^2. Result 4 does not apply: \displaystyle \frac{\partial}{\partial x}\left(3x^{2/3}\right) = 2x^{-1/3} does not exist at x = 0.

Flow lines can end in finite time. Result 4 gives a flow line on some open interval, not on all of \mathbb{R}. The field \langle 1 + x^2, y \rangle is C^1 on all of \mathbb{R}^2, yet its flow line through (0, 1) at t = 0 is \langle \tan t, e^t \rangle, which runs off to infinity as \displaystyle t \to \frac{\pi}{2}. In Example 2 the flow line reaches the edge of the domain at \displaystyle t = \frac94.

Where this goes next

Integrating \mathbf{F}(\mathbf{r}(t)) \cdot \mathbf{r}'(t) along a curve gives the work a force field does, the subject of Line integrals (extending Work as an integral). Gradient fields are exactly the fields whose line integrals depend only on the endpoints; Conservative fields and potential functions tests for them, and Example 6 is a first obstruction. How much a field spreads out or turns is measured in Divergence and curl and related to boundary flow by Green's theorem. Linear fields \mathbf{x}' = A\mathbf{x} are solved with eigenvalues in Linear systems of differential equations.

Exercises

Evaluating fields

  1. Let \mathbf{F}(x, y) = \langle y^2, x - y \rangle. Find \mathbf{F}(2, -1) and its length, and every point where \mathbf{F} = \mathbf{0}.
  2. Let \mathbf{F}(x, y, z) = \langle yz, xz, xy \rangle. Find \mathbf{F}(1, 2, 3) and its length, and show that \mathbf{F} = \nabla(xyz).
  3. Find every point where \mathbf{F}(x, y) = \langle x^2 - y, x + y - 2 \rangle is \mathbf{0}.
  4. At which points other than the origin is \mathbf{F}(x, y) = \langle x, 4y \rangle parallel to the position vector \langle x, y \rangle?

Flow lines

  1. Show that e^t\langle \cos t, \sin t \rangle is a flow line of the field of Example 1, and find the flow line of that field through (0, 2) at t = 0.
  2. Find the flow line of \mathbf{F}(x, y) = \langle 1, y \rangle through (0, 3) at t = 0, and the curve it traces.
  3. Find the flow line of \mathbf{F}(x, y) = \langle 3x, y \rangle through (2, 1) at t = 0, and an equation in x and y for the curve it traces.
  4. For the field of Example 2, find the flow line through (0, 6, 0) at t = 0, and the time at which it reaches the origin.
  5. In Example 3, Priya sets a second float at (0, 3) at t = 0. When is it 1 meter from the center, and through what angle has it turned by then?

Conserved quantities and rescaling

  1. Let \mathbf{F}(x, y) = \langle x + 2y, -2x - y \rangle and g(x, y) = x^2 + xy + y^2. Show that g is constant along every flow line of \mathbf{F}, and name the curve containing the flow line through (1, 0).
  2. Let \mathbf{F}(x, y) = \langle y, -x - x^3 \rangle. Find a function constant along its flow lines, and the largest x-coordinate of any point on its level curve through (0, 2).
  3. Explain why the flow line of \mathbf{G}(x, y) = \langle xy^2, y^3 \rangle through (3, 4) at t = 0 runs along the ray from the origin through (3, 4), away from the origin.

Proofs

  1. Let \mathbf{F} be continuous on \mathbb{R}^2 with \mathbf{F}(\mathbf{x}) \cdot \mathbf{x} < 0 for every \mathbf{x} \ne \mathbf{0}. Prove that along any flow line, \lVert \mathbf{r}(t) \rVert is strictly decreasing on every interval on which \mathbf{r}(t) \ne \mathbf{0}.
  2. Prove that if \mathbf{r} : I \to D is a flow line of \mathbf{F}, then \mathbf{q}(t) = \mathbf{r}(-t) is a flow line of -\mathbf{F}.
  3. Let \mathbf{c} be a constant vector of \mathbb{R}^n. Prove, without Result 4, that every flow line of the constant field \mathbf{F} = \mathbf{c} has the form \mathbf{r}(t) = \mathbf{a} + t\mathbf{c}.
  4. Let H have continuous first partials on an open set D \subseteq \mathbb{R}^2, and let \mathbf{F} = \langle H_y, -H_x \rangle. Prove that \mathbf{F} is perpendicular to \nabla H at every point and that H is constant along every flow line of \mathbf{F}.
  5. Prove that \mathbf{F}(x, y) = \langle 8y, -2x \rangle is not the gradient of a differentiable function on any open set containing the ellipse x^2 + 4y^2 = 4.
Answers
  1. \langle 1, 3 \rangle, of length \displaystyle \sqrt{10}. Only the origin: y^2 = 0 forces y = 0, and then x - y = 0 forces x = 0.
  2. \langle 6, 3, 2 \rangle, of length \displaystyle \sqrt{49} = 7. The partials of xyz in x, y and z are yz, xz and xy.
  3. (1, 1) and (-2, 4). The first component gives y = x^2, and then x^2 + x - 2 = (x + 2)(x - 1) = 0.
  4. Exactly the points of the two axes. Nonzero vectors \langle a, b \rangle and \langle c, d \rangle are parallel when ad - bc = 0 (Computing a determinant), here xy - 4yx = -3xy = 0; and \mathbf{F} \ne \mathbf{0} away from the origin.
  5. With \langle x, y \rangle = e^t\langle \cos t, \sin t \rangle, the derivative is e^t\langle \cos t - \sin t, \sin t + \cos t \rangle, which is \langle x - y, x + y \rangle. Through (0, 2): 2e^t\langle -\sin t, \cos t \rangle, checked the same way, and by Result 4 the only one.
  6. \langle t, 3e^t \rangle, tracing y = 3e^x: x' = 1 gives x = t, and y' = y gives y = 3e^t.
  7. \langle 2e^{3t}, e^t \rangle, on the curve x = 2y^3 with y > 0, since 2e^{3t} = 2(e^t)^3.
  8. \langle 0, (216 - 12t)^{1/3}, 0 \rangle for t < 18, reaching the origin at t = 18. As in Example 2, \ell^3 = \ell(0)^3 - 12t with \ell(0) = 6.
  9. At t = 40\ln 3 \approx 43.9 seconds, after turning 5\ln 3 \approx 5.49 radians. Its path is \displaystyle 3e^{-t/40}\left\langle -\sin\frac{t}{8}, \cos\frac{t}{8} \right\rangle, checked as in Example 3; 3e^{-t/40} = 1 gives t = 40\ln 3, and the angle is \displaystyle \frac{t}{8}.
  10. \nabla g = \langle 2x + y, x + 2y \rangle, so \nabla g \cdot \mathbf{F} is (2x + y)(x + 2y) plus (x + 2y)(-2x - y), which is 0, and Result 2(a) applies. Since g(1, 0) = 1, the flow line lies on the ellipse x^2 + xy + y^2 = 1.
  11. \displaystyle H(x, y) = \frac12 y^2 + \frac12 x^2 + \frac14 x^4, with \nabla H \cdot \mathbf{F} = (x + x^3)y - y(x + x^3) = 0. H(0, 2) = 2, and on H = 2, \displaystyle \frac12 x^2 + \frac14 x^4 \le 2, that is (x^2 + 4)(x^2 - 2) \le 0, so \displaystyle x \le \sqrt2, with equality at \displaystyle (\sqrt2, 0). The largest is \displaystyle \sqrt2.
  12. \mathbf{G} = y^2\langle x, y \rangle, and y^2 is continuous and positive on the open half plane y > 0. \langle 3e^t, 4e^t \rangle is a flow line of \langle x, y \rangle through (3, 4), running out along the ray, so by Result 3, \mathbf{G} has a flow line through (3, 4) at t = 0 along the same ray in the same direction; \mathbf{G} is C^1, so by Result 4 it is the only one.
  13. Let J be an interval on which \mathbf{r}(t) \ne \mathbf{0}, and k(t) = \mathbf{r}(t) \cdot \mathbf{r}(t). By the dot product rule (Vector-valued functions), k'(t) = 2\,\mathbf{r}(t) \cdot \mathbf{F}(\mathbf{r}(t)) < 0 on J. For s < t in J the mean value theorem gives k(t) - k(s) = k'(c)(t - s) < 0, so k is strictly decreasing, and so is \displaystyle \lVert \mathbf{r} \rVert = \sqrt{k}.
  14. \mathbf{q} is defined on the open interval of t with -t \in I, and by the chain rule \mathbf{q}'(t) = -\mathbf{r}'(-t) = -\mathbf{F}(\mathbf{r}(-t)) = (-\mathbf{F})(\mathbf{q}(t)).
  15. Let \mathbf{r} : I \to \mathbb{R}^n be a flow line and \mathbf{w}(t) = \mathbf{r}(t) - t\mathbf{c}. Then \mathbf{w}' = \mathbf{c} - \mathbf{c} = \mathbf{0} on I, so each component of \mathbf{w} is constant by the mean value theorem: \mathbf{w}(t) = \mathbf{a}, and \mathbf{r}(t) = \mathbf{a} + t\mathbf{c}.
  16. \nabla H \cdot \mathbf{F} = H_x H_y + H_y(-H_x) = 0 at every point. H has continuous partials, so it is differentiable (Result 2 of Tangent planes and linear approximation), and \mathbf{F} is continuous; Result 2(a) then makes H constant along every flow line.
  17. \mathbf{r}(t) = \langle 2\cos 4t, -\sin 4t \rangle lies on the ellipse, since 4\cos^2 4t + 4\sin^2 4t = 4, and \mathbf{r}'(t) = \langle -8\sin 4t, -4\cos 4t \rangle is \langle 8y, -2x \rangle at \mathbf{r}(t): a flow line. It has \displaystyle \mathbf{r}(0) = \mathbf{r}\left(\frac{\pi}{2}\right) = (2, 0) but \displaystyle \mathbf{r}\left(\frac{\pi}{8}\right) = (0, -1). \mathbf{F} is continuous, so Example 6 rules out \mathbf{F} = \nabla f on any open set containing the ellipse.
math.wiki · CC-BY-SA-4.0