Calculus › Section C · Derivatives › Topic C.1

The derivative and the tangent line

One limit captures both the slope of a tangent line and the speed of anything that changes. That limit is the derivative, and it is the engine of the entire first half of this course.

From one point to every point

In Section B you computed the slope of a tangent line at one particular point by taking a limit of secant slopes. That was hard-won and point-by-point. The next move is the one that makes calculus a subject: do it for every point at once. Feed the whole function into the limit-machinery and receive back a second function, one that reports the slope of the original curve wherever you ask. That second function is the derivative. Velocity, marginal cost, the growth rate of a population, the steepness of a hill: all of these are derivatives in different clothing, and the rest of Section C is about computing them quickly.

The definition

The derivative of \(f\) at \(x\) is defined by

f'(x) = \lim_{h \to 0} \frac{f(x + h) - f(x)}{h}

provided this limit exists. Everything here is familiar: the fraction is the slope of a secant line through the points at \(x\) and \(x + h\), and the limit shrinks the second point into the first, leaving the exact slope of the tangent line. With that slope in hand, the tangent line at \((a, f(a))\) is \(y - f(a) = f'(a)(x - a)\). When the limit exists at a point, \(f\) is called differentiable there. When it fails, and we will see precisely where it fails, there is no tangent slope at that point.

Worked example: the derivative of x²

Let \(f(x) = x^2\). Build the difference quotient: \(f(x + h) - f(x) = (x + h)^2 - x^2 = 2xh + h^2\). Divide by \(h\), which is legal because \(h \ne 0\) inside the limit: we get \(2x + h\). Now let \(h \to 0\): the quotient approaches \(2x\). So \(\frac{d}{dx} x^2 = 2x\), a whole slope-reporting function from one limit. Check it against the graph: at \(x = -3\) the parabola falls steeply, and \(2x = -6\), negative, as it should; at the bottom, \(x = 0\), the slope is \(0\). At \(x = 1\) the slope is \(2\), so the tangent line through \((1, 1)\) is \(y - 1 = 2(x - 1)\), that is \(y = 2x - 1\).

Worked example: the derivative of 1/x

Take \(f(x) = \frac{1}{x}\). The numerator of the difference quotient is \(\frac{1}{x + h} - \frac{1}{x} = \frac{x - (x + h)}{x(x + h)} = \frac{-h}{x(x + h)}\). Dividing by \(h\) cancels that \(-h\):

\frac{f(x + h) - f(x)}{h} = \frac{-1}{x(x + h)}

As \(h \to 0\) the denominator approaches \(x^2\), so \(f'(x) = -\frac{1}{x^2}\). The sign deserves a look: it is negative at every point, exactly matching the graph of the reciprocal, which falls along each of its two branches. The derivative did not just give numbers; it explained the shape.

Costumes: an equivalent form and the notations

The change of variable \(x = a + h\) rewrites the definition with no \(h\) at all:

f'(a) = \lim_{x \to a} \frac{f(x) - f(a)}{x - a}

the same secant slopes approaching \(a\) directly, handy when the factor \((x - a)\) cancels neatly. The derivative also wears several notations: Lagrange's prime \(f'(x)\); Leibniz's \(\frac{dy}{dx}\) and \(\frac{d}{dx}f(x)\), which keep the variables visible and so carry units; and the physicists' dot \(\dot{s}\) for time rates. At a point you will write \(f'(3)\) or \(\left.\frac{dy}{dx}\right|_{x=3}\). Different costumes on the same limit; recognise the difference quotient however it is dressed.

Reading the derivative off a graph

You can sketch \(f'\) without any formula at all, straight from the graph of \(f\). Where \(f\) rises, its slopes are positive, so \(f' > 0\); where \(f\) falls, \(f' < 0\). At a smooth peak or valley the tangent is horizontal, so \(f' = 0\). Steep stretches of \(f\) translate to large \(|f'|\), gentle stretches to values near zero. A good drill: trace a curve with a finger and narrate 'positive, smaller, zero, negative' as the slope tilts. This two-way reading between a function's shape and its derivative's graph becomes a core skill in Section D.

Worked example: a growth rate with units

A town's population is modelled by \(P(t) = 1000 + 20t + t^2\), with \(t\) in years from now. Use the definition to find the growth rate. The quotient is

\frac{P(t + h) - P(t)}{h} = \frac{20h + 2th + h^2}{h} = 20 + 2t + h

so \(P'(t) = 20 + 2t\) people per year. Right now the town is growing at 20 people a year; in ten years, at \(t = 10\), it will be growing at \(40\) people per year. Units carry the meaning. Likewise, if \(C(x)\) is the cost in dollars of making \(x\) items, \(C'(x)\) is in dollars per item: the marginal cost, roughly what one more item costs.

Where derivatives fail

The limit in the definition can fail in four recognisable ways. At a corner the secants from the left and right settle on different slopes, like \(|x|\) at \(0\), where every left quotient is \(-1\) and every right quotient is \(1\). Cusps are sharper corners, where the left and right secant slopes race to opposite infinities. A vertical tangent, like the cube root graph at \(0\), has secant slopes exploding without bound. And any discontinuity, a break, hole or jump, kills the derivative immediately: you cannot have a single tangent slope where the function itself is not even holding together.

Differentiable means continuous

That last observation has a famous upgrade to a theorem: if \(f\) is differentiable at \(a\), then \(f\) is continuous at \(a\). The reason is one line. The quotient \(\frac{f(a + h) - f(a)}{h}\) approaches a finite number, while \(h\) approaches \(0\), so the numerator \(f(a + h) - f(a)\), which is \(h\) times that quotient, must approach \(0\). That is exactly continuity at \(a\). The converse fails, and \(|x|\) is the counterexample to remember: continuous at \(0\) but not differentiable there. Differentiability is the stricter, smoother property.

Estimating derivatives from data

Real-world data rarely arrives as a formula; it arrives as a table. To estimate a rate at a recorded point, use the symmetric difference quotient, the slope across the interval straddling the point:

f'(a) \approx \frac{f(a + h) - f(a - h)}{2h}

It averages the forward and backward looks and usually beats a one-sided estimate. If a city's population was 48,000 in 2018 and 52,400 in 2022, the growth rate around 2020 is roughly \(\frac{52\,400 - 48\,000}{4} = 1\,100\) people per year. Notice the denominator is \(2h\), the full width of the window, not half of it.

Common mistakes to avoid

The recurring slips: writing \(f(x) + h\) where the definition needs \(f(x + h)\); forgetting to subtract \(f(x)\), so the quotient becomes \(\frac{f(x+h)}{h}\); plugging in \(h = 0\) before simplifying, which lands on the meaningless \(\frac{0}{0}\); believing continuity implies differentiability, with \(|x|\) as the standing counterexample; using \(h\) instead of \(2h\) in the symmetric estimate; and reporting a rate without units. Finally, the derivative at a point is a number, while the derivative as a function is a formula: keep those two straight when a question asks for each.

Key takeaways

The derivative \(f'(x) = \lim_{h \to 0} \frac{f(x+h) - f(x)}{h}\) is the limit of secant slopes: the slope of the tangent line and the instantaneous rate of change at \(x\). Its units are output per input, which is what makes it a velocity, a marginal cost, a growth rate. Read \(f'\) from the graph of \(f\) by sign and steepness. Differentiability implies continuity, but corners, cusps, vertical tangents and breaks are never differentiable. From tables, estimate with the symmetric quotient over the full window. Next topic: the rules that make these computations seconds instead of minutes.

Flashcards

Flashcard

Derivative definition

f′(x) = lim(h→0) [f(x+h) − f(x)]/h

The limit of secant slopes.

Test your understanding

Question 1 of 5

Using the limit definition, what is the derivative of f(x) = x² + 2x?

Question 2 of 5

If C(x) is the cost in dollars of making x items and C′(50) = 12, what does that mean?

Question 3 of 5

Why does |x| fail to be differentiable at 0?

Question 4 of 5

If f is differentiable at x = a, what must also be true?

Question 5 of 5

A city had 48,000 people in 2018 and 52,400 in 2022. What is the best estimate of the population's growth rate in 2020?

Standards: 2

← Previous topicNext topic →