Root finding and numerical integration
引入| English | 释义 | 拼音 |
|---|---|---|
| trapezium rule/trəˈpiːzɪəm ruːl/ | 梯形法则 | tī xíng fǎ zé |
| ordinates | 纵坐标 | zòng zuò biāo |
| width | 宽度 | kuān dù |
| convex/kɒnˈveks/ | 下凸的 | xià tū de |
| concave/kɒnˈkeɪv/ | 上凸的 | shàng tū de |
Jupiter's path on a clay tablet, 2016
- In 2016, the historian Mathieu Ossendrijver showed that Babylonian astronomers, between 350 and 50 BC, found how far Jupiter moved across the sky by working out the area of a trapezium under a speed–time graph. That is the idea behind the trapezium rule, about 1,400 years before it appears in European books.
- Many integrals, such as $\displaystyle\int_0^2 \sqrt{1 + x^3}\,dx$, have no neat antiderivative. The trapezium rule 梯形法则 gives a good estimate from a few $y$-values.
Strips and ordinates
- Split $a \le x \le b$ into $n$ strips of equal width 宽度 $h = \dfrac{b - a}{n}$. The $y$-values at the strip edges, $y_0, y_1, \dots, y_n$, are the ordinates 纵坐标.
- $n$ strips need $n + 1$ ordinates: 4 strips on $0 \le x \le 2$ give $h = 0.5$ and $x = 0, 0.5, 1, 1.5, 2$.
You use 6 strips on $1 \le x \le 4$. What are $h$ and the number of ordinates?
难度 1/5 · 今日首次完成可获 100 金币及少量额外奖励
$h = \tfrac{4 - 1}{6} = 0.5$, and $n$ strips need $n + 1$ edges.
The rule
- Each strip becomes a trapezium with area $\tfrac{h}{2}(y_{i} + y_{i+1})$. Adding them gives
- $\displaystyle\int_a^b y\,dx \approx \frac{h}{2}\left[y_0 + y_n + 2(y_1 + y_2 + \dots + y_{n-1})\right]$ — the end ordinates once, the middle ones twice.
Root finding and numerical integration
For equal strip width h, the trapezium estimate is h/2 times the sum of the two endpoint heights plus twice the internal heights
Compare the model with the worked case and explain one change.
Use the trapezium rule with 2 strips to estimate $\displaystyle\int_0^2 x^2\,dx$.
难度 2/5 · 今日首次完成可获 140 金币及少量额外奖励
$h = 1$: $\tfrac{1}{2}[0 + 4 + 2(1)] = 3$.
Worked example: a curve with no neat integral
Use the trapezium rule with 4 strips to estimate $\displaystyle\int_0^2 \sqrt{1 + x^3}\,dx$, giving your answer to 3 s.f.
- Width: $h = \dfrac{2 - 0}{4} = 0.5$.
- Table: at $x = 0, 0.5, 1, 1.5, 2$, $y = 1, 1.0607, 1.4142, 2.0917, 3$.
- Rule: $\dfrac{0.5}{2}\left[1 + 3 + 2(1.0607 + 1.4142 + 2.0917)\right] = 0.25 \times 13.1332 = 3.28$ (3 s.f.).
Use 4 strips to estimate $\displaystyle\int_0^1 \frac{1}{1 + x^2}\,dx$, to 4 d.p.
难度 2/5 · 今日首次完成可获 140 金币及少量额外奖励
$h = 0.25$; $y = 1, 0.9412, 0.8, 0.64, 0.5$; $0.125[1.5 + 2(2.3812)] = 0.7828$.
Over or under?
- If the curve is convex 下凸的 (bends upwards), each straight top lies above it, so the rule overestimates: the true value above is about 3.24.
- If the curve is concave 上凸的 (bends downwards), the tops lie below it, so the rule underestimates.
Why does the trapezium rule overestimate $\displaystyle\int_0^2 x^2\,dx$?
难度 1/5 · 今日首次完成可获 100 金币及少量额外奖励
Chords of a convex curve lie above it. Increasing alone does not decide it.
Improving the estimate
- More strips make each straight top closer to the curve, so the error shrinks: halving $h$ cuts it to about a quarter.
- For $\displaystyle\int_0^2 2^x\,dx$, 4 strips give 4.371 against the exact 4.328; 8 strips give about 4.339.
Doubling the number of strips usually makes the trapezium estimate more accurate.
难度 1/5 · 今日首次完成可获 100 金币及少量额外奖励
Narrower strips follow the curve more closely; the error falls to about a quarter.
Put the steps of a trapezium-rule answer in order.
难度 3/5 · 今日首次完成可获 200 金币及少量额外奖励
Width first, then the table, then the rule.
Counting strips as ordinates. "Use 4 strips" means 5 $y$-values and $h = \tfrac{b - a}{4}$. Students who use $h = \tfrac{b - a}{5}$, or who double the first and last ordinates, get the wrong answer even with a correct table.
Find $h = \tfrac{b - a}{n}$, tabulate the $n + 1$ ordinates, then use $\tfrac{h}{2}[\text{ends} + 2 \times \text{middles}]$. A curve that bends upwards gives an overestimate, one that bends downwards an underestimate, and more strips give a better answer.
Use the trapezium rule with 4 strips to estimate $\displaystyle\int_0^2 2^x\,dx$. The exact value is 4.328 (3 d.p.). Explain why your estimate is too big.
Rule plus reliability.