6.4 Arc Length of a Curve and Surface Area

Arc length from the Pythagorean theorem on an infinitesimal triangle, the formula in either variable, why so few arc length integrals are elementary, the surface area of a solid of revolution, and the frustum's slant height.

Subject: Calculus I · 71 slides · symbolic lesson

Open the interactive version of this deck

What this lesson covers

The lesson, slide by slide

1. Arc Length and Surface Area

Title

Calculus I · Section 6.4

Measuring along a curve instead of under it

2. What this lesson gives you

Objectives

Every application so far has measured something across the curve: the area under it, the volume it sweeps. This lesson measures something along it: how far you would walk if the curve were a path. One right triangle produces the formula, and the integrals it produces are the hardest in the chapter.

OpenStax Calculus Volume 1, §6.4 Arc Length of a Curve and Surface Area §6.4, pp. 582-593 — learning objectives 6.4.1 to 6.4.3

Everything in this chapter so far has measured a region: the area between two curves, the volume of a solid, the shell of a rotated region. All of those look across the curve. This lesson turns and looks along it, and asks how far you would walk if the graph were a footpath.

The derivation takes one right triangle and about five lines, and it produces a formula you could have guessed. What you could not have guessed is how badly behaved the resulting integrals are. Most of them have no antiderivative you can write down, so an important part of the skill here is recognising early which ones will work and which need a machine.

The second half revolves the curve and measures the surface it sweeps. The pieces are no longer discs or shells but bands, and their width is measured along the curve, not across it. That single fact is the source of both the square root in the formula and the most common error in the section.

3. Before anything new: what is the representative piece?

Warm-up

Sections 6.1 to 6.3 all ran the same way: describe one representative piece, then integrate it. The pieces were vertical strips, then circular slabs, then cylindrical shells.

Discussion prompt

If you want the length of a curve rather than an area or a volume, what should the representative piece be, and what would you measure about it? Answer before reading on.

Write your answer before reading the reveal, even if it is only a phrase. The routine you have been following for three sections is worth naming out loud: pick the piece, measure the piece, integrate the piece.

For a length the piece has to be a piece of the curve itself, and the trick that makes it measurable is that a short enough piece of a smooth curve is indistinguishable from a straight segment. Straight segments you can measure, because the Pythagorean theorem measures them.

Notice how little is new. The only genuinely fresh idea in this whole section is that a chord approximates a curve, and everything after that is bookkeeping with a Riemann sum you have seen many times.

4. The formula from one triangle

Section

Part 1

5. A curve has no length until you approximate it

Concept

Figure (svg): Two copies of the parabola y equals x squared on the interval from 0 to 2. The left copy has two straight chords joining three points on it, total length 4.576; the right copy has eight shorter chords hugging the curve, total length 4.642.

Every chord cuts a corner, so every polygon is shorter than the curve. Adding more chords pushes the total up toward the true length.

There is no direct way to measure a curved path. What you can measure is a straight chord, so cut the interval into pieces, join consecutive points on the curve, and add up the chords.

\[ L_n = \sum_{i=1}^{n} \text{length of the } i\text{th chord} \]

Every chord cuts a corner, so each of these totals is too small. Refining the partition lifts the total, and the arc length is the limit it rises to.

OpenStax Calculus Volume 1, §6.4 Arc Length of a Curve and Surface Area §6.4, p. 582 — Figure 6.37

Look at the two panels. On the left the parabola is replaced by two straight chords, and their total length is 4.576. On the right eight chords hug the curve much more closely and total 4.642. The true length is 4.6468.

The important feature is that the numbers increase. A straight chord is the shortest route between its two endpoints, so replacing a piece of curve by its chord always loses length. Every polygon undercuts the curve, and refining the partition can only add length back.

That one-sidedness is different from the rectangles of Chapter 5, which could err in either direction. Here any chord sum you compute is a guaranteed lower bound, which will give you a free check on every answer in this lesson.

6. Smooth: the condition the formula needs

Concept

smooth — A function is smooth on an interval when it is differentiable there and its derivative is continuous.

Area only needed the function to be integrable. Arc length asks for more, because the formula integrates an expression built from the derivative, and that expression has to be integrable in its turn.

\[ f \text{ smooth} \;\Longrightarrow\; f' \text{ continuous} \;\Longrightarrow\; \sqrt{1+\left[f'\right]^2} \text{ integrable} \]

A corner, where the derivative jumps, breaks the argument below; you would split the curve at the corner and add the pieces.

OpenStax Calculus Volume 1, §6.4 Arc Length of a Curve and Surface Area §6.4, p. 582 — the definition of a smooth function

Read the hypothesis carefully, because it is stronger than anything the earlier applications asked for. Area needed the function to be integrable. Volume needed not much more. Arc length needs the function to be differentiable and the derivative itself to be continuous.

The reason appears at the end of the derivation: the thing you integrate is built out of the derivative, so the derivative has to be well enough behaved to be integrated in its turn. A function with a corner, like the absolute value, is not smooth at the corner.

That is not a disaster in practice. You split the curve at the corner, compute the length of each smooth piece, and add. Length is additive along a curve, so nothing is lost.

7. Why length is harder than area

Intuition

Compare what the two questions need to know about the curve. For the area under it, the height at each point is enough: the strip is a rectangle whose width is the step in x.

\[ \text{strip} = f(x)\,dx \qquad\text{against}\qquad \text{piece of curve} = \sqrt{1+\left[f'(x)\right]^{2}}\,dx \]

For the length, the height is irrelevant and the direction is everything: a steep piece of curve covers more distance across the same step. So the derivative enters the integrand, gets squared, and then sits under a root.

That is the whole reason this section's integrals are so much worse than the last three sections'. Nothing about the curve got harder; the question started asking about the boundary rather than the interior.

This slide is worth pausing on, because it explains the difficulty of the whole section in one comparison. To find the area under a curve you need to know how high it is at each point. The strip's width is the step in x, and the curve's steepness has no effect on it.

To find the length you need the opposite information. The height is irrelevant; what matters is the direction the curve is heading, because a steep piece covers far more distance over the same horizontal step. So the derivative has to be in the integrand, and because distance is a hypotenuse, it enters squared and under a root.

Keep this in mind whenever an integral in this section looks impossible. The curve did not become harder. The question changed from one about the region to one about its boundary, and boundaries are measured with hypotenuses.

8. One chord, two legs

Picture it

Figure (svg): A rising curve with one subinterval marked off; a straight chord joins the two curve points, and dashed horizontal and vertical legs below it form a right triangle with legs labelled delta x and delta y.

The chord is the hypotenuse of a right triangle whose legs are the horizontal change and the vertical change over one subinterval.

This is the whole derivation in a picture. Over one subinterval the horizontal change is the same for every piece, and the vertical change is whatever the curve does there.

OpenStax Calculus Volume 1, §6.4 Arc Length of a Curve and Surface Area §6.4, pp. 582-583 — Figures 6.37 and 6.38

Find the three pieces in the picture before moving on: the curve, the chord drawn between two of its points, and the dashed right triangle underneath the chord. The chord is the hypotenuse of that triangle.

The horizontal leg is the same for every subinterval, because the partition is regular, so it deserves the single name delta x. The vertical leg is whatever the curve happens to do on that subinterval, so it carries the index. Some of those vertical changes can be negative, and the squaring in the next step is what makes that harmless.

The whole derivation is now visible: measure this hypotenuse, add up all of them, and let the pieces shrink. There is no further geometric idea in the section.

9. Deriving the arc length formula

Worked example

Turn the chord picture into an integral, one move at a time.

Partition the interval into n equal pieces

Why: A regular partition gives every chord the same horizontal leg.

\[ \Delta x = \frac{b-a}{n}, \qquad a = x_0 < x_1 < \cdots < x_n = b \]

Measure one chord by Pythagoras

Why: The legs are the horizontal and vertical changes across that subinterval.

\[ \text{chord}_i = \sqrt{(\Delta x)^2 + (\Delta y_i)^2} \]

Factor the horizontal leg out of the root

Why: Both terms carry a factor of the horizontal step squared.

\[ \text{chord}_i = \Delta x\,\sqrt{1 + \left(\frac{\Delta y_i}{\Delta x}\right)^{2}} \]

Name that quotient with the Mean Value Theorem

Why: On a smooth function the average slope over the subinterval is the actual slope somewhere inside it.

\[ \frac{\Delta y_i}{\Delta x} = f'(x_i^{*}), \qquad x_i^{*} \in [x_{i-1}, x_i] \]

Add the chords

Why: The total length of the polygon is the sum of the pieces.

\[ L \approx \sum_{i=1}^{n} \sqrt{1 + \left[f'(x_i^{*})\right]^{2}}\,\Delta x \]

Recognise a Riemann sum and take the limit

Why: Sample points times a common width: exactly a Riemann sum for the function inside.

\[ L = \lim_{n\to\infty}\sum_{i=1}^{n} \sqrt{1 + \left[f'(x_i^{*})\right]^{2}}\,\Delta x = \int_a^{b}\sqrt{1+\left[f'(x)\right]^{2}}\,dx \]

Check it on a straight line

Why: For a line of slope m the integrand is constant, and the answer is the Pythagorean distance between the endpoints, which is right.

\[ f(x) = mx+c: \quad L = \int_a^b\sqrt{1+m^2}\,dx = (b-a)\sqrt{1+m^2} \]

OpenStax Calculus Volume 1, §6.4 Arc Length of a Curve and Surface Area §6.4, p. 583 — Theorem 6.4 and its derivation

Follow the algebra with your eye on the picture. The first two lines are pure Pythagoras. The third is the only real trick: pull the horizontal step out of the square root. Because it is squared inside, pulling it out divides the vertical change by it, and what is left inside is one plus a quotient of changes.

That quotient is an average slope, and the Mean Value Theorem says a smooth function attains its average slope somewhere inside the subinterval. Naming that point lets you replace a difference quotient with an actual derivative, which is what turns the sum into a Riemann sum with a sample point in each piece.

Once it is a Riemann sum, the limit is an integral by definition, and the formula is complete. The final line is the check you should do with any new formula: apply it to a case you already know. A line of constant slope gives a constant integrand and the answer is the width times the stretch factor, which is exactly the Pythagorean distance between the endpoints.

10. Reading Theorem 6.4 piece by piece

Notation

Annotate

On: \( L = \int_a^{b} \sqrt{1+\left[f'(x)\right]^{2}}\,dx \)

  • The horizontal leg, squared, after the step in x has been factored out. Drop it and you are measuring only the rise, not the distance travelled.
  • The vertical leg, squared: the slope times the step. It is the DERIVATIVE that is squared, never the function.
  • The hypotenuse of the little triangle. It is also the reason these integrals are hard: a root of a polynomial rarely has an elementary antiderivative.
  • The horizontal step. Because it was factored out of the root, the whole integrand is a length per unit of x: a stretch factor.

OpenStax Calculus Volume 1, §6.4 Arc Length of a Curve and Surface Area §6.4, p. 583 — equation 6.7

Work through the annotations in order, and connect each one to a leg of the triangle. The one is the horizontal leg, the derivative squared is the vertical leg, and the square root over both of them is the hypotenuse.

The second annotation is where marks are lost. It is the derivative that gets squared, not the function. If you find yourself writing the function squared under the root, you have written down the disc-volume integrand from Section 6.2 by mistake.

The third annotation is the honest warning. The square root is the reason you will spend more time in this section deciding whether an integral is doable than actually doing it. That is not a flaw in the formula; it is what measuring a hypotenuse costs.

11. The integrand is a stretch factor

Concept

Figure (svg): The curve of the square root of one plus m squared, plotted against slope m from 0 to 4, rising from 1 and straightening into a line; dots mark the values at slopes 0, 1, 2 and 3.

A horizontal step of one becomes a slanted step of this length. On flat ground the factor is one; on a slope of one it is already the square root of two.

Read the formula as a rate. For each unit you travel horizontally, the distance you actually cover along the curve is the square root of one plus the slope squared.

\[ ds = \sqrt{1+\left[f'(x)\right]^{2}}\,dx \]

On flat ground the factor is one and length equals width. On a slope of one it is already the square root of two, about 1.414, and it keeps climbing: steep ground costs distance.

Read the graph as a conversion rate. The horizontal axis is the slope of the curve at a point, and the vertical axis is how much real distance you cover for each unit of horizontal travel there.

At slope zero the rate is one: flat ground, distance equals width. At slope one it is already about 1.414, so a forty five degree climb costs you forty per cent more walking. At slope three it is about 3.162, barely more than the slope itself, and the dashed line shows why: for steep curves the distance is almost all rise.

This reading also explains the shape of every arc length answer you will get. The length always exceeds the width of the interval, and it exceeds it by more when the curve spends time being steep.

12. The differential of arc length

Picture it

Figure (svg): A large right triangle: the horizontal leg is labelled dx, the vertical leg is labelled f prime of x times dx, and the hypotenuse, drawn over a faint piece of curve, is labelled ds.

Shrink the subinterval and the chord becomes the differential of arc length: one leg is the step in x, the other is the slope times that step.

Shrink the subinterval until the chord and the curve are indistinguishable. The two legs become the differentials, and their hypotenuse is what the integral adds up.

This picture is the previous derivation with the limit already taken. The subinterval has shrunk until the chord and the curve lie on top of each other, and the triangle's legs have become the differentials.

Give the hypotenuse its own name, the differential of arc length, because it will appear again in the surface formula and again in Calculus II and III for parametric and polar curves. In every one of those settings, the only thing that changes is how the two legs are described.

The practical value of the name is that both formulas of this section are short once you have it: a length is the integral of the differential of arc length, and a surface is the integral of a circumference times the same thing.

13. Build the integrand

Fill the middle

The arc length differential, with one piece removed.

Fill in the blanks

ds = \sqrt1} + \left[f'(x)\right]^___}\,dx

Why: The missing piece is 1: the horizontal leg squared, divided by the step squared, which is why it comes out as a bare one. Without it a horizontal segment would have length zero.

Fill the blank before revealing. If you hesitated, go back to the triangle: the square root is over the sum of two squared legs, and the horizontal one was divided out.

The answer is a bare one, and the reason it looks so plain is that the horizontal leg was factored out of the root and turned into the differential outside it. A missing one is the single most frequent error in this section, and it is easy to catch: with the one removed, a horizontal segment would have zero length.

14. Where do the chord sums land?

Estimation

\[ y = x^2 \text{ on } [0,2]: \quad L_2 = 4.576, \; L_4 = 4.627, \; L_8 = 4.642 \]

Predict first

The straight-line distance from (0,0) to (2,4) is 4.472. Given the three polygon totals above, where does the true arc length sit?

  • Below 4.472
  • Between 4.472 and 4.642
  • Just above 4.642, near 4.647
  • Well above 5

Correct: Just above 4.642, near 4.647

Why: Every polygon undercuts the curve, so the totals increase toward the answer and never pass it. They are converging, and the integral gives 4.6468. Halving the piece width roughly quarters the error, which is why eight chords are already good to three decimal places.

\[ \int_0^{2}\sqrt{1+4x^{2}}\,dx = 4.6468 \]

Commit to an option before you reason it through. Then use the structure rather than the arithmetic: the chord sums increase, they never pass the true value, and they are already crowded together near 4.64.

The answer is that the true length is just above the eight-chord total, at 4.6468. Notice also how the errors behave: doubling the number of chords cuts the error by roughly a factor of four each time, which is why eight chords already agree with the answer in three decimal places.

If you chose the option below 4.472, look again at why that is impossible. The straight-line distance between the endpoints is the shortest possible route, so no curve joining them can be shorter.

15. What does the formula say about a horizontal line?

Prediction

Predict first

Apply the arc length formula to the constant function whose graph is the segment from (1,3) to (7,3). What comes out?

  • 6, the width of the interval
  • 0, because nothing changes
  • 3, the height
  • 18, the area under it

Correct: 6, the width of the interval

Why: The derivative is zero, so the integrand is the square root of one, which is one, and the integral of one from 1 to 7 is 6. That is the length of the segment, as it must be. Answering zero means the one in the radicand was dropped, which is the single most common slip in this section.

Answer before revealing. This is the cheapest possible test of whether you have the formula right, and it takes ten seconds.

The derivative of a constant function is zero, the integrand becomes the square root of one, which is one, and the integral of one over an interval is its width. Six, and that is the length of the segment.

If you chose zero, you integrated the derivative instead of the whole integrand, which means the one went missing. Use this test whenever you are unsure of a formula you have half-remembered: a horizontal segment must come out with its own width.

16. Trap: integrating the wrong thing

Trap

The trap

Two wrong integrands for the length of the curve below.

\[ y = 2x^{3/2} \text{ on } [0,1] \]

\[ \int_0^1 3x^{1/2}\,dx = 2 \]

\[ \int_0^1 \sqrt{1+\left[2x^{3/2}\right]^2}dx \]

The fix

The first drops the one and integrates the slope alone, which measures total rise, not distance. The second squares the function where the derivative belongs.

\[ L = \int_0^1\sqrt{1+9x}\,dx = 2.268 \]

The rise is 2 and the true length is 2.268.

Both wrong lines here have been written by real students, and they fail in opposite directions. The first integrates the slope alone. What it computes is the total rise of the curve, two in this case, which happens to be a perfectly meaningful number but is not a length.

The second squares the function where the derivative belongs. That is the disc-volume pattern from Section 6.2 leaking into this section, and it produces a number with no meaning at all.

The defence against both is to say the integrand out loud as you write it: one, plus the derivative squared, all under a root. Three words and the error cannot survive.

17. Computing a length

Section

Part 2

18. Example 6.18: a three-halves power

Worked example

Find the length of the graph over the given interval, to three decimal places.

\[ f(x) = 2x^{3/2} \quad\text{on}\quad [0,1] \]

Differentiate and square

Why: The power rule drops the exponent to one half, and squaring removes the root.

\[ f'(x) = 3x^{1/2}, \qquad \left[f'(x)\right]^{2} = 9x \]

Write the integral

Why: Add one under the root and integrate over the interval.

\[ L = \int_0^{1}\sqrt{1+9x}\,dx \]

Substitute

Why: The radicand is linear, so a single substitution clears it.

\[ u = 1+9x, \quad du = 9\,dx, \quad u: 1 \to 10 \]

Integrate in u

Why: The power rule with exponent one half.

\[ L = \frac19\int_1^{10}\sqrt{u}\,du = \frac19\cdot\frac23\left[u^{3/2}\right]_1^{10} \]

Evaluate

Why: Ten to the three halves is ten times the square root of ten.

\[ L = \frac{2}{27}\left[10\sqrt{10}-1\right] \approx 2.268 \]

Figure (svg): The curve y equals two x to the three halves from the origin to the point one comma two, with a dashed straight chord joining the same two endpoints.

The arc must be longer than the straight chord joining its ends, and only a little longer here because the curve is gentle.

Check against the chord

Why: The endpoints are the origin and the point one comma two, a straight-line distance of the square root of five, about 2.236. The arc is longer, and only slightly, which matches a gentle curve.

\[ \sqrt{5} \approx 2.236 < 2.268 \]

OpenStax Calculus Volume 1, §6.4 Arc Length of a Curve and Surface Area §6.4, pp. 583-584 — Example 6.18

This is the model solution for the section, and it is worth copying the order of the lines. Differentiate, square, add one, integrate. The whole method is in those four verbs.

The exponent three halves is not an accident. Differentiating drops it to one half, and squaring a one-half power gives a first power, so the radicand comes out linear. A linear radicand is the friendliest thing that can happen: one substitution and a power rule finish it.

Take the check seriously, because you will use it on every arc length from here on. The endpoints are the origin and the point one comma two, so the chord has length the square root of five, about 2.236. Your answer must be a little more than that, and 2.268 is. Any answer below the chord is wrong by geometry alone, before you look at the algebra.

19. Checkpoint 6.18: the same power, a smaller coefficient

Worked example

\[ f(x) = \tfrac43 x^{3/2} \quad\text{on}\quad [0,1] \]

Differentiate and square

Why: The three halves and the four thirds cancel to a clean two.

\[ f'(x) = 2x^{1/2}, \qquad \left[f'(x)\right]^{2} = 4x \]

Write the integral

Why: Same shape as the last one, with a four instead of a nine.

\[ L = \int_0^{1}\sqrt{1+4x}\,dx \]

Substitute and integrate

Why: With u equal to one plus four x, the differential brings out a quarter.

\[ L = \frac14\cdot\frac23\left[u^{3/2}\right]_1^{5} = \frac16\left[5\sqrt5 - 1\right] \]

Evaluate

Why: Five root five is about 11.180.

\[ L \approx 1.697 \]

Check against the chord

Why: The curve runs from the origin to the point one comma four thirds, a straight-line distance of 1.667, just under the arc length, as it must be.

\[ \sqrt{1+\tfrac{16}{9}} \approx 1.667 < 1.697 \]

OpenStax Calculus Volume 1, §6.4 Arc Length of a Curve and Surface Area §6.4, p. 584 — Checkpoint 6.18

Do this one yourself before reading the steps; it is the same problem with different numbers, which makes it the perfect test of whether the method stuck.

The coefficient four thirds is chosen so that the derivative comes out as exactly twice the square root of x. Squaring gives four x, and the radicand is again linear. When you meet a three-halves power with an odd-looking coefficient, that coefficient is usually there to make the arithmetic clean.

The chord check works as before: from the origin to the point one comma four thirds is a distance of about 1.667, just below the answer 1.697. Both this example and the previous one are gentle curves, so the arc barely exceeds the chord.

20. Every length sits between a chord and a staircase

Intuition

Two free bounds are available on any arc length, and checking them costs seconds. For a curve that only rises, the straight chord is the shortest route between the endpoints and the two legs walked separately are the longest.

\[ \sqrt{(\Delta x)^{2}+(\Delta y)^{2}} \;\le\; L \;\le\; \Delta x + \Delta y \]

For the last example the legs are one and four thirds, so the length must lie between 1.667 and 2.333, and 1.697 does. For the one before it the bounds were 2.236 and 3, around 2.268.

Get into the habit before you trust any arc length you compute, including a numerical one. An answer below the chord is impossible, and one above the staircase means a slip in the algebra.

Here is the checking habit, stated once so you can apply it everywhere. Any curve that only rises is trapped between two easy routes from one endpoint to the other: the straight diagonal, which is the shortest, and the two legs walked separately, which is the longest.

The upper bound is worth a moment. Walking across and then up covers the horizontal change plus the vertical change, and no rising curve can wander more than that, because it never doubles back. Together the two bounds usually pin an answer down to within a few per cent.

Use both, especially on numerical answers, where you have no algebra to inspect. If a calculator hands you a length below the chord you typed the integrand wrongly, and if it hands you one above the staircase you almost certainly squared the function instead of the derivative.

21. Before integrating: look at the radicand

Step zero

\[ \text{(a) } y = x^{3/2} \qquad \text{(b) } y = x^{2} \qquad \text{(c) } y = \frac{x^3}{3}+\frac{1}{4x} \]

Discussion prompt

For each curve, square the derivative and add one. Which radicands can you integrate with the tools you already have, and what is special about them?

Write your answers before revealing, because the habit this slide is building is worth more than any single computation: look at the radicand before you commit to a method.

There are exactly two lucky cases. Either the radicand is linear, which happens when squaring the derivative kills a fractional power, or the radicand is a perfect square, which happens when the curve was designed backwards from a square. In the second case the root disappears entirely and you integrate an ordinary expression.

Everything else needs a substitution beyond this course, and recognising that in the first thirty seconds saves you from a long and hopeless search for an antiderivative. Exam questions in this section are almost always one of the two lucky cases, so spotting the pattern is also a practical shortcut.

22. Why so few arc lengths are elementary

Concept

Figure (svg): Left: the radicand one plus nine x plotted on the interval from 0 to 1, a straight line rising from 1 to 10. Right: the radicand one plus cosine squared x on the interval from 0 to pi, a wave dipping from 2 down to 1 and back to 2.

What decides whether an arc length is elementary is the shape of the radicand, not the shape of the curve.

The integrand is a square root of one plus a squared derivative. For almost every function you can write down, that expression has no elementary antiderivative at all.

\[ \int\sqrt{1+\cos^{2}x}\,dx \quad\text{is an elliptic integral, not an elementary one} \]

This is not a gap in your technique. It is a theorem that no combination of powers, roots, exponentials, logarithms and trigonometric functions differentiates to that integrand.

OpenStax Calculus Volume 1, §6.4 Arc Length of a Curve and Surface Area §6.4, p. 584 — the remark before Example 6.19

The two panels show what actually decides the difficulty, and it is not how complicated the curve looks. On the left, a three-halves power produces the straight line one plus nine x as a radicand, and one substitution finishes it. On the right, a plain sine produces one plus cosine squared, which is a gentle wave, and that innocent-looking wave defeats every technique of elementary calculus.

This is not a gap in your training. It is a theorem that no finite combination of powers, roots, exponentials, logarithms and trigonometric functions has a derivative equal to that integrand. Mathematicians defined a new family of functions, the elliptic integrals, precisely to give such lengths a name.

So the honest workflow in this section has two exits: finish it by hand when the radicand cooperates, and otherwise set the integral up carefully and evaluate it numerically. Setting it up correctly is the graded skill; the arithmetic afterwards belongs to the machine.

23. Example 6.19: a parabola, done numerically

Worked example

The set-up is routine; the antiderivative is not available in this course.

\[ f(x) = x^{2} \quad\text{on}\quad [1,3] \]

Differentiate and square

Why: The derivative is linear, so its square is a quadratic.

\[ f'(x) = 2x, \qquad \left[f'(x)\right]^{2} = 4x^{2} \]

Write the integral

Why: One plus four x squared, under a root.

\[ L = \int_1^{3}\sqrt{1+4x^{2}}\,dx \]

Recognise the obstacle

Why: A root of a sum of squares needs a trigonometric substitution, which belongs to Calculus II; here, use technology.

\[ \sqrt{1+4x^2}: \text{ no elementary antiderivative at this stage} \]

Evaluate numerically

Why: A calculator or computer algebra system gives five decimal places.

\[ L \approx 8.26815 \]

Check against the chord

Why: The endpoints are one comma one and three comma nine, so the chord has length the square root of sixty eight, about 8.246. The arc is a little longer, which fits a curve that bends gently over this stretch.

\[ \sqrt{2^2+8^2} = \sqrt{68} \approx 8.246 < 8.268 \]

OpenStax Calculus Volume 1, §6.4 Arc Length of a Curve and Surface Area §6.4, p. 584 — Example 6.19

The parabola is about the simplest curve you can name, and its arc length is already out of reach at this stage. Let that sink in, because it corrects the impression textbook exercises tend to give.

The set-up is three lines and costs nothing: the derivative is two x, its square is four x squared, and the integral of the square root of one plus four x squared runs from 1 to 3. An integral of that shape is done with a trigonometric substitution in Calculus II and produces an inverse hyperbolic function; here you report the numerical value.

The check is the chord again. From one comma one to three comma nine is a distance of about 8.246, and the answer 8.268 sits just above it. That agreement tells you the integrand was right, which is the part a numerical evaluation cannot verify for you.

24. Checkpoint 6.19: the length of one sine arch

Worked example

\[ f(x) = \sin x \quad\text{on}\quad [0,\pi] \]

Differentiate and square

Why: The derivative of sine is cosine.

\[ f'(x) = \cos x, \qquad \left[f'(x)\right]^{2} = \cos^{2}x \]

Write the integral

Why: One plus cosine squared, under a root.

\[ L = \int_0^{\pi}\sqrt{1+\cos^{2}x}\,dx \]

Evaluate numerically

Why: This is the classical elliptic integral; a computer gives four decimal places.

\[ L \approx 3.8202 \]

Figure (svg): The sine curve over the interval from 0 to pi with two straight chords joining the origin, the peak and the far zero, forming a triangle under the arch.

Even a two-chord polygon is within three per cent of the arc, yet the exact length of one sine arch cannot be written with the functions of this course.

Check with a two-chord polygon

Why: Joining the origin to the peak and the peak to the far zero gives twice the square root of one plus pi squared over four, about 3.724, safely below the arc and within three per cent of it.

\[ 2\sqrt{\left(\tfrac{\pi}{2}\right)^{2}+1} \approx 3.724 < 3.820 \]

OpenStax Calculus Volume 1, §6.4 Arc Length of a Curve and Surface Area §6.4, p. 584 — Checkpoint 6.19

This is the classical example of a length that cannot be written down. One arch of a sine curve, the most familiar shape in trigonometry, has length 3.8202 and no closed form in elementary terms.

The set-up is two lines, and you should be able to produce them without hesitation: the derivative of sine is cosine, so the radicand is one plus cosine squared. Then stop and report the numerical value rather than hunting for an antiderivative.

The picture gives the check. Two straight chords, from the origin up to the peak and down to the far zero, have total length about 3.724. That is already within three per cent of the arc, which shows how quickly chord approximations close in, and it confirms that 3.820 is the right order of size.

25. Exercise 176: a radicand that is a perfect square

Worked example

The exercise sets are stocked with curves built backwards so that the root disappears.

\[ y = \frac{x^{3}}{3}+\frac{1}{4x} \quad\text{on}\quad [1,3] \]

Differentiate

Why: A power rule on each term; the second term is one quarter of x to the minus one.

\[ y' = x^{2} - \frac{1}{4x^{2}} \]

Square it

Why: The cross term is minus twice x squared times one over four x squared, which is exactly minus one half.

\[ \left[y'\right]^{2} = x^{4} - \frac12 + \frac{1}{16x^{4}} \]

Add one

Why: The minus one half becomes plus one half, and the three terms are a perfect square.

\[ 1+\left[y'\right]^{2} = x^{4} + \frac12 + \frac{1}{16x^{4}} = \left(x^{2}+\frac{1}{4x^{2}}\right)^{2} \]

Take the root and integrate

Why: Both terms are positive on the interval, so the root is the bracket itself.

\[ L = \int_1^{3}\left(x^{2}+\frac{1}{4x^{2}}\right)dx = \left[\frac{x^{3}}{3}-\frac{1}{4x}\right]_1^{3} \]

Evaluate

Why: Nine minus one twelfth, less one third minus one quarter.

\[ L = \frac{107}{12}-\frac{1}{12} = \frac{53}{6} \approx 8.833 \]

Check against the chord

Why: The endpoints are one comma seven twelfths and three comma one hundred nine twelfths, giving a chord of about 8.732, just under the arc length.

\[ \sqrt{2^{2}+8.5^{2}} \approx 8.732 < 8.833 \]

OpenStax Calculus Volume 1, §6.4 Arc Length of a Curve and Surface Area §6.4, p. 592 — Exercise 176

Watch what happens in the squaring step, because this is the second of the two lucky patterns and it is entirely deliberate on the author's part. Squaring a difference of two terms produces a cross term, and here that cross term is exactly minus one half, a constant.

Adding one therefore turns the minus one half into plus one half, which is precisely the cross term you would get from squaring the sum of the same two terms. So the radicand is a perfect square, the root vanishes, and what is left is an ordinary power integral.

Curves like this one are built backwards from the answer, which is why they look so contrived. Recognising the pattern is still useful: whenever a curve is a sum of a power and a reciprocal power with an odd coefficient, try the squaring and look for the constant cross term. The answer here is fifty three sixths, and the chord below it is about 8.732.

26. Which of these can you finish by hand?

Sorting

Sort into buckets

Square the derivative, add one, and sort by whether the resulting integral is within reach of Calculus I.

Finish it by hand
y = 5x on [0,2]; y = 2x^(3/2) on [0,1]; y = x²/8 − ln x on [1,2]
Set it up, then use technology
y = x² on [1,3]; y = e^x on [0,1]; y = ln x on [1,3]
hand
A line gives a constant integrand. The three-halves power gives a linear radicand, cleared by one substitution. The third is another built-backwards curve: its radicand is (x/4 + 1/x)², a perfect square.
tech
All three leave a root of a sum of squares: 1 + 4x², 1 + e^(2x), 1 + 1/x². Each needs a trigonometric or hyperbolic substitution from Calculus II, so at this stage you set them up and evaluate numerically.

Sort these before checking, and sort them by doing the first two steps of each in your head rather than by how the curve looks.

Three of them cooperate. A line gives a constant integrand. The three-halves power gives a linear radicand. The third one, with a logarithm in it, is another backwards-built perfect square, and you can spot it by the constant cross term when you square the derivative.

The other three all leave a root over a sum of squares, and each of those needs a trigonometric or hyperbolic substitution from Calculus II. The point is not that they are impossible, but that at this stage the correct answer is a clean set-up and a numerical value, not a page of failed attempts.

27. Slide the exponent and watch the integrand

Tweak it

Parameter explorer

The curve shown is the arc length integrand for y = x raised to the power p. Slide p. Where does the integrand stay near one, and where does it blow up?

\[ \sqrt{1+\left[p\,x^{{p}-1}\right]^{2}} \]

  • p — from 0.4 to 3: exponent p

Move the exponent slowly from the low end upward and watch the left-hand edge of the plot. Below one the integrand shoots up near the origin, because the curve is leaving the origin almost vertically and every horizontal step buys an enormous amount of travel.

At exactly one the curve is a straight line and the integrand is flat, the constant square root of two. Above one the integrand starts at one, because the curve leaves the origin flat, and then rises as the curve steepens further along.

Two things to take away. First, a tall spike in the integrand is the warning sign of a vertical tangent, and that is what the next part of the lesson removes by switching variables. Second, the area under this curve is the arc length, so you can see directly why steeper curves are longer.

28. Choosing the variable

Section

Part 3

29. The same derivation, sliced the other way

Concept

Figure (svg): A curve climbing steeply to the right, described as x equals g of y; one horizontal slab is marked off between two heights, with a chord joining the two curve points and dashed legs labelled delta y for the vertical change and delta x for the horizontal change.

Slice the vertical axis instead. The same triangle appears, with the vertical change now the one you control.

A curve can be described as a function of either variable. Partition the vertical axis instead, and the chord's vertical leg is the one you control.

\[ \text{chord}_i = \sqrt{(\Delta x_i)^{2}+(\Delta y)^{2}} = \Delta y\,\sqrt{1+\left(\frac{\Delta x_i}{\Delta y}\right)^{2}} \]

\[ L = \int_c^{d}\sqrt{1+\left[g'(y)\right]^{2}}\,dy \]

OpenStax Calculus Volume 1, §6.4 Arc Length of a Curve and Surface Area §6.4, p. 585 — Theorem 6.5 and Figure 6.39

Nothing here is new; it is the same triangle with the roles of the legs exchanged. This time you partition the vertical axis, so the vertical change is the controlled one and the horizontal change is whatever the curve does.

Factor the vertical step out of the root instead of the horizontal one, and the quotient inside becomes the derivative of x with respect to y. The formula that comes out has exactly the same shape, with every x and y swapped.

Why bother having two versions? Because a curve is one set of points but has two descriptions, and the two integrals can differ enormously in difficulty. The next few slides are about choosing.

30. Reading Theorem 6.5 piece by piece

Notation

Annotate

On: \( L = \int_c^{d} \sqrt{1+\left[g'(y)\right]^{2}}\,dy \)

  • The curve solved for x: the input is now the height, the output the horizontal position.
  • The derivative dx/dy, not dy/dx. Mixing the two is the standard error here; the whole integral must speak one variable.
  • Limits on the HEIGHT, not on x. Converting an interval in x to one in y is usually the first line of the solution.
  • The roles of the legs have swapped, but the triangle is the same triangle, so the formula's shape is unchanged.

OpenStax Calculus Volume 1, §6.4 Arc Length of a Curve and Surface Area §6.4, p. 585 — equation 6.8

Read the annotations with the previous theorem next to them in your mind. Three of the four are word-for-word the same idea; only the variable's name has changed.

The second annotation is the one that costs marks. The derivative in this formula is the rate at which the horizontal position changes with height, not the familiar slope. They are reciprocals of each other, so using the wrong one is not a small error.

The third matters in practice more than it looks. You will often be handed endpoints as points in the plane, and the limits of a vertical integral are their heights. Converting them is usually the first line of the solution, and forgetting to convert is a silent error that still produces a number.

31. Example 6.20: a cubic in y

Worked example

\[ g(y) = 3y^{3} \quad\text{on}\quad [1,2] \]

Differentiate with respect to y and square

Why: The power rule, then square the nine.

\[ g'(y) = 9y^{2}, \qquad \left[g'(y)\right]^{2} = 81y^{4} \]

Write the integral

Why: One plus eighty one y to the fourth, under a root, integrated in y.

\[ L = \int_1^{2}\sqrt{1+81y^{4}}\,dy \]

Recognise the obstacle

Why: A quartic under a root: no elementary antiderivative, so evaluate numerically.

\[ L \approx 21.0277 \]

Check against the chord

Why: The endpoints are three comma one and twenty four comma two, a chord of length the square root of four hundred and forty two, about 21.0238. The curve is so steep here that it is barely longer than its chord.

\[ \sqrt{21^{2}+1^{2}} \approx 21.0238 < 21.0277 \]

OpenStax Calculus Volume 1, §6.4 Arc Length of a Curve and Surface Area §6.4, p. 585 — Example 6.20

This example exists to make the vertical version routine before it becomes useful. The function is already given in terms of the height, so there is nothing to rewrite: differentiate with respect to the height, square, add one, integrate in the same variable.

The radicand is one plus eighty one times the fourth power of the height, which is neither linear nor a perfect square, so the answer is numerical. That is the ordinary case, as you know by now.

The check is delightful here. The two endpoints are three comma one and twenty four comma two, so the chord is about 21.0238 and the arc is 21.0277. Those agree to three decimal places, because over this interval the curve is so steep that it is essentially a straight line at a very sharp angle.

32. Checkpoint 6.20: a hyperbola branch

Worked example

\[ g(y) = \frac{1}{y} \quad\text{on}\quad [1,4] \]

Differentiate with respect to y

Why: The derivative of y to the minus one.

\[ g'(y) = -\frac{1}{y^{2}}, \qquad \left[g'(y)\right]^{2} = \frac{1}{y^{4}} \]

Write the integral

Why: The square kills the minus sign, as it always does.

\[ L = \int_1^{4}\sqrt{1+\frac{1}{y^{4}}}\,dy \]

Evaluate numerically

Why: The radicand is not a perfect square and not linear.

\[ L \approx 3.1502 \]

Check with two bounds

Why: The curve runs from one comma one to one quarter comma four, a chord of about 3.092; and the length cannot exceed the two legs of that triangle taken separately, three quarters plus three.

\[ 3.092 < 3.150 < 3.75 \]

OpenStax Calculus Volume 1, §6.4 Arc Length of a Curve and Surface Area §6.4, p. 585 — Checkpoint 6.20

Try this before reading. The function is a reciprocal, given as a function of the height, so once again no rewriting is needed.

The derivative is negative, which frightens some people, but the formula squares it immediately and the sign disappears. A decreasing curve has exactly the same length as its mirror image, and the formula knows that.

Both bounds are available here, so use them: the chord between the endpoints is about 3.092 and the two legs walked separately total 3.75. The answer 3.150 sits comfortably between them, so the set-up was right even though the integral itself had to go to a machine.

33. A vertical tangent in one variable is a flat one in the other

Picture it

Figure (svg): Left: y equals x to the two thirds plotted against x, standing up vertically at the origin. Right: the same points read as x equals y to the three halves, plotted against y, leaving the origin horizontally.

One set of points, two descriptions. The derivative blows up at the origin on the left and is zero on the right.

At the origin the left description has a tangent that stands straight up, so the derivative does not exist there and the integral is improper. The right description of the same points has a horizontal tangent, and nothing goes wrong at all.

Compare the two panels carefully, because they show the same curve. On the left the input is the horizontal coordinate, and at the origin the curve rises vertically, so the slope is unbounded and the arc length integral is improper there.

On the right the input is the height, and the very same set of points leaves the origin flat. The derivative is zero at the origin, the integrand is a well-behaved function, and nothing is improper at all.

This is why the second version of the theorem exists. A vertical tangent is not a property of the curve; it is a property of the description you chose. Choosing the other description makes the problem ordinary, and it costs you nothing but one line of algebra.

34. Exercise 172: switching to make the integral finite

Worked example

Find the length of the curve from the point one comma one to the point eight comma four.

\[ y = x^{2/3} \]

Try it in x first

Why: The derivative has a negative exponent, which is awkward and unbounded at the origin.

\[ \frac{dy}{dx} = \frac23 x^{-1/3}, \quad 1+\frac{4}{9}x^{-2/3} \]

Solve the curve for x

Why: Raise both sides to the three halves.

\[ x = y^{3/2} \]

Convert the limits

Why: The endpoints one comma one and eight comma four have heights one and four.

\[ y: 1 \to 4 \]

Differentiate with respect to y and square

Why: Now the exponent is positive.

\[ \frac{dx}{dy} = \frac32 y^{1/2}, \qquad \left[\frac{dx}{dy}\right]^{2} = \frac94 y \]

Integrate: the radicand is linear

Why: Substitute u equal to one plus nine y over four.

\[ L = \int_1^{4}\sqrt{1+\frac94 y}\,dy = \frac{8}{27}\left[\left(1+\frac94y\right)^{3/2}\right]_1^{4} \]

Evaluate

Why: At the top the bracket is ten, at the bottom thirteen quarters.

\[ L = \frac{8}{27}\left[10^{3/2}-\left(\tfrac{13}{4}\right)^{3/2}\right] \approx 7.634 \]

Check against the chord

Why: The straight-line distance from one comma one to eight comma four is the square root of fifty eight, about 7.616, just below the arc; and a numerical evaluation of the original integral in x gives 7.634 as well.

\[ \sqrt{7^{2}+3^{2}} \approx 7.616 < 7.634 \]

OpenStax Calculus Volume 1, §6.4 Arc Length of a Curve and Surface Area §6.4, p. 592 — Exercise 172

Notice the shape of this solution: the first step is a failed attempt, deliberately shown. Starting in the horizontal variable gives a derivative with a negative fractional exponent, which is unbounded at the origin and unpleasant everywhere else.

Solving the curve for the horizontal coordinate turns the two thirds power into a three halves power, and you already know what a three halves power does when you square its derivative: it gives a linear radicand. What was an improper integral with no elementary antiderivative is now a one-substitution exercise.

Two things must change together when you switch, and forgetting either is the classic error. The derivative must be taken with respect to the new variable, and the limits must be the values of that variable at the endpoints, so heights one and four rather than positions one and eight.

35. Find the error: two variables in one integral

Error analysis

Annotate

On: \( x = y^{3/2} \;\Longrightarrow\; L = \int_1^{4}\sqrt{1+\left[\tfrac23 x^{-1/3}\right]^{2}}\,dy \)

  • Correctly solved for x, so the plan to integrate in y is fine.
  • This is dy/dx, left over from the first attempt. Integrating in y requires dx/dy, which is (3/2)y^(1/2).
  • 1 to 4 are heights, which is right for a dy integral — but they are being used with an integrand written in x.
  • Integrand, derivative, differential and limits must all name the same variable. Switching some but not all leaves an expression with no meaning.

The repaired integral is the one from the previous slide, and it gives 7.634.

Look at the expression and try to name the error before revealing the annotations. Everything about it looks plausible, which is exactly what makes it dangerous.

The mistake is that the derivative was carried over from the earlier attempt in the horizontal variable, while the differential and the limits belong to the vertical one. The result is an expression that cannot be evaluated meaningfully, even though it will happily produce a number if you feed it to a calculator.

The rule to remember is simple and covers every integral you will ever write: the integrand, the derivative inside it, the differential and both limits must all refer to one variable. When you switch, switch all four.

36. Which variable would you integrate in?

Sorting

Sort into buckets

For each curve, decide which description keeps the derivative finite over the whole piece.

Integrate in x
y = x² from (0,0) to (2,4); x = √y from (0,0) to (2,4); y = x³ from (0,0) to (1,1)
Integrate in y
y = x^(2/3) from (0,0) to (8,4); x = y² from (0,0) to (4,2)
x
These already have bounded dy/dx on the whole piece: 2x, and 3x², and for x = √y the rewrite y = x² again gives 2x. Nothing forces a switch.
y
y = x^(2/3) has a vertical tangent at the origin, so dy/dx is unbounded there; rewritten as x = y^(3/2) the derivative is (3/2)√y, which is fine. x = y² is already a function of y with derivative 2y.

Sort these by looking for one thing only: does the curve ever stand up vertically over the piece in question? Everything else is secondary.

Two of them force a switch. The two thirds power has a vertical tangent at the origin, so integrating in the horizontal variable is improper there. The curve given as the square of the height is already a function of the height and would need a square root to be rewritten, so leave it as it is.

For the rest there is nothing to gain from switching, and switching would only add work. The decision is not about which variable looks nicer; it is about which one keeps the derivative finite over the whole piece.

37. Does switching change the answer?

Prediction

Predict first

A curve is described both as y = f(x) and as x = g(y). You compute the length both ways. What should you expect?

  • The same number, since it is the same set of points
  • A larger number in y, since dy is larger
  • Different numbers; the description matters
  • The same only for straight lines

Correct: The same number, since it is the same set of points

Why: Arc length is a property of the curve, not of how you describe it. The two integrals are related by the substitution y equal to f of x, which turns one into the other exactly. What changes is only how hard the integral is to evaluate, which is the whole reason for choosing.

Commit to an option first. The instinct that the two descriptions might give different lengths is common and worth confronting directly.

Arc length is a property of the set of points, not of the algebra you use to describe them. The two integrals are related by a substitution that turns one into the other exactly, so they are the same number written two ways.

What does change is the difficulty. One version can be improper while the other is routine, or one can have a linear radicand while the other has an ugly fractional-power one. That is all the choice is ever about, and Exercise 172 is a clean example of the difference.

38. Surface area of a solid of revolution

Section

Part 4

39. Revolve the chords, not the curve

Concept

Figure (svg): The curve y equals the square root of x from 1 to 4 with its mirror image below the horizontal axis; a narrow vertical slice between them is shaded, standing for one band of the surface, and an arrow shows the curve being swept around the axis.

One band: a circle of radius f(x) swept along a slanted piece of curve of length ds. Its area is the circumference times that slant.

Take the same chord approximation and spin it about the axis. Each chord sweeps a band, and each band is a slice of a cone with its tip cut off.

frustum — The piece of a cone left when a smaller cone is cut off the top by a plane parallel to the base.

Adding the bands' areas and letting the pieces shrink will give the surface area, exactly as adding the chords gave the length.

OpenStax Calculus Volume 1, §6.4 Arc Length of a Curve and Surface Area §6.4, pp. 586-587 — Figures 6.41 and 6.42

Here is the step that makes surface area possible: do not try to revolve the curve, revolve the chords. You already know the chords, and you know their lengths.

Each chord is a slanted straight segment, so spinning it about the axis sweeps out a cone with its tip cut off, which is called a frustum. The picture shows one such band standing on the shaded slice. Its radius is the height of the curve, and its width, measured on the surface, is the slant of the chord.

The plan from here mirrors Part 1 exactly. Find one band's area, add the bands, let the pieces shrink. All that is missing is a formula for the lateral area of a frustum, and the next two slides supply it.

40. A cone unrolled is a flat sector

Picture it

Figure (svg): Left: a cone standing on its base, with the base radius labelled r, the dashed vertical height labelled h and the sloping side labelled s. Right: the same cone cut open and laid flat, a sector of a circle of radius s whose curved edge has length two pi r.

Cut the cone along one slant line and flatten it: a sector of radius s whose rim is the base circle. Its area is pi times r times s.

Cut a cone up one slant line and lay it flat: you get a sector of a circle whose radius is the slant height and whose curved edge is the base circle. The sector's area is half the rim times the radius, which is where the cone formula comes from.

\[ \text{lateral area} = \tfrac12(2\pi r)(s) = \pi r s \]

This is the picture to remember whenever you doubt the cone formula. Cut the cone along one straight line from tip to base and flatten it. What you get is a sector of a circle whose radius is the slant height, because that is the distance from the tip to the rim.

The curved edge of the sector is the base circle of the cone, so its length is the circumference of the base. A sector's area is half its arc length times its radius, exactly as a triangle's is half its base times its height, and that gives the cone's lateral area as pi times the radius times the slant.

Notice which length appears and which does not. The vertical height never enters, because you never travel up the axis while walking on the surface. The whole of the next part of this lesson is an elaboration of that observation.

41. Deriving the frustum's lateral area

Worked example

A frustum is the difference of two cones. Let the radii be r one at the wide end and r two at the narrow end, with slant l between them, and let s be the full cone's slant.

Set up the similar triangles

Why: The small cone cut off and the whole cone have the same shape.

\[ \frac{r_2}{r_1} = \frac{s-l}{s} \]

Cross-multiply and collect the s terms

Why: Expand the right side and move the two s terms together.

\[ r_2 s = r_1 s - r_1 l \;\Longrightarrow\; r_1 l = (r_1-r_2)s \]

Solve for the full slant

Why: Divide by the difference of the radii.

\[ s = \frac{r_1 l}{r_1-r_2} \]

Write the frustum as a difference of cones

Why: Whole cone minus the tip, each by the cone formula.

\[ S = \pi r_1 s - \pi r_2 (s-l) \]

Substitute the slant and combine over one denominator

Why: Everything is a multiple of l over the difference of the radii.

\[ S = \frac{\pi l\left(r_1^{2}-r_2^{2}\right)}{r_1-r_2} \]

Factor the difference of squares

Why: The denominator cancels, leaving a sum.

\[ S = \pi (r_1+r_2)\,l \]

Figure (svg): A frustum of a cone drawn with an elliptical top of radius r two and an elliptical base of radius r one, its sloping side labelled l, and dashed lines continuing the sides up to the apex of the whole cone, whose slant from apex to base is labelled s.

A frustum is a whole cone minus the small cone at the tip, and its two slant lengths are what the similar triangles relate.

Check the two extreme cases

Why: Equal radii give a cylinder, whose side really is its circumference times its height; a zero top radius gives the cone formula back.

\[ r_1=r_2=r: \; S = 2\pi r l, \qquad r_2 = 0: \; S = \pi r_1 l \]

OpenStax Calculus Volume 1, §6.4 Arc Length of a Curve and Surface Area §6.4, pp. 587-588 — the frustum derivation

Follow this derivation once with a pencil, then you will never need to memorise the result. The idea is simply that a frustum is a big cone with a small cone removed, and the two cones are similar.

Similarity gives one equation relating the two radii and the two slant lengths, and solving it for the full slant is what lets you eliminate the unknown tip. After that the algebra is just putting two terms over a common denominator, and the difference of squares in the numerator cancels the denominator entirely.

The result is beautifully simple: pi times the sum of the two radii times the slant between them. Check it on both extreme cases as the last step does. Equal radii give the side of a cylinder, and a zero top radius gives the cone formula back. A formula that survives both limits is almost certainly right.

42. From one band to the integral

Concept

Apply the frustum formula to the band over one subinterval. Its slant is the chord you measured in Part 1, and its two radii are the curve's heights at the ends.

\[ S_i = \pi\left(f(x_{i-1})+f(x_i)\right)\,\Delta x\sqrt{1+\left[f'(x_i^{*})\right]^{2}} \]

The average of the two heights is itself a value of the function somewhere in the subinterval, by the Intermediate Value Theorem, so the sum of the bands is a Riemann sum after all, and its limit is the integral below.

\[ S_i = 2\pi f(x_i^{**})\,\Delta x\sqrt{1+\left[f'(x_i^{*})\right]^{2}} \]

\[ S = \int_a^{b}2\pi f(x)\sqrt{1+\left[f'(x)\right]^{2}}\,dx \]

OpenStax Calculus Volume 1, §6.4 Arc Length of a Curve and Surface Area §6.4, pp. 588-589 — Theorem 6.6 and equation 6.9

Apply the frustum formula to one band and look at what appears. The sum of the two radii is the sum of the function's values at the two ends of the subinterval, and the slant is the chord you measured in Part 1.

There is one delicate point, and the book is honest about it. The expression involves the function at two different sample points, the average height at one and the slope at another, so it is not literally a Riemann sum. Because the function is continuous, the average of the two end heights is itself a value of the function somewhere in the subinterval, and as the pieces shrink the two sample points are squeezed together.

What comes out is a clean statement worth memorising in words rather than symbols: the surface integrand is the circumference of the circle a point traces, multiplied by the arc length integrand. Circumference times slant width, integrated.

43. Reading Theorem 6.6 piece by piece

Notation

Annotate

On: \( S = \int_a^{b} 2\pi f(x)\sqrt{1+\left[f'(x)\right]^{2}}\,dx \)

  • The circumference of the circle one point of the curve traces. The function must be nonnegative here: it is a radius.
  • The whole arc length integrand, unchanged. The band's width along the surface is ds, never dx.
  • Circumference times slant length: the area of a thin band, the same 'unroll it flat' idea as the cone.
  • Same formula with x and y exchanged: the radius becomes g(y) and the differential dy.

Notice first what is already familiar. The square root is the entire arc length integrand from Part 1, unchanged. All that is new is the factor in front of it.

That factor is a circumference, which is why the function must be nonnegative: it is being used as a radius. If a curve dips below the axis you either reflect it or split the integral, because a negative radius is meaningless.

The last annotation gives you the second form for free. Revolving about the vertical axis with the curve written as a function of the height changes nothing structurally; the radius becomes the horizontal coordinate and the differential becomes the step in height.

44. Example 6.21: a square root revolved about the x-axis

Worked example

\[ f(x) = \sqrt{x} \quad\text{on}\quad [1,4], \quad\text{about the } x\text{-axis} \]

Differentiate and square

Why: The derivative of the square root is one over twice the square root.

\[ f'(x) = \frac{1}{2\sqrt{x}}, \qquad \left[f'(x)\right]^{2} = \frac{1}{4x} \]

Write the surface integral

Why: Radius times the arc length factor, times two pi.

\[ S = \int_1^{4}2\pi\sqrt{x}\sqrt{1+\frac{1}{4x}}\,dx \]

Multiply the two roots into one

Why: The square root of x times the root of one plus one over four x is the root of the product.

\[ S = \int_1^{4}2\pi\sqrt{x+\frac14}\,dx \]

Substitute

Why: With u equal to x plus a quarter the differential is unchanged.

\[ u = x+\tfrac14, \quad u: \tfrac54 \to \tfrac{17}{4} \]

Integrate and evaluate

Why: The power rule with exponent one half, then clear the quarters.

\[ S = 2\pi\left[\tfrac23u^{3/2}\right]_{5/4}^{17/4} = \frac{\pi}{6}\left[17\sqrt{17}-5\sqrt5\right] \approx 30.846 \]

Check against two cylinders

Why: The curve's radius runs from 1 to 2 over a length of 3, so the surface must lie between the side of a cylinder of radius one and that of radius two: between 18.85 and 37.70. It does.

\[ 2\pi(1)(3) = 18.85 < 30.846 < 37.70 = 2\pi(2)(3) \]

OpenStax Calculus Volume 1, §6.4 Arc Length of a Curve and Surface Area §6.4, pp. 589-590 — Example 6.21

This is the surface example worth knowing by heart, because it shows the one simplification that makes surface integrals tractable. The radius is a square root and the arc length factor is another square root, and the product of two square roots is a single square root of the product.

Watch that merging step closely. The square root of x times the square root of one plus one over four x gives the square root of x plus one quarter, and suddenly the integral is an ordinary power rule with a shift. Without that merger you would be facing a product of a root and a function, which is the usual hopeless case.

The check uses two cylinders. The radius of the surface runs from one to two over a horizontal length of three, so the answer must lie between the side area of the thin cylinder and that of the fat one, which is between about 18.85 and 37.70. The value 30.85 sits inside, and closer to the top, which fits a radius that spends most of the interval above 1.5.

45. What multiplies the arc length factor?

Fill the middle

The surface integral for the previous example, with the radius removed.

Fill in the blanks

S = \int_1^\sqrt{x} 2\pi\,___\,\sqrt______}\,dx

Why: The missing factor is the square root of x, the function itself: the radius of the circle each point sweeps. Two pi times it is the circumference, and the root factor is the slant width of the band.

Fill the blank before revealing, and say the answer as a phrase rather than as a formula: the distance from the axis to the curve.

Here that distance is the function itself, the square root of x, because the axis of revolution is the horizontal one and the curve's height is its distance from that axis. Two pi times it is the circumference of the circle the point traces as the curve spins.

If you wrote the arc length factor again, or the differential, look back at the band in the earlier picture. One factor is around the band and one is across it, and the area of the band is their product.

46. Checkpoint 6.21: a shifted square root

Worked example

\[ f(x) = \sqrt{1-x} \quad\text{on}\quad [0,\tfrac12], \quad\text{about the } x\text{-axis} \]

Differentiate and square

Why: The chain rule contributes the minus one, which the square removes.

\[ f'(x) = \frac{-1}{2\sqrt{1-x}}, \qquad \left[f'(x)\right]^{2} = \frac{1}{4(1-x)} \]

Write the integral and merge the roots

Why: Radius times factor again collapses to a single root.

\[ S = \int_0^{1/2}2\pi\sqrt{(1-x)+\tfrac14}\,dx = \int_0^{1/2}2\pi\sqrt{\tfrac54-x}\,dx \]

Integrate

Why: The inside decreases, so the antiderivative carries a minus sign.

\[ S = 2\pi\left[-\tfrac23\left(\tfrac54-x\right)^{3/2}\right]_0^{1/2} \]

Evaluate

Why: Five quarters to the three halves less three quarters to the three halves.

\[ S = \frac{4\pi}{3}\left[\left(\tfrac54\right)^{3/2}-\left(\tfrac34\right)^{3/2}\right] \approx 3.133 \]

Check against a cylinder

Why: The radius runs from 1 down to about 0.707 over a length of one half, so a cylinder of the average radius, about 0.854, has side area 2.68, and the extra slant pushes the true value a little above it.

\[ 2\pi(0.854)\left(\tfrac12\right) \approx 2.68 < 3.133 \]

OpenStax Calculus Volume 1, §6.4 Arc Length of a Curve and Surface Area §6.4, p. 590 — Checkpoint 6.21

Do this one before reading the solution. It is Example 6.21 with the curve reflected and shifted, so the same merging trick has to work.

The chain rule produces a minus sign in the derivative, and the squaring removes it, as always. After that the radius and the arc length factor merge into a single square root again, this time of five quarters minus x, and the integral is a power rule with a minus sign from the chain rule going the other way.

The check compares the answer with a cylinder of the average radius. The curve's radius runs from one down to about 0.707 over a horizontal length of one half, and a cylinder of the average radius has side area about 2.68. The true value 3.133 is larger, which is right, because the band widths are slants and slants are longer than horizontal steps.

47. The radius is the distance from the axis

Concept

The factor two pi multiplies a radius, and a radius is measured from the axis of revolution to the point. Which formula you write depends on the axis, not on which letter happens to name the function.

\[ \text{about } x\text{-axis, in } x: \; r = f(x) \qquad \text{about } y\text{-axis, in } y: \; r = g(y) \]

\[ \text{about } y\text{-axis, but integrating in } x: \; r = x \]

The third case is the one that catches people, and the lamp-base exercise later in this lesson is exactly it: the curve is revolved about the vertical axis while the natural variable is still the horizontal one, so the radius is the horizontal coordinate itself.

OpenStax Calculus Volume 1, §6.4 Arc Length of a Curve and Surface Area §6.4, p. 589 — the two forms of Theorem 6.6

This slide fixes the one thing people most often get wrong once both forms of the theorem are on the table. The factor two pi multiplies a radius, and a radius always means the distance from the axis of revolution.

In the first two cases the radius is the function, because the curve's output is measured perpendicular to the axis. The third case is the awkward one: the axis is vertical but you are integrating in the horizontal variable, and then the radius is the horizontal coordinate itself, not the function's value.

You will meet that third case in the lamp-base exercise later in this lesson. Whenever you set up a surface integral, ask the question in words first: how far is this point from the axis? Then write that distance, whatever letter it happens to involve.

48. Example 6.22: revolving about the y-axis

Worked example

The interval is given in the vertical variable, so the whole problem must be written in that variable.

\[ y = \sqrt[3]{3x}, \quad 0 \le y \le 2, \quad\text{about the } y\text{-axis} \]

Solve for x

Why: Cube both sides and divide by three.

\[ x = g(y) = \frac{y^{3}}{3} \]

Differentiate with respect to y and square

Why: The power rule on the cube.

\[ g'(y) = y^{2}, \qquad \left[g'(y)\right]^{2} = y^{4} \]

Write the surface integral

Why: The radius is now the horizontal distance g(y).

\[ S = \int_0^{2}2\pi\frac{y^{3}}{3}\sqrt{1+y^{4}}\,dy \]

Substitute

Why: Let u be one plus y to the fourth; its differential supplies the cube.

\[ u = 1+y^{4}, \quad du = 4y^{3}dy, \quad u: 1 \to 17 \]

Integrate and evaluate

Why: The two thirds from the power rule and the twelfth from the constants combine.

\[ S = \frac{2\pi}{3}\cdot\frac14\cdot\frac23\left[u^{3/2}\right]_1^{17} = \frac{\pi}{9}\left[17^{3/2}-1\right] \approx 24.118 \]

Check against a cone

Why: The profile runs from the axis out to a radius of eight thirds at the top, so a straight cone through those two points has lateral area about 27.9; the real profile bulges inward, so a smaller answer is right.

\[ \pi\left(\tfrac83\right)\sqrt{\left(\tfrac83\right)^{2}+2^{2}} \approx 27.9 > 24.118 \]

OpenStax Calculus Volume 1, §6.4 Arc Length of a Curve and Surface Area §6.4, pp. 590-591 — Example 6.22

Read the problem statement carefully before doing anything. The interval is given in terms of the height, and the axis is the vertical one, so every part of the set-up has to be written in the height variable.

Solving the cube root for the horizontal coordinate is the first line, and it turns the awkward root into a plain cube, which is much easier to differentiate. From there the pattern is exactly as before: square the derivative, add one, multiply by two pi times the radius, integrate.

The substitution works because the derivative of one plus the fourth power supplies the cube that is already sitting in front. The check compares the answer with the cone through the same two endpoints, about 27.9, and the true surface is smaller because the profile bulges inward rather than running straight.

49. Checkpoint 6.22: a zone of a sphere

Worked example

\[ g(y) = \sqrt{9-y^{2}}, \quad y \in [0,2], \quad\text{about the } y\text{-axis} \]

Differentiate and square

Why: The chain rule on the square root.

\[ g'(y) = \frac{-y}{\sqrt{9-y^{2}}}, \qquad \left[g'(y)\right]^{2} = \frac{y^{2}}{9-y^{2}} \]

Add one over a common denominator

Why: The y squared terms cancel in the numerator.

\[ 1+\frac{y^{2}}{9-y^{2}} = \frac{9}{9-y^{2}} \]

Multiply by the radius

Why: The two square roots cancel exactly, leaving a constant.

\[ g(y)\sqrt{1+\left[g'(y)\right]^{2}} = \sqrt{9-y^{2}}\cdot\frac{3}{\sqrt{9-y^{2}}} = 3 \]

Integrate the constant

Why: The hard part vanished; what is left is the area of a rectangle.

\[ S = \int_0^{2}2\pi(3)\,dy = 12\pi \approx 37.699 \]

Check against the sphere's total area

Why: This is a band of a sphere of radius 3, whose whole surface is four pi times nine, that is 36 pi. The band spans two of the six units of height, and two sixths of 36 pi is 12 pi: Archimedes' result that equal heights give equal areas.

\[ \frac{2}{6}\left(36\pi\right) = 12\pi \]

OpenStax Calculus Volume 1, §6.4 Arc Length of a Curve and Surface Area §6.4, p. 591 — Checkpoint 6.22

This is the prettiest calculation in the section, so do not rush the middle step. The curve is a quarter of a circle of radius three, and revolving it about the vertical axis gives a band of a sphere.

When you add one to the squared derivative over a common denominator, the numerator collapses to nine, so the arc length factor is three divided by the same square root that is the radius. Multiplying radius by factor cancels the roots completely and leaves the constant three. The integral of a constant is trivial, and the answer is twelve pi.

The check is Archimedes' theorem, proved in the third century before the common era and still surprising. The area of a band of a sphere depends only on its height, not on where it sits, so a band two units tall on a sphere of radius three always has area twelve pi, whether it is at the equator or near the pole.

50. Equal heights, equal areas

Picture it

Figure (svg): A quarter circle of radius 3 running from the point three comma zero up to the point zero comma three, with dashed horizontal lines at heights 1 and 2; the arc between heights 0 and 1 and the arc between heights 2 and 3 are drawn in bold, and both are marked as zones of the same swept area.

Both bold arcs span one unit of height, and both sweep out exactly the same area when the quarter circle is revolved about the vertical axis: six pi.

The narrow arc low down is short but sweeps a wide circle; the long arc near the top sweeps a small one. The two effects cancel exactly, and every band of the same height on a sphere has the same area.

Look at the two bold arcs. The one near the bottom is short, because the circle is nearly vertical there, but it sweeps a wide circle. The one near the top is long, because the circle is nearly horizontal, but it sweeps a small one.

The two effects cancel exactly, which is what the algebra on the previous slide showed when the radius and the arc length factor multiplied to a constant. Both bands have area six pi.

This is worth remembering as a sanity check and as a fact in its own right. It is also why maps that preserve area, such as the Lambert cylindrical projection, work by projecting each point of the globe straight out onto a surrounding cylinder.

51. Three ways a surface integral turns out to be doable

Intuition

A surface integrand is an arc length integrand with a function stuck in front of it, so it is harder again. The examples that work are not random: each one is a case where the extra factor helps rather than hurts.

Away from those three patterns, expect to set the integral up and hand it to a machine, which is exactly what several exercises in this section ask you to do.

By now you have seen several surface integrals succeed, and it is worth collecting the reasons, because they are the only reasons you will meet in this course.

The first is the merging of two square roots, which happens when the profile is itself a square root, as in Example 6.21. The second is a constant arc length factor, which happens exactly when the profile is straight, so every cone and every cylinder is easy. The third is cancellation, which happens when the profile is circular and the same expression appears above and below.

Outside those three patterns, expect the integral to need a machine. That is not a failure of method; several of this section's exercises say so explicitly. What is graded is whether your radius and your arc length factor are right.

52. What each factor is doing

Matching

Match the pairs

  • a. 2π f(x)
  • b. √(1 + [f′(x)]²) dx
  • c. f(x) dx
  • d. π [f(x)]² dx
  • w. the circumference of the circle one point traces
  • x. the slant width of one band
  • y. the area of a thin strip under the curve
  • z. the volume of a thin disc

Why: The first two multiply to give a band's area, and that product is the surface integrand. The last two are the integrands of Sections 6.1 and 6.2, neither of which carries a square root, because neither measures anything along the curve.

Match these before revealing, and read each expression as a physical quantity rather than as algebra.

The first two are the two directions of a band: around it and across it. Their product is the band's area, which is the surface integrand. Saying this out loud once is usually enough to stop you from ever writing the surface integral without the root.

The last two are the integrands of Sections 6.1 and 6.2, and neither carries a root. That is the pattern of the whole chapter: quantities measured across the curve use the horizontal step, and quantities measured along it use the slant.

53. Trap: the arc length factor dropped

Trap

The trap

A surface computed with the width of the strip instead of its slant.

\[ S = \int_1^{4}2\pi\sqrt{x}\,dx \]

\[ S = \frac{28\pi}{3} \approx 29.32 \]

The fix

The band is slanted, so its width along the surface is the arc length differential, never the horizontal step.

\[ S = \int_1^{4}2\pi\sqrt{x}\sqrt{1+\tfrac{1}{4x}}\,dx \]

\[ S \approx 30.85 \]

The missing factor is at least one, so this error always undercuts: about five per cent here.

This is the most common error in the second half of the section, and it is quiet, because the wrong answer looks entirely reasonable.

The mistake is to treat the band as if it were a cylinder of width equal to the horizontal step. It is not: it is slanted, so its width measured on the surface is the arc length differential. Leaving out the factor replaces a hypotenuse by a leg.

Because the missing factor is never less than one, this error always undercuts the true area. Here it loses about five per cent, but on a steep profile it can lose most of the surface. If you ever produce a surface area with no square root anywhere in the working, check whether the profile really was a straight vertical line.

54. Four integrands, two of them with roots

Comparison

Comparison matrix

quantityintegrandcarries a root?
area under a curvef(x)no
volume by discsπ[f(x)]²no
arc length√(1 + [f′(x)]²)yes
surface of revolution2πf(x)√(1 + [f′(x)]²)yes

Anything measured along the curve inherits the square root; anything measured across it does not. That single observation predicts which of these problems will be elementary before you set any of them up.

Fill the blanks, then read the table as a single statement about this chapter rather than as four separate formulas.

The top two rows measure things across the curve. A strip's width and a disc's thickness are both the horizontal step, and the curve's steepness does not affect them, so no root appears and the integrands are as easy as the function is.

The bottom two rows measure things along the curve, so both carry the slant factor and both inherit all of its difficulty. This is why a solid can have an elementary volume and a surface area that no one can write down, and it lets you predict the difficulty of a problem before you set it up.

55. Why the slant height

Section

Part 5

56. A surface spans along the boundary, not across it

Concept

Figure (svg): A cone of base radius 3 and height 4 drawn in cross-section, with the right triangle inside it: horizontal leg 3, vertical leg 4, and hypotenuse 5 marked along the sloping side.

The surface runs along the hypotenuse, never up the axis, so the slant height is the length that belongs in the formula.

The band's width is measured on the surface itself, along the sloping side, and that side is the hypotenuse of a triangle whose legs are the horizontal step and the rise.

\[ s = \sqrt{r^{2}+h^{2}} \;\ge\; h \]

Using the vertical height instead is the same mistake as dropping the root from the integrand: both replace a hypotenuse by a leg, and both always give an answer that is too small.

The triangle in the picture is the whole idea. Walk from the rim of the cone up to the tip along the outside, and you travel the hypotenuse, five units here, not the vertical leg of four.

That is the same statement as the one about band widths, dressed in geometry rather than in calculus. The surface lives along the sloping side, so the length that belongs in any surface formula is the slant.

Notice also that the slant is always at least as long as the height, with equality only when the radius is zero. So mistaking one for the other can only ever lose area, which matches what the trap slide showed about the missing root factor.

57. The cone's lateral surface, from the integral

Worked example

Derive the formula used in the frustum argument, this time from Theorem 6.6.

\[ y = \frac{R}{h}x \quad\text{on}\quad [0,h], \quad\text{about the } x\text{-axis} \]

Differentiate and square

Why: The line has a constant slope.

\[ f'(x) = \frac{R}{h}, \qquad \left[f'(x)\right]^{2} = \frac{R^{2}}{h^{2}} \]

Write the integral

Why: The root is a constant and comes outside.

\[ S = 2\pi\sqrt{1+\frac{R^{2}}{h^{2}}}\int_0^{h}\frac{R}{h}x\,dx \]

Integrate the linear part

Why: Half the square of the upper limit, times the slope.

\[ \int_0^{h}\frac{R}{h}x\,dx = \frac{R h}{2} \]

Simplify the constant root

Why: Put the two terms over h squared and take the root of the top.

\[ \sqrt{1+\frac{R^{2}}{h^{2}}} = \frac{\sqrt{h^{2}+R^{2}}}{h} = \frac{s}{h} \]

Combine

Why: The h in the denominator cancels one of the h's from the integral.

\[ S = 2\pi\cdot\frac{s}{h}\cdot\frac{Rh}{2} = \pi R s \]

Check against the unrolled sector

Why: Flattening the cone gave a sector of radius s whose rim is two pi R, with area half the rim times the radius: the same pi R s, from pure geometry.

\[ \tfrac12(2\pi R)(s) = \pi R s \]

OpenStax Calculus Volume 1, §6.4 Arc Length of a Curve and Surface Area §6.4, p. 587 — the lateral surface area of a cone

This example closes a loop. The cone formula was used to derive the frustum formula, the frustum formula was used to derive the surface integral, and now the surface integral reproduces the cone formula. Nothing circular has happened, because the cone formula came from elementary geometry, but it is a strong consistency check on the whole chain.

The calculation itself is easy for the reason given a few slides back: the profile is straight, so the arc length factor is a constant and comes straight out of the integral. What is left is the integral of a linear function.

Watch the simplification of the constant root, since it is where the slant appears. Putting the two terms over the square of the height and taking the root of the numerator gives the slant divided by the height, and the height then cancels. That cancellation is the algebraic version of the unrolled sector.

58. What the wrong height costs: a 3-4-5 cone

Worked example

A cone of base radius 3 and height 4. Compute the lateral surface correctly, then with the vertical height in place of the slant.

Find the slant

Why: The classic right triangle.

\[ s = \sqrt{3^{2}+4^{2}} = 5 \]

Compute the correct area

Why: Pi times radius times slant.

\[ S = \pi(3)(5) = 15\pi \approx 47.12 \]

Compute the wrong one

Why: The same formula with the height substituted for the slant.

\[ S_{\text{wrong}} = \pi(3)(4) = 12\pi \approx 37.70 \]

Express the gap

Why: The ratio of the two is the ratio of height to slant.

\[ \frac{12\pi}{15\pi} = \frac45 = 0.8 \]

Check the limiting shapes

Why: For a very tall thin cone the height and slant are nearly equal and the error vanishes; for a very flat one the slant is much longer and the error is enormous, as it must be, since a flat cone is almost a disc.

\[ R \ll h: \; \frac{h}{s}\to 1, \qquad R \gg h: \; \frac{h}{s}\to 0 \]

Numbers make an abstract warning concrete. With radius three and height four, the slant is five, so the correct lateral surface is fifteen pi and the version using the height is twelve pi.

That is a twenty per cent loss, and it is no accident that the ratio of the two answers is exactly the ratio of the height to the slant. The error scales with how far the cone is from being a cylinder.

The final step is the one to remember for estimating. A tall thin cone has a slant barely longer than its height, so the mistake hardly shows. A wide flat cone has a slant far longer than its height, and the mistake becomes enormous, which makes sense, because a very flat cone is almost a disc and has plenty of surface but almost no height at all.

59. Trap: the vertical height used as the slant

Trap

The trap

Surface of a cone, radius 3, height 4:

\[ S = \pi r h = 12\pi \]

Wrong. The number 4 measures the axis, not the surface.

The fix

Walk up the outside of the cone and you travel 5, not 4. A cylinder is the one shape where the two agree, which is why cylinders are a poor place to learn the difference.

\[ S = \pi r s = 15\pi \]

This is the same error as the missing root factor, wearing different clothes. Both replace a distance measured along the surface with a distance measured across it.

The reason it survives so long is that the first surface most people meet is a cylinder, where the slant and the height are the same number. A cylinder can never expose the mistake, which is exactly why it is the wrong shape to test yourself on.

Use a cone instead. You know its lateral area independently, from the unrolled sector, so it will tell you immediately whether the formula you have written is the right one.

60. Exercise 207: the base of a lamp

Real world

Figure (svg): Two quarter circles facing each other, one from the point one comma one down to two comma zero and its mirror image on the left, forming the profile of a low dome; a dashed vertical line marks the axis they are revolved about.

The lamp's profile is a quarter circle standing a unit away from the axis, so every point of it sweeps a circle of a different radius.

Discussion prompt

A lamp base is made by revolving the quarter circle y = √(2x − x²) about the vertical axis, for x from 1 to 2. Set up the surface integral and evaluate it. Watch what the radius is here.

OpenStax Calculus Volume 1, §6.4 Arc Length of a Curve and Surface Area §6.4, p. 593 — Exercise 207

Set this up yourself before revealing, and give the radius question a full ten seconds of thought, because it is the entire point of the exercise.

The profile is a quarter circle whose centre is at one, so it never touches the axis of revolution: the lamp base is a dome with a flat top and a wide skirt. Because the axis is vertical while the convenient variable is horizontal, the radius of the circle each point sweeps is the horizontal coordinate itself, not the height of the curve.

The arc length factor simplifies beautifully. Adding one to the squared derivative over a common denominator gives one over the same expression under the root, so the whole integrand becomes the horizontal coordinate divided by the root, and the substitution that moves the centre of the circle to zero turns it into the integral of one plus a sine. The answer, about 16.15, is the area of material the lamp shade needs.

61. How much surface does a cylinder miss?

Estimation

Predict first

The curve y = √x on [1,4] revolved about the x-axis gave 30.85. A cylinder of the average radius 1.5 over the same length has side area about 28.3. Why is the true surface larger?

  • Because the curve is longer than the interval
  • Because the average radius is too small
  • Because two pi was left out
  • It is not: 28.3 is the correct value

Correct: Because the curve is longer than the interval

Why: The cylinder's side is circumference times the horizontal length 3. The real band widths add up to the arc length of the curve, which is about 3.17, longer than 3. Averaging the radius is a reasonable approximation; replacing slant by horizontal width is the part that always loses area.

Commit to an option before reasoning. The comparison is a useful one because the cylinder gets the radius roughly right and still comes out low.

The cylinder's side is its circumference times the horizontal length three. The real surface is built from bands whose widths add up to the arc length of the curve, which is about 3.17, so the cylinder is short in the one direction that the square root controls.

The lesson generalises. Averaging the radius is usually a decent approximation and can err in either direction, but replacing the slant by the horizontal step is a systematic error in one direction only, and it is the one that costs you area every time.

62. When do the two heights agree?

Counterexample

Discussion prompt

Name the one shape for which using the vertical height in place of the slant height gives the right answer, and say what that tells you about testing your own formulas.

Answer before revealing, and then take the second half of the question seriously, because it is a lesson about testing formulas in general.

A cylinder is the only shape where the vertical height and the slant height coincide, and that is because its profile is horizontal, so the derivative is zero and the arc length factor is exactly one. Any test case with a zero derivative is blind to a missing derivative term.

This is a habit worth carrying well beyond this section. When you check a formula, choose a case where every term you are unsure of actually contributes something. Here a cone is the right test, because you know the answer independently and the slant genuinely differs from the height.

63. Putting it together

Section

Part 6

64. Pattern: length or surface, start to finish

Pattern

Figure (svg): A flow diagram: decide whether length or surface is wanted, choose the variable whose derivative stays finite, build the integrand, then either integrate exactly or go numerical.

Two decisions come before any integration: which quantity, and which variable.
  1. Decide what is being measured. Length along the curve, or the area of the surface it sweeps.
  2. Choose the variable so the derivative stays finite: no vertical tangents in the variable you integrate in.
  3. Differentiate, square, add one, take the root. That is the arc length factor, and it appears in both formulas.
  4. For a surface, multiply by two pi times the radius, which is the distance from the axis, not always the function's value.
  5. Look at the radicand before integrating. Linear or a perfect square means you can finish by hand; otherwise set it up and go numerical.
  6. Check every answer against the chord below it, or a cylinder or cone around it.

This is the order to work in, and the first two steps are decisions, not calculations. Deciding what is measured and in which variable fixes everything that follows, and both are cheap to get right and expensive to get wrong.

The middle of the procedure never changes: differentiate, square, add one, take the root. Say those four words as you write them and the one will never go missing. A surface then attaches two pi times the distance from the axis, and that distance is a question about geometry, not about which letter the function has.

The last two steps are the ones students skip. Looking at the radicand before integrating tells you within seconds whether this is a by-hand problem or a numerical one. Checking the answer against a chord, a cylinder or a cone catches almost every mistake this lesson can produce.

65. Put a surface area calculation in order

Ranking

Put in order

Order the steps for the surface generated by revolving a curve about an axis.

  1. Write the curve in the variable you will integrate in
  2. Differentiate in that variable and square the derivative
  3. Add one and take the square root: the arc length factor
  4. Multiply by two pi times the distance from the axis
  5. Integrate between the limits, in that same variable
  6. Check the answer against a cylinder or cone

Why: The choice of variable comes first because it fixes what the derivative, the differential and the limits all mean. The radius is attached after the arc length factor is built, and the check is not optional: it catches a dropped root immediately.

Drag the steps into order before checking. The one genuinely debatable position is the radius, and it goes after the arc length factor because the factor is built entirely out of the derivative.

The choice of variable has to be first, because it decides what the derivative means, what the differential is, and what the limits count. Every step after it inherits that choice, and switching halfway is the error you saw a few slides ago.

The check is on the list deliberately. A dropped root factor or a radius measured from the wrong axis both produce plausible-looking numbers, and the only cheap way to catch them is to compare with a cylinder or a cone you can compute in your head.

66. Check: the arc length integrand

Check

Check your understanding

Which integral gives the length of y = x³ on the interval from 0 to 2?

  • A. the integral from 0 to 2 of √(1 + 9x⁴) dx (correct)
  • B. the integral from 0 to 2 of √(1 + x⁶) dx
  • C. the integral from 0 to 2 of √(1 + 3x²) dx
  • D. the integral from 0 to 2 of 3x² dx

Answer: A

Why: The derivative is 3x², and the formula squares the derivative: (3x²)² is 9x⁴. Add one, take the root, integrate in x.

Why B tempts people
That squares the function instead of the derivative; x³ squared is x⁶.
Why C tempts people
The derivative was inserted without being squared.
Why D tempts people
That is the integral of the derivative alone, which measures total rise, not distance travelled.

This question tests one thing: whether what gets squared is the derivative or the function. The derivative of the cube is three x squared, and squaring that gives nine x to the fourth.

The choice with x to the sixth squares the function instead, which is the disc-volume pattern. The choice with three x squared forgot to square at all, and the last one integrates the derivative alone, which measures total rise rather than distance travelled.

67. Check: choosing the variable

Check

Check your understanding

For which curve is it clearly better to integrate with respect to y?

  • A. y = x² from (0,0) to (2,4)
  • B. y = x^(1/3) from (0,0) to (8,2) (correct)
  • C. y = 3x + 1 from (0,1) to (5,16)
  • D. y = e^x from (0,1) to (1,e)

Answer: B

Why: The cube root has a vertical tangent at the origin, so dy/dx is unbounded there and the integral in x is improper. Rewritten as x = y³ the derivative is 3y², which is finite on the whole piece.

Why A tempts people
Its derivative 2x is finite everywhere on the piece, so there is nothing to gain.
Why C tempts people
A line: the integrand is a constant in either variable.
Why D tempts people
The exponential is smooth with a bounded derivative on this interval; switching gives a logarithm and no simplification.

Scan the four options for a vertical tangent, and ignore everything else. Only the cube root has one, at the origin, where its slope is unbounded.

Rewritten with the height as the input, that curve becomes the cube of the height, whose derivative is a perfectly ordinary quadratic. The improper integral has become a routine one, and that is the whole reason for the second form of the theorem.

The other three are smooth with bounded derivatives on the given pieces, so switching would add work and remove nothing.

68. Check: the surface integrand

Check

Check your understanding

The curve x = g(y) on the interval from c to d is revolved about the y-axis. Which integral gives the surface area?

  • A. the integral from c to d of 2π g(y) √(1 + [g′(y)]²) dy (correct)
  • B. the integral from c to d of 2π y √(1 + [g′(y)]²) dy
  • C. the integral from c to d of π [g(y)]² dy
  • D. the integral from c to d of 2π g(y) dy

Answer: A

Why: The radius of the circle a point sweeps is its distance from the y-axis, which is g(y), and the band's width is the arc length differential in y.

Why B tempts people
That uses the height as the radius; the distance from the y-axis is the horizontal coordinate.
Why C tempts people
That is the volume by discs, which measures across the curve, not along it.
Why D tempts people
The arc length factor is missing, so this measures band widths horizontally and always comes out too small.

The axis here is the vertical one, so ask the radius question in words: how far is a point of the curve from the vertical axis? It is the horizontal coordinate, which for this curve is the value of the function.

The option using the height as the radius measures distance from the wrong axis. The option with the square and a single pi is the disc-volume integrand from Section 6.2. The option without the root drops the slant width, which always undercuts.

69. Explain the square root

Explain it to yourself

Discussion prompt

In two sentences, explain to someone who has just finished Section 6.2 why the arc length and surface formulas carry a square root when the area and volume formulas do not.

Write your two sentences before revealing. Being able to say this cleanly is a good test of whether the section has landed, because it is the one idea that connects all four formulas of the chapter.

The distinction to make is between across and along. Strips and discs are measured across the curve, so their width is the step in the variable and the steepness is irrelevant. Lengths and bands are measured along the curve, so their width is the hypotenuse of the little triangle, which is the step multiplied by the stretch factor.

If your explanation mentions the Pythagorean theorem and the word hypotenuse, you have it. If it only says that the formula has a root in it, go back to the differential triangle and try again.

70. Exit ticket

Exit ticket

\[ y = \frac{x^{2}}{2}-\frac{\ln x}{4} \quad\text{on}\quad [1,2] \]

Discussion prompt

Square the derivative and add one. Show that this radicand is a perfect square, then finish the arc length, and say in one line what the surface integral about the x-axis would need in addition.

This one problem uses the main skills of the lesson: spotting a perfect square, finishing an elementary arc length, and thinking one step ahead about the surface version.

Square the derivative and look for the constant cross term. Here the two terms multiply to one quarter, so twice their product is one half with a minus sign, and adding one turns it into plus one half. That is the signature of a backwards-built curve, and the root disappears.

The integral that remains is a power plus a reciprocal, giving three halves plus a quarter of the natural logarithm of two, about 1.673. Check it against the chord if you like. The last part of the question is the important one: the surface integral about the horizontal axis multiplies that same bracket by the function, which drags a logarithm into a product with a rational function and puts the integral out of reach by hand.

71. Recap

Recap

quantityintegrandwhen it is elementary
length, in x√(1 + [f′(x)]²)the radicand is linear or a perfect square
length, in y√(1 + [g′(y)]²)same test, after solving the curve for x
surface about the x-axis2πf(x)√(1 + [f′(x)]²)when the radius merges with the root
surface about the y-axis2πg(y)√(1 + [g′(y)]²)same, with the roles exchanged

\[ ds = \sqrt{1+\left[f'(x)\right]^{2}}\,dx = \sqrt{1+\left[g'(y)\right]^{2}}\,dy \]

One triangle produced all four rows. The next section leaves geometry for physics, applying the same slice-and-integrate discipline to work, force and pressure.

Stewart, Calculus: Early Transcendentals 8e, §8.1 Arc Length §8.1, pp. 544-550 — arc length in Stewart, with the same derivation

Four formulas, one triangle. Every row of the table is the hypotenuse of the same little triangle, described in a different variable or multiplied by a circumference. If you remember the triangle, you can rebuild any row of the table on the spot.

The third column is the practical one. An arc length is elementary when the radicand is linear or a perfect square, and a surface is elementary when the radius merges with the root, when the profile is straight, or when the roots cancel. Outside those cases, a correct set-up and a numerical value is the complete answer.

Two habits are worth carrying forward. Choose the variable that keeps the derivative finite, and check every answer against something you know: a chord below a length, a cylinder or a cone around a surface. Section 6.5 keeps the same slice-and-integrate discipline but turns to physics, where the piece being measured is work rather than distance.

Sources

  1. OpenStax Calculus Volume 1, §6.4 Arc Length of a Curve and Surface Area — Strang & Herman, OpenStax / Rice University, CC BY-NC-SA 4.0, pp. 582-593
  2. Stewart, Calculus: Early Transcendentals 8e, §8.1 Arc Length — James Stewart, Cengage Learning, 2016, pp. 544-550
  3. Stewart, Calculus: Early Transcendentals 8e, §8.2 Area of a Surface of Revolution — James Stewart, Cengage Learning, 2016, pp. 551-557

Want this taught 1-on-1? Alexander tutors Calculus I — $55/session, free consultation.

Book on Wyzant · Text (657) 465-8108