In this section, we extend differentiation and integration to vector-valued functions. This is straightforward because the techniques of single-variable calculus carry over with little change. What is new and important, however, is the geometric interpretation of the derivative as a tangent vector. We describe this later in the section.
The first step is to define limits of vector-valued functions.
A vector-valued function \({\bf r}(t)\) approaches the limit \({\bf u}\) (a vector) as \(t\) approaches \(t_0\) if \(\displaystyle{\lim_{t\to t_0}\|{{\bf r}(t)-{\bf u}}\|=0}\). In this case, we write \[ \lim_{t\to t_0}{\bf r}(t) = {\bf u} \]
We can visualize the limit of a vector-valued function as a vector \({\bf r}(t)\) “moving” toward the limit vector \({\bf u}\) (Figure 13.8). According to the next theorem, vector limits may be computed componentwise.
\({\bf r}(t)=\langle x(t), y(t), z(t)\rangle\) approaches a limit as \(t\to t_0\) if and only if each component approaches a limit, and in this case, \[ \lim_{t\to t_0}{\bf r}(t) = \langle \lim_{t\to t_0}x(t), \lim_{t\to t_0}y(t), \lim_{t\to t_0}z(t)\rangle\tag{1} \]
Let \({\bf u} = \langle a,b,c\rangle\) and consider the square of the length \[ \|{{\bf r}(t)-{\bf u}}\|^2 = (x(t)-a)^2+(y(t)-b)^2+(z(t)-c)^2\tag{2} \]
The term on the left approaches zero if and only if each term on the right approaches zero (because these terms are nonnegative). It follows that \(\|{{\bf r}(t)-{\bf u}}\|\) approaches zero if and only if \(|x(t)-a|\), \(|y(t)-b|\), and \(|z(t)-c|\) tend to zero. Therefore, \({\bf r}(t)\) approaches a limit \({\bf u}\) as \(t\to t_0\) if and only if \(x(t)\), \(y(t)\), and \(z(t)\) converge to the components \(a\), \(b\), and \(c\).
The Limit Laws of scalar functions remain valid in the vector-valued case. They are verified by applying the Limit Laws to the components.
Calculate \(\displaystyle{\lim_{t\to 3}{\bf r}(t)}\), where \(\displaystyle{{\bf r}(t) = \langle t^2, 1-t, t^{-1}\rangle}\).
Solution By Theorem 1, \[ \lim_{t\to 3}{\bf r}(t) = \lim_{t\to 3}\langle t^2, 1-t, t^{-1}\rangle = \langle \lim_{t\to 3}t^2, \lim_{t\to 3}(1-t), \lim_{t\to 3} t^{-1}\rangle = \langle 9, -2, \frac13\rangle \]
732
Continuity of vector-valued functions is defined in the same way as in the scalar case. A vector-valued function \({\bf r}(t)=\langle x(t), y(t), z(t)\rangle\) is continuous at \(t_0\) if \[ \lim_{t\to t_0}\,{\bf r}(t) = {\bf r}(t_0) \]
By Theorem 1, \({\bf r}(t)\) is continuous at \(t_0\) if and only if the components \(x(t)\), \(y(t)\), \(z(t)\) are continuous at \(t_0\).
We define the derivative of \({\bf r}(t)\) as the limit of the difference quotient: \[ \boxed{\bbox[#FAF8ED,5pt]{{\bf r}'(t) = \dfrac{d}{dt}{\bf r}(t) = \lim_{h\to 0} \frac{{\bf r}(t+h) - {\bf r}(t)}{h}}}\tag{3} \]
In Leibniz notation, the derivative is written \(\displaystyle{d{\bf r}/{dt}\).
We say that \({\bf r}(t)\) is differentiable at \(t\) if the limit in Eq.(3) exists. Notice that the components of the difference quotient are difference quotients: \[ \lim_{h\to 0} \frac{{\bf r}(t+h) - {\bf r}(t)}{h} = \lim_{h\to 0} \langle \frac{x(t+h)-x(t)}h, \frac{y(t+h)-y(t)}h, \frac{z(t+h) -z(t)}h\rangle \] and by Theorem 1, \({\bf r}(t)\) is differentiable if and only if the components are differentiable. In this case, \({\bf r}'(t)\) is equal to the vector of derivatives \(\langle x'(t), y'(t), z'(t)\rangle\).
A vector-valued function \({\bf r}(t)=\langle x(t), y(t), z(t)\rangle\) is differentiable if and only if each component is differentiable. In this case, \[ {\bf r}'(t) = \frac{d}{dt}{\bf r}(t) = \langle x'(t), y'(t), z'(t)\rangle \]
By Theorems 1 and 2, vector-valued limits and derivatives are computed “componentwise,” so they are not more difficult to compute than ordinary limits and derivatives.
Here are some vector-valued derivatives, computed componentwise: \[ \frac{d}{dt} \langle t^2, t^3, \sin t\rangle = \langle 2t, 3t^2, \cos t\rangle,\qquad \frac{d}{dt} \langle \cos t, -1, e^{2t}\rangle = \langle -\sin t, 0, 2e^{2t}\rangle \] Higher-order derivatives are defined by repeated differentiation: \[ {\bf r}”(t)=\frac{d}{dt}\,{\bf r}'(t),\quad {\bf r}”'(t)=\frac{d}{dt}\, {\bf r}”(t),\quad \dots \]
Calculate \({\bf r}''(3)\), where \({\bf r}(t)=\langle \ln t, t, t^2\rangle\).
Solution We perform the differentiation componentwise: \begin{align*} {\bf r}'(t) =\frac{d}{dt}\,\langle \ln t, t, t^2\rangle =\langle t^{-1},1,2t\rangle\\ {\bf r}”(t) =\frac{d}{dt}\,\langle t^{-1},1,2t\rangle =\langle -t^{-2},0,2\rangle \end{align*} Therefore, \({\bf r}''(3) =\langle -\frac19,0,2\rangle\).
The differentiation rules of single-variable calculus carry over to the vector setting.
733
Assume that \({\bf r}(t)\), \({\bf r}_1(t)\), and \({\bf r}_2(t)\) are differentiable. Then
Each rule is proved by applying the differentiation rules to the components. For example, to prove the Product Rule (we consider vector-valued functions in the plane, to keep the notation simple), we write \[ f(t){\bf r}(t) = f(t)\langle x(t), y(t)\rangle = \langle f(t)x(t), f(t)y(t)\rangle \] Now apply the Product Rule to each component: \begin{align*} \frac{d}{dt} f(t){\bf r}(t) &= \langle \frac{d}{dt}f(t)x(t), \frac{d}{dt}f(t)y(t)\rangle\\ &= \langle f'(t)x(t)+f(t)x'(t), f'(t)y(t) + f(t)y'(t)\rangle\\ &= \langle f'(t)x(t), f'(t)y(t)\rangle + \langle f(t)x'(t), f(t)y'(t)\rangle \\ &= f'(t)\langle x(t), y(t) \rangle + f(t)\langle x'(t), y'(t)\rangle = f'(t){\bf r}(t)+f(t){\bf r}'(t) \end{align*} The remaining proofs are left as exercises (Exercises 69–70).
Let \({\bf r}(t)=\langle t^2, 5t, 1\rangle\) and \(f(t)= e^{3t}\). Calculate:
Solution We have \({\bf r}'(t)=\langle 2t, 5 , 0\rangle\) and \(f'(t)=3e^{3t}\).
There are three different Product Rules for vector-valued functions. In addition to the rule for the product of a scalar function \(f(t)\) and a vector-valued function \({\bf r}(t)\) stated above, there are Product Rules for the dot and cross products. These rules are very important in applications, as we will see.
734
Assume that \({\bf r}_1(t)\) and \({\bf r}_2(t)\) are differentiable. Then \[ \begin{eqnarray} &&\textrm{Dot Products:}& \frac{d}{dt} \big({\bf r}_1(t)\cdot{\bf r}_2(t)\big) &=& {\bf r}_1(t)\cdot {\bf r}_2'(t)+{\bf r}_1'(t)\cdot {\bf r}_2(t)\tag{4}\\ &&\textrm{Cross Products:}& \frac{d}{dt} \big({\bf r}_1(t)\times {\bf r}_2(t)\big) &=& \bigl[{\bf r}_1(t)\times {\bf r}_2'(t)\bigr] + \bigl[{\bf r}_1'(t)\times {\bf r}_2(t)\bigr]\tag{5} \end{eqnarray} \]
We verify Eq. (4) for vector-valued functions in the plane. If \({\bf r}_1(t)=\langle x_1(t), y_1(t)\rangle \) and \({\bf r}_2(t)=\langle x_2(t), y_2(t)\rangle \), then \begin{align*} \frac{d}{dt} \bigl({\bf r}_1(t)\cdot{\bf r}_2(t)\bigr) &= \frac{d}{dt} \big(x_1(t)x_2(t)+y_1(t)y_2(t)\big)\\ &=x_1(t)x_2'(t)+x_1'(t)x_2(t)+y_1(t)y_2'(t)+y_1'(t)y_2(t)\\ &= \big(x_1(t)x_2'(t)+ y_1(t)y_2'(t) \big) + \big(x_1'(t)x_2(t)+y_1'(t)y_2(t) \big)\\ &= {\bf r}_1(t)\cdot{\bf r}_2'(t) + {\bf r}_1'(t)\cdot{\bf r}_2(t) \end{align*}
The proof of Eq. (5) is left as an exercise (Exercise 71).
Order is important in the Product Rule for cross products. The first term in Eq. (5) must be written as \[{\bf r}_1(t)\times {\bf r}_2'(t) \] not \({\bf r}_2'(t)\times {\bf r}_1(t)\). Similarly, the second term is \({\bf r}_1'(t)\times {\bf r}_2(t)\). Why is order not a concern for dot products?
In the next example and throughout this chapter, all vector-valued functions are assumed differentiable, unless otherwise stated.
Prove the formula \(\displaystyle{\frac{d}{dt} \bigl({\bf r}(t)\times{\bf r}'(t)\bigr) = {\bf r}(t)\times{\bf r}”(t)}\).
Solution By the Product Formula for cross products, \[ \frac{d}{dt} \bigl({\bf r}(t)\times{\bf r}'(t)\bigr) = {\bf r}(t)\times{\bf r}”(t) + \underbrace{{\bf r}'(t)\times{\bf r}'(t)}_{\textrm{Equals \({\bf 0}\)}} = {\bf r}(t)\times{\bf r}”(t) \]
Here, \({\bf r}'(t)\times{\bf r}'(t) ={\bf 0}\) because the cross product of a vector with itself is zero.
The derivative vector \({\bf r}'(t_0)\) has an important geometric property: It points in the direction tangent to the path traced by \({\bf r}(t)\) at \(t=t_0\).
To understand why, consider the difference quotient, where \(\Delta {\bf r} = {\bf r}(t_0+ h)-{\bf r}(t_0)\) and \(\Delta t = h\) with \(h \neq 0\): \[ \frac{\Delta {\bf r}}{\Delta t} = \frac{{\bf r}(t_0+h)-{\bf r}(t_0)}{h}\tag{6} \]
The vector \(\Delta {\bf r}\) points from the head of \({\bf r}(t_0)\) to the head of \({\bf r}(t_0+h)\) as in Figure 13.9. The difference quotient \(\Delta {\bf r}/\Delta t\) is a scalar multiple of \(\Delta {\bf r}\) and therefore points in the same direction [Figure 13.9].
As \(h=\Delta t\) tends to zero, \(\Delta {\bf r}\) also tends to zero but the quotient \(\Delta {\bf r}/\Delta t\) approaches a vector \({\bf r}'(t_0)\), which, if nonzero, points in the direction tangent to the curve. Figure 13.10 illustrates the limiting process. We refer to \({\bf r}'(t_0)\) as the tangent vector or the velocity vector at \({\bf r}(t_0)\).
The tangent vector \(\displaystyle{{\bf r}'(t_0)}\) (if it is nonzero) is a direction vector for the tangent line to the curve. Therefore, the tangent line has vector parametrization: \[ \boxed{\bbox[#FAF8ED,5pt]{\textrm{Tangent line at \({\bf r}(t_0)\):}\qquad {\bf L}(t) = {\bf r}(t_0) + t\,{\bf r}'(t_0)}}\tag{7} \]
Although it has been our convention to regard all vectors as based at the origin, the tangent vector \({\bf r}'(t)\) is an exception; we visualize it as a vector based at the terminal point of \({\bf r}(t)\). This makes sense because \({\bf r}'(t)\) then appears as a vector tangent to the curve ( Figure 13.10).
735
Plot \({\bf r}(t) = \langle \cos t, \sin t, 4\cos^2t\rangle\) together with its tangent vectors at \(t = \frac{\pi}4\) and \(\frac{3\pi}2\). Find a parametrization of the tangent line at \(t = \frac{\pi}4\).
Solution The derivative is \(\displaystyle{ {\bf r}'(t) = \langle -\sin t, \cos t, -8\cos t\sin t\rangle}\), and thus the tangent vectors at \(t = \frac{\pi}4\) and \(\frac{3\pi}2\) are \[ {\bf r}'\left(\frac{\pi}4\right) = \langle -\frac {\sqrt 2}2, \frac{\sqrt 2}2, -4\rangle,\qquad {\bf r}'\left(\frac{3\pi}2\right) = \langle 1, 0, 0\rangle \]
Figure 13.11 shows a plot of \({\bf r}(t)\) with \({\bf r}'\big(\tfrac{\pi}4\big)\) based at \({\bf r}\big(\tfrac{\pi}4\big)\) and \({\bf r}'\big(\tfrac{3\pi}2\big)\) based at \({\bf r}\big(\tfrac{3\pi}2\big)\). At \(t=\frac{\pi}4\), \({\bf r}\big(\frac{\pi}4\big) = \langle \frac{\sqrt 2}2, \frac{\sqrt 2}2, 2 \rangle\) and thus the tangent line is parametrized by \[ {\bf L}(t) = {\bf r}\left(\frac{\pi}4\right) + t\,{\bf r}'\left(\frac{\pi}4\right) = \langle \frac{\sqrt 2}2, \frac{\sqrt 2}2, 2 \rangle + t \langle -\frac {\sqrt 2}2, \frac{\sqrt 2}2, -4\rangle \]
There are some important differences between vector- and scalar-valued derivatives. The tangent line to a plane curve \(y=f(x)\) is horizontal at \(x_0\) if \(f'(x_0)=0\). But in a vector parametrization, the tangent vector \({\bf r}'(t_0)=\langle x'(t_0), y'(t_0)\rangle\) is horizontal and nonzero if \(y'(t_0) = 0\) but \(x'(t_0)\ne 0\).
The function \[ {\bf r}(t) = \langle t - \sin t, 1-\cos t\rangle \] traces a cycloid. Find the points where:
736
Solution The tangent vector is \({\bf r}'(t) = \langle 1 - \cos t, \sin t\rangle\). The \(y\)-component of \({\bf r}'(t)\) is zero if \(\sin t = 0\)—that is, if \(t = 0, \pi, 2\pi, \ldots.\) We have \begin{eqnarray*} {\bf r}(0) &=& \langle 0,0\rangle,& {\bf r}'(0) &=& \langle 1-\cos 0,\sin 0 \rangle = \langle 0,0\rangle\quad\textrm{(zero vector)}\\ {\bf r}( \pi ) &=& \langle \pi,2\rangle,& {\bf r}'(\pi) &=& \langle 1-\cos \pi,\sin \pi \rangle = \langle 2,0\rangle\quad\textrm{(horizontal)} \end{eqnarray*}
By periodicity, we conclude that \({\bf r}'(t)\) is nonzero and horizontal for \(t =\pi, 3\pi, 5\pi,\dots\) and \({\bf r}'(t)={\bf 0}\) for \(t=0, 2\pi, 4\pi,\dots\) (Figure 13.12).
The cycloid in Figure 13.12 has sharp points called cusps at points where \(x= 0, 2\pi, 4\pi,\dots\). If we represent the cycloid as the graph of a function \(y=f(x)\), then \(f'(x)\) does not exist at these points. By contrast, the vector derivative \({\bf r}'(t)=\langle 1-\cos t, \sin t\rangle\) exists for all \(t\), but \({\bf r}'(t)={\bf 0}\) at the cusps. In general, \({\bf r}'(t)\) is a direction vector for the tangent line whenever it exists, but we get no information about the tangent line (which may or may not exist) at points where \({\bf r}'(t)={\bf 0}\).
The next example establishes an important property of vector-valued functions that will be used in Sections 13.4–13.6.
Prove that if \({\bf r}(t)\) has constant length, then \({\bf r}(t)\) is orthogonal to \({\bf r}'(t)\).
Solution By the Product Rule for Dot Products, \[ \frac{d}{dt} \|{{\bf r}(t)}\|^2 = \frac{d}{dt} \bigl({\bf r}(t)\cdot{\bf r}(t)\bigr) = {\bf r}(t)\cdot{\bf r}'(t)+{\bf r}'(t)\cdot{\bf r}(t) = 2{\bf r}(t)\cdot{\bf r}'(t) \]
This derivative is zero because \(\displaystyle{\|{{\bf r}(t)}}\|\) is constant. Therefore \displaystyle{{\bf r}(t)\cdot{\bf r}'(t)=0}, and \({\bf r}(t)\) is orthogonal to \({\bf r}'(t)\) [or \({\bf r}'(t)=\mathbf{0}\)].
The result of Example 7 has a geometric explanation. A vector parametrization \({\bf r}(t)\) consisting of vectors of constant length \(R\) traces a curve on the surface of a sphere of radius \(R\) with center at the origin (Figure 13.13). Thus \({\bf r}'(t)\) is tangent to this sphere. But any line that is tangent to a sphere at a point \(P\) is orthogonal to the radial vector through \(P\), and thus \({\bf r}(t)\) is orthogonal to \({\bf r}'(t)\).
The integral of a vector-valued function can be defined in terms of Riemann sums as in Chapter 5. We will define it more simply via componentwise integration (the two definitions are equivalent). In other words, \[ \int_a^b {\bf r}(t)\, dt = \langle\, \int_a^b x(t)\,dt, \int_a^b y(t)\,dt, \int_a^b z(t)\,dt\rangle \]
The integral exists if each of the components \(x(t), y(t), z(t)\) is integrable. For example, \[ \int_0^{\pi} \langle 1, t,\sin t \rangle\, dt = \langle\, \int_0^{\pi}1\,dt, \int_0^{\pi} t\,dt,\int_0^{\pi} \sin t \,dt \rangle = \langle \pi , \frac12\pi^2 ,2 \rangle \]
Vector-valued integrals obey the same linearity rules as scalar-valued integrals (see Exercise 72).
An antiderivative of \({\bf r}(t)\) is a vector-valued function \({\bf R}(t)\) such that \({\bf R}'(t) = {\bf r}(t)\). In the single-variable case, two functions \(f_1(x)\) and \(f_2(x)\) with the same derivative differ by a constant. Similarly, two vector-valued functions with the same derivative differ by a constant vector (i.e., a vector that does not depend on \(t\)). This is proved by applying the scalar result to each component of \({\bf r}(t)\).
737
If \({\bf R}_1(t)\) and \({\bf R}_2(t)\) are differentiable and \({\bf R}_1'(t) = {\bf R}_2'(t)\), then \[ {\bf R}_1(t)={\bf R}_2(t)+c \] for some constant vector \(c\).
The general antiderivative of \({\bf r}(t)\) is written \[ \int {\bf r}(t)\,dt = {\bf R}(t)+c \] where \(c=\langle c_1, c_2, c_3\rangle\) is an arbitrary constant vector. For example, \[ \int \langle 1, t,\sin t \rangle\, dt =\langle t , \frac12t^2 , -\cos t \rangle +c= \langle t +c_1, \frac12t^2 +c_2, -\cos t +c_3\rangle \]
If \({\bf r}(t)\) is continuous on \([a,b]\), and \({\bf R}(t)\) is an antiderivative of \({\bf r}(t)\), then \[ \int_a^b {\bf r}(t)\, dt = {\bf R}(b)-{\bf R}(a) \]
The path of a particle satisfies \[ \frac{d{\bf r}}{dt} = \langle 1-6\sin 3t, \dfrac15t \rangle \] Find the particle’s location at \(t=4\) if \({\bf r}(0) = \langle 4,1\rangle\).
Solution The general solution is obtained by integration: \[ {\bf r}(t) = \int \langle 1-6\sin 3t, \dfrac15t \rangle\, dt = \langle t+2\cos 3t, \dfrac1{10}t^2\rangle + {\bf c} \]
The initial condition \({\bf r}(0)= \langle 4, 1\rangle\) gives us \[ {\bf r}(0) = \langle 2 , 0 \rangle + {\bf c} = \langle 4, 1\rangle \]
Therefore, \(\displaystyle{{\bf c} = \langle 2, 1\rangle}\) and (Figure 13.14) \[ {\bf r}(t) = \langle t+2\cos 3t, \dfrac1{10}t^2 \rangle+ \langle 2, 1\rangle= \langle t+2\cos 3t + 2, \dfrac1{10}t^2+1\rangle \]
The particle’s position at \(t=4\) is \[ {\bf r}(4)=\langle 4+2\cos 12 + 2, \dfrac1{10}(4^2)+1\rangle \approx \langle 7.69, 2.6\rangle \]
738