5.1 Approximating Areas

Sigma notation and the four summation formulas, approximating the area under a curve with rectangles, the left, right and midpoint rules, upper and lower sums bracketing the answer, and the limit of Riemann sums that defines area exactly.

Subject: Calculus I · 73 slides · symbolic lesson

Open the interactive version of this deck

What this lesson covers

The lesson, slide by slide

1. Approximating Areas

Title

Calculus I · Section 5.1

Rectangles, sums, and the limit that makes them exact

2. What this lesson gives you

Objectives

Derivatives measured rates. This chapter turns to the other great question of calculus: how much is accumulated, which begins as the area under a curve. There is no geometry formula for a curved top, so you will build the area out of rectangles and let their number grow.

OpenStax Calculus Volume 1, §5.1 Approximating Areas §5.1, p. 438 — learning objectives 5.1.1 to 5.1.3

For two chapters you have been asking how fast something changes. This chapter asks the reverse question: if you know how fast something changes, how much of it piles up? Distance from speed, water from a flow rate, total cost from a marginal cost. Every one of those turns out to be an area under a curve.

The trouble is that geometry only measures regions with straight sides or circular arcs. So this lesson does what Archimedes did two thousand years ago: it fills the region with rectangles, adds up their areas, and then asks what happens as the rectangles get thinner and more numerous.

Adding up many rectangles means writing long sums, so the lesson starts with sigma notation and a few formulas that collapse those sums. By the end you will compute an area exactly, as a limit, with nothing but algebra.

3. Before anything new: areas you already know

Warm-up

Discussion prompt

Find the area under the line y = x between x = 0 and x = 4. Then say why the same question for y = x² cannot be answered the same way.

Write your answer down before you reveal it. The first part is a triangle, and you have known its area formula for years: half the base times the height, which gives 8.

The second part is the real question of this chapter. The region under a parabola is bounded by the axis, a vertical line and a curve. There is no triangle, rectangle or circle hiding in it, and no formula in any geometry book covers it.

So a new idea is needed, and it is surprisingly simple: approximate the region by shapes you can measure, and then make the approximation perfect by taking a limit. That is the same move you made with derivatives, where secant slopes were approximations and the limit made them exact.

4. Sigma notation

Section

Part 1

5. Archimedes' plan: fill the region with measurable shapes

Concept

Figure (svg): Three unit circles, each with a regular polygon drawn inside it: a hexagon, a 12-sided polygon and a 48-sided polygon. The polygon areas, 2.598, 3.000 and 3.133, creep up toward the circle's area, pi, about 3.1416.

More, thinner pieces leave less uncovered. Archimedes filled the circle with shapes he could measure and watched the total close in on the area.

Cover the curved region with pieces whose areas you know, add them, then use more and smaller pieces. The totals close in on the true area. Adding many pieces needs a compact way to write long sums.

OpenStax Calculus Volume 1, §5.1 Approximating Areas §5.1, p. 438 — the method of exhaustion

Look at the three circles. Each has a regular polygon inside it, and a polygon's area is easy: cut it into triangles from the centre. With six sides the polygon misses a visible amount of the circle; with forty-eight sides you can barely see the gap.

The numbers under the pictures tell the same story. The areas climb, 2.598, then 3.000, then 3.133, closing in on pi, about 3.1416. Archimedes used exactly this method of exhaustion to trap pi between two fractions.

You will do the same with rectangles under a curve. The price of more, smaller pieces is that you must add many numbers, which is why the next slides give you a notation for long sums.

6. A long sum in one symbol

Concept

Adding the whole numbers from one to twenty takes a line of dots to write out:

\[ 1 + 2 + 3 + 4 + 5 + \cdots + 19 + 20 \]

Sigma notation packs it into one symbol. The Greek capital sigma means add up; the letter underneath counts from its starting value to the value on top.

\[ \sum_{i=1}^{20} i = 1 + 2 + 3 + \cdots + 20 \]

OpenStax Calculus Volume 1, §5.1 Approximating Areas §5.1, p. 438 — sigma (summation) notation

Writing the numbers from one to twenty with a string of dots is clumsy, and it hides the rule. Sigma notation says the same thing in one symbol and states the rule plainly: take the index, which here is the term itself, and add its values from 1 to 20.

Read the capital sigma as the words add up. Underneath it the index is introduced along with its starting value; on top sits its stopping value. To the right is the formula for a single term, written in terms of the index.

Whenever you meet a sum written this way and feel unsure, expand the first two or three terms and the last one. That habit turns the notation back into ordinary addition and catches most mistakes.

7. Reading sigma notation piece by piece

Notation

Annotate

On: \( \sum_{i=1}^{n} a_i = a_1 + a_2 + \cdots + a_n \)

  • Add up. Each value of the index produces one term, and all the terms are added.
  • The index and where it starts. Any letter works: i, j, k, m and n are the usual ones.
  • Where the index stops, including that value. From 1 to n there are exactly n terms.
  • The rule for a term. Substitute the index into it: the rule 3ⁱ with i = 2 gives 9.
  • The index disappears from the answer: a sum over i and the same sum over k are the same number.

OpenStax Calculus Volume 1, §5.1 Approximating Areas §5.1, p. 438 — the index as a dummy variable

Step through the notes one at a time. The most important one is the last: the index is a dummy variable. It exists only to count the terms, and it is gone from the final number. A sum over i and the same sum over k are identical.

Count the terms carefully. From 1 to n there are n terms, but from 3 to 6 there are four, not three. The count is the top minus the bottom, plus one, and forgetting the plus one is a classic slip.

Notice also that the term rule can be anything: a power, a fraction, a function value. Later in this lesson the term will be the area of one rectangle, a height times a width, and the sum will be a whole approximation to an area.

8. Example 5.1(a): powers of three

Worked example

Write in sigma notation and evaluate the sum of the terms three to the i, for i from 1 to 5.

Write the sum

Why: The rule is three to the i; the index runs from 1 to 5.

\[ \sum_{i=1}^{5} 3^i \]

Expand it

Why: Substitute i equal to 1, 2, 3, 4 and 5 in turn.

\[ \sum_{i=1}^{5} 3^i = 3 + 3^2 + 3^3 + 3^4 + 3^5 \]

Evaluate each power

Why: Each term is three times the one before.

\[ = 3 + 9 + 27 + 81 + 243 \]

Add

Why: Five terms.

\[ = 363 \]

Check with the geometric-sum formula

Why: First term 3, ratio 3, five terms: the finite geometric formula gives the same total.

\[ \frac{3(3^5 - 1)}{3 - 1} = \frac{3 \cdot 242}{2} = 363 \]

OpenStax Calculus Volume 1, §5.1 Approximating Areas §5.1, pp. 438-439 — Example 5.1a

The first step is translation. The phrase the sum of the terms three to the i for i from 1 to 5 becomes a sigma with the rule three to the i, bottom limit 1 and top limit 5.

Then expand. Substitute each value of the index in turn and write the terms out; with only five terms this is the quickest way to the answer. The powers of three are 3, 9, 27, 81 and 243, which add to 363.

The check uses a formula you met in precalculus: a finite geometric sum equals the first term times the ratio to the power of the number of terms, minus one, all over the ratio minus one. It gives 363 too. When a sum has a formula, checking a hand computation against it costs seconds.

9. Example 5.1(b): name the pattern

Fill the middle

The book's second part runs the other way: you are given the terms and must find the rule. Every denominator is a perfect square.

Fill in the blanks

1 + \frac14 + \frac19 + \frac1i^2 + \frac1___ = \sum____^___ \frac______} }

Why: The denominators are 1, 4, 9, 16 and 25, the squares of 1 to 5, so the term is one over i squared with i running from 1 to 5. The first term, 1, is one over one squared, so it fits the rule too.

OpenStax Calculus Volume 1, §5.1 Approximating Areas §5.1, p. 439 — Example 5.1b

Now the translation runs backwards: you see the terms and must find the rule. Look at the denominators, 1, 4, 9, 16 and 25. They are the squares of 1, 2, 3, 4 and 5.

So the term is one over i squared, and the index runs from 1 to 5. Do not skip the first term: 1 looks different from the others, but it is one over one squared and fits the rule.

This skill matters more than it seems. When you set up a Riemann sum later, you will write out a few rectangle areas and then have to spot the rule that generates all of them. That is exactly this exercise.

10. Checkpoint 5.1: an index that starts at three

Prediction

\[ \sum_{i=3}^{6} 2^i \]

Predict first

Evaluate the sum of the terms 2 to the i for i = 3, 4, 5, 6.

  • 120
  • 126
  • 64
  • 36

Correct: 120

Why: The index starts at 3, so the terms are 8, 16, 32 and 64, which add to 120. Starting at 1 instead would add 2 and 4 and give 126; 36 comes from doubling i instead of raising 2 to the power i.

OpenStax Calculus Volume 1, §5.1 Approximating Areas §5.1, p. 439 — Checkpoint 5.1

Commit to an option before you reveal. The trap here is the starting value. The index begins at 3, so the first term is two cubed, which is 8, not 2.

The four terms are 8, 16, 32 and 64, and they add to 120. If you got 126, you started at 1 and included 2 and 4. If you got 36, you doubled the index instead of using it as an exponent.

A quick check for any sum with a starting value other than 1: count the terms. From 3 to 6 is four terms, so your expansion should show exactly four numbers.

11. Five rules for rearranging a sum

Concept

Sums of the same length can be split, merged and rescaled. Each rule is ordinary arithmetic written in sigma notation.

rulein symbols
a constant added n times (5.1)Σ c = nc
a constant factor comes out (5.2)Σ caᵢ = c Σ aᵢ
a sum splits (5.3)Σ (aᵢ + bᵢ) = Σ aᵢ + Σ bᵢ
a difference splits (5.4)Σ (aᵢ − bᵢ) = Σ aᵢ − Σ bᵢ
the range splits at m (5.5)Σ from 1 to n = Σ from 1 to m + Σ from m + 1 to n

\[ \sum_{i=1}^{n} c = \underbrace{c + c + \cdots + c}_{n \text{ terms}} = nc \]

OpenStax Calculus Volume 1, §5.1 Approximating Areas §5.1, p. 439 — Rule: Properties of Sigma Notation

These five rules let you break a complicated sum into simple pieces. None of them is new mathematics; each is the ordinary rules of arithmetic, commutativity, associativity and the distributive law, written for a long sum.

The one that trips people up is the first. A constant summed from 1 to n is not the constant; it is the constant added n times, which is n times the constant. The line under the table shows it with the terms written out, so you can see the n copies.

The last rule, splitting at m, is the one that lets you handle a sum that starts at 5 or 11: write it as a sum from 1 minus the part you do not want.

12. Why a constant factors out

Concept

The proof of rule 5.2 is the distributive law, written out:

\[ \sum_{i=1}^{n} c a_i = c a_1 + c a_2 + \cdots + c a_n \]

\[ = c\left(a_1 + a_2 + \cdots + a_n\right) \]

\[ = c \sum_{i=1}^{n} a_i \]

Rule 5.3 is the same move with regrouping: collect all the a terms, then all the b terms.

\[ \sum_{i=1}^{n} (a_i + b_i) = (a_1 + \cdots + a_n) + (b_1 + \cdots + b_n) \]

OpenStax Calculus Volume 1, §5.1 Approximating Areas §5.1, pp. 439-440 — proof of properties 2 and 3

Follow the three lines. The first writes every term out, each with its factor c. The second pulls the c out of all of them at once, which is the distributive law. The third rewrites the bracket as a sum again.

Rule 5.3 is proved the same way. Write the terms out in pairs, then use the fact that addition can be done in any order to collect all the a terms together and all the b terms together.

Why prove something so obvious? Because the same pattern, write it out, rearrange, rewrite as sigma, is how you will justify every manipulation of a sum you are unsure about. There is one thing no rule lets you do: split a product. The sum of a times b is not the sum of a times the sum of b.

13. Equal or unequal?

Sorting

Sort into buckets

Exercise 1 and some variations: is each pair of sums equal or unequal?

Equal
Σ from i=1 to 10 of i, and Σ from k=1 to 10 of k; Σ from i=1 to 10 of i, and Σ from i=6 to 15 of (i − 5); Σ from i=1 to 10 of i(i − 1), and Σ from j=0 to 9 of (j + 1)j; Σ from i=1 to 10 of 2i, and 2 times Σ from i=1 to 10 of i
Unequal
Σ from i=1 to 10 of i², and (Σ from i=1 to 10 of i)²; Σ from i=1 to 10 of 3, and 3
eq
Renaming the index changes nothing; shifting the index (i − 5 from 6 to 15, or j = i − 1) lists the same terms; a constant factor comes out by rule 5.2.
ne
The sum of the squares is 385 but the square of the sum is 55² = 3025. Adding 3 ten times gives 30, not 3: rule 5.1 says nc.

OpenStax Calculus Volume 1, §5.1 Approximating Areas §5.1, p. 450 — Exercise 1

Sort every pair before you check. For the pairs that look different but are equal, expand both and compare the actual terms. Shifting the index, as in the second and third pairs, changes the labels but produces the same list of numbers.

The two unequal pairs are the two most common mistakes with sums. The sum of the squares is not the square of the sum: 385 against 3025. And a constant summed ten times is ten times the constant, not the constant.

The last pair, with the factor 2, is equal by rule 5.2. The difference between it and the squares pair is worth saying out loud: a constant factor can come out of a sum, but a power cannot be moved outside it.

14. Trap: a constant summed once

Trap

The trap

Splitting a sum, and then treating the constant piece as a single term:

\[ \sum_{i=1}^{200} (i^2 + 9) = \sum_{i=1}^{200} i^2 + 9 \]

Wrong. The nine is added two hundred times.

The fix

Rule 5.1 says a constant summed from 1 to n is n times the constant. Here that piece is worth 1800, not 9.

\[ \sum_{i=1}^{200} 9 = 200 \cdot 9 = 1800 \]

\[ \sum_{i=1}^{200} (i^2 + 9) = \sum_{i=1}^{200} i^2 + 1800 \]

This is the most frequent error in the whole lesson, and it hides inside correct-looking algebra. The split is fine; the mistake is reading the sum of nine as just nine.

Say the sum out loud: add nine, once for each i from 1 to 200. That is two hundred nines, which is 1800. If you ever find a constant alone inside a sigma, replace it immediately by the number of terms times the constant.

You will see this piece again in a moment, in Example 5.2(a), where the 1800 is one of the three numbers that make up the answer.

15. Three formulas that collapse a sum

Concept

For the sums that come up again and again, the whole string of terms collapses to a short formula in n.

\[ \sum_{i=1}^{n} i = \frac{n(n+1)}{2} \]

\[ \sum_{i=1}^{n} i^2 = \frac{n(n+1)(2n+1)}{6} \]

\[ \sum_{i=1}^{n} i^3 = \frac{n^2(n+1)^2}{4} \]

These three, with the five rules, are the whole toolkit for this lesson: they turn a sum of a thousand rectangles into one expression in n.

OpenStax Calculus Volume 1, §5.1 Approximating Areas §5.1, p. 440 — Rule: Sums and Powers of Integers

These formulas are the reason sigma notation is useful rather than just tidy. A sum with n terms takes n additions; the formula takes a few multiplications, whether n is 10 or ten million.

Test each one at n equal to 2 before trusting it: one plus two is 3, one plus four is 5, one plus eight is 9. Plug 2 into each formula and you get 3, 5 and 9. Checking a formula at a small n is a habit that will save you on exams.

Notice the size of each: the first grows like n squared over 2, the second like n cubed over 3, the third like n to the fourth over 4. Those fractions, one half, one third, one quarter, will reappear as areas under x, x squared and x cubed.

16. Why the first formula is true: two staircases

Picture it

Figure (svg): A 5 by 6 grid of squares. A blue staircase of columns with 1, 2, 3, 4 and 5 squares fills the bottom-left; an orange upside-down staircase with 5, 4, 3, 2 and 1 squares fills the rest. Together they make a 5 by 6 rectangle of 30 squares.

Two copies of the staircase fit together into a rectangle n wide and n plus 1 tall, so one staircase is half of n times n plus 1.

Stack a staircase of one, two, three, four and five squares against an upside-down copy. They lock into a rectangle five wide and six tall, so one staircase holds half of thirty squares.

\[ 2\sum_{i=1}^{n} i = n(n+1) \quad\Longrightarrow\quad \sum_{i=1}^{n} i = \frac{n(n+1)}{2} \]

Look at the grid. The blue squares form a staircase with one square in the first column, two in the second, up to five. That staircase holds one plus two plus three plus four plus five squares.

Now the orange squares: the same staircase, turned upside down and slid into the gap. Together they fill a rectangle five columns wide and six rows tall, which holds thirty squares. One staircase is half of that, fifteen.

Nothing in the picture depends on five. For any n the two staircases make a rectangle n by n plus 1, so the sum of the first n whole numbers is n times n plus 1, over 2. The story goes that the young Gauss used this pairing to add the numbers from 1 to 100 in seconds.

17. Reading the sum-of-squares formula

Notation

Annotate

On: \( \sum_{i=1}^{n} i^2 = \frac{n(n+1)(2n+1)}{6} \)

  • n terms, 1 + 4 + 9 + … + n². Adding them one by one takes n steps.
  • Three factors, each about n, 1 × n and 2n: the product grows like 2n³.
  • Always divides exactly: among n, n + 1 and 2n + 1 there is a factor of 2 and a factor of 3.
  • Roughly n³/3 for large n. That one-third will turn up again as the area under x².

\[ n = 3: \quad 1 + 4 + 9 = 14 = \frac{3 \cdot 4 \cdot 7}{6} \]

Go through the notes, and then check the small case under the formula: one plus four plus nine is 14, and three times four times seven over six is also 14.

The divisibility note answers a fair worry: why is the formula always a whole number when there is a six underneath? Among any two consecutive numbers one is even, and among n, n plus 1 and 2n plus 1 one is always a multiple of three.

The size note is the one to remember for later. For large n the formula behaves like n cubed over 3. When you compute the area under x squared exactly, that one third is exactly where the answer's one third comes from.

18. Example 5.2(a): a squared shift, two hundred terms

Worked example

Write in sigma notation and evaluate the sum of the terms i minus 3, squared, for i from 1 to 200.

\[ \sum_{i=1}^{200} (i-3)^2 \]

Expand the square

Why: The formulas only know powers of i, so multiply it out.

\[ = \sum_{i=1}^{200} \left(i^2 - 6i + 9\right) \]

Split the sum

Why: Rules 5.3 and 5.4.

\[ = \sum_{i=1}^{200} i^2 - \sum_{i=1}^{200} 6i + \sum_{i=1}^{200} 9 \]

Pull out the constant and count the nines

Why: Rule 5.2 on the middle sum, rule 5.1 on the last.

\[ = \sum_{i=1}^{200} i^2 - 6\sum_{i=1}^{200} i + 9(200) \]

Apply the power-sum formulas

Why: With n equal to 200.

\[ = \frac{200(201)(401)}{6} - 6\left[\frac{200(201)}{2}\right] + 1800 \]

Evaluate each piece

Why: Keep the three numbers separate until the end.

\[ = 2{,}686{,}700 - 120{,}600 + 1800 \]

Combine

Why: Subtract, then add.

\[ = 2{,}567{,}900 \]

Check by shifting the index instead

Why: Let j equal i minus 3: j runs from minus 2 to 197. The terms for j equal to minus 2, minus 1 and 0 are 4, 1 and 0, and the rest is one sum of squares.

\[ 5 + \sum_{j=1}^{197} j^2 = 5 + \frac{197(198)(395)}{6} = 5 + 2{,}567{,}895 = 2{,}567{,}900 \]

OpenStax Calculus Volume 1, §5.1 Approximating Areas §5.1, pp. 440-441 — Example 5.2a

The formulas know only powers of i, not powers of i minus 3, so the first move is to expand the square. After that the example is a chain of rules: split the sum, pull out the six, count the nines.

Keep the three pieces separate when you substitute n equal to 200. The sum of squares is 2,686,700; six times the sum of the integers is 120,600; the nines give 1800. Combining at the very end keeps arithmetic errors local and easy to find.

The check takes a different road to the same place. Instead of expanding, rename the index: j equal to i minus 3 runs from minus 2 to 197. The first three terms are 4, 1 and 0, and the rest is a plain sum of squares up to 197. Two independent routes giving 2,567,900 is strong evidence.

19. Example 5.2(b): cubes minus squares

Worked example

Evaluate the sum of i cubed minus i squared, for i from 1 to 6.

Split the difference

Why: Rule 5.4.

\[ \sum_{i=1}^{6}\left(i^3 - i^2\right) = \sum_{i=1}^{6} i^3 - \sum_{i=1}^{6} i^2 \]

Apply the cube and square formulas

Why: Both with n equal to 6.

\[ = \frac{6^2(7)^2}{4} - \frac{6(7)(13)}{6} \]

Evaluate the numerators

Why: Thirty-six times forty-nine, and six times seven times thirteen.

\[ = \frac{1764}{4} - \frac{546}{6} \]

Divide

Why: Both divisions are exact.

\[ = 441 - 91 = 350 \]

Check by listing the six terms

Why: Each term is i squared times i minus 1.

\[ 0 + 4 + 18 + 48 + 100 + 180 = 350 \]

OpenStax Calculus Volume 1, §5.1 Approximating Areas §5.1, p. 441 — Example 5.2b

Rule 5.4 splits this into two sums you have formulas for. Substitute n equal to 6 into each: the cube formula gives 1764 over 4, which is 441, and the square formula gives 546 over 6, which is 91.

With only six terms, the direct check is cheap and genuinely independent. Each term, i cubed minus i squared, factors as i squared times i minus 1, so the terms are 0, 4, 18, 48, 100 and 180. They add to 350.

For six terms the formula was hardly worth it. The point is that the same four lines would handle six million terms without any extra work.

20. Example 5.3: the first ten cubes

Worked example

Find the sum of the values of f(x), equal to x cubed, over the integers 1 to 10.

Write the sum

Why: The values are f(1), f(2), and so on up to f(10).

\[ \sum_{i=1}^{10} i^3 \]

Use the cube formula

Why: With n equal to 10.

\[ = \frac{10^2(10+1)^2}{4} \]

Evaluate

Why: One hundred times one hundred twenty-one, over four.

\[ = \frac{100(121)}{4} = 3025 \]

Check against the sum of the first ten integers

Why: The cube formula is the square of the first formula, so the sum of cubes equals the square of 55.

\[ \left(\sum_{i=1}^{10} i\right)^2 = 55^2 = 3025 \]

OpenStax Calculus Volume 1, §5.1 Approximating Areas §5.1, p. 441 — Example 5.3

The book phrases this as the sum of the values of a function over the integers from 1 to 10. That phrasing is worth noticing, because a Riemann sum is exactly this: a sum of function values, each multiplied by a width.

The cube formula gives 100 times 121 over 4, which is 3025.

The check uses a curiosity: the cube formula is the square of the formula for the sum of the integers. So the sum of the first ten cubes equals the square of the sum of the first ten integers, 55 squared, which is 3025. That is a coincidence of these particular formulas, not a rule about sums, so do not try it with squares.

21. Checkpoint 5.2: one hundred values of four plus three i

Fill the middle

Fill in the blanks

\sum_100^5050(4 + 3i) = 4 \cdot 15550 + 3 \cdot ___ = ___

Why: The constant 4 is added 100 times (rule 5.1), and the sum of i from 1 to 100 is 100 times 101 over 2, which is 5050. So the total is 400 plus 15150, which is 15550.

OpenStax Calculus Volume 1, §5.1 Approximating Areas §5.1, p. 441 — Checkpoint 5.2

Split the sum into a constant part and a part in i. The constant 4 is added one hundred times, so that piece is 4 times 100. The other piece is 3 times the sum of i from 1 to 100.

The sum of the first hundred integers is 100 times 101 over 2, which is 5050. Three times that is 15150, and adding 400 gives 15550.

If your answer came out as 15154, you added the 4 only once: that is the trap from a few slides ago. If it came out as 15550 on the first try, the rules are becoming automatic, which is what you need for Part 5.

22. Exercises 4 to 7: sums built from two known sums

Matching

\[ \sum_{i=1}^{100} a_i = 15, \qquad \sum_{i=1}^{100} b_i = -12 \]

Match the pairs

  • p. Σ (aᵢ + bᵢ)
  • q. Σ (aᵢ − bᵢ)
  • r. Σ (3aᵢ − 4bᵢ)
  • s. Σ 2aᵢ
  • w. 3
  • x. 27
  • y. 93
  • z. 30

Why: Split and factor with rules 5.2 to 5.4, then substitute: 15 + (−12) = 3; 15 − (−12) = 27; 3(15) − 4(−12) = 45 + 48 = 93; 2(15) = 30. You never need the individual terms.

OpenStax Calculus Volume 1, §5.1 Approximating Areas §5.1, p. 451 — Exercises 4 to 7

You know only the two totals, 15 and minus 12, and nothing about the individual terms. The point of these exercises is that you do not need them: the rules let you compute each new sum from the two totals alone.

Split, factor, substitute. Three times the a sum minus four times the b sum is 45 minus minus 48, which is 93. Be careful with the minus signs; the b total is negative, so subtracting it adds.

This is the same reasoning you will use in Section 5.2 with integrals, where the book gives you the values of two integrals and asks for a combination. The rules for sums become the rules for integrals.

23. Checkpoint 5.3: twenty odd numbers

Estimation

\[ \sum_{k=1}^{20} (2k + 1) \]

Predict first

Evaluate the sum. Estimate first: twenty terms, running from 3 up to 41.

  • 440
  • 400
  • 420
  • 460

Correct: 440

Why: Split it: 2 times the sum of k from 1 to 20 is 2 times 210, which is 420, and the constant 1 added twenty times is 20. Total 440. As an estimate, twenty terms averaging 22 (the middle of 3 and 41) also give 440.

OpenStax Calculus Volume 1, §5.1 Approximating Areas §5.1, p. 441 — Checkpoint 5.3

Estimate before computing. The terms run from 3 to 41 in equal steps, so their average is the average of the first and last, which is 22. Twenty terms averaging 22 give 440.

The rules give the same: twice the sum of the first twenty integers is 420, and the constant 1 added twenty times is 20. Total 440.

That average-of-the-ends trick works for any sum whose terms go up by the same amount each time. It is the staircase picture again: pair the first term with the last, the second with the second-to-last, and every pair has the same total.

24. Rectangles under a curve

Section

Part 2

25. The region, and the question

Concept

Figure (svg): A curve y equals f of x above the x-axis, with the region under it between the vertical lines x equals a and x equals b shaded and labelled A.

The target: the area A trapped between the graph, the x-axis and the two vertical lines. No formula from geometry measures a curved top.

Take a function that is continuous and never negative on the closed interval from a to b. The area A is bounded by the graph above, the x-axis below, and the vertical lines at a and b.

Rectangles have a formula, height times width. The plan: cut the interval into strips, stand a rectangle on each, and add their areas.

OpenStax Calculus Volume 1, §5.1 Approximating Areas §5.1, pp. 441-442 — Figure 5.2

Here is the object the rest of the lesson is about. The shaded region sits under the graph of a function, above the x-axis, between the vertical lines at a and b. Call its area A.

Two assumptions are built in for now. The function is continuous, so the top edge has no gaps, and it is never negative, so the region really lies above the axis. Section 5.2 will drop the second assumption.

A rectangle is the simplest shape with a known area, height times width. If you stand thin rectangles side by side under the curve, their total area should be close to A, and closer the thinner they are. The next slides make that precise.

26. Cutting the interval into equal pieces

Concept

Figure (svg): A number line from a to b cut into six equal pieces by the points x sub 0 up to x sub 6, each piece of width delta x; the third piece, from x sub 2 to x sub 3, is highlighted.

Seven points make six pieces. The i-th piece runs from x sub i minus 1 to x sub i, so its left end has the smaller index.

\[ \Delta x = \frac{b-a}{n}, \qquad x_i = a + i\,\Delta x, \quad i = 0, 1, \ldots, n \]

Partition — A set of points from a to b, in increasing order, that cuts the interval into subintervals. When all the subintervals have the same width it is a regular partition.

OpenStax Calculus Volume 1, §5.1 Approximating Areas §5.1, p. 442 — Definition of a partition

Look at the number line. Seven points, numbered from zero to six, cut the interval into six pieces of equal width. The width is the length of the interval divided by the number of pieces, and it is called delta x.

Each point sits a whole number of widths from a: the point numbered i is a plus i times delta x. So the first point is a itself and the last is b. The highlighted piece, the third, runs from point 2 to point 3.

That indexing is the thing to get right. The i-th piece runs from point i minus 1 to point i. Its left end carries the smaller number. Every formula in the next few slides follows from that one convention.

27. The left-endpoint rule

Concept

On each strip, stand a rectangle whose height is the function's value at the LEFT end of the strip. The i-th strip runs from the point numbered i minus 1 to the point numbered i, so its left end has the smaller index.

\[ \text{area of the } i\text{th rectangle} = f(x_{i-1})\,\Delta x \]

\[ L_n = f(x_0)\Delta x + f(x_1)\Delta x + \cdots + f(x_{n-1})\Delta x = \sum_{i=1}^{n} f(x_{i-1})\,\Delta x \]

OpenStax Calculus Volume 1, §5.1 Approximating Areas §5.1, pp. 442-443 — Rule: Left-Endpoint Approximation (5.6)

Stand a rectangle on each piece and let its height be the function's value at the left end of the piece. The i-th rectangle then has height f at point i minus 1 and width delta x.

Adding all n rectangles gives the left-endpoint approximation, written L with a subscript n. In the long form you can see which heights are used: from point 0 up to point n minus 1. The height at b is never used.

That is the whole rule: a sum of heights times widths. The only choices are how many pieces, and where in each piece the height is measured.

28. The right-endpoint rule

Concept

The same strips, but each rectangle now takes its height at the RIGHT end of its strip. The heights shift along by one point.

\[ R_n = f(x_1)\Delta x + f(x_2)\Delta x + \cdots + f(x_n)\Delta x = \sum_{i=1}^{n} f(x_i)\,\Delta x \]

The two sums share every height except one at each end: the left sum uses the height at a and not at b, the right sum the height at b and not at a.

OpenStax Calculus Volume 1, §5.1 Approximating Areas §5.1, p. 443 — Rule: Right-Endpoint Approximation (5.7)

The right-endpoint rule makes the other obvious choice: measure each height at the right end of the piece. Now the heights run from point 1 up to point n, so the height at a is never used.

Compare the two long forms and you will see they share every height except the two at the ends. That observation is small but powerful: it will tell you exactly how far apart the two approximations are, without computing either.

Which one is better? Neither, in general. For a rising curve the left rule sits under the curve and the right rule pokes above it; for a falling curve they swap. The examples show both.

29. Figure 5.5: six rectangles under half of x squared

Worked example

Approximate the area under f(x) equal to x squared over 2 on the interval from 0 to 3 with six strips, both ways.

Find the width

Why: Three units cut into six equal pieces.

\[ \Delta x = \frac{3 - 0}{6} = 0.5 \]

List the partition points

Why: Start at 0 and step by one half.

\[ x_i = 0, 0.5, 1, 1.5, 2, 2.5, 3 \]

Left endpoints: the heights

Why: f at 0 up to 2.5; the height at 3 is not used.

\[ f(x_{i-1}) = 0, 0.125, 0.5, 1.125, 2, 3.125 \]

Left sum

Why: Multiply each height by one half and add.

\[ L_6 = 0.5(0 + 0.125 + 0.5 + 1.125 + 2 + 3.125) = 0.5(6.875) = 3.4375 \]

Right endpoints: the heights

Why: Now f at 0.5 up to 3; the height at 0 is dropped.

\[ f(x_i) = 0.125, 0.5, 1.125, 2, 3.125, 4.5 \]

Right sum

Why: The same width.

\[ R_6 = 0.5(11.375) = 5.6875 \]

Figure (svg): Two panels of the curve y equals x squared over 2 on 0 to 3, each with six rectangles of width one half. Left: heights from left endpoints, every rectangle under the curve. Right: heights from right endpoints, every rectangle poking above it.

Same curve, same six strips; only the height rule changes. For a rising curve the left rule sits under it and the right rule sticks out above it.

Check the gap between them

Why: The two lists share five heights. The right list adds f(3), 4.5, and drops f(0), 0, so the sums differ by one half times 4.5.

\[ R_6 - L_6 = 0.5\left(f(3) - f(0)\right) = 0.5(4.5) = 2.25 = 5.6875 - 3.4375 \]

OpenStax Calculus Volume 1, §5.1 Approximating Areas §5.1, pp. 443-444 — Figure 5.5

Follow the steps in order: width first, then the list of points, then the heights, then the sum. Making the list of points explicit is what stops you using the wrong heights.

Look at the two panels. On the left every rectangle sits under the rising curve, so 3.4375 is too small. On the right every rectangle sticks up above it, so 5.6875 is too big. The true area is somewhere between, and Part 5 will show it is exactly 4.5.

The check is worth understanding. The two lists of heights share five values; the right list has f of 3 where the left list has f of 0. So the sums differ by the width times f of 3 minus f of 0, which is 2.25, and that is exactly the difference of the two numbers you computed.

30. Example 5.4: x squared on zero to two, with four rectangles

Worked example

Use left- and right-endpoint approximations with n equal to 4 for the area under f(x) equal to x squared on the interval from 0 to 2.

Find the width

Why: Two units, four strips.

\[ \Delta x = \frac{2 - 0}{4} = 0.5 \]

Left heights

Why: At 0, 0.5, 1 and 1.5.

\[ f(0) = 0, \; f(0.5) = 0.25, \; f(1) = 1, \; f(1.5) = 2.25 \]

Left sum

Why: Width times each height, added.

\[ L_4 = 0(0.5) + 0.25(0.5) + 1(0.5) + 2.25(0.5) = 1.75 \]

Right heights

Why: At 0.5, 1, 1.5 and 2.

\[ f(0.5) = 0.25, \; f(1) = 1, \; f(1.5) = 2.25, \; f(2) = 4 \]

Right sum

Why: The same width again.

\[ R_4 = 0.25(0.5) + 1(0.5) + 2.25(0.5) + 4(0.5) = 3.75 \]

Figure (svg): Two panels of y equals x squared on 0 to 2 with four rectangles of width one half. Left: left-endpoint heights 0, 0.25, 1 and 2.25. Right: right-endpoint heights 0.25, 1, 2.25 and 4.

The first left rectangle has height zero and vanishes; the last right rectangle reaches all the way up to 4. That is where most of the gap between 1.75 and 3.75 comes from.

Check with the end heights

Why: The right sum trades f(0) for f(2), so the two must differ by one half times four.

\[ R_4 - L_4 = 0.5\left(f(2) - f(0)\right) = 0.5(4) = 2 = 3.75 - 1.75 \]

OpenStax Calculus Volume 1, §5.1 Approximating Areas §5.1, pp. 444-445 — Example 5.4, Figures 5.6 and 5.7

This is the same procedure with a different curve and four strips. The width is one half, the left heights are the squares of 0, one half, 1 and one and a half, and the right heights shift along by one point.

In the figure notice how lopsided the error is. The first left rectangle has height zero, so it contributes nothing, while the last right rectangle is 4 tall and alone adds 2 to the right sum. With so few strips, both estimates are poor.

The check again uses the end heights. Swapping f of 0, which is 0, for f of 2, which is 4, changes the sum by one half times 4, which is 2. And indeed 3.75 minus 1.75 is 2. When the check fails, you have almost certainly used a wrong height.

31. Checkpoint 5.4: one over x on one to two

Worked example

Sketch and compute the left- and right-endpoint approximations for f(x) equal to one over x on the interval from 1 to 2, with n equal to 4.

Width and points

Why: One unit, four strips.

\[ \Delta x = 0.25, \qquad x_i = 1, 1.25, 1.5, 1.75, 2 \]

Heights at the points

Why: One over each point.

\[ f(x_i) = 1, \; 0.8, \; 0.6667, \; 0.5714, \; 0.5 \]

Left sum

Why: The first four heights.

\[ L_4 = 0.25(1 + 0.8 + 0.6667 + 0.5714) \approx 0.7595 \]

Right sum

Why: The last four heights.

\[ R_4 = 0.25(0.8 + 0.6667 + 0.5714 + 0.5) \approx 0.6345 \]

Figure (svg): Two panels of y equals 1 over x on 1 to 2 with four rectangles of width one quarter. Left: left-endpoint heights, each rectangle above the falling curve. Right: right-endpoint heights, each rectangle below it.

The curve falls, so the roles swap: now the left rule overestimates and the right rule underestimates.

Check the gap and the true value

Why: The gap must be a quarter of f(1) minus f(2); and the true area, the natural log of 2 from Chapter 6, must lie between the two sums.

\[ L_4 - R_4 = 0.25(1 - 0.5) = 0.125, \qquad 0.6345 < \ln 2 \approx 0.6931 < 0.7595 \]

OpenStax Calculus Volume 1, §5.1 Approximating Areas §5.1, p. 445 — Checkpoint 5.4

Here the curve falls instead of rising, and the roles of the two rules swap. In the left panel each rectangle takes its height at the left end, where one over x is largest, so every rectangle pokes above the curve.

Keep four decimal places in the heights. Rounding each to one decimal would move the answer in the second decimal place, and then the check would not work.

The check has two parts. The gap should be a quarter times f of 1 minus f of 2, which is 0.125, and it is. And the true area, which Chapter 6 will show is the natural logarithm of 2, about 0.6931, must lie between the two sums. It does, and it sits almost in the middle, which is a hint about the midpoint rule to come.

32. Four, eight, thirty-two rectangles

Picture it

Figure (svg): Three panels of the curve y equals x minus 1 cubed plus 4 on 0 to 2, filled by 4, 8 and 32 left-endpoint rectangles. The white gaps between the rectangles and the curve shrink from panel to panel.

Thinner rectangles hug the curve more closely, and the left sums climb toward 8: 7.5, then 7.75, then 7.9375.
nLₙRₙgap Rₙ − Lₙ
47.58.51
87.758.250.5
327.93758.06250.125

Both columns close in on 8, and the gap between them halves each time the number of strips doubles. The book's Table 5.1 rounds the last row to 7.94 and 8.06.

OpenStax Calculus Volume 1, §5.1 Approximating Areas §5.1, pp. 445-448 — Figures 5.8 to 5.13 and Table 5.1

Look at the three panels from left to right. With four rectangles there are obvious white gaps between the tops of the rectangles and the curve. With eight the gaps are smaller, and with thirty-two you can barely see them.

The table puts numbers on that. The left sums rise, 7.5, 7.75, 7.9375, and the right sums fall, 8.5, 8.25, 8.0625. Both are heading to 8. The book's Table 5.1 rounds the last row to two decimals.

Now look at the last column. The gap between the right and left sums halves each time n doubles. That is the gap formula at work: the width halves, and the end heights never change, so the gap is always the width times the same number, here 2.

33. Slide n and watch the rectangles' tops

Tweak it

Parameter explorer

The staircase is the tops of the right-endpoint rectangles for x² on [0, 2]. Slide n. How many rectangles until the staircase is hard to tell from the parabola?

\[ R_{n} = \sum_{i=1}^{{n}} \left(\frac{2i}{{n}}\right)^2 \cdot \frac{2}{{n}} \]

  • n — from 1 to 40: rectangles n

The staircase you see is the outline of the right-endpoint rectangles for x squared on 0 to 2; the rectangles themselves stand under it. With n equal to 1 there is one step, at height 4, and it is a terrible approximation.

Slide n up slowly. Each step hugs the parabola more closely, and the readout shows the sum being computed. By twenty or so the steps are small, and by forty the staircase is hard to tell from a smooth curve.

The question to ask yourself is not whether the staircase ever becomes the parabola; it never does, for any finite n. The question is whether the areas settle down to one number as n grows. That is a limit, and it is the definition of the area.

34. The midpoint rule

Concept

Why commit to an end at all? The midpoint rule takes each height at the middle of its strip, where the curve is neither at its highest nor at its lowest.

\[ M_n = \sum_{i=1}^{n} f\left(\frac{x_{i-1} + x_i}{2}\right)\Delta x \]

It is still a sum of rectangles, just with a different height rule, and for smooth curves it is usually far more accurate than either end.

The left and right rules both measure the height at an extreme of the strip, so on a rising or falling curve they are guaranteed to be wrong in one direction. The midpoint rule measures in the middle instead.

The formula is the same sum of heights times widths; only the sample point changes, to the average of the two ends. For the i-th strip that is a plus i minus one half, times delta x.

Why should the middle be better? Over a short strip a smooth curve is nearly a straight line. For a straight line, the midpoint height times the width is exactly the area of the strip, because the triangle it misses on one side equals the triangle it adds on the other.

35. The midpoint rule on x squared

Worked example

Approximate the area under x squared on 0 to 2 with four midpoint rectangles, and compare with Example 5.4.

Midpoints of the four strips

Why: Halfway across each strip of width one half.

\[ 0.25, \; 0.75, \; 1.25, \; 1.75 \]

Heights at the midpoints

Why: Square each one.

\[ 0.0625, \; 0.5625, \; 1.5625, \; 3.0625 \]

Add the heights

Why: Four heights.

\[ 0.0625 + 0.5625 + 1.5625 + 3.0625 = 5.25 \]

Multiply by the width

Why: Delta x is one half.

\[ M_4 = 0.5(5.25) = 2.625 \]

Figure (svg): The curve y equals x squared on 0 to 2 with four rectangles whose heights are taken at the midpoints 0.25, 0.75, 1.25 and 1.75; a dot marks each midpoint on the curve, and each rectangle's top crosses the curve there.

Each rectangle's top crosses the curve at the middle of its strip: the sliver it misses on one side is partly paid back by the sliver it adds on the other.

Check against the exact area

Why: Part 5 of this lesson shows the exact area is eight thirds. The midpoint error is about a twentieth of the left or right error.

\[ \left|\tfrac83 - 2.625\right| \approx 0.042 \quad\text{vs}\quad \left|\tfrac83 - 1.75\right| \approx 0.917, \; \left|\tfrac83 - 3.75\right| \approx 1.083 \]

The steps are the same as for the left and right rules: find the sample points, find the heights, add, multiply by the width. The midpoints are a quarter, three quarters, one and a quarter, and one and three quarters.

Look at the figure. Each dot sits on the curve at the middle of its strip, and each rectangle's top crosses the curve there. The rectangle overshoots on the right half of the strip and undershoots on the left half, and the two slivers nearly cancel.

The check shows how dramatic that cancellation is. With the same four strips, the midpoint rule is off by about 0.04, while the left and right rules were off by about 0.9 and 1.1. You will meet the reason again in Calculus II, where numerical integration is studied properly.

36. The gap between right and left, exactly

Concept

Figure (svg): The curve y equals x squared on 0 to 2 with four right-endpoint rectangles; the part of each right rectangle above the matching left rectangle is coloured. Beside the graph the four coloured slivers are stacked into one column of width one half and total height 4.

Right sum minus left sum is exactly the four coloured slivers. Slide them sideways into one column: width delta x, height f(2) minus f(0), which is 4.

Write both sums out. Every height from the first interior point to the last appears in both, so they cancel, and only the two end heights survive.

\[ R_n - L_n = \Delta x\left[f(x_1) + \cdots + f(x_n)\right] - \Delta x\left[f(x_0) + \cdots + f(x_{n-1})\right] \]

\[ R_n - L_n = \Delta x\left[f(b) - f(a)\right] \]

OpenStax Calculus Volume 1, §5.1 Approximating Areas §5.1, p. 455 — Exercise 55

The coloured slivers in the picture are the parts of the right rectangles that stick up above the left rectangles. Each sliver is one strip wide and as tall as the rise of the curve across that strip.

Now slide all the slivers sideways into one column. Their heights stack up to the total rise of the curve from a to b, which is f of b minus f of a. The column is delta x wide, so its area is delta x times f of b minus f of a.

The algebra says the same thing: write the two sums out, and every height except the two end ones appears in both and cancels. This is Exercise 55 in the book. It gives you a free check on every pair of endpoint sums, and it shows the gap shrinks to zero as n grows.

37. Find the error: the missing width

Error analysis

Annotate

On: \( L_4 = f(0) + f(0.5) + f(1) + f(1.5) = 0 + 0.25 + 1 + 2.25 = 3.5 \)

  • A left sum for x² on [0, 2] with four strips, as it often appears on quizzes. The heights are right.
  • Heights were added, not areas. Each rectangle is only 0.5 wide, so every height must be multiplied by Δx.
  • 3.5 is almost the height of the tallest rectangle times the full width 2 (that is 4.5). An estimate that big for a region inside a 2 by 4 box should raise a flag.
  • L₄ = 0.5 × 3.5 = 1.75, the value from Example 5.4.

Read through the notes. The heights in the line are all correct, and so is their total. The only mistake is that a rectangle's area is height times width, and the widths never appeared.

The warning-sign note gives you a way to catch this without redoing the problem. The whole region sits inside a box 2 wide and 4 tall. An area estimate of 3.5 would fill almost half that box, but the region under the parabola is visibly much less.

A simple habit prevents the error entirely: factor the width out first. Write L sub 4 as one half times the sum of the heights. Then the width is on the page before any adding starts, and it cannot be forgotten.

38. Upper and lower sums

Section

Part 3

39. Upper and lower sums

Concept

Figure (svg): The curve y equals sine x on 0 to pi cut into three strips. Tall outlined rectangles reach the highest point of the curve on each strip, which for the middle strip is the peak at pi over 2, inside the strip rather than at an end; short green rectangles reach only the lowest point on each strip.

An upper sum uses the tallest point of each strip and a lower sum the shortest. On the middle strip the tallest point is the peak, which neither endpoint rule would ever pick.

Upper sum — A Riemann sum in which each rectangle's height is the largest value of the function on its strip. Every rectangle covers the curve, so the sum is an overestimate.

Lower sum — A Riemann sum in which each height is the smallest value of the function on its strip. Every rectangle fits under the curve, so the sum is an underestimate.

\[ \text{lower sum} \le A \le \text{upper sum} \]

OpenStax Calculus Volume 1, §5.1 Approximating Areas §5.1, p. 449 — upper and lower sums

Look at the sine arch in the picture. On each of the three strips the tall outlined rectangle reaches the highest point of the curve on that strip, and the short green rectangle reaches only the lowest point.

Those two choices of height are guaranteed to bracket the area. Rectangles at the highest point cover everything under the curve on that strip, so their total is an upper sum; rectangles at the lowest point fit underneath, so their total is a lower sum.

Notice where the highest points are. On the outer strips the highest point is the inner end and the lowest is the outer end, so neither rule picks the same end on both. On the middle strip the curve peaks at pi over 2, inside the strip, where no endpoint rule would ever look. For curves that turn around you must look inside each strip.

40. When the endpoints do the job

Concept

If the function only rises, or only falls, across the whole interval, then its largest and smallest values on each strip sit at the strip's ends. Upper and lower sums become endpoint sums.

f on [a, b]lower sumupper sum
increasingLₙ (left ends)Rₙ (right ends)
decreasingRₙ (right ends)Lₙ (left ends)
neitherfind each strip's minimumfind each strip's maximum

For a curve that turns around, like the sine arch in the last picture, you must look inside each strip.

OpenStax Calculus Volume 1, §5.1 Approximating Areas §5.1, p. 449 — upper and lower sums for monotone f

For most of the examples in this section the function only rises or only falls, and then finding the highest and lowest points is easy: they sit at the ends of every strip.

Read the table as a decision. First decide whether the function is increasing or decreasing on the whole interval, using the derivative if the graph is not obvious. Then the table tells you which endpoint sum is the lower sum and which is the upper.

The last row is the warning. If the function turns around inside the interval, neither endpoint rule is guaranteed to be an upper or lower sum, and you need each strip's actual maximum and minimum.

41. Example 5.5: a lower sum for ten minus x squared

Worked example

Find a lower sum for f(x) equal to 10 minus x squared on the interval from 1 to 2, with n equal to 4.

Width and strips

Why: One unit, four strips.

\[ \Delta x = \tfrac14; \quad [1, 1.25], [1.25, 1.5], [1.5, 1.75], [1.75, 2] \]

Decide which end is lowest

Why: The derivative is negative for x positive, so f falls on the whole interval and each strip's lowest point is its right end.

\[ f'(x) = -2x < 0 \;\Longrightarrow\; \text{lower sum} = R_4 \]

Heights at the right ends

Why: Ten minus each square.

\[ f(1.25) = 8.4375, \; f(1.5) = 7.75, \; f(1.75) = 6.9375, \; f(2) = 6 \]

Add and multiply by the width

Why: The heights add to 29.125.

\[ 0.25(8.4375 + 7.75 + 6.9375 + 6) = 0.25(29.125) = 7.28125 \]

Figure (svg): The curve y equals 10 minus x squared from 0 to 2.2, falling from 10. On 1 to 2 four rectangles of width one quarter take their heights at right endpoints: 8.4375, 7.75, 6.9375 and 6, all under the curve.

On a falling curve the right end of each strip is its lowest point, so right-endpoint rectangles all fit underneath: a lower sum.

Check with the upper sum and the gap formula

Why: The upper sum uses left ends; the two must differ by a quarter of f(1) minus f(2), which is a quarter of 3.

\[ L_4 - R_4 = 0.25(9 - 6) = 0.75 \;\Longrightarrow\; \text{upper} = 7.28125 + 0.75 = 8.03125 \]

OpenStax Calculus Volume 1, §5.1 Approximating Areas §5.1, p. 449 — Example 5.5, Figure 5.14

The question asks for a lower sum, not a left or right sum, so the first real step is deciding which ends are lowest. The derivative, minus 2x, is negative for every x in the interval, so the function falls, and each strip's lowest point is its right end.

The rest is arithmetic you know. The four right-end heights add to 29.125, and a quarter of that is 7.28125, which the book rounds to 7.28. In the figure every rectangle sits under the curve.

The check computes the matching upper sum for free. Swapping the right-end height at 2 for the left-end height at 1 raises the sum by a quarter times 3, so the upper sum must be 8.03125. The next slide computes it directly and confirms.

42. Checkpoint 5.5: the matching upper sum

Worked example

Find an upper sum for the same function on the same interval with n equal to 4, and sketch it.

Choose the ends

Why: f falls, so each strip's highest point is its left end.

\[ \text{upper sum} = L_4 \]

Heights at the left ends

Why: f(1) is 9; the other three are shared with the lower sum.

\[ f(1) = 9, \; f(1.25) = 8.4375, \; f(1.5) = 7.75, \; f(1.75) = 6.9375 \]

Add and multiply by the width

Why: The heights add to 32.125.

\[ L_4 = 0.25(32.125) = 8.03125 \]

Figure (svg): The same curve y equals 10 minus x squared with four rectangles on 1 to 2 whose heights come from left endpoints: 9, 8.4375, 7.75 and 6.9375, each rectangle poking above the curve.

The sketch Checkpoint 5.5 asks for: every rectangle takes the tallest point of its strip, the left end, so every one overshoots.

Check that the true area is trapped

Why: Section 5.3 will give the exact area as twenty-three thirds. It lies inside the bracket, and the upper sum matches the previous slide's prediction.

\[ 7.28125 < \tfrac{23}{3} \approx 7.6667 < 8.03125 \]

OpenStax Calculus Volume 1, §5.1 Approximating Areas §5.1, p. 449 — Checkpoint 5.5

The function falls, so the tallest point of each strip is its left end, and the upper sum is the left-endpoint sum. Three of its four heights are shared with the lower sum; only f of 1, which is 9, is new.

The figure is the sketch the checkpoint asks for. Every rectangle's top left corner sits on the curve and the rest of the top is above it. Compare it with the lower-sum picture on the previous slide: the same strips, each rectangle one step taller.

The true area is trapped between 7.28125 and 8.03125. Section 5.3 will give the exact value, twenty-three thirds, about 7.667, and it lands inside the bracket. That is what a bracket buys you: a guarantee, not just a guess.

43. Example 5.6: a lower sum for sine

Worked example

Find a lower sum for f(x) equal to sine x on the interval from 0 to pi over 2, with n equal to 6.

Width

Why: A quarter turn cut into six strips.

\[ \Delta x = \frac{\pi/2}{6} = \frac{\pi}{12} \]

Decide which end is lowest

Why: Sine increases from 0 to pi over 2, so each strip's lowest point is its left end.

\[ \text{lower sum} = L_6 = \sum_{i=0}^{5} \sin\left(\frac{i\pi}{12}\right)\frac{\pi}{12} \]

Heights at the left ends

Why: Sines of 0, 15, 30, 45, 60 and 75 degrees.

\[ 0, \; 0.2588, \; 0.5, \; 0.7071, \; 0.8660, \; 0.9659 \]

Add the heights

Why: Six values.

\[ 0 + 0.2588 + 0.5 + 0.7071 + 0.8660 + 0.9659 = 3.2978 \]

Multiply by the width

Why: Pi over 12 is about 0.2618.

\[ L_6 = 3.2978 \cdot 0.2618 \approx 0.863 \]

Figure (svg): The curve y equals sine x from 0 to pi, with six rectangles of width pi over 12 on 0 to pi over 2 whose heights come from left endpoints; each sits under the rising curve.

Sine rises all the way from 0 to pi over 2, so the left end of each strip is its lowest point and the left sum is the lower sum.

Check with the gap formula

Why: The upper sum, R sub 6, trades sine 0 for sine of pi over 2, so it is larger by pi over 12 times 1. The exact area, 1, found in Section 5.3, sits between them.

\[ R_6 = 0.863 + \frac{\pi}{12}(1 - 0) \approx 1.125, \qquad 0.863 < 1 < 1.125 \]

OpenStax Calculus Volume 1, §5.1 Approximating Areas §5.1, p. 450 — Example 5.6, Figure 5.15

Sine rises from 0 at the origin to 1 at pi over 2, so on this interval the lowest point of every strip is its left end, and the lower sum is the left-endpoint sum.

The width is pi over 12, which is 15 degrees. The heights are the sines of 0, 15, 30, 45, 60 and 75 degrees; the familiar ones, one half and root 2 over 2 and root 3 over 2, appear among them. Their total, times the width, is about 0.863.

The check uses the gap formula once more. The upper sum is the right-endpoint sum, and it exceeds the lower sum by the width times sine of pi over 2 minus sine 0, which is pi over 12. So the upper sum is about 1.125, and the exact area, 1, sits between.

44. Checkpoint 5.6: the upper sum for sine

Prediction

\[ f(x) = \sin x \text{ on } \left[0, \tfrac{\pi}{2}\right], \quad n = 6 \]

Predict first

Which endpoints give the upper sum here, and what is it, to three decimals?

  • Right endpoints: about 1.125
  • Left endpoints: about 0.863
  • Right endpoints: about 1.000
  • Midpoints: about 1.003

Correct: Right endpoints: about 1.125

Why: Sine rises on this interval, so each strip's highest point is its right end. The right sum adds sin of pi over 2, which is 1, and drops sin 0, so it equals the lower sum plus pi over 12: about 0.863 + 0.262 = 1.125. The midpoint sum, 1.003, is a better estimate but not an upper sum.

OpenStax Calculus Volume 1, §5.1 Approximating Areas §5.1, p. 450 — Checkpoint 5.6

Choose before you reveal. The function is the same rising sine as in Example 5.6, so the upper sum uses the other end of each strip: the right end.

You do not need to add six new sines. The right sum drops the height sine 0 and adds the height sine of pi over 2, so it is the lower sum plus pi over 12 times 1, about 0.863 plus 0.262, which is 1.125.

The midpoint option is a good distractor. The midpoint sum, about 1.003, is much closer to the true area, 1, but it is not an upper sum: it is not guaranteed to be above the area, and closeness is not what the question asked for.

45. Trap: left endpoints always give the lower sum

Trap

The trap

A rule remembered from rising curves, applied to a falling one:

\[ f(x) = 10 - x^2 \text{ on } [1, 2] \]

\[ \text{lower sum} = L_4 = 8.03125 \; ? \]

Wrong. That is the upper sum.

The fix

On a falling curve the LEFT end of each strip is the highest point. The lower sum uses right ends. Check the direction first, with the derivative if you must.

\[ f' < 0 \;\Longrightarrow\; \text{lower} = R_n \]

\[ R_4 = 7.28125 < 8.03125 = L_4 \]

The slogan left is low comes from rising curves, where it is true, and it gets applied without checking to falling ones, where it is backwards.

For ten minus x squared on 1 to 2 the curve falls, so the left end of each strip is its highest point and the left sum, 8.03125, is the upper sum. The lower sum is the right sum, 7.28125.

The fix is one line at the start of every upper-or-lower problem: say whether the function is increasing or decreasing on the interval, and why. If the graph is not obvious, the sign of the derivative settles it.

46. When neither endpoint rule is an upper or lower sum

Counterexample

Discussion prompt

Find a function and an interval where the left sum is neither the upper sum nor the lower sum. Use the sine arch on [0, π] with n = 3 from the upper-and-lower picture, and give the numbers.

Try it before you reveal: you need a curve that turns around, so that some strip has its highest or lowest point in an awkward place. The sine arch on 0 to pi is the natural candidate.

With three strips, both endpoint sums come out equal, about 1.814, because the arch is symmetric. The lower sum, using each strip's smallest value, is about 0.907, and the upper sum is about 2.861. The endpoint sums land strictly between.

So for a curve that turns around, an endpoint sum is just an estimate with no guarantee of direction. The true area here is 2, and 1.814 happens to be close, but nothing about the method promised that.

47. How many strips for a tight bracket?

Estimation

\[ f(x) = x^2 \text{ on } [0, 2]: \quad R_n - L_n = \frac{2}{n}(4 - 0) = \frac{8}{n} \]

Predict first

The true area is trapped between Lₙ and Rₙ. How many strips make that bracket narrower than 0.01?

  • 80
  • 800
  • 8,000
  • 80,000

Correct: 800

Why: The bracket's width is exactly 8 over n, from the gap formula. Eight over n is at most 0.01 when n is at least 800. With n = 6 the bracket is still 2.037 to 3.370, more than a whole unit wide.

Guess first. The gap formula gives the width of the bracket exactly: the width of a strip times the total rise of the curve. For x squared on 0 to 2 the rise is 4 and the strip width is 2 over n, so the bracket is 8 over n wide.

To make 8 over n smaller than a hundredth, n must be at least 800. That is a lot of rectangles for two decimal places, and it shows why the endpoint rules are rarely used for serious computation.

It also shows something about the midpoint rule: with only four strips it was already within 0.04. Guarantees and accuracy are different things; the bracket is slow, but it is certain.

48. The general Riemann sum

Section

Part 4

49. Any point in the strip will do

Concept

Figure (svg): A wavy curve on 0 to 6 cut into six strips of width 1. In each strip a sample point x-star is marked at a different place, a dot sits on the curve above it, and the rectangle's height is the curve's height at that dot.

Any point in each strip may set the height. The dashed stems show where each sample was taken; no two strips need to agree.

Nothing forces the height to come from an end. Pick any point in each strip, call it x-star with the strip's number, and use the function's value there.

\[ A \approx \sum_{i=1}^{n} f(x_i^*)\,\Delta x, \qquad x_i^* \in [x_{i-1}, x_i] \]

OpenStax Calculus Volume 1, §5.1 Approximating Areas §5.1, p. 448 — Forming Riemann Sums

Look at the figure. In each strip a dot marks a sample point chosen somewhere in the strip, and the rectangle's height is the curve's height above that dot. The sample points are in different places in different strips, and that is allowed.

The sample point in strip i is written x sub i with a star. The star means you may choose it anywhere in the strip. Left, right and midpoint rules are three particular, systematic choices.

A sum of this form, function values at sample points times widths, is called a Riemann sum. Every approximation in this lesson has been one.

50. Reading the Riemann sum

Notation

Annotate

On: \( \sum_{i=1}^{n} f(x_i^*)\,\Delta x \)

  • The sample point: any point of the i-th strip [xᵢ₋₁, xᵢ]. The star says 'your choice'.
  • The rectangle's height. Left, right and midpoint sums are three particular choices.
  • The common width (b − a)/n of a regular partition.
  • After Bernhard Riemann, the 19th-century mathematician who made the idea precise.

OpenStax Calculus Volume 1, §5.1 Approximating Areas §5.1, p. 448 — Definition of a Riemann sum

Step through the notes. The formula looks identical to the left and right sums; the only change is the star, which says the sample point is yours to choose.

Think of it as a template with one slot. Put the left end in the slot and you get L sub n; the right end, R sub n; the middle, M sub n; the highest point on each strip, an upper sum. Every rule in this section fills the same slot differently.

Bernhard Riemann made the idea precise in the 1850s, and the definite integral you will meet in Section 5.2 is still defined this way today.

51. The area is a limit

Concept

Area under the curve — For a continuous, nonnegative function on a closed interval, the limit of its Riemann sums as the number of strips grows without bound.

\[ A = \lim_{n \to \infty} \sum_{i=1}^{n} f(x_i^*)\,\Delta x \]

Two subtleties. This is a limit of a sum whose number of terms grows, and for now the familiar limit techniques apply to it. And for a continuous function the limit exists and is the same for every choice of sample points, although the proof is beyond this course.

OpenStax Calculus Volume 1, §5.1 Approximating Areas §5.1, pp. 448-449 — Definition of the area under a curve

This is the definition the whole lesson has been building toward. The area under a continuous, nonnegative curve is not an approximation; it is the number the Riemann sums approach as the strips get thinner.

The first subtlety is that this limit is unusual: as n grows, both the number of terms and each term change. You cannot take the limit of each term separately, as a trap later in the lesson shows. For now, the limit techniques you know will work once the sum is written as a formula in n.

The second subtlety is the choice of sample points. For a continuous function it does not matter: every choice gives the same limit. Without that fact, area would not be well defined.

52. Left and right sums close in from both sides

Picture it

Figure (svg): Two rows of dots for n from 1 to 40: the right sums of x squared on 0 to 2 falling from 8 toward a dashed line at 8 over 3, and the left sums rising from 0 toward the same line.

Every left sum is too small and every right sum too big, and both close in on the same number. That common limit is the area.

For x squared on 0 to 2, every left sum is too small and every right sum too big. As n grows the two rows of dots squeeze together onto one height, eight thirds, and any other choice of sample point is trapped between them.

The upper row of dots is the right sums for x squared on 0 to 2, starting at 8 when n is 1 and falling. The lower row is the left sums, starting at 0 and rising. The dashed line is at eight thirds.

Every other Riemann sum on the same strips has each height between the left and right heights, because x squared rises. So every Riemann sum lies between the two rows of dots.

As n grows the two rows squeeze together onto the dashed line, and anything trapped between them is squeezed onto it too. That is the squeeze theorem at work, and it is why the choice of sample point cannot change the limit here.

53. Exercise 24: a left sum in sigma notation

Fill the middle

Write the left-endpoint sum with thirty strips for x squared on the interval from 1 to 2, without evaluating it. The index starts at zero so the sample point is simply one plus i times the width.

Fill in the blanks

L_29 = \sum_1/30^___} } \left(1 + \frac______\right)^2 \cdot ___

Why: The width is (2 − 1)/30 = 1/30. Left endpoints are x₀ to x₂₉, so with the index starting at 0 it stops at 29: thirty terms. Starting at 0 and stopping at 30 would give 31 rectangles, the last one outside the interval.

OpenStax Calculus Volume 1, §5.1 Approximating Areas §5.1, p. 451 — Exercise 24

The width is the interval length over the number of strips: one over thirty. That fills the second blank.

With the index starting at 0, the sample points are one plus zero thirtieths, one plus one thirtieth, and so on. The left endpoints stop one strip before the end, at point 29, so the top limit is 29. Count the terms: 0 to 29 is thirty terms, one per strip.

Starting the index at 0 is a legitimate choice that makes left sums look clean. If you prefer starting at 1, write i minus 1 in place of i in the sample point and stop at 30. Both describe the same thirty rectangles.

54. Three rules on the same four strips

Comparison

Fill in the table for x squared on 0 to 2 with four strips; the exact area is eight thirds, about 2.667.

Comparison matrix

rulevalueerror (value − 8/3)
left, L₄1.75−0.917
right, R₄3.75+1.083
midpoint, M₄2.625−0.042
average of L₄ and R₄2.75+0.083

Fill in the blanks from what you have already computed: the right sum from Example 5.4 and the midpoint sum from the midpoint slide. The error column is the value minus eight thirds.

Compare the errors. Left and right are wrong by nearly a whole unit, in opposite directions. The midpoint is wrong by only 0.042. The average of the left and right sums, 2.75, is better than either, but still twice as far off as the midpoint.

That average has a name, the trapezoidal rule, because it is the same as replacing each rectangle by a trapezoid. It and the midpoint rule are the two workhorses of numerical integration, which Calculus II studies in detail.

55. Computing an area exactly

Section

Part 5

56. The plan: a formula in n, then the limit

Concept

The limit of Riemann sums cannot be taken while the sum is a string of terms whose length is n. First collapse the sum to a formula in n with the rules of Part 1, and only then let n grow.

\[ R_n = \sum_{i=1}^{n} f(a + i\Delta x)\,\Delta x \;\longrightarrow\; \text{closed form in } n \;\longrightarrow\; \lim_{n\to\infty} \]

Right endpoints are the usual choice, since the sample point is simply a plus i times delta x.

Here is the order of operations for an exact area. You cannot take a limit of a sum while it is still a list of n terms, because both the terms and their number change with n.

So you use the tools from Part 1 first: write the sample points, write the heights in terms of i, pull out everything that does not depend on i, and replace each remaining sum by its formula. What is left is an ordinary expression in n, and its limit is a limit at infinity of the kind you met in Section 4.6.

Right endpoints are the usual choice because the sample point, a plus i times delta x, needs no correction. The next four slides carry out the plan for three different curves.

57. The area under x squared on zero to two, exactly

Worked example

Find the area under f(x) equal to x squared on 0 to 2 as the limit of right-endpoint sums.

Width and sample points

Why: Two units in n strips.

\[ \Delta x = \frac{2}{n}, \qquad x_i = \frac{2i}{n} \]

Write the right sum

Why: Height times width for each strip.

\[ R_n = \sum_{i=1}^{n} \left(\frac{2i}{n}\right)^2 \frac{2}{n} \]

Pull the constants out

Why: Everything except i squared is constant in the sum.

\[ R_n = \frac{8}{n^3}\sum_{i=1}^{n} i^2 \]

Apply the sum-of-squares formula

Why: Now the sum is a formula in n.

\[ R_n = \frac{8}{n^3} \cdot \frac{n(n+1)(2n+1)}{6} = \frac{4(n+1)(2n+1)}{3n^2} \]

Take the limit

Why: Top and bottom are both degree two; the leading coefficients are 8 and 3.

\[ A = \lim_{n\to\infty} \frac{4(2n^2 + 3n + 1)}{3n^2} = \frac{8}{3} \]

Check against Example 5.4

Why: At n equal to 4 the formula must give the right sum found earlier, and at a thousand strips it is close to eight thirds.

\[ R_4 = \frac{4(5)(9)}{3(16)} = \frac{180}{48} = 3.75, \qquad R_{1000} \approx 2.6707 \]

Follow the plan step by step. The width is 2 over n and the right endpoints are 2i over n. Squaring the endpoint and multiplying by the width gives 8 i squared over n cubed for each rectangle.

Everything except i squared is constant in the sum, so it comes out front, and the sum that remains is exactly the sum-of-squares formula. After that the right sum is an ordinary fraction in n.

The limit is a limit at infinity of a rational function with numerator and denominator both of degree two, so it is the ratio of leading coefficients, 8 over 3. The check matters: plugging n equal to 4 into your formula must give 3.75 from Example 5.4. If it does not, the formula is wrong, whatever the limit looks like.

58. The same area from the left

Worked example

Repeat with left endpoints, to see that the choice of sample point does not change the limit.

Write the left sum

Why: Sample points x sub i minus 1, which is two times i minus 1, over n.

\[ L_n = \sum_{i=1}^{n} \left(\frac{2(i-1)}{n}\right)^2 \frac{2}{n} = \frac{8}{n^3}\sum_{j=0}^{n-1} j^2 \]

Use the formula up to n minus 1

Why: The j equal to 0 term is zero; replace n by n minus 1 in the formula.

\[ L_n = \frac{8}{n^3} \cdot \frac{(n-1)n(2n-1)}{6} = \frac{4(n-1)(2n-1)}{3n^2} \]

Take the limit

Why: Leading coefficients 8 and 3 again.

\[ \lim_{n\to\infty} L_n = \frac{8}{3} \]

Check against Example 5.4

Why: At n equal to 4.

\[ L_4 = \frac{4(3)(7)}{3(16)} = \frac{84}{48} = 1.75 \]

This slide repeats the computation with left endpoints to show the definition's promise in action: a different choice of sample points, the same limit.

The one new step is the index shift. Left endpoints bring in i minus 1, so it is cleanest to rename it j, running from 0 to n minus 1. The j equal to 0 term is zero, and the rest is the sum-of-squares formula with n minus 1 in place of n.

The resulting formula differs from the right sum's, but its leading behaviour is the same, and the limit is again eight thirds. The check at n equal to 4 recovers 1.75 from Example 5.4.

59. Table 5.1 explained: the cubic's area is exactly 8

Worked example

Find the right-endpoint sum for f(x) equal to x minus 1, cubed, plus 4 on 0 to 2 as a formula in n, and explain the book's table.

Width and sample points

Why: Two units in n strips.

\[ \Delta x = \frac{2}{n}, \qquad f(x_i) = \left(\frac{2i}{n} - 1\right)^3 + 4 \]

Expand the cube

Why: Binomial expansion of the bracket.

\[ \left(\frac{2i}{n} - 1\right)^3 = \frac{8i^3}{n^3} - \frac{12i^2}{n^2} + \frac{6i}{n} - 1 \]

Sum each piece with its formula

Why: Cubes, squares, integers, and the constant minus 1 added n times.

\[ \sum_{i=1}^{n}\left(\frac{2i}{n} - 1\right)^3 = \frac{2(n+1)^2}{n} - \frac{2(n+1)(2n+1)}{n} + 3(n+1) - n \]

Simplify

Why: The first two terms combine to minus 2 times n plus 1.

\[ = -2(n+1) + 3(n+1) - n = 1 \]

Assemble the right sum

Why: Add the constant 4 times n, then multiply by the width.

\[ R_n = \frac{2}{n}\left(1 + 4n\right) = 8 + \frac{2}{n} \]

Take the limit

Why: The correction dies away.

\[ A = \lim_{n\to\infty}\left(8 + \frac{2}{n}\right) = 8 \]

Check against Table 5.1

Why: The formula reproduces every right sum in the table; by the gap formula the left sums are 8 minus 2 over n.

\[ R_4 = 8.5, \; R_8 = 8.25, \; R_{32} = 8.0625; \quad L_n = R_n - \frac{2}{n}(5 - 3) = 8 - \frac{2}{n} \]

OpenStax Calculus Volume 1, §5.1 Approximating Areas §5.1, p. 448 — Table 5.1

The book's table showed the left and right sums of this cubic approaching 8, but only as a numerical pattern. This slide proves it, and finds a formula for every right sum at once.

The cube expands into four powers of i: cubed, squared, to the first, and a constant. Each gets its formula, and then a small miracle happens in the simplification: everything collapses to 1. That reflects the cubic's symmetry about x equal to 1, where its bend on the left cancels its bend on the right.

So the right sum is 8 plus 2 over n, and its limit is 8. Check it against the table: n equal to 4, 8 and 32 give 8.5, 8.25 and 8.0625, the book's values exactly. The gap formula then gives the left sums as 8 minus 2 over n, which matches the table's other column.

60. Trap: the limit taken term by term

Trap

The trap

Every term of the sum below goes to zero, so:

\[ \lim_{n\to\infty}\sum_{i=1}^{n}\frac{i^2}{n^3} = \sum \lim_{n\to\infty}\frac{i^2}{n^3} = 0 \]

Wrong. The number of terms is growing too.

The fix

The terms shrink but there are more and more of them. Collapse the sum first, then take the limit.

\[ \frac{1}{n^3}\sum_{i=1}^{n} i^2 = \frac{(n+1)(2n+1)}{6n^2} \]

\[ \lim_{n\to\infty} \frac{(n+1)(2n+1)}{6n^2} = \frac13 \]

Each term, i squared over n cubed, does go to zero as n grows with i fixed. The wrong line then adds up a lot of zeros and gets zero. The mistake is that the number of terms is growing at the same time.

Here is a way to feel why it fails. The largest term, with i equal to n, is one over n, which is small. But there are n terms, and many of them are nearly that big, so the total does not vanish.

The fix is the order of operations from the plan slide: collapse the sum to a formula in n first, and only then take the limit. Done that way, the limit is one third, the same one third that appeared in the size of the sum-of-squares formula.

61. The region of Figure 5.5, exactly

Worked example

Find the exact area under f(x) equal to x squared over 2 on 0 to 3, the region approximated by L sub 6 and R sub 6 in Part 2.

Width and sample points

Why: Three units in n strips.

\[ \Delta x = \frac{3}{n}, \qquad x_i = \frac{3i}{n} \]

Write the right sum

Why: Half the square of each sample point, times the width.

\[ R_n = \sum_{i=1}^{n} \frac12\left(\frac{3i}{n}\right)^2 \frac{3}{n} \]

Pull the constants out

Why: One half times nine times three is twenty-seven halves.

\[ R_n = \frac{27}{2n^3}\sum_{i=1}^{n} i^2 \]

Apply the formula and simplify

Why: Twenty-seven over twelve reduces to nine over four.

\[ R_n = \frac{27}{2n^3} \cdot \frac{n(n+1)(2n+1)}{6} = \frac{9(n+1)(2n+1)}{4n^2} \]

Take the limit

Why: Leading coefficients 18 and 4.

\[ A = \lim_{n\to\infty}\frac{9(2n^2 + 3n + 1)}{4n^2} = \frac{18}{4} = 4.5 \]

Check against the six-strip sums

Why: At n equal to 6 the formula gives the right sum from Part 2, and 4.5 lies between the two six-strip sums.

\[ R_6 = \frac{9(7)(13)}{4(36)} = \frac{819}{144} = 5.6875, \qquad 3.4375 < 4.5 < 5.6875 \]

This returns to the region you estimated with six strips in Part 2 and finds its area exactly. The width is 3 over n and the right endpoints are 3i over n.

Collecting constants, one half times nine times three gives twenty-seven halves over n cubed, in front of the sum of i squared. The formula then reduces the right sum to 9 times n plus 1 times 2n plus 1, over 4n squared.

The limit is the ratio of leading coefficients, 18 over 4, which is 4.5. The check closes the loop: at n equal to 6 the formula gives 5.6875, the value you computed by hand in Part 2, and 4.5 sits between the two six-strip sums, as it must.

62. Before summing: which formula will you need?

Step zero

\[ \text{area under } f(x) = x^3 \text{ on } [0, 1] \]

Discussion prompt

Before writing a single term: what is the width, what power of i will appear, and which of the three formulas will collapse the sum? Then finish it.

Plan before you compute. The width is one over n, the right endpoints are i over n, and cubing an endpoint puts i cubed in the numerator. So the cube formula is the one that will collapse the sum.

Carry it out and the right sum is n plus 1, squared, over 4n squared, which tends to one quarter. Together with one half for the area under x on 0 to 1 and one third for x squared, a pattern is emerging.

The check at n equal to 4 gives 25 over 64, about 0.39, above one quarter. That is right, because x cubed rises, and right sums on a rising curve overshoot.

63. Where the method runs out

Intuition

The exact method worked because x squared and x cubed turn into sums of powers of i, and there are formulas for those. Try sine x on 0 to pi over 2 and the right sum is a sum of sines of multiples of an angle; the method needs a closed form for that, which is not among the three.

\[ R_n = \frac{\pi}{2n}\sum_{i=1}^{n}\sin\left(\frac{i\pi}{2n}\right) \quad \text{(no power-sum formula applies)} \]

Section 5.3 will replace the whole limit with a two-line computation, the Fundamental Theorem of Calculus. Riemann sums stay as the definition of what that computation means.

The exact method has a hidden requirement: after substituting the sample points, the sum must be one you have a formula for. Polynomials give sums of powers of i, and the three formulas cover those up to cubes.

Sine breaks this. Its right sum is a sum of sines of equally spaced angles, and none of the three formulas applies. There is a trigonometric identity that handles this particular sum, but for most functions there is nothing at all.

That is the motivation for Section 5.3. The Fundamental Theorem of Calculus computes the same limit using an antiderivative evaluated at two points, and it works for any continuous function with a known antiderivative. Riemann sums remain the definition of what is being computed.

64. Rate times time: carbon dioxide

Real world

Figure (svg): A bar chart of five bars, each ten years wide starting at 1964, 1974, 1984, 1994 and 2004, with heights 1.07, 1.34, 1.40, 1.87 and 2.07 parts per million per year.

Rate times time is amount, so each bar's area is that decade's increase, and the total area is the total increase.

Discussion prompt

Exercise 38 gives the average yearly rise in atmospheric carbon dioxide for five decades. Estimate the total rise from 1964 to 2013, and say what the rectangles in the chart are.

OpenStax Calculus Volume 1, §5.1 Approximating Areas §5.1, p. 452 — Exercise 38, Table 5.2

Look at the chart before answering. Each bar is ten years wide, and its height is a rate, parts per million per year. Width times height is years times parts per million per year, which is parts per million: an amount.

So each bar's area is the rise in carbon dioxide during that decade, and the total area is the total rise from 1964 to 2013, about 77.5 parts per million.

This is the reason areas matter in calculus. Whenever you know a rate over time, the area under the rate graph is the total change. The bars are a Riemann sum with a width of ten years, and finer data would give thinner bars and a better total.

65. Putting it together

Section

Part 6

66. Pattern: from rectangles to an exact area

Pattern

Figure (svg): A flow diagram of five boxes joined by arrows: find delta x; write the sample point x sub i; write f of x sub i in terms of i; pull constants out and use the sum formulas; take the limit as n goes to infinity.

Four algebra steps turn the sum into a formula in n; only then is the limit taken.
  1. Width first: delta x is b minus a, over n.
  2. Sample points: xᵢ = a + iΔx for right ends; subtract one Δx for left ends, half a Δx for midpoints.
  3. Write the sum with f(xᵢ) expanded into powers of i; for an estimate, stop here and add the numbers.
  4. Collapse the sum with the rules 5.1 to 5.5 and the three power-sum formulas.
  5. Only now take the limit as n goes to infinity; check the formula at a small n against a direct sum.

The flow diagram is the order to work in, and the list says what each step involves. The first three steps are all you need for an estimate: once the sum is written with numbers, add it up.

For an exact area, keep going. Collapse the sum with the rules and formulas, then take the limit. The rule that matters most is in the middle of the diagram: a closed form in n first, and the limit only at the very end.

Always finish by checking your formula at a small n against a direct sum. It is the single best way to find an algebra slip before it reaches the answer.

67. Put the exact-area computation in order

Ranking

Put in order

Order the steps for finding the area under x² on [0, 3] exactly.

  1. Set Δx = 3/n and xᵢ = 3i/n
  2. Write Rₙ = Σ (3i/n)² · (3/n)
  3. Pull out 27/n³, leaving Σ i²
  4. Replace Σ i² by n(n + 1)(2n + 1)/6
  5. Let n → ∞ to get 9

Why: Width and sample points come first because the sum is built from them. Constants must be pulled out before the formula can be matched, and the limit waits until the sum is a closed form: 27(n + 1)(2n + 1)/(6n²) → 54/6 = 9.

Order the steps before checking. Everything depends on the width and the sample points, so they come first. The sum is then written from them.

The constants must come out before the formula can be recognised: until the 27 over n cubed is in front, the sum does not look like the sum of i squared.

The limit is last, and it gives 9, which is one third of 3 cubed. Compare with eight thirds for the interval 0 to 2: the area under x squared from 0 to b is b cubed over 3. Section 5.3 will explain that pattern.

68. Check: a left sum

Check

Check your understanding

What is L₄ for f(x) = 2 − |x| on [−2, 2] (Exercise 20)?

  • A. 4 (correct)
  • B. 3
  • C. 2
  • D. 8

Answer: A

Why: The width is 1 and the left endpoints are −2, −1, 0 and 1, with heights 0, 1, 2 and 1. So L₄ = 1(0 + 1 + 2 + 1) = 4. The region is a triangle of base 4 and height 2, area 4, and R₄ is also 4: the errors on the two sides of the peak cancel.

Why B tempts people
This leaves out one of the four heights; there are four strips, so four heights.
Why C tempts people
This multiplies by a width of 0.5, but the interval has length 4 and n = 4, so each strip is 1 wide.
Why D tempts people
This doubles the answer, as if the width were 2; four strips of width 1 cover the interval.

OpenStax Calculus Volume 1, §5.1 Approximating Areas §5.1, p. 451 — Exercise 20

Choose an answer before revealing. The width is 1 because the interval has length 4 and there are 4 strips, and the left endpoints are minus 2, minus 1, 0 and 1.

The heights of two minus the absolute value of x at those points are 0, 1, 2 and 1, so the left sum is 4. The graph is a tent, and the region under it is a triangle with base 4 and height 2, whose area is also 4.

The right sum is 4 too. On the rising half the left rule undershoots and on the falling half it overshoots, by the same amount, so the errors cancel. Exercise 20 asks you to notice exactly that.

69. Check: which sum is the lower sum

Check

Check your understanding

For f(x) = 1/x on [1, 3] with n = 4, which Riemann sum is the lower sum?

  • A. The left-endpoint sum
  • B. The right-endpoint sum (correct)
  • C. The midpoint sum
  • D. It depends on the number of strips

Answer: B

Why: 1/x decreases on [1, 3], so on each strip the smallest value is at the right end. The right-endpoint sum is the lower sum for every n.

Why A tempts people
The left end is where a decreasing function is largest: that is the upper sum.
Why C tempts people
The midpoint value is neither the smallest nor the largest on a strip of a decreasing function.
Why D tempts people
For a monotone function the extreme values always sit at the ends of each strip, whatever n is.

Commit to an answer first. The only fact you need is whether one over x rises or falls on 1 to 3. It falls: bigger x, smaller fraction.

For a falling function the smallest value on each strip is at the right end, so the right-endpoint sum is the lower sum. That holds for four strips or four thousand.

If you chose the left sum, you used the rising-curve slogan on a falling curve, the trap from Part 3. Say the direction out loud before choosing an end.

70. Check: a limit of a sum

Check

Check your understanding

What is the limit, as n → ∞, of (8/n³) Σ from i=1 to n of i²?

  • A. 0
  • B. 8/3 (correct)
  • C. 8
  • D. The limit does not exist

Answer: B

Why: Collapse first: (8/n³) · n(n + 1)(2n + 1)/6 = 4(n + 1)(2n + 1)/(3n²), which tends to 8/3. It is the right-endpoint sum for x² on [0, 2], the area found in Part 5.

Why A tempts people
This takes the limit of each term, i²/n³ → 0, while ignoring that the number of terms grows too.
Why C tempts people
This treats the sum of squares as if it were about n³, not n³/3; the formula's leading term is 2n³/6.
Why D tempts people
The closed form is a ratio of two quadratics in n, which always has a finite limit.

Answer before revealing. Recognise the expression first: it is the right-endpoint sum for x squared on 0 to 2, written with the constants already pulled out.

Replace the sum of squares by its formula, and the expression becomes 4 times n plus 1 times 2n plus 1, over 3n squared. Its limit is the ratio of leading coefficients, eight thirds.

The distractor zero comes from taking the limit term by term. Each term does go to zero, but the number of terms grows, and the total settles at eight thirds, not zero.

71. Explain why any sample point gives the same area

Explain it to yourself

Discussion prompt

For an increasing function, explain in two or three sentences why every Riemann sum on the same partition lies between Lₙ and Rₙ, and why that forces all of them to the same limit.

Write your two or three sentences before you look. This is the reasoning behind the definition's claim that the choice of sample point does not matter, at least for an increasing function.

The first idea is trapping. Wherever the sample point is in a strip, an increasing function's value there lies between its values at the two ends. Add over all strips and the whole Riemann sum lies between the left sum and the right sum.

The second idea is squeezing. The gap between the left and right sums is the width times the total rise, which goes to zero as the strips get thinner. Anything trapped between two sums with the same limit must share that limit.

72. Exit ticket

Exit ticket

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

Discussion prompt

Find R₃ with three strips, then find the exact area as a limit of right sums, and check your answer with geometry.

Do all three parts before revealing. The right sum with three strips uses heights at 1, 2 and 3, which are 3, 5 and 7, each times a width of 1, giving 15.

For the exact area, the width is 3 over n and the right endpoints are 3i over n. The sum splits into a multiple of the sum of i and a constant piece, which is 3. The formula becomes 9 times n plus 1, over n, plus 3, and its limit is 12.

The geometry check is a trapezoid with parallel sides 1 and 7 and width 3, area 12. For a straight line the right sum overshoots by exactly half the gap formula, half of 1 times 6, which is 3, and 15 minus 12 is indeed 3.

73. Recap

Recap

ideaformulause it for
sigma notationΣ aᵢ from i = 1 to nwriting long sums compactly
power sumsn(n+1)/2, n(n+1)(2n+1)/6, n²(n+1)²/4collapsing a sum to a formula in n
left / right / midpointΣ f(sample point)Δxestimating an area
upper / lower sumsmax / min of f on each stripa guaranteed bracket
gap formulaRₙ − Lₙ = Δx(f(b) − f(a))how wide the bracket is
arealimit of Riemann sumsthe exact value

\[ A = \lim_{n\to\infty}\sum_{i=1}^{n} f(x_i^*)\,\Delta x \]

Next, Section 5.2 names this limit the definite integral, and lets the function go negative.

Stewart, Calculus: Early Transcendentals 8e, §5.1 Areas and Distances §5.1, pp. 366-377 — the same material in Stewart

Stewart, Calculus: Early Transcendentals 8e, Appendix E — Sigma Notation Appendix E, pp. A34-A38 — sigma notation in Stewart

The table is the lesson in six rows. Sigma notation and the power-sum formulas are the tools; left, right and midpoint sums are estimates; upper and lower sums are guaranteed brackets; the gap formula tells you how tight a bracket is; and the limit of Riemann sums is the exact area.

The single formula under the table is the one to carry forward. Every other result in this chapter, including the Fundamental Theorem, is a statement about that limit.

In Section 5.2 the limit gets its name, the definite integral, and the function is allowed to dip below the axis, where rectangles count negatively. Stewart's Section 5.1 and his appendix on sigma notation cover this same material if you want a second explanation.

Sources

  1. OpenStax Calculus Volume 1, §5.1 Approximating Areas — Strang & Herman, OpenStax / Rice University, CC BY-NC-SA 4.0, pp. 438-455
  2. Stewart, Calculus: Early Transcendentals 8e, §5.1 Areas and Distances — James Stewart, Cengage Learning, 2016, pp. 366-377
  3. Stewart, Calculus: Early Transcendentals 8e, Appendix E — Sigma Notation — James Stewart, Cengage Learning, 2016, pp. A34-A38

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

Book on Wyzant · Text (657) 465-8108