The derivative as a function in its own right, sketching the graph of f prime from the graph of f, the Lagrange and Leibniz notations, the theorem that differentiability implies continuity and its useful contrapositive, the three ways a derivative fails, and higher-order derivatives with their interpretation.
Subject: Calculus I · 69 slides · symbolic lesson
Open the interactive version of this deck
Title
Calculus I · Section 3.2
From one slope at one point to a slope at every point
Objectives
Section 3.1 found the slope of a curve at one point at a time. This lesson keeps the point as a letter, so one computation gives the slope everywhere at once: a new function.
OpenStax Calculus Volume 1, §3.2 The Derivative as a Function §3.2, p. 203 — learning objectives 3.2.1 to 3.2.5
In Section 3.1 you found the slope of a curve at one chosen point: pick a number, build the difference quotient, take the limit. It works, but if you wanted the slope at fifty points you would have to do it fifty times.
This lesson makes one small change with large consequences. Leave the point as a letter, x, all the way through the computation. The answer is then a formula in x, and that formula is a new function: the derivative function. Plug in any number and it gives the slope there.
Once the derivative is a function, you can graph it, compare its graph with the original, ask where it exists, and even take its derivative. Those are the other four parts of the lesson.
Warm-up
Discussion prompt
From Section 3.1: write the limit that defines the derivative of f at a point a, then use it to find the slope of y = x² at a = 3.
Write your answer before revealing it. If the definition does not come back to you straight away, that is exactly what this warm-up is for; the rest of the lesson uses it on every slide.
The computation for x squared at 3 follows four moves: substitute 3 plus h, expand, cancel the h in the denominator against an h in the numerator, and only then let h go to zero. The answer, 6, is the slope of the tangent at the point where x is 3.
Notice that 6 is twice 3. Hold on to that; in two slides you will see that it is not a coincidence.
Section
Part 1
Concept
Figure (svg): The parabola y equals x squared with short tangent lines drawn at x equals minus 1, 0, 1 and 2; dashed drop lines lead from each point of tangency to its slope written under the axis: minus 2, 0, 2, 4.
Repeat the warm-up at a equal to minus 1, 0, 1 and 2 and the slopes come out minus 2, 0, 2 and 4. Each time the answer is twice the point.
\[ f(x) = x^2 \quad\Longrightarrow\quad \text{slope at } x = 2x \]
Instead of computing slopes one point at a time, do the computation once with a general x. The result is a new function that hands you the slope anywhere.
OpenStax Calculus Volume 1, §3.2 The Derivative as a Function §3.2, p. 203 — introduction
Look at the picture. Four short tangent lines touch the parabola at four different places, and the number under each one is its slope. On the left the tangent tilts down, at the bottom it is flat, and on the right it tilts up more and more steeply.
Read the slopes against the positions: at minus 1 the slope is minus 2, at 0 it is 0, at 1 it is 2, at 2 it is 4. Every slope is twice its position. That is a rule, a function of x, and it holds at every point of the parabola, not just these four.
The idea of this section is to get that rule directly from one computation instead of guessing it from a table of examples.
Concept
\[ f'(x) = \lim_{h\to 0}\frac{f(x+h) - f(x)}{h} \]
derivative function — The function f prime whose value at x is the limit above. Its domain is exactly the set of x where that limit exists (equation 3.9).
OpenStax Calculus Volume 1, §3.2 The Derivative as a Function §3.2, p. 203 — Definition and equation 3.9
The formula is the same limit as in Section 3.1, with the fixed point a replaced by the letter x. Everything you know about computing that limit carries over unchanged.
Pay attention to the phrase about the domain. The derivative function is defined only at those x where the limit exists. For most functions you meet that is almost everywhere, but not always, and you will see in a few slides that the square root loses the point zero.
The three bullet points give the vocabulary. Differentiable at a point means the limit exists there. Differentiable on an open set means it exists at every point of that set. A differentiable function, with no qualifier, has a derivative at every point of its domain.
Notation
Annotate
On: \( f'(x) = \lim_{h\to 0}\frac{f(x+h) - f(x)}{h} \)
Step through the notes and match each one to part of the formula. The most important one is the second: inside the limit, x is treated as a constant. Only h is moving. That is why the answer still contains x but no h.
The numerator is a rise and h is a run, so the fraction is the slope of a secant line between the point at x and a neighbouring point. Letting h shrink slides the neighbour in toward x, and the secant slopes settle on the tangent slope.
The last note warns you that the derivative can have a smaller domain than the function. Every time you compute a derivative, ask where your final formula makes sense.
Worked example
Find the derivative function from the definition.
\[ f(x) = x^2 - 2x \]
Substitute into the definition
Why: Replace x by x plus h in every place it appears.
\[ f'(x) = \lim_{h\to 0}\frac{\left[(x+h)^2 - 2(x+h)\right] - \left(x^2 - 2x\right)}{h} \]
Expand the brackets
Why: Square the binomial and distribute the minus 2.
\[ = \lim_{h\to 0}\frac{x^2 + 2xh + h^2 - 2x - 2h - x^2 + 2x}{h} \]
Simplify the numerator
Why: Everything without an h cancels, as it always must.
\[ = \lim_{h\to 0}\frac{2xh - 2h + h^2}{h} \]
Factor out h
Why: Every surviving term carries at least one h.
\[ = \lim_{h\to 0}\frac{h(2x - 2 + h)}{h} \]
Cancel the h
Why: Allowed because h is not zero inside the limit.
\[ = \lim_{h\to 0}(2x - 2 + h) \]
Take the limit
Why: Now h can simply be set to zero.
\[ f'(x) = 2x - 2 \]
Check with a difference quotient
Why: At x equal to 3 the formula gives 4; a secant with h equal to 0.001 gives 4.001.
\[ \frac{f(3.001) - f(3)}{0.001} = \frac{3.004001 - 3}{0.001} = 4.001 \approx 4 = f'(3) \]
OpenStax Calculus Volume 1, §3.2 The Derivative as a Function §3.2, p. 204 — Example 3.12
This is the template you will follow for every polynomial. Substitute x plus h everywhere, being careful to bracket the whole expression so the subtraction of f of x hits every term.
After expanding, every term without an h cancels. That is not luck: the terms without h are exactly f of x, and you subtracted f of x. If they do not cancel in your own work, you have made an algebra slip.
Then factor h from what remains, cancel it against the denominator, and let h go to zero. The check at the end is a habit worth keeping: pick one value of x, compute a secant slope with a tiny h, and see that it lands close to your formula. Here 4.001 against 4 is convincing.
Fill the middle
Fill in the missing lines of the derivative of x squared from the definition.
Fill in the blanks
\frac2xh + h^22x + h = \frac2x}___ = ___ \;\xrightarrow___\; f'(x) = ___
Why: Expanding gives x squared plus 2xh plus h squared; the x squared cancels, leaving 2xh plus h squared. Dividing by h leaves 2x plus h, and as h goes to zero that becomes 2x, which matches the tangent slopes on the first figure.
OpenStax Calculus Volume 1, §3.2 The Derivative as a Function §3.2, p. 204 — Checkpoint 3.6
Type each missing piece before checking. The first blank is the numerator after the x squared terms cancel; the second is what is left after dividing by h; the third is the limit.
If your first blank still contains x squared, go back to the expansion. The square of x plus h has three terms, and the first of them cancels with the x squared you subtract.
The answer, 2x, confirms the pattern you read from the tangent lines on the first figure. One computation has replaced infinitely many.
Worked example
Find the derivative function of the square root.
\[ f(x) = \sqrt{x} \]
Substitute into the definition
Why: Direct substitution of h equal to zero gives zero over zero, so work is needed.
\[ f'(x) = \lim_{h\to 0}\frac{\sqrt{x+h} - \sqrt{x}}{h} \]
Multiply by the conjugate over itself
Why: The conjugate turns a difference of roots into a difference of squares.
\[ = \lim_{h\to 0}\frac{\sqrt{x+h} - \sqrt{x}}{h}\cdot\frac{\sqrt{x+h} + \sqrt{x}}{\sqrt{x+h} + \sqrt{x}} \]
Multiply the numerators
Why: The roots square away; leave the denominator factored.
\[ = \lim_{h\to 0}\frac{(x+h) - x}{h\left(\sqrt{x+h} + \sqrt{x}\right)} = \lim_{h\to 0}\frac{h}{h\left(\sqrt{x+h} + \sqrt{x}\right)} \]
Cancel the h
Why: The troublesome zero over zero is gone.
\[ = \lim_{h\to 0}\frac{1}{\sqrt{x+h} + \sqrt{x}} \]
Evaluate the limit
Why: Set h to zero; the two roots are equal.
\[ f'(x) = \frac{1}{2\sqrt{x}} \]
Check at x equal to 4
Why: The formula gives one quarter; a secant with h equal to 0.01 gives 0.24984.
\[ \frac{\sqrt{4.01} - 2}{0.01} \approx 0.24984 \approx 0.25 = \frac{1}{2\sqrt4} \]
OpenStax Calculus Volume 1, §3.2 The Derivative as a Function §3.2, p. 203 — Example 3.11
With a square root, substituting h equal to zero straight away gives zero over zero, and expanding does nothing because roots do not expand. The tool from Chapter 2 is the conjugate: multiply the top and bottom by the sum of the two roots.
The point of that move is the difference of squares. On top, the square roots disappear and leave x plus h minus x, which is just h. Leave the bottom factored, because you want that h to cancel, not to be multiplied out.
After cancelling, h can be set to zero and the two roots become equal, giving one over two root x. The numerical check at x equal to 4 gives 0.24984 against the exact 0.25. Notice already that the formula divides by the root of x, so it makes no sense at zero.
Picture it
Figure (svg): The curve y equals root x rising from the origin and flattening, together with its derivative one over two root x, which shoots up near x equals 0 and then decays toward zero while staying positive.
The root always rises, so its derivative is always positive. It rises less steeply as x grows, so the derivative decreases. At zero the tangent line is vertical, and the derivative's graph shoots up to meet it.
\[ \lim_{x\to 0^+} f'(x) = +\infty, \qquad f'(0) \text{ undefined} \]
OpenStax Calculus Volume 1, §3.2 The Derivative as a Function §3.2, p. 205 — Figure 3.12
Compare the two curves. The root curve climbs everywhere, so its tangent slopes are all positive, and indeed the derivative curve lies entirely above the axis.
The root curve also flattens as x grows: it climbs, but more and more gently. So its slopes shrink, and the derivative curve falls toward zero without ever reaching it.
Near zero the story reverses. The root curve leaves the origin going straight up, with a vertical tangent, so the slopes become enormous and the derivative shoots up to infinity. At zero itself there is no slope to report, and the derivative is undefined.
Trap
After Example 3.11:
\[ f(x) = \sqrt x, \; x \ge 0 \]
\[ \Longrightarrow\; f'(0) = \frac{1}{2\sqrt 0} \]
Wrong. That is one over zero.
The root is defined at zero, but its derivative is not: the tangent there is vertical. The domain of the derivative is worked out separately, by asking where the limit exists.
\[ \text{dom } f = [0, \infty) \]
\[ \text{dom } f' = (0, \infty) \]
It is natural to assume that wherever you can evaluate a function, you can evaluate its derivative. The square root is the standard example where that fails. It is perfectly defined at zero, but the formula for its derivative asks you to divide by zero there.
The graph on the previous slide explains why: at zero the tangent line is vertical, and a vertical line has no slope. So when a question asks for the domain of the derivative, work it out from the derivative's own formula and from the definition, not by copying the domain of the function.
Worked example
Use the definition to find the derivative function.
\[ f(x) = \frac{9}{x} \]
Substitute into the definition
Why: The numerator is a difference of two fractions.
\[ f'(x) = \lim_{h\to 0}\frac{1}{h}\left(\frac{9}{x+h} - \frac{9}{x}\right) \]
Combine over a common denominator
Why: The common denominator is x times x plus h.
\[ \frac{9}{x+h} - \frac{9}{x} = \frac{9x - 9(x+h)}{x(x+h)} = \frac{-9h}{x(x+h)} \]
Divide by h and cancel
Why: The h in the numerator meets the h outside.
\[ f'(x) = \lim_{h\to 0}\frac{-9}{x(x+h)} \]
Take the limit
Why: Set h to zero.
\[ f'(x) = -\frac{9}{x^2}, \quad x \ne 0 \]
Check at x equal to 3
Why: The formula gives minus 1; a secant with h equal to 0.001 gives about minus 0.99967.
\[ \frac{9/3.001 - 3}{0.001} \approx -0.99967 \approx -1 = -\frac{9}{3^2} \]
OpenStax Calculus Volume 1, §3.2 The Derivative as a Function §3.2, p. 212 — Exercise 61
When the function is a fraction, the difference in the numerator is a difference of fractions, and the first job is to combine them over a common denominator. That is where the h appears as a factor.
Once the numerator is minus 9 h over x times x plus h, dividing by h cancels it cleanly, and letting h go to zero leaves minus 9 over x squared.
The minus sign makes sense on a graph: nine over x falls everywhere on each branch, so every tangent slope is negative. The derivative is undefined at zero, but so is the function, so here the domains agree. The check at 3 gives minus 0.99967 against the exact minus 1.
Tweak it
Parameter explorer
The curve is the difference quotient of √x, drawn as a function of x for a fixed step h. Slide h toward 0. Which curve does it settle onto?
\[ \frac{\sqrt{x+{h}}-\sqrt{x}}{{h}} \]
\[ h \to 0: \quad \frac{\sqrt{x+h} - \sqrt x}{h} \to \frac{1}{2\sqrt x} \]
The slider draws the difference quotient of the root function, with h fixed, as a function of x. For each x it is the slope of the secant from x to x plus h.
Start with a large h. The curve is low and flat, because long secants average the slope over a wide stretch. Now drag h toward zero. The curve rises and bends, and near zero it starts to climb steeply.
At the smallest h the curve is indistinguishable from one over two root x, the derivative you just computed. This is the definition in motion: the derivative function is what the whole family of difference quotients settles into as the step shrinks.
Matching
Each limit is the derivative of some f at some point a. Match it to the pair.
Match the pairs
Why: Read the pattern f(a + h) − f(a): the part added to h is a, and the subtracted constant is f(a). For example, 16 is 2 to the fourth, and minus 1 is cos π, so plus 1 subtracts it. The x-to-the-fourth limit equals 32, and a secant with h = 0.001 gives 32.024.
OpenStax Calculus Volume 1, §3.2 The Derivative as a Function §3.2, p. 213 — Exercises 68, 70, 71, 73
Here the definition is run backwards. Each limit is some f at some point a written in the form f of a plus h minus f of a, all over h. Your job is to recognise f and a.
Look for what is added to h: that is a. Then check that the constant being subtracted is f of a. Sixteen is two to the fourth; one is one to the two thirds; minus one is cosine of pi, which is why the numerator shows plus one; and one is e to the zero.
Recognising a limit as a derivative is a genuinely useful skill, because once you know the rules of Section 3.3 you will be able to evaluate such limits in one line.
Section
Part 2
Concept
The derivative's graph is not a copy of the function's graph. Its height at each x records how steep f is at that x.
| on the graph of f | on the graph of f′ |
|---|---|
| rising (tangent slopes up) | above the axis: f′(x) > 0 |
| falling (tangent slopes down) | below the axis: f′(x) < 0 |
| horizontal tangent (peak, valley) | on the axis: f′(x) = 0 |
| steeper | farther from the axis |
| vertical tangent | f′ undefined, blows up |
OpenStax Calculus Volume 1, §3.2 The Derivative as a Function §3.2, p. 205 — Graphing a Derivative
This table is the dictionary for sketching. The left column describes what you see on the graph of f; the right column says what that means for the graph of f prime.
The trap to avoid is copying shapes. A peak on f does not become a peak on f prime; it becomes a zero, because the tangent at a peak is flat. The place where f prime has its own peak is where f is climbing most steeply.
Steepness shows up as distance from the axis. A curve that rises steeply gives a large positive derivative; one that rises gently gives a small positive derivative.
Picture it
Figure (svg): Top panel: the parabola f of x equals x squared minus 2x, drawn red where it falls (left of x equals 1) and green where it rises, with its lowest point at (1, minus 1). Bottom panel: the line f prime of x equals 2x minus 2, shaded red below the axis left of 1 and green above the axis right of 1. A dashed line at x equals 1 joins the two panels.
Line the two graphs up on the same horizontal scale and read down each column. The valley of the parabola at x equal to 1 is the zero of the line.
\[ f'(x) < 0 \text{ for } x < 1, \qquad f'(1) = 0, \qquad f'(x) > 0 \text{ for } x > 1 \]
OpenStax Calculus Volume 1, §3.2 The Derivative as a Function §3.2, pp. 205-206 — Figure 3.13
Stacking the graphs, with the same horizontal scale, lets you read them column by column. Run your eye down the dashed line at x equal to 1: the parabola has its lowest point there, and the line crosses zero.
To the left of 1 the parabola falls, drawn in red, and the derivative line sits below the axis in the red region. To the right the parabola rises and the line is above the axis. The farther you go from 1, the steeper the parabola and the farther the line is from the axis.
Whenever you sketch a derivative, draw it underneath the function in exactly this way.
Worked example
Given only the graph of f (top panel), sketch the graph of f prime.
Mark the horizontal tangents
Why: The valley and the peak have flat tangent lines, so the derivative is zero there.
\[ f'(-2) = 0, \qquad f'(3) = 0 \]
Where f rises, put f′ above the axis
Why: Between the valley and the peak the curve climbs.
\[ f'(x) > 0 \text{ on } (-2, 3) \]
Where f falls, put f′ below the axis
Why: Before the valley and after the peak the curve descends.
\[ f'(x) < 0 \text{ on } (-\infty, -2) \text{ and } (3, \infty) \]
Place the highest point of f′
Why: The climb is steepest roughly halfway between the valley and the peak.
\[ \text{max of } f' \text{ near } x = \tfrac12 \]
Figure (svg): Top panel: a smooth curve with a low point at (minus 2, 2) and a high point at (3, 6), falling again after 3. Bottom panel: its derivative, negative left of minus 2, positive between minus 2 and 3 with a peak near x equals 0.5, and negative after 3. Dashed lines at x equals minus 2 and 3 run through both panels.
Check the sketch against the tangent lines
Why: Pick x equal to 0: the tangent to f there slopes up clearly, and the sketch of f prime is above the axis there. At x equal to 6 the curve falls, and the sketch is below.
\[ f'(0) > 0 \;\checkmark, \qquad f'(6) < 0 \;\checkmark \]
OpenStax Calculus Volume 1, §3.2 The Derivative as a Function §3.2, pp. 206-207 — Example 3.13
Here there is no formula, only a picture, so the sketch must come from the dictionary. Start with the easiest facts: the flat points. The valley at minus 2 and the peak at 3 have horizontal tangents, so the derivative is zero at both.
Next fill in the signs. Between the valley and the peak the curve climbs, so the derivative is positive there; outside that stretch the curve falls, so the derivative is negative. That already fixes the shape: below, up through zero at minus 2, a hump, back down through zero at 3.
Finally place the top of the hump where the climb is steepest, around the middle. The graph shown is a smooth model with the same features as the book's; the book's sketch is qualitative, and only these features matter. The check tests two points against the tangent lines.
Sorting
Sort into buckets
Use the graph of f in Example 3.13 (valley at x = −2, peak at x = 3). Sort each point by the sign of f′ there.
For each point, imagine laying a ruler along the tangent to the graph of f there. Does it tilt up, lie flat, or tilt down? That decides the bucket.
The two flat points are the valley and the peak. The points at 0 and 2.5 are on the climb between them, so the tangent tilts up. The point at minus 3.5 is on the way down into the valley and the point at 6 is on the way down from the peak.
Notice that you never used the height of f. The point at 3 is the highest point of the graph, and the derivative there is zero, not large.
Fill the middle
\[ f(x) = x^2 - 4 \]
Fill in the blanks
The derivative is f′(x) = 2x. It is zero at x = 0, where the parabola has its lowest point, and the graph of f′ is above the x-axis on the interval (0, ∞).
Why: Subtracting 4 shifts the parabola down but does not change any slope, so the derivative is 2x, as for x squared. 2x is positive exactly when x is positive, which is where the parabola rises.
OpenStax Calculus Volume 1, §3.2 The Derivative as a Function §3.2, p. 207 — Checkpoint 3.7
Fill each blank before checking. The shortcut is to notice that subtracting 4 moves the parabola down without tilting it, so the derivative must be the same as for x squared.
The derivative 2x is zero only at the origin, where the parabola has its lowest point, and positive exactly when x is positive, which is where the parabola rises.
If you want to be sure, run the definition: the minus 4 appears in both f of x plus h and f of x, and cancels.
Prediction
\[ f(x) = 3x - 5 \]
Predict first
Before computing: what does the graph of f′ look like?
Correct: A horizontal line at height 3
Why: A line has the same slope, 3, at every point, so the derivative is the constant function 3. The definition confirms it: the quotient is [3(x + h) − 5 − (3x − 5)]/h = 3h/h = 3. The −5 only moves the line up or down, and the derivative cannot see that.
Figure (svg): Top panel: two parallel lines, y equals 3x minus 5 and y equals 3x plus 1. Bottom panel: their shared derivative, the horizontal line at height 3.
Commit to an option before revealing. Two of the wrong options come from confusing the height of a line with its slope.
A line has the same slope everywhere, so its derivative is a constant, and the graph of a constant is a horizontal line. The minus 5 decides where the line crosses the vertical axis, which has nothing to do with how steep it is.
The figure makes this concrete: two parallel lines, one shifted up by six, share the same horizontal derivative. That is your first sight of a general fact: the derivative cannot see vertical shifts.
Worked example
Find the derivative of the cubic from the definition, then use it to explain the shape of the graph.
\[ f(x) = x^3 - 3x \]
Expand the cube in f of x plus h
Why: Use the binomial pattern for a cube.
\[ (x+h)^3 = x^3 + 3x^2h + 3xh^2 + h^3 \]
Form the numerator
Why: The x cubed and the minus 3x cancel.
\[ f(x+h) - f(x) = 3x^2h + 3xh^2 + h^3 - 3h \]
Divide by h
Why: Every term has a factor of h.
\[ \frac{f(x+h) - f(x)}{h} = 3x^2 + 3xh + h^2 - 3 \]
Take the limit
Why: Both terms containing h vanish.
\[ f'(x) = 3x^2 - 3 \]
Find where f′ is zero
Why: Factor the difference of squares.
\[ 3(x-1)(x+1) = 0 \iff x = \pm 1 \]
Figure (svg): Top panel: the cubic x cubed minus 3x with a local high point at (minus 1, 2) and a local low point at (1, minus 2). Bottom panel: the upward parabola 3x squared minus 3, crossing zero at minus 1 and 1 and dipping to minus 3 at x equals 0.
Check against the graph
Why: At minus 1 the cubic has a local high point of height 2, at 1 a low point of height minus 2, and in between it falls, just where the parabola dips below zero.
\[ f(-1) = 2, \quad f(1) = -2, \quad f'(0) = -3 < 0 \]
OpenStax Calculus Volume 1, §3.2 The Derivative as a Function §3.2, p. 212 — the skill Exercises 64 to 67 practise
This cubic is a good one to practise on because it has both a hump and a dip. The only new algebra is the cube of a binomial, which has four terms with coefficients 1, 3, 3, 1.
After the cancellation and division, the limit is 3x squared minus 3. Setting it to zero gives x equal to 1 and minus 1, the two places where the cubic has a horizontal tangent.
The figure checks the whole calculation. The cubic's local high point sits above the parabola's left zero, its low point above the right zero, and on the stretch between them, where the cubic falls, the parabola dips below the axis.
Intuition
The dictionary also runs in reverse. Given only the graph of the derivative, you can recover the shape of the function.
\[ f'(x) = 2x - 2 \;\Longrightarrow\; f(x) = x^2 - 2x + C \text{ for any constant } C \]
What you cannot recover is the height: every vertical shift of f has the same derivative, so the shape is determined but not where it sits.
Homework in this section also asks you to go the other way: given the graph of the derivative, sketch the function. The same dictionary works in reverse, because the sign of the derivative tells you whether the function rises or falls.
Pay attention to the difference between crossing zero and touching zero. If the derivative crosses from positive to negative, the function climbs and then descends: a peak. If it only touches zero and stays positive, the function merely pauses, as x cubed does at the origin.
The one thing the derivative cannot tell you is height. Any vertical shift of a function has the same derivative, so the graph you draw is correct up to sliding it up or down. The formula line shows this with a plus C. To pin the height down you need one extra piece of information, such as a single point on the graph.
Error analysis
Annotate
On: \( f(x) = x^3, \quad f'(0) = 0 \;\Longrightarrow\; \text{f turns around at } x = 0 \)
Read the claim, decide whether you believe it, then step through the notes. The derivative calculation is correct; the conclusion drawn from it is not.
A zero of the derivative means a horizontal tangent, nothing more. For x cubed, the derivative is positive on both sides of zero, so the curve rises, pauses for an instant, and rises again. A turning point needs the derivative to change sign.
Keep x cubed as your standard example whenever you are tempted to equate a zero derivative with a peak or a valley. You will make this precise in Chapter 4.
Section
Part 3
Concept
Four ways of writing the same derivative, for a function written as y equal to f of x:
\[ f'(x), \qquad \frac{dy}{dx}, \qquad y', \qquad \frac{d}{dx}\big(f(x)\big) \]
For Example 3.12 every one of these says the same thing:
\[ y = x^2 - 2x: \quad y' = 2x - 2, \quad \frac{dy}{dx} = 2x - 2, \quad \frac{d}{dx}\left(x^2 - 2x\right) = 2x - 2 \]
The value at a point has its own Leibniz form, a vertical bar with the point underneath.
\[ f'(a) = \frac{dy}{dx}\Big|_{x=a} \]
OpenStax Calculus Volume 1, §3.2 The Derivative as a Function §3.2, p. 204 — notation
Two notations are in use, and you need to read and write both. Lagrange's prime notation is compact and names the function: f prime of x. Leibniz's notation, dy over dx, names the two variables and the fact that one is changing with respect to the other.
The third line shows all of them saying the same thing for Example 3.12. The notation d over dx in front of an expression is an instruction: differentiate this with respect to x.
Evaluating at a point also has two forms. In Lagrange's notation you simply write the number in the brackets. In Leibniz's notation you write the derivative and put a vertical bar after it, with the point underneath.
Picture it
Figure (svg): A rising curve with a point at x equals a and a second point at a plus delta x. A secant line joins them; a dashed horizontal run marked delta x and a dashed vertical rise marked delta y form a right triangle under it. A second, steeper line is the tangent at a.
A small change in x, written delta x, produces a change in y, written delta y. Their ratio is the slope of the secant. Leibniz wrote the limit of that ratio with the letters d in place of the deltas.
OpenStax Calculus Volume 1, §3.2 The Derivative as a Function §3.2, p. 205 — Figure 3.11
Look at the right triangle under the secant line. Its horizontal side is the change in x, delta x, and its vertical side is the change in y, delta y. The ratio of the two is the slope of the secant.
Now imagine the second point sliding along the curve toward the first. The triangle shrinks, and the secant swings toward the dashed tangent line. Leibniz wrote the limit of the ratio with the letter d in place of the delta, a reminder that it came from tiny changes.
This is why Leibniz's notation is popular in physics and engineering: it keeps visible that a derivative is a rate, change in output per change in input, and it carries units naturally.
Notation
Annotate
On: \( \frac{dy}{dx} = \lim_{\Delta x\to 0}\frac{\Delta y}{\Delta x} \)
OpenStax Calculus Volume 1, §3.2 The Derivative as a Function §3.2, p. 204 — Leibniz notation
Step through the notes. Delta x is the same thing as the h in the definition, and delta y is the same thing as f of x plus h minus f of x. So this formula is the definition again, written in Leibniz's letters.
The last note is the careful one. Although dy over dx looks like a fraction, it is one symbol, the limit of fractions. In later chapters you will see that it often behaves like a fraction, which makes it a powerful bookkeeping device, but you should not cancel the d's or treat dy as a number yet.
Comparison
For the parabola of Example 3.12, fill in the missing entries.
\[ y = x^2 - 2x \]
Comparison matrix
| meaning | Lagrange | Leibniz | value |
|---|---|---|---|
| the derivative function | f′(x) | dy/dx | 2x − 2 |
| slope at x = 3 | f′(3) | dy/dx at x = 3 | 4 |
| slope at x = 1 | f′(1) | dy/dx at x = 1 | 0 |
| height at x = 3 | f(3) | y at x = 3 | 3 |
Fill each blank before checking. The table separates three different objects that students often blur together: the derivative function, the derivative at a point, and the value of the function itself.
At x equal to 3 the function x squared minus 2x has height 3 but slope 4. At x equal to 1 its slope is 0, since that is the bottom of the parabola. Height and slope are different questions with different answers.
Trap
Asked for the slope of x squared at 3:
\[ f(3) = 9 \]
\[ f'(3) = \frac{d}{dx}(9) = 0 \]
Wrong. The 9 is a constant; of course its derivative is 0.
Differentiate first, then substitute. The derivative function needs x to still be a variable when you take the limit; plugging in first throws away all the information about slope.
\[ f'(x) = 2x \]
\[ f'(3) = 6 \]
The wrong line looks efficient: evaluate first, then differentiate. But once you have substituted 3, you are holding the number 9, and a number does not change, so its derivative is zero. All information about slope has been thrown away.
The correct order is always: find the derivative as a function of x, then substitute the point. The notation helps if you read it carefully. f prime of 3 means the derivative function evaluated at 3; d over dx of f of 3 means the derivative of a constant.
Real world
Discussion prompt
A company's profit is P(x) thousand dollars when x units are sold (Exercise 98), and the temperature is T(x) °F at height x feet (Exercise 97). What do P′(x), P′(30) = 5 and T′(1000) = −0.1 mean? Give units.
OpenStax Calculus Volume 1, §3.2 The Derivative as a Function §3.2, p. 214 — Exercises 97 and 98
Write your interpretation, with units, before revealing. The rule is that a derivative is measured in output units per input unit, because it is a limit of rise over run.
Profit is in thousands of dollars and x counts items, so the derivative is in thousands of dollars per item. A derivative of 5 at 30 items means that the 31st item adds about five thousand dollars of profit. The word about matters: the derivative describes the rate at exactly 30, and the rate may drift over the next unit.
For temperature against height, the derivative is in degrees per foot. A value of minus 0.1 at 1000 feet means the air there cools by about a tenth of a degree for each foot you climb.
Section
Part 4
Concept
Theorem 3.1 — Let f be a function and a a point in its domain. If f is differentiable at a, then f is continuous at a.
\[ f'(a) \text{ exists} \;\Longrightarrow\; \lim_{x\to a} f(x) = f(a) \]
The proof is one algebraic trick: multiply and divide the change in f by the change in x.
OpenStax Calculus Volume 1, §3.2 The Derivative as a Function §3.2, p. 207 — Theorem 3.1
This theorem connects this chapter to the continuity of Chapter 2. It says that having a derivative is a stronger property than being continuous: every differentiable function is automatically continuous.
Intuitively, if the graph has a tangent line at a point, the graph cannot jump or have a hole there, because a jump would make the secant slopes blow up. The proof on the next slide turns that intuition into three lines of limit laws.
Concept
The proof and the examples in this part use a second, equivalent form of the derivative at a point. Put x equal to a plus h, so h is x minus a, and h goes to zero exactly when x goes to a.
\[ f'(a) = \lim_{h\to 0}\frac{f(a+h) - f(a)}{h} \]
\[ f'(a) = \lim_{x\to a}\frac{f(x) - f(a)}{x - a} \]
Both are the slope of the secant from the fixed point at a to a moving point, written with a different name for the moving point. Use whichever makes the algebra shorter.
OpenStax Calculus Volume 1, §3.2 The Derivative as a Function §3.2, p. 207 — the form used in the proof of Theorem 3.1
You will meet the derivative at a point written in two ways, and it is worth seeing once that they are the same. In the first form the moving point is a plus h and the step h shrinks to zero. In the second the moving point is simply called x, and it slides toward a.
The substitution between them is x equal to a plus h. Then x minus a is h, and x approaching a is the same thing as h approaching zero, so the two limits are the same limit with the letters renamed.
The second form is convenient when the point a is zero, as in the next three examples, and it is the form the proof of Theorem 3.1 needs, because it puts x minus a in the denominator where it can be multiplied back out.
Concept
Assume the derivative exists, written in the equivalent form with x approaching a:
\[ f'(a) = \lim_{x\to a}\frac{f(x) - f(a)}{x - a} \]
Add and subtract f(a), then multiply and divide by x minus a:
\[ \lim_{x\to a} f(x) = \lim_{x\to a}\left(\frac{f(x) - f(a)}{x - a}\cdot(x - a) + f(a)\right) \]
Split with the limit laws, which is allowed because each piece has a limit:
\[ = \left(\lim_{x\to a}\frac{f(x) - f(a)}{x - a}\right)\left(\lim_{x\to a}(x - a)\right) + \lim_{x\to a} f(a) \]
\[ = f'(a)\cdot 0 + f(a) = f(a) \]
So the limit equals the value, which is continuity at a. The finite number f prime of a times zero is what makes it work; an infinite or missing slope would break the step.
OpenStax Calculus Volume 1, §3.2 The Derivative as a Function §3.2, p. 207 — proof of Theorem 3.1
The trick is to write f of x in a form that contains the difference quotient. Add and subtract f of a, then multiply and divide the difference by x minus a. Nothing has changed in value, but now the expression contains the quotient whose limit is f prime of a.
The limit laws let you split the limit of a product into a product of limits, but only because both limits exist. The quotient tends to the finite number f prime of a, and x minus a tends to zero. Their product tends to zero, and what remains is f of a.
Notice exactly where the hypothesis was used. If the derivative were infinite, as it is for the cube root, the product would be infinity times zero, and the argument would collapse. That is why a finite derivative is required.
Concept
\[ \text{not continuous at } a \;\Longrightarrow\; \text{not differentiable at } a \]
That is the same theorem read backwards, and it saves work: a jump or a hole rules out a derivative at once.
\[ \text{continuous at } a \;\not\Longrightarrow\; \text{differentiable at } a \]
The converse is false. The next three examples are continuous everywhere and still have no derivative at 0, each for a different reason.
OpenStax Calculus Volume 1, §3.2 The Derivative as a Function §3.2, pp. 207-209 — summary points 1 to 4
Any true implication gives you its contrapositive for free: if the conclusion fails, the hypothesis must fail too. So a function that is not continuous at a point cannot be differentiable there. Spotting a jump or a hole saves you from computing any limit.
The converse, continuous implies differentiable, is a different statement, and it is false. Continuity only says the graph has no breaks. Differentiability asks for more: a single, finite tangent slope.
The book gives three continuous functions that fail in three different ways. Learn them as a set: a corner, a vertical tangent, and endless oscillation.
Worked example
Show that the absolute value is not differentiable at 0, though it is continuous there.
\[ f(x) = |x| \]
Write the difference quotient at 0
Why: Use the x-approaching-a form with a equal to 0.
\[ \frac{f(x) - f(0)}{x - 0} = \frac{|x|}{x} \]
Take the limit from the left
Why: For negative x the absolute value is minus x.
\[ \lim_{x\to 0^-}\frac{|x|}{x} = \lim_{x\to 0^-}\frac{-x}{x} = -1 \]
Take the limit from the right
Why: For positive x the absolute value is x.
\[ \lim_{x\to 0^+}\frac{|x|}{x} = \lim_{x\to 0^+}\frac{x}{x} = 1 \]
Compare the one-sided limits
Why: They disagree, so the two-sided limit does not exist.
\[ -1 \ne 1 \;\Longrightarrow\; f'(0) \text{ does not exist} \]
Figure (svg): The V-shaped graph of y equals the absolute value of x. Each branch is extended through the origin as a dashed line, showing that the left branch has slope minus 1 and the right branch slope 1, meeting at a sharp corner at the origin.
Check with two small steps
Why: With h equal to 0.01 the quotient is 1; with h equal to minus 0.01 it is minus 1, however small h gets.
\[ \frac{|0.01|}{0.01} = 1, \qquad \frac{|-0.01|}{-0.01} = -1 \]
OpenStax Calculus Volume 1, §3.2 The Derivative as a Function §3.2, p. 208 — Figure 3.14
The difference quotient for the absolute value at zero simplifies to the absolute value of x over x. That fraction is minus 1 for every negative x and 1 for every positive x. It never changes as x gets closer to zero; it simply depends on the side.
So the two one-sided limits exist but disagree, and a two-sided limit does not exist. In the picture, the left branch comes in with slope minus 1 and the right branch leaves with slope 1. No single line can be tangent to both.
The numerical check shows that shrinking the step does not help: the quotient is exactly 1 or exactly minus 1, forever.
Worked example
Show that the cube root is not differentiable at 0.
\[ f(x) = \sqrt[3]{x} \]
Write the difference quotient at 0
Why: The value at 0 is 0.
\[ \frac{f(x) - f(0)}{x - 0} = \frac{\sqrt[3]{x}}{x} \]
Simplify the powers
Why: x to the one third over x is x to the minus two thirds.
\[ \frac{x^{1/3}}{x} = x^{-2/3} = \frac{1}{\sqrt[3]{x^2}} \]
Take the limit
Why: The denominator is positive on both sides and shrinks to zero.
\[ \lim_{x\to 0}\frac{1}{\sqrt[3]{x^2}} = +\infty \]
Conclude
Why: An infinite limit is not a number, so there is no derivative; the tangent is vertical.
\[ f'(0) \text{ does not exist} \]
Figure (svg): The curve y equals the cube root of x, passing through the origin with a vertical tangent there. Three secant lines through the origin, to the points at x equals 1, 0.3 and 0.05, get steeper and steeper, with slopes 1, 2.2 and 7.4.
Check with a small step
Why: With h equal to 0.001 (or minus 0.001) the quotient is already 100, and a tenfold smaller h multiplies it by about 4.6.
\[ \frac{\sqrt[3]{0.001}}{0.001} = \frac{0.1}{0.001} = 100 \]
OpenStax Calculus Volume 1, §3.2 The Derivative as a Function §3.2, p. 208 — Figure 3.15
This time the one-sided limits agree, but on infinity. The quotient simplifies to one over the cube root of x squared, which is positive on both sides and grows without bound as x approaches zero.
Look at the secant lines in the figure. As the second point moves toward the origin, the secants get steeper: slopes 1, then 2.2, then 7.4, and they keep going. Their limit is the vertical line, and a vertical line has no slope.
Note the contrast with the proof of Theorem 3.1. The cube root is continuous at zero, so a vertical tangent does not break continuity; it only breaks differentiability.
Worked example
The function below is continuous at 0 (it is squeezed between y equal to x and y equal to minus x). Show that it has no derivative there.
\[ f(x) = \begin{cases} x\sin\left(\frac1x\right) & x \ne 0 \\ 0 & x = 0 \end{cases} \]
Write the difference quotient at 0
Why: The x in front cancels the x in the denominator.
\[ \frac{f(x) - f(0)}{x - 0} = \frac{x\sin(1/x)}{x} = \sin\left(\frac1x\right) \]
Ask whether the limit exists
Why: As x shrinks, one over x runs through every multiple of pi, so the sine keeps sweeping from minus 1 to 1.
\[ \lim_{x\to 0}\sin\left(\frac1x\right) \text{ does not exist} \]
Figure (svg): The graph of x sin of one over x for x between minus 0.1 and 0.1, oscillating faster and faster as it approaches the origin, squeezed between the dashed lines y equals x and y equals minus x.
Check with two sequences of points
Why: Points where one over x is pi over 2 plus a full turn give quotient 1; points where it is 3 pi over 2 plus a full turn give minus 1, and both sequences approach 0.
\[ x = \frac{1}{\pi/2 + 2\pi k}: \; 1, \qquad x = \frac{1}{3\pi/2 + 2\pi k}: \; -1 \]
OpenStax Calculus Volume 1, §3.2 The Derivative as a Function §3.2, pp. 208-209 — Figure 3.16
This function is continuous at zero because the factor x squeezes it: it always stays between the dashed lines y equal to x and y equal to minus x, both of which pass through the origin.
The difference quotient at zero, however, simplifies to sine of one over x, and that oscillates between minus 1 and 1 infinitely often as x approaches zero. The secant lines from the origin keep swinging back and forth between slope 1 and slope minus 1 and never settle.
The check exhibits two sequences of points marching to zero, one on which the quotient is always 1 and one on which it is always minus 1. Two different limits along two paths means no limit at all.
Picture it
Figure (svg): Two magnified windows, each 0.2 wide. Left: y equals x squared near x equals 1 looks like a straight line of slope 2. Right: y equals the absolute value of x near 0 is still a sharp V.
A differentiable function is locally linear: magnify it enough near a point and it looks like its tangent line. A corner never straightens, which is the picture behind the definition failing.
Here is a practical way to see differentiability. In the left window, the parabola near x equal to 1 is magnified until the window is only two tenths wide, and it looks like a straight line of slope 2. That is its tangent line.
In the right window the absolute value is magnified by the same amount, and it still looks exactly like a V. You could zoom a million times and the corner would still be there.
Differentiable functions are locally linear: they look like lines under a microscope. Every failure of the derivative is a failure of this zoom test.
Sorting
Sort into buckets
For each function and point, decide how the derivative fails, or whether it exists.
OpenStax Calculus Volume 1, §3.2 The Derivative as a Function §3.2, p. 213 — Exercises 74 to 77
Sort each item, then compare with the rationales. For each one, ask first whether the function is continuous at the point, then whether the one-sided slopes agree and are finite.
Two items deserve extra care. The function x to the two thirds has a cusp: its one-sided slopes are infinite but with opposite signs, so it is a sharper version of a corner, not a vertical tangent. The function x times the absolute value of x looks as if it should have a corner, but its difference quotient is the absolute value of h, which goes to zero, so it is differentiable with derivative zero.
The step function is caught before any limit is needed: it jumps, so by the contrapositive of Theorem 3.1 it cannot be differentiable.
Concept
At the joint of a piecewise function, the two sides follow different rules, so compute the slope from each side separately.
\[ f'_-(a) = \lim_{h\to 0^-}\frac{f(a+h) - f(a)}{h}, \qquad f'_+(a) = \lim_{h\to 0^+}\frac{f(a+h) - f(a)}{h} \]
The derivative exists exactly when both one-sided limits exist, are finite, and agree.
\[ f'(a) \text{ exists} \iff f'_-(a) = f'_+(a) \text{ (both finite)} \]
For the absolute value at zero they were minus 1 and 1: a corner. For a piecewise function, check continuity first, then compare the two one-sided slopes.
OpenStax Calculus Volume 1, §3.2 The Derivative as a Function §3.2, p. 210 — Example 3.14's method
A piecewise function obeys one formula to the left of the joint and another to the right. A difference quotient with h negative uses the left formula, and one with h positive uses the right formula. So the natural thing is to take the two one-sided limits separately.
The two-sided limit exists exactly when both one-sided limits exist and agree, just as in Chapter 2. For derivatives there is one extra requirement: the common value must be a finite number, because an infinite slope is a vertical tangent, not a derivative.
The order of work matters. Check continuity at the joint first. If the pieces do not meet, the function has no derivative there and you are finished. If they do meet, compute the left slope and the right slope and compare them.
Worked example
Show that f is continuous at 1 but has no derivative there.
\[ f(x) = \begin{cases} 2x & x \le 1 \\ \frac{2}{x} & x > 1 \end{cases} \]
Check continuity
Why: Both formulas give 2 at x equal to 1, and f(1) itself is 2.
\[ \lim_{x\to 1^-} 2x = 2, \quad \lim_{x\to 1^+}\frac2x = 2, \quad f(1) = 2 \]
Left-hand difference quotient
Why: For h negative, f(1 + h) uses the 2x rule.
\[ \lim_{h\to 0^-}\frac{2(1+h) - 2}{h} = \lim_{h\to 0^-}\frac{2h}{h} = 2 \]
Right-hand difference quotient
Why: For h positive, f(1 + h) uses the 2 over x rule; combine the fractions.
\[ \frac{\frac{2}{1+h} - 2}{h} = \frac{2 - 2(1+h)}{h(1+h)} = \frac{-2}{1+h} \]
Let h shrink from the right
Why: The denominator goes to 1.
\[ \lim_{h\to 0^+}\frac{-2}{1+h} = -2 \]
Compare
Why: The one-sided slopes disagree: a corner.
\[ 2 \ne -2 \;\Longrightarrow\; f'(1) \text{ does not exist} \]
Figure (svg): The graph of 2x for x up to 1 rising to the point (1, 2), then 2 over x falling from that point. Dashed extensions show the incoming slope 2 and the outgoing slope minus 2.
Check with numbers
Why: h equal to minus 0.01 gives exactly 2; h equal to 0.01 gives minus 1.980, heading for minus 2.
\[ \frac{f(0.99) - 2}{-0.01} = 2, \qquad \frac{f(1.01) - 2}{0.01} \approx -1.980 \]
OpenStax Calculus Volume 1, §3.2 The Derivative as a Function §3.2, p. 213 — Exercise 77
For a piecewise function, always check continuity first, because if it fails you are done. Here both pieces reach 2 at x equal to 1, so the function is continuous.
Then compute the two one-sided difference quotients, each using the rule that applies on its own side. From the left the rule 2x gives slope 2. From the right, combining fractions gives minus 2 over 1 plus h, which tends to minus 2.
The picture shows what that means: the graph arrives climbing at slope 2 and leaves falling at slope minus 2, a corner. The numbers in the check confirm both one-sided slopes.
Trap
A common wrong line:
\[ |x| \text{ is continuous at } 0 \]
\[ \Longrightarrow\; |x| \text{ is differentiable at } 0 \]
Wrong. The theorem only runs one way.
Differentiable implies continuous. Continuous implies nothing about the derivative. The absolute value is the standard counterexample: continuous at 0, slopes minus 1 and 1 on the two sides.
\[ \text{differentiable} \Rightarrow \text{continuous} \]
\[ \text{continuous} \not\Rightarrow \text{differentiable} \]
This error comes from remembering that differentiability and continuity are linked, but not the direction of the link. The theorem is a one-way street.
The absolute value is the counterexample to carry around: its graph is unbroken, so it is continuous at zero, but the corner means it has no derivative there. Any time you feel like concluding differentiable from continuous, test your argument on it.
Counterexample
Discussion prompt
Write down one formula for a function that is continuous at every real number but not differentiable at x = 1 or at x = 3. Justify both properties in a sentence.
Write a formula before revealing. The hint is that you already know a function with a corner at zero; you need to move copies of that corner to 1 and to 3.
Shifting the absolute value gives corners wherever you like, and adding two continuous functions gives a continuous function. Each corner survives the addition because the other term is smooth near it and cannot cancel a sudden change of slope.
The one-sided slopes, minus 2 then 0 at 1, and 0 then 2 at 3, show that the graph is a flat-bottomed valley with two creases.
Worked example
The track follows a parabola for x below minus 10 and a line afterwards. Find b and c so that f is continuous and differentiable at minus 10.
\[ f(x) = \begin{cases} \frac{1}{10}x^2 + bx + c & x < -10 \\ -\frac14 x + \frac52 & x \ge -10 \end{cases} \]
Match heights at the joint
Why: The left limit must equal f(minus 10), which the line gives as 5.
\[ \frac{1}{10}(-10)^2 - 10b + c = 5 \;\Longrightarrow\; c = 10b - 5 \]
Left-hand difference quotient
Why: Substitute c equal to 10b minus 5, then multiply top and bottom by 10.
\[ \lim_{x\to -10^-}\frac{\frac{1}{10}x^2 + bx + 10b - 10}{x + 10} = \lim_{x\to -10^-}\frac{x^2 + 10bx + 100b - 100}{10(x+10)} \]
Factor by grouping
Why: The numerator has x plus 10 as a factor.
\[ x^2 + 10bx + 100b - 100 = (x + 10)(x - 10 + 10b) \]
Cancel and take the limit
Why: Set x to minus 10.
\[ \lim_{x\to -10^-}\frac{x - 10 + 10b}{10} = \frac{-20 + 10b}{10} = b - 2 \]
Right-hand difference quotient
Why: The line has slope minus one quarter.
\[ \lim_{x\to -10^+}\frac{-\frac14 x + \frac52 - 5}{x + 10} = \lim_{x\to -10^+}\frac{-(x + 10)}{4(x + 10)} = -\frac14 \]
Set the slopes equal and solve
Why: Then recover c from the height condition.
\[ b - 2 = -\frac14 \;\Longrightarrow\; b = \frac74, \quad c = 10\cdot\frac74 - 5 = \frac{25}{2} \]
Figure (svg): The toy track: a parabolic piece for x less than minus 10 joining, at the point (minus 10, 5), a straight ramp of slope minus one quarter that reaches the axis at x equals 10. At the joint the two pieces share both height and slope.
Check height and slope at the joint
Why: The parabola gives 10 minus 17.5 plus 12.5, which is 5; its slope, one fifth of x plus b, is minus 2 plus 1.75.
\[ \tfrac{1}{10}(100) - \tfrac{70}{4} + \tfrac{25}{2} = 5, \qquad \tfrac{2}{10}(-10) + \tfrac74 = -\tfrac14 \]
OpenStax Calculus Volume 1, §3.2 The Derivative as a Function §3.2, pp. 209-210 — Example 3.14
The engineering requirement, a smooth ride, translates into two mathematical conditions at the joint: the pieces must meet (continuity) and they must meet at the same slope (differentiability). Two conditions, two unknowns.
Continuity gives a linear relation between b and c. For the slope, the book computes the left-hand difference quotient directly, which needs a factoring by grouping; the factor x plus 10 is guaranteed to appear because continuity has already been imposed.
Setting the left slope b minus 2 equal to the line's slope minus one quarter gives b equal to seven quarters, and then c is twenty-five halves. The check substitutes both back and confirms height 5 and slope minus one quarter at the joint, which is exactly what the picture shows.
Tweak it
Parameter explorer
Here c = 10b − 5, so the track is always continuous at x = −10. Slide b. For which b does the joint stop being a corner?
\[ b = {b}, \quad \text{left slope } b - 2 = {b} - 2 \]
\[ \text{smooth} \iff b - 2 = -\tfrac14 \iff b = 1.75 \]
In this slider c is tied to b by the continuity condition, so the two pieces always meet at the point minus 10, 5. What changes is the angle at which the parabola arrives.
Try b equal to one half: the track comes in steeply and then kinks. Push b up and the kink softens; overshoot and it kinks the other way. At exactly 1.75 the parabola arrives at the line's slope and the joint becomes invisible.
This is the difference between the two conditions made visible. Continuity is a matter of meeting; differentiability is a matter of meeting without a crease.
Fill the middle
\[ f(x) = \begin{cases} ax + b & x < 3 \\ x^2 & x \ge 3 \end{cases} \]
Fill in the blanks
\text9 3a + b = 6, \quad \text-9 a = ___, \quad b = ___
Why: Continuity at 3 needs the line to reach 3 squared, so 3a + b = 9. The parabola's slope at 3 is 2 times 3 = 6, and the line's slope is a, so a = 6. Then b = 9 − 18 = −9. Check: the line 6x − 9 is the tangent line to x squared at x = 3.
OpenStax Calculus Volume 1, §3.2 The Derivative as a Function §3.2, p. 210 — Checkpoint 3.8
Fill the three blanks. Use the same two conditions as the toy track: equal heights at 3, then equal slopes at 3.
The parabola has height 9 and slope 6 at x equal to 3, using the derivative 2x you found earlier. The line has height 3a plus b and slope a. Matching slopes gives a equal to 6 immediately, and then the height condition gives b equal to minus 9.
There is a nice way to see the answer: the line 6x minus 9 is the tangent line to the parabola at 3. Gluing a curve smoothly to a line always means using the tangent line.
Section
Part 5
Concept
The derivative is a function, so it has a derivative of its own: the second derivative. Keep going for the third, fourth, and so on.
\[ f''(x), \quad f'''(x), \quad f^{(4)}(x), \quad \ldots, \quad f^{(n)}(x) \]
\[ y'', \quad y''', \quad y^{(4)}, \quad \ldots, \quad y^{(n)} \]
\[ \frac{d^2y}{dx^2}, \quad \frac{d^3y}{dx^3}, \quad \frac{d^4y}{dx^4}, \quad \ldots, \quad \frac{d^ny}{dx^n} \]
For motion: the derivative of position is velocity, and the derivative of velocity is acceleration.
OpenStax Calculus Volume 1, §3.2 The Derivative as a Function §3.2, pp. 210-211 — Higher-Order Derivatives
Since the derivative is a function, nothing stops you from applying the definition to it. The result is the second derivative, and you can keep going. These are the higher-order derivatives.
There are three parallel notations. With primes, you add a prime per derivative up to three, then switch to a number in brackets, because four primes are unreadable. In Leibniz notation, the order appears as a power of d on top and of dx on the bottom.
The most important interpretation is motion. Position differentiates to velocity, the rate of change of position, and velocity differentiates to acceleration, the rate of change of velocity.
Intuition
The second derivative answers a different question from the first. The first asks how fast the quantity is changing; the second asks how fast that speed of change is itself changing.
| quantity | first derivative | second derivative |
|---|---|---|
| position (m) | velocity (m/s) | acceleration (m/s²) |
| price level | inflation (price change per year) | whether inflation is rising or falling |
| water in a tank (L) | flow rate (L/min) | whether the flow is speeding up or easing off |
Falling inflation means prices are still rising, only more slowly: the second derivative is negative while the first is still positive. Reading one rung too low turns that into prices falling, which is false.
The first derivative tells you how fast something is changing. The second derivative tells you whether that rate is itself growing or shrinking. These are different questions, and mixing them up is one of the most common errors in reading real data.
Inflation is the clearest everyday example. Inflation is the rate at which prices rise. When a news report says inflation is falling, prices are still going up, only more slowly than before. The first derivative is still positive; the second is negative.
The same shape appears in the other rows of the table. A car can be moving forwards while braking: positive velocity, negative acceleration. A tank can be filling while the tap is being turned down. Always ask which rung of the ladder a statement is about before drawing a conclusion.
Notation
Annotate
On: \( \frac{d}{dx}\left(\frac{dy}{dx}\right) = \frac{d^2y}{dx^2} \)
OpenStax Calculus Volume 1, §3.2 The Derivative as a Function §3.2, p. 211 — the Leibniz form of higher derivatives
The symbol for the second derivative looks strange until you read it as shorthand. Apply the instruction d over dx to the first derivative, dy over dx, and gather up the pieces: two d's on top, and dx twice on the bottom.
The placement of the little 2 matters. On top it sits on the d, not on the y, because y is not being squared. On the bottom it sits after the dx, because the unit per x appears twice. That is why acceleration comes out in metres per second per second.
Worked example
\[ f(x) = 2x^2 - 3x + 1 \]
Form the difference quotient for f
Why: Expand 2 times x plus h squared and minus 3 times x plus h.
\[ \frac{\left[2(x+h)^2 - 3(x+h) + 1\right] - \left(2x^2 - 3x + 1\right)}{h} = \frac{4xh + 2h^2 - 3h}{h} \]
Cancel h and take the limit
Why: Factor h out of the numerator.
\[ f'(x) = \lim_{h\to 0}(4x + 2h - 3) = 4x - 3 \]
Now apply the definition to f′
Why: The same limit, with f prime in place of f.
\[ f''(x) = \lim_{h\to 0}\frac{\left[4(x+h) - 3\right] - (4x - 3)}{h} \]
Simplify
Why: Only the 4h survives.
\[ = \lim_{h\to 0}\frac{4h}{h} = \lim_{h\to 0} 4 \]
Take the limit
Why: A constant stays put.
\[ f''(x) = 4 \]
Check the first derivative numerically
Why: At x equal to 2 the formula 4x minus 3 gives 5; a secant with h equal to 0.001 gives 5.002, with the extra 0.002 being exactly 2h.
\[ \frac{f(2.001) - f(2)}{0.001} = 5.002 \approx 5 = f'(2) \]
OpenStax Calculus Volume 1, §3.2 The Derivative as a Function §3.2, p. 211 — Example 3.15
The first derivative is a routine application of the definition, the same moves as Example 3.12. The expansion gives 4xh plus 2h squared minus 3h, and the limit is 4x minus 3.
The second derivative is the definition applied again, this time to the line 4x minus 3. A line's difference quotient is its slope, so the answer is the constant 4.
The check tests the first derivative numerically. The secant slope at 2 is 5.002, and the extra 0.002 is exactly 2 times h, the leftover term that the limit removes.
Prediction
\[ f(x) = x^2 \]
Predict first
You already know f′(x) = 2x. What is f″(x)?
Correct: 2
Why: Differentiate 2x: the quotient is [2(x + h) − 2x]/h = 2h/h = 2. The first derivative is the line 2x, and a line's derivative is its constant slope. A third derivative would be 0.
OpenStax Calculus Volume 1, §3.2 The Derivative as a Function §3.2, p. 211 — Checkpoint 3.9
Commit to an answer before revealing. The most common wrong answer is 2x, which is the first derivative, not the second.
Two applications of the definition: x squared becomes 2x, and the line 2x becomes its constant slope 2. A constant second derivative says the slope of the parabola changes at a steady rate, going up by 2 for every unit of x.
Worked example
A particle's position (in metres) after t seconds is given below. Find its acceleration.
\[ s(t) = 3t^2 - 4t + 1 \]
Velocity is the derivative of position
Why: Form the difference quotient in t.
\[ s'(t) = \lim_{h\to 0}\frac{3(t+h)^2 - 4(t+h) + 1 - \left(3t^2 - 4t + 1\right)}{h} \]
Simplify and take the limit
Why: The numerator is 6 t h plus 3 h squared minus 4 h; cancel one h.
\[ v(t) = s'(t) = \lim_{h\to 0}(6t + 3h - 4) = 6t - 4 \]
Acceleration is the derivative of velocity
Why: Differentiate the line.
\[ a(t) = \lim_{h\to 0}\frac{\left[6(t+h) - 4\right] - (6t - 4)}{h} = \lim_{h\to 0}\frac{6h}{h} = 6 \]
State it with units
Why: Metres per second, per second.
\[ a(t) = 6 \text{ m/s}^2 \]
Figure (svg): Three panels over t from 0 to 2 seconds: the position parabola 3t squared minus 4t plus 1 dipping to its lowest point at t equals two thirds; the velocity line 6t minus 4 crossing zero at t equals two thirds; and the acceleration, a horizontal line at 6.
Check with a second difference
Why: The second difference of s over h squared estimates the second derivative; with h equal to 0.1 around t equal to 1 it gives 6 exactly, as it must for a quadratic.
\[ \frac{s(1.1) - 2s(1) + s(0.9)}{0.1^2} = 6 \]
OpenStax Calculus Volume 1, §3.2 The Derivative as a Function §3.2, pp. 211-212 — Example 3.16
This is the physical reading of Example 3.15. Position is a quadratic in time, so velocity, its derivative, is a line, and acceleration, the derivative of velocity, is a constant: 6 metres per second squared.
Look at the three panels in the figure. The position parabola bottoms out at two thirds of a second, which is where the velocity line crosses zero: the particle stops for an instant and turns around. The acceleration line is flat at 6 the whole time.
The check uses the second difference, the change of the change, divided by h squared. For a quadratic it gives the second derivative exactly, and here it gives 6.
Worked example
\[ s(t) = t^3 \]
Expand the cube
Why: Binomial pattern.
\[ (t+h)^3 - t^3 = 3t^2h + 3th^2 + h^3 \]
Divide by h and let h shrink
Why: Velocity.
\[ v(t) = \lim_{h\to 0}(3t^2 + 3th + h^2) = 3t^2 \]
Differentiate the velocity
Why: Expand 3 times t plus h squared.
\[ \frac{3(t+h)^2 - 3t^2}{h} = \frac{6th + 3h^2}{h} = 6t + 3h \]
Take the limit
Why: Acceleration.
\[ a(t) = 6t \]
Check at t equal to 1
Why: A secant of the velocity with h equal to 0.001 gives 6.003, close to a(1) equal to 6.
\[ \frac{3(1.001)^2 - 3}{0.001} = 6.003 \approx 6 = a(1) \]
OpenStax Calculus Volume 1, §3.2 The Derivative as a Function §3.2, p. 212 — Checkpoint 3.10
The first derivative needs the cube of a binomial, as in the cubic example. After cancelling and taking the limit, the velocity is 3t squared.
Differentiating the velocity is the same as finding the derivative of x squared, multiplied by 3, so the acceleration is 6t. Unlike Example 3.16, this acceleration changes with time: the particle speeds up faster and faster.
The numerical check at t equal to 1 gives 6.003, close to the exact 6.
Worked example
\[ f(x) = x + \frac1x \]
Difference quotient for f
Why: The x terms give h; the fractions combine as in Exercise 61.
\[ \frac{f(x+h) - f(x)}{h} = \frac{h + \frac{1}{x+h} - \frac1x}{h} = 1 - \frac{1}{x(x+h)} \]
First derivative
Why: Let h go to zero.
\[ f'(x) = 1 - \frac{1}{x^2} \]
Difference quotient for f′
Why: The 1s cancel; combine the two fractions over x squared times x plus h squared.
\[ \frac{f'(x+h) - f'(x)}{h} = \frac{1}{h}\left(\frac{1}{x^2} - \frac{1}{(x+h)^2}\right) = \frac{2x + h}{x^2(x+h)^2} \]
Second derivative
Why: Set h to zero and simplify.
\[ f''(x) = \frac{2x}{x^4} = \frac{2}{x^3} \]
Check at x equal to 2
Why: The formulas give 0.75 and 0.25; secants with h equal to 0.001 give 0.75012 and 0.24981.
\[ f'(2) = 0.75 \approx 0.75012, \qquad f''(2) = 0.25 \approx 0.24981 \]
OpenStax Calculus Volume 1, §3.2 The Derivative as a Function §3.2, p. 213 — Exercise 83
This one combines two earlier ideas. The x part contributes 1 to the derivative, like any line of slope 1, and the one over x part is handled by combining fractions, as in Exercise 61.
The second derivative needs one more round of combining fractions. The difference of squares in the numerator factors as h times 2x plus h, and that h cancels. Setting h to zero leaves 2x over x to the fourth, which is 2 over x cubed.
Both formulas pass a numerical check at x equal to 2: 0.75 and 0.25 for the exact values, 0.75012 and 0.24981 from secants.
Estimation
| t (s) | 0 | 1 | 2 | 3 | 4 | 5 |
|---|---|---|---|---|---|---|
| h (m) | 0 | 2 | 4 | 13 | 25 | 32 |
Predict first
Using the table, estimate h′(3), the rocket's upward speed at t = 3 s.
Correct: About 10.5 m/s
Why: The left slope is (13 − 4)/1 = 9 m/s and the right slope is (25 − 13)/1 = 12 m/s. Their average, 10.5 m/s, is the slope of the secant from t = 2 to t = 4: (25 − 4)/2. The height 13 is not a speed, and 13/3 ≈ 4.3 is the average speed since launch, not the speed at 3 s.
Figure (svg): The six table points of rocket height against time, from (0, 0) to (5, 32). Around t equals 3, a secant from the point at t equals 2 to the point at t equals 4 is drawn; its slope, 10.5, estimates the rocket's speed at t equals 3.
OpenStax Calculus Volume 1, §3.2 The Derivative as a Function §3.2, p. 215 — Exercises 100 and 101
Choose an estimate before revealing. With only a table, you cannot take a limit, so you use the closest secants available.
The secant to the left of 3 has slope 9 metres per second and the one to the right has slope 12. Neither is centred on 3, so averaging them is a better estimate, and that average is exactly the slope of the secant across the point, from 2 to 4: 10.5 metres per second.
The distractors are worth understanding. 13 is the height, not a speed. 13 divided by 3 is the average speed since launch, which is much lower because the rocket started slowly.
Error analysis
Annotate
On: \( s(t) = t^3 - 3t^2 + 2, \quad a(1) = 0 \;\Longrightarrow\; \text{the particle is at rest at } t = 1 \)
Decide what is wrong before revealing the notes. The calculus is correct; the reading of it is one rung off.
At rest means the velocity is zero. The acceleration being zero means only that the velocity is, for that instant, not changing. At t equal to 1 this particle is moving backwards at 3 metres per second, and at that moment its velocity has stopped decreasing and is about to start increasing.
Whenever you interpret a derivative, name the quantity it is the rate of. The second derivative is the rate of change of the rate of change.
Section
Part 6
Pattern
Figure (svg): A flow diagram: start with f of x; form f of x plus h minus f of x over h; simplify until a factor h cancels; take the limit as h goes to 0 to get f prime of x; then feed f prime back in to get f double prime. Side notes: record the domain of f prime, and read its sign against the rising and falling of f.
This is the procedure the whole lesson has used. The diagram shows that it is the same four moves every time; only the algebra in the middle changes, depending on whether you are expanding a polynomial, combining fractions, or multiplying by a conjugate.
Two habits are easy to forget. State the domain of the derivative separately, because it can lose points where the function has a vertical tangent or a corner. And after finding the derivative, read its sign against the graph of the function as a sanity check.
To get higher derivatives, feed the answer back in at the start.
Ranking
Put in order
Order the steps of finding f′(x) from the definition for f(x) = √x.
Why: The conjugate step exists only to make the numerator a multiple of h, and the limit can be taken only after h is cancelled. The domain comes last because it is read off the finished formula: division by √x rules out x = 0.
Drag the steps into order before checking. The step people most often misplace is the limit: it can only be taken after the h has been cancelled, because before that, setting h to zero gives zero over zero.
The domain goes last because it is read from the finished formula. The root in the denominator must be positive, so zero is excluded, even though the root function itself is defined there.
Check
Check your understanding
Using the definition, what is f′(x) for f(x) = 5x − x² (Exercise 58)?
Answer: A
Why: The quotient is [5(x + h) − (x + h)² − 5x + x²]/h = (5h − 2xh − h²)/h = 5 − 2x − h, which tends to 5 − 2x.
Run the definition in your head or on paper. The linear term contributes its slope, 5, and the square contributes minus 2x, from the middle term of the expansion.
Each distractor is a specific algebra slip: forgetting the 2 in 2xh, keeping the x in 5x, or dropping the 5h. If you chose one, find that slip in your own working.
Check
Check your understanding
A smooth function f rises on (−∞, 1), has a peak at x = 1, and falls on (1, ∞). Which describes f′?
Answer: A
Why: The sign of f′ follows whether f rises or falls, and a smooth peak has a horizontal tangent, so f′ crosses from positive to negative at 1.
Translate each feature using the dictionary: rising means a positive derivative, falling means negative, a smooth peak means zero.
The option that gives f prime its own peak at 1 is the shape-copying error. At the top of a hill the ground is level, so the slope there is zero, not large.
Check
Check your understanding
Which statement is true?
Answer: B
Why: B is the contrapositive of Theorem 3.1, so it is exactly as true as the theorem: a derivative at a forces continuity at a.
Only one of these statements follows from Theorem 3.1, and it is the contrapositive. The other three are all versions of reading the theorem backwards.
The absolute value refutes the first and third options, and the cube root refutes the last. Being continuous never guarantees a derivative.
Explain it to yourself
Discussion prompt
f(x) = x² − 4 and g(x) = x² + 7 have the same derivative, 2x. Explain in two sentences why, once using the definition and once using the graphs.
Write your two sentences before revealing. The goal is to be able to explain one fact in two languages, algebra and geometry.
Algebraically, the constant appears in both f of x plus h and f of x, so the subtraction removes it before the limit is even taken. Geometrically, adding a constant slides the whole graph straight up, and sliding does not tilt any tangent line.
The consequence will matter a great deal in Chapter 4: knowing the derivative of a function tells you the function only up to an added constant.
Exit ticket
\[ f(x) = \sqrt{x - 6} \quad \text{(Exercise 60)} \]
Discussion prompt
Find f′(x) from the definition, give the domains of f and f′, and say what the graph of f does at x = 6.
OpenStax Calculus Volume 1, §3.2 The Derivative as a Function §3.2, p. 212 — Exercise 60
This combines the whole first part of the lesson in one problem. Use the conjugate exactly as in Example 3.11; the minus 6 rides along inside both roots.
The result is one over two times the root of x minus 6, which is the root derivative shifted six units right. The function is defined from 6 onwards, but its derivative only for x strictly bigger than 6.
At 6 the graph starts off vertically, just as the root function does at zero, and a vertical tangent has no slope.
Recap
| idea | what to remember |
|---|---|
| derivative function | f′(x) = lim [f(x + h) − f(x)]/h, with x held fixed; its domain is worked out separately |
| graph of f′ | height of f′ = slope of f: rising ↔ positive, falling ↔ negative, flat ↔ zero |
| notation | f′(x) = y′ = dy/dx = d/dx f(x); f′(a) = dy/dx at x = a |
| Theorem 3.1 | differentiable ⇒ continuous; the converse is false |
| three failures | corner (|x|), vertical tangent (∛x), oscillation (x sin(1/x)); plus any discontinuity |
| higher derivatives | f″ = (f′)′; position → velocity → acceleration |
\[ f'(x) = \lim_{h\to 0}\frac{f(x+h) - f(x)}{h} \]
Next, Section 3.3 turns the patterns you have seen here (2x for x squared, 3x squared for x cubed) into rules, so you rarely need the limit again.
Stewart, Calculus: Early Transcendentals 8e, §2.8 The Derivative as a Function §2.8, pp. 152-164 — the same material in Stewart
The derivative function is the Section 3.1 limit with the point left as a letter. Its graph records the steepness of the original graph, and its domain has to be found on its own.
Two notations describe it, and each has its strengths: Lagrange's is compact, Leibniz's shows the rate and its units. Differentiability is stronger than continuity; a function can be continuous and still fail to have a derivative through a corner, a vertical tangent, or oscillation. And derivatives can be taken again, giving acceleration from position.
Computing every derivative from the limit is slow. Section 3.3 turns the patterns you have seen here into rules, so that the derivative of a polynomial takes one line.
Want this taught 1-on-1? Alexander tutors Calculus I — $55/session, free consultation.