Why the derivative of the area under a curve is the height of the curve, how that turns area problems into integration problems, and how to find areas of loops given parametrically or in polar coordinates.
Up to now integration has been a purely formal game: find a function with a given derivative. Here it acquires its most important meaning. Hardy shows that the area under the curve y=ϕ(x), viewed as a function of the right-hand end x, has derivative ϕ(x). So areas can be computed by integration, and every integral we have learned to find becomes a formula for an area. This is the first form of the fundamental theorem of calculus. Hardy is careful to say what he is assuming: here “area” is taken on trust, with its common-sense properties; Chapter VII defines it properly.
The idea
Let Φ(x) be the area under the curve y=ϕ(x) from 0 to x (Hardy’s region ONPP0, with ϕ>0). Move x to x+h. The area grows by a thin strip NN′P′P. If the strip were a rectangle, its area would be hϕ(x), and {Φ(x+h)−Φ(x)}/h would be exactly ϕ(x). It is not quite a rectangle: there is a small sliver PRP′ between the curve and the horizontal line through P. The whole question is whether that sliver is small compared with h.
ExploreThe strip, the rectangle and the sliver
Try this.Slide ∣h∣ down from 1 to 0.001 and watch the readout: compare {Φ(x+h)−Φ(x)}/h with ϕ(x). Switch on “Zoom on the strip” and do it again. Try h negative, and try the steadily rising curve.
ϕ(x)=1.3+0.55sin2.3x+0.25sin6x, area Φ(x) from 0 to x
xy
Φ(x), area so far
rectangle h·φ(x)
sliver PRP′ (signed)
φ(x) ± λ(h)
Φ(x)=2.607. Strip: hΦ(x+h)−Φ(x)=0.78962, while ϕ(x)=0.9744. The difference μ(h)=−0.185 is bounded by λ(h)=0.424, and both shrink to 0 with h.
What to notice
The sliver (vermillion) always lies inside the green band of half-width λ(h), the largest distance of the arc PP′ from the horizontal through P, so its area is at most ∣h∣λ(h). As h→0, λ(h)→0 because ϕ is continuous, so the quotient tends to ϕ(x). In the zoomed view the curve over the strip looks flatter and flatter, and the strip becomes a rectangle. The curve need not rise or fall steadily; that is why Hardy uses λ(h) rather than the height k of the step.
Precise statements
Hardy assumes that a region such as ONPP0 has an area, a positive number, and that areas behave as common sense says: the area of a region cut into two pieces is the sum of their areas, a region inside another has smaller area, and a rectangle has area base times height.
TheoremThe ordinate is the derivative of the areaHardy: §148
Let ϕ be continuous and positive, and let Φ(x) be the area bounded by the curve y=ϕ(x), the axis of x and the ordinates at 0 and x. Then Φ is continuous and differentiable, and
Φ′(x)=ϕ(x).
In Hardy’s words, the ordinate of the curve is the derivative of the area, and the area is the integral of the ordinate. This gives a rule:
CorollaryComputing an areaHardy: §148
To find the area under y=ϕ(x) between the ordinates x=x1 and x=x2 (x1<x2), find any integral Φ of ϕ; the area is Φ(x2)−Φ(x1). If the curve lies below the axis, this difference is negative, and its absolute value is the area.
The rule holds because two integrals of ϕ differ by a constant (§130), so the area function, which vanishes at x1, is Φ(x)−Φ(x1). When the curve crosses the axis, Φ(x2)−Φ(x1) is a signed area: parts below the axis count negatively.
ExploreSigned area: sinx from 0 to 2π
Try this.Drag x from 0 to about 6.28. Watch the lower curve, F(x)=∫0xsintdt=1−cosx.
F(x)=∫axsintdt, a=0
tf(t)
xF(x)
f(t)
counts +
counts −
F(x)
x (drag)
F(4)=∫04sintdt=1.6536. Slope of F at x: F′(x)=f(x)=−0.7568.
What to notice
F rises to 2 at x=π, then falls back to 0 at x=2π: the part of the curve below the axis is counted negatively. The total geometric area between the curve and the axis is 2+2=4; it is found by integrating separately over [0,π] and [π,2π] (Hardy’s Ex. LIV. 7).
Loops given by parameters
Many curves are best described by x=ϕ(t), y=ψ(t). If x increases with t, the substitution rule turns the area under the curve into ∫ydtdxdt (Hardy’s Ex. LIV. 8). For a closed loop there is a neater formula, which measures the area as the region swept out by the radius from the origin: the radius moving from P(t) to P(t+dt) sweeps a thin triangle of area about 21(xdy−ydx).
PropositionArea of a loopHardy: Examples LIV, 9
If a closed curve x=ϕ(t), y=ψ(t) (with continuous derivatives), forming a single loop that no parallel to either axis meets in more than two points, is described once anticlockwise as t goes from t0 to t1, its area is the increase from t0 to t1 of any of
−∫ydtdxdt,∫xdtdydt,21∫(xdtdy−ydtdx)dt.
In polar coordinates r=f(θ), the area swept between the radii θ1 and θ2 is the increase of 21∫r2dθ (Ex. LIV. 18).
ExploreSweeping out a loop
Try this.Drag the slider to trace the loop of the folium x3+y3=3xy and watch the swept area grow towards 23. Then try the ellipse, the figure-of-eight loop and the cardioid.
x=1+t33t,y=1+t33t2,0≤t<∞
xy
region swept so far
radius OP
origin O
Swept so far: 21∫(xdy−ydx)≈0.3429. For the whole loop the exact area is 23a2=1.5.
What to notice
The radius from the origin sweeps exactly the region inside the loop when the origin is on or inside the loop and every ray from it meets the curve once more, as here. The readout adds up the thin triangles 21(xdy−ydx); at 100% it agrees with the exact area to the digits shown.
Proofs
ProofThe ordinate is the derivative of the area
Strategy
Compare the strip Φ(x+h)−Φ(x) with the rectangle hϕ(x). The difference is the sliver PRP′, whose area is at most ∣h∣ times the largest vertical distance λ(h) of the arc PP′ from the horizontal through P. Continuity of ϕ makes λ(h)→0.
Split the strip (take h>0)
By the additivity of area, Φ(x+h)−Φ(x) is the area of the strip NN′P′P, and this equals hϕ(x)+(PRP′), where (PRP′) is the area between the arc and the line y=ϕ(x), counted negatively where the arc is below the line. For h<0 the same holds with the strip on the left.
Bound the sliver
Let λ(h) be the largest value of ∣ϕ(t)−ϕ(x)∣ for t between x and x+h. The sliver lies in a rectangle of width ∣h∣ and height λ(h) (above or below the line), so ∣(PRP′)∣≤∣h∣λ(h). Write (PRP′)=hμ(h); then ∣μ(h)∣≤λ(h), and
Φ(x+h)−Φ(x)=h{ϕ(x)+μ(h)}.
λ(h)→0
Let ε>0. By continuity of ϕ at x there is δ>0 with ∣ϕ(t)−ϕ(x)∣<ε whenever ∣t−x∣<δ. So λ(h)≤ε when ∣h∣<δ.
Continuity and the derivative
From the last two steps, Φ(x+h)−Φ(x)→0 as h→0, so Φ is continuous; and hΦ(x+h)−Φ(x)=ϕ(x)+μ(h)→ϕ(x), so Φ′(x)=ϕ(x).
∎
The argument is rigorous given the existence of areas with the stated properties. The gap is exactly that assumption, which Chapter VII fills by defining the area (the definite integral) as a limit of sums. Note also what the proof gives for free: since Φ is an integral of ϕ, every continuous ϕhas an integral. (The theorem assumed ϕ>0; on a closed interval a continuous ϕ is bounded, so ϕ+K is positive for a suitable constant K, and if Φ is an integral of ϕ+K then Φ−Kx is an integral of ϕ.) This is the existence theorem promised in §130.
ProofArea of a loop: the triangle formula
Strategy
Cut the loop by vertical lines into an upper and a lower arc, use “area under the upper arc minus area under the lower arc”, and convert each to an integral in t; then average with the analogous horizontal computation.
Upper minus lower
Going anticlockwise, the lower arc is traversed with x increasing and the upper arc with x decreasing. The area under an arc traversed with x increasing is the increase of ∫ydtdxdt (substitution, §136); traversed backwards it gives minus that area.
Add
So the increase of ∫ydtdxdt over the whole circuit is (area under lower) − (area under upper) =−(area of the loop). This is the first formula.
The other two
The same argument with the roles of x and y exchanged gives the second (the sign flips because going anticlockwise is clockwise in the (y,x) plane). The third is their average.
Polar form
With x=rcosθ, y=rsinθ and θ as parameter, xdθdy−ydθdx=r2(cos2θ+sin2θ)=r2, the terms in dr/dθ cancelling.
∎
Worked examples and exercises
ExerciseUnder a parabola●○○Hardy: Examples LIV, 1
Find the area bounded by the parabola y=x2/4a (a>0), the axis of x and the ordinate x=ξ>0.
Hint
An integral of x2/4a is x3/12a; choose the constant so that it vanishes at x=0.
Solution
Φ(x)=x3/12a, with Φ(0)=0, so the area is ξ3/12a. Note that it is one third of the rectangle ξ⋅ξ2/4a with the same base and height, as Archimedes found.
ExerciseThe ellipse●●○Hardy: Examples LIV, 2
Show that the area of the ellipse x2/a2+y2/b2=1 is πab, (i) from y=b1−x2/a2 and (ii) from the parametric form x=acost, y=bsint.
Hint
(i) Use ∫α2−x2dx=21xα2−x2+21α2arcsin(x/α) from the lesson on parts.
Hint
(ii) xdtdy−ydtdx=ab.
Solution
(i) The upper half has area ab[21xa2−x2+21a2arcsinax]−aa=ab⋅21a2π=21πab. Double it.
(ii) 21∫02π(acost⋅bcost+bsint⋅asint)dt increases by 21ab⋅2π=πab. The widget’s ellipse has a=2, b=1, area 2π.
ExerciseThe loop of the folium●●○Hardy: Examples LIV, 11
Find the area of the loop of x3+y3=3axy.
Hint
Put y=tx: x=3at/(1+t3), y=3at2/(1+t3). The loop is described once as t goes from 0 to ∞.
Hint
xdy−ydx=x2d(y/x)=x2dt.
Solution
Since y/x=t, we have xdtdy−ydtdx=x2dtd(xy)=x2=(1+t3)29a2t2. An integral is 21∫(1+t3)29a2t2dt=−2(1+t3)3a2. Its increase from t=0 to t→∞ is 0−(−23a2)=23a2. (Strictly, we take the limit as t→∞; the widget traces the loop for a=1.)
Prove that the area of a loop of x=asin2t, y=asint is 34a2.
Hint
One loop is 0≤t≤π. Compute xy˙−yx˙ and simplify with sin2t=2sintcost, cos2t=2cos2t−1.
Solution
xy˙−yx˙=a2(sin2tcost−2sintcos2t)=a2(2sintcos2t−4sintcos2t+2sint)=2a2sint(1−cos2t)=2a2sin3t. So the area is the increase of a2∫sin3tdt from 0 to π. Since sin3t=sint−cos2tsint, an integral is −cost+31cos3t, which increases by (1−31)−(−1+31)=34. The area is 34a2.
Find the area enclosed by the limaçon r=a+bcosθ when a≥b>0.
Hint
Use 21∫r2dθ over 0≤θ≤2π, and cos2θ=21(1+cos2θ).
Solution
21r2=21a2+abcosθ+41b2(1+cos2θ). Over a full turn the cosine terms contribute nothing, so the area is 2π(21a2+41b2)=π(a2+21b2). For a=b (the cardioid of the widget, with a=b=1) this is 23πa2. If a<b the curve has an inner loop, which the formula counts twice; Hardy asks you to distinguish this case.
Check yourself
Check yourselfEnter a number
What is the total area between y=sinx and the axis of x for 0≤x≤2π, counting every part positively?
Why
From 0 to π: −cosπ+cos0=2. From π to 2π the integral gives −2, an area of 2 below the axis. Total 4. The single computation −cos2π+cos0=0 gives the signed area.
Check yourselfChoose one
In the proof that Φ′(x)=ϕ(x), which property of ϕ makes the sliver negligible?
Why
The sliver is at most ∣h∣λ(h), and λ(h)→0 is exactly continuity at x. Monotonicity is not needed (Hardy’s Fig. 42a shows a curve that rises and falls), nor differentiability. Positivity only keeps the region above the axis.
Check yourselfTrue or false?
The area under y=[x] (the integral part) from 0 to x is a differentiable function of x.
Why
False. At an integer the curve jumps, so λ(h) does not tend to 0, and the one-sided difference quotients of the area tend to the two different heights. The area function has corners at the integers. The theorem needs continuity of ϕ at the point.
Check yourselfEnter a number
What is the area under y=x2 from 0 to 1?
Why
Φ(x)=x3/3 with Φ(0)=0, so the area is 31.
Hardy’s language
Thus the ordinate of the curve is the derivative of the area, and the area is the integral of the ordinate.
— G. H. Hardy, §148
Summary
Summary
The area Φ(x) under a continuous curve satisfies Φ′(x)=ϕ(x): the strip is a rectangle hϕ(x) plus a sliver of area at most ∣h∣λ(h), and λ(h)→0 by continuity.
So areas are found by integration: Φ(x2)−Φ(x1) for any integral Φ of ϕ. Parts below the axis count negatively.
The proof assumes that areas exist; Chapter VII proves it. Granted that, every continuous function has an integral.
For a loop traced once anticlockwise, the area is the increase of 21∫(xy˙−yx˙)dt; in polar coordinates, of 21∫r2dθ.