From a Scalar Difference Quotient to a Vector Difference Quotient
For a scalar-valued function, the derivative at a point describes the limit of its difference quotient. A vector-valued function has a difference quotient too: its output is a vector, so the quotient records the change in every component at once.
This definition matters when a function describes a point moving through the plane or space. The vector difference quotient measures change in position per unit change in the parameter. Its limit, when it exists, is the derivative vector.
Let \(\mathbf{r}:I\to\mathbb{R}^n\), where \(I\) is an interval and \(a\) is an interior point of \(I\). Write the function in components as \(\mathbf{r}(t)=\langle r_1(t),r_2(t),\ldots,r_n(t)\rangle\). The point \(a\) is interior so that values of \(t\) in the domain are available on both sides of \(a\).
Vector subtraction and division by the scalar \(t-a\) are performed component by component. Thus the difference quotient is \[ \left\langle \frac{r_1(t)-r_1(a)}{t-a}, \frac{r_2(t)-r_2(a)}{t-a}, \ldots, \frac{r_n(t)-r_n(a)}{t-a} \right\rangle. \] The denominator is excluded at \(t=a\), but the limit asks what this vector approaches as \(t\) gets close to \(a\).
What It Means for a Vector Limit to Exist
For a vector \(\mathbf{v}=\langle v_1,\ldots,v_n\rangle\), its Euclidean norm is \(\|\mathbf{v}\|=\sqrt{v_1^2+\cdots+v_n^2}\). A vector-valued function approaches \(\mathbf{L}\) as \(t\to a\) when the distance between its value and \(\mathbf{L}\), measured by this norm, approaches zero.
In particular, the difference quotient approaches \(\mathbf{L}\) precisely when \(\left\|\dfrac{\mathbf{r}(t)-\mathbf{r}(a)}{t-a}-\mathbf{L}\right\|\to 0\). This definition gives a precise meaning to a vector limit; it is not enough for the components to look close for a few selected values of \(t\).
Why the Component Test Works
The result follows from two simple comparisons with the Euclidean norm. Each component's distance is no larger than the whole vector's distance. In the other direction, if all finitely many component distances are small, then the sum of their squares is small.
Let \(q_i(t)=\dfrac{r_i(t)-r_i(a)}{t-a}\), and suppose \(q_i(t)\to L_i\) for every \(i\). Set \(\mathbf{L}=\langle L_1,\ldots,L_n\rangle\).
Given \(\varepsilon>0\), convergence of each \(q_i\) gives a neighborhood of \(a\) in which \(|q_i(t)-L_i|<\varepsilon/\sqrt{n}\). There are finitely many components, so take the smallest of these neighborhood widths. Then all component bounds hold together when \(t\) is sufficiently close to \(a\).
In that common neighborhood, \(\| \langle q_1(t),\ldots,q_n(t)\rangle-\mathbf{L}\|^2 =\sum_{i=1}^{n}|q_i(t)-L_i|^2 <n(\varepsilon/\sqrt{n})^2=\varepsilon^2\). Therefore the vector norm is less than \(\varepsilon\), so the vector quotient converges to \(\mathbf{L}\).
If the vector quotient converges to \(\mathbf{L}\), then for each component \(i\), \(|q_i(t)-L_i|\le \|\langle q_1(t),\ldots,q_n(t)\rangle-\mathbf{L}\|\). The right side approaches zero, so each component quotient converges to \(L_i\). This proves both directions.
The finite number of components is important in the argument: it lets us choose one neighborhood where all component estimates hold. For the finite-dimensional vector functions used in AP Calculus, the theorem means that differentiability can be checked by examining each scalar difference quotient.
Worked Examples
Worked Example: A Polynomial Vector Function
Let \(\mathbf{r}(t)=\langle t^2,3t-1\rangle\). Find \(\mathbf{r}'(2)\) directly from the definition.
First evaluate the function at \(2\): \(\mathbf{r}(2)=\langle 2^2,3(2)-1\rangle=\langle 4,5\rangle\). For \(t\ne2\), the difference quotient is
Taking \(t\to2\) gives \(\mathbf{r}'(2)=\langle 2+2,3\rangle=\langle4,3\rangle\). To check, the first component quotient \(t+2\) approaches \(4\), and the second is constantly \(3\); thus the vector quotient approaches exactly \(\langle4,3\rangle\).
Worked Example: A Component with an Absolute Value
Let \(\mathbf{r}(t)=\langle t^2,|t|^3\rangle\). Determine whether \(\mathbf{r}\) is differentiable at \(0\), using the difference quotient.
Since \(\mathbf{r}(0)=\langle0,0\rangle\), for \(h\ne0\) the quotient at \(0\) is \[ \frac{\mathbf{r}(h)-\mathbf{r}(0)}{h} =\left\langle\frac{h^2}{h},\frac{|h|^3}{h}\right\rangle =\langle h,h|h|\rangle. \] As \(h\to0\), the first component \(h\to0\). Also, \(|h|<1\) sufficiently close to zero, so \(|h|h|=|h|^2\to0\); hence \(h|h|\to0\). Therefore \(\mathbf{r}'(0)=\langle0,0\rangle\).
The check is direct: the norm of the quotient is \(\sqrt{h^2+h^2|h|^2}\), which approaches \(0\) as \(h\to0\). Thus the full vector quotient, not merely its separate entries, has the claimed limit.
Worked Example: A Component That Prevents Differentiability
Let \(\mathbf{r}(t)=\langle|t|,t^2\rangle\). Test differentiability at \(0\).
Here \(\mathbf{r}(0)=\langle0,0\rangle\), so for \(h\ne0\), \[ \frac{\mathbf{r}(h)-\mathbf{r}(0)}{h} =\left\langle\frac{|h|}{h},h\right\rangle. \] If \(h>0\), then \(|h|/h=1\), so the quotient is \(\langle1,h\rangle\) and approaches \(\langle1,0\rangle\) from the right. If \(h<0\), then \(|h|/h=-1\), so it is \(\langle-1,h\rangle\) and approaches \(\langle-1,0\rangle\) from the left.
The two limits differ. Equivalently, the first component difference quotient has no two-sided limit. By the component test, \(\mathbf{r}'(0)\) does not exist, even though the second component quotient \(h\) does converge to \(0\). The check is that a two-sided vector limit would have to give the same limit from both sides, which these two results do not.
A Common Pitfall: Checking Only One Side
The derivative at an interior point is defined by a two-sided limit. In the preceding example, using only \(h>0\) would produce \(\langle1,0\rangle\). Calling this the derivative would be incorrect: the values for \(h<0\) approach \(\langle-1,0\rangle\), not \(\langle1,0\rangle\).
A related error is to decide that a vector derivative exists because most components have limits. Every component difference quotient must have a finite two-sided limit. One component that fails is enough to rule out the vector limit, as the example shows.
Why the Definition Is Useful
The difference quotient gives a common definition for vector-valued functions, regardless of whether their components are polynomials, absolute-value expressions, or other scalar functions. The component theorem makes the definition practical: calculate or analyze each scalar quotient, then collect the resulting limits into a vector.
It also prevents an important conceptual mistake. A vector derivative is not a collection of unrelated answers; it is one limit in \(\mathbb{R}^n\). The equivalence theorem explains exactly when that single vector limit exists and how its components are determined. At the AP level, this provides the foundation for interpreting derivatives of parametric position functions without relying on a separate vector-specific notion of differentiation.
Check Your Understanding
For each question, use the difference quotient definition and distinguish a two-sided limit from a one-sided limit.
- Write the difference quotient for \(\mathbf{r}(t)=\langle r_1(t),r_2(t),r_3(t)\rangle\) at \(t=a\).
- State the component test for existence of a vector-valued derivative.
- For \(\mathbf{r}(t)=\langle t^2,2t\rangle\), find \(\mathbf{r}'(1)\) directly from the difference quotient.
- For \(\mathbf{r}(t)=\langle|t|,t^3\rangle\), does the derivative at \(0\) exist? Explain using its component quotients.
- Why does convergence of each of finitely many component quotients imply convergence of the vector quotient?
