Mathematics · Book 3 · Bachelor Year 1

University Mathematics — Year 1

University Mathematics — Year 1 · Bachelor Year 1

24Plane Curves

The calculus of Chapters 14 and 16 was built for functions y=f(x)y = f(x); most curves of geometry and mechanics — trajectories, circles rolled on circles, orbits — refuse that form and come instead as parametrized curves t(x(t),y(t))t \mapsto (x(t), y(t)) or as polar curves r=r(θ)r = r(\theta). This classical chapter studies both, and closes with the conics.

24.1 Parametrized curves

Definition 24.1

A parametrized curve is a map f ⁣:IR2f \colon I \to \R^2, tM(t)=(x(t),y(t))t \mapsto M(t) = (x(t), y(t)), with x,yx, y of class C1C^1 (at least) on the interval II. The velocity vector is f(t)=(x(t),y(t))f'(t) = (x'(t), y'(t)); the point M(t0)M(t_0) is regular when f(t0)(0,0)f'(t_0) \neq (0,0), and the tangent there is the line through M(t0)M(t_0) directed by f(t0)f'(t_0).

Why ff' directs the tangent. By Taylor–Young componentwise, M(t)=M(t0)+(tt0)f(t0)+o(tt0)M(t) = M(t_0) + (t - t_0) f'(t_0) + o(t - t_0): the chord direction M(t)M(t0)tt0\frac{M(t) - M(t_0)}{t - t_0} tends to f(t0)f'(t_0). The regularity hypothesis is what makes the limit a direction: if f(t0)=0f'(t_0) = 0, the display collapses to M(t)=M(t0)+o(tt0)M(t) = M(t_0) + o(t - t_0) and says nothing about how the curve leaves the point — whence the separate treatment of singular points in the method below, where the first nonvanishing derivative takes over the role of ff'. Note also that the tangent line is a geometric object: any reparametrization changes ff' by a nonzero scalar factor and leaves the line unchanged.

Method 24.2 (Studying a parametrized curve)

  1. Reduce the domain using symmetries: relations between M(t)M(-t), M(t+T)M(t + T), … and M(t)M(t) (reflections, translations, periodicity), then draw only the reduced part.
  2. Variations: tabulate signs of xx' and yy' jointly; the curve moves right/left as xx', up/down as yy'.
  3. Remarkable points: horizontal tangent (y=0xy' = 0 \neq x'), vertical tangent (x=0yx' = 0 \neq y'); at a singular point (f=0f' = 0), examine higher derivatives for the tangent direction.
  4. Asymptotic behavior at the ends of II, then sketch.

Example 24.3 (Asymptotic branches, worked)

x(t)=tx(t) = t, y(t)=t+1ty(t) = t + \dfrac1t on (0,+)\intoo{0}{+\infty}. As t0+t \to 0^{+}: x0x \to 0 while y+y \to +\infty — the curve climbs along the vertical asymptote x=0x = 0 (finite limit for one coordinate, infinite for the other). As t+t \to +\infty: both coordinates blow up, so test a line: the difference

y(t)x(t)=1t0+y(t) - x(t) = \frac1t \longrightarrow 0^{+}

exhibits the oblique asymptote y=xy = x, approached from above. In between, y=11t2y' = 1 - \frac1{t^2} vanishes at t=1t = 1: the point (1,2)(1, 2) is the low point of the branch, and x=1>0x' = 1 > 0 throughout, so the curve always advances rightward. One arc, two asymptotes, one minimum: a complete picture from three computations — with the general recipe visible underneath: when x,yx, y \to \infty together, examine y/xy/x for a candidate slope (here 1\to 1), then y(slope)xy - (\text{slope})\,x for the intercept and the side of approach.

Example 24.4 (The astroid)

x(t)=cos3tx(t) = \cos^3 t, y(t)=sin3ty(t) = \sin^3 t. Symmetries: M(t+2π)=M(t)M(t + 2\pi) = M(t) (study a period); M(t)M(-t) is the reflection of M(t)M(t) in the xx-axis; M(πt)M(\pi - t) in the yy-axis; M(π2t)M(\frac\pi2 - t) in the diagonal y=xy = x: it suffices to study t[0,π4]t \in \intcc{0}{\frac\pi4} and unfold.

Velocity: f(t)=3sintcost(cost,sint)f'(t) = 3\sin t\cos t\,(-\cos t, \sin t). On (0,π2)\intoo{0}{\frac\pi2} all points are regular with tangent directed by (cost,sint)(-\cos t, \sin t); at t=0t = 0 (the point (1,0)(1,0)) the velocity vanishes: a cusp, where the curve reverses along the tangent direction (1,0)(±)(-1, 0)\cdot(\pm) — by symmetry the four cusps sit at (±1,0),(0,±1)(\pm1, 0), (0, \pm1).

Example 24.5 (A figure-eight, studied in full)

x(t)=sintx(t) = \sin t, y(t)=sin2ty(t) = \sin 2t (a Lissajous curve). Symmetries: M(t+π)=(x(t),y(t))M(t + \pi) = (-x(t), y(t)) (reflection in the yy-axis), M(t)=(x(t),y(t))M(-t) = (-x(t), -y(t)) (central symmetry), M(πt)=(x(t),y(t))M(\pi - t) = (x(t), -y(t)) (reflection in the xx-axis): it suffices to study t[0,π2]t \in \intcc{0}{\frac\pi2} and unfold. Variations: x=cost0x' = \cos t \geq 0 throughout, while y=2cos2ty' = 2\cos 2t is positive before t=π4t = \frac\pi4 and negative after: the arc climbs rightward to the summit (22, 1)\bigl(\frac{\sqrt2}2,\ 1\bigr) at t=π4t = \frac\pi4 (horizontal tangent), then descends rightward to (1,0)(1, 0) at t=π2t = \frac\pi2, where x=0yx' = 0 \neq y': vertical tangent. Double point: M(0)=M(π)=(0,0)M(0) = M(\pi) = (0,0), with two different velocities

f(0)=(1, 2),f(π)=(1, 2):f'(0) = (1,\ 2), \qquad f'(\pi) = (-1,\ 2) :

two regular branches crossing at the origin at distinct angles — a double point, not a singular point: each passage is perfectly smooth, the two passages merely share their location. The whole curve is the figure-eight below.

The Lissajous curve ( t, 2t): a double point at the origin, where two regular branches cross with velocities (1, 2) and (-1, 2), and horizontal tangents at the four summits. Symmetry reduced all the work to a quarter-period.
The Lissajous curve (sint,sin2t)(\sin t, \sin 2t): a double point at the origin, where two regular branches cross with velocities (1,2)(1, 2) and (1,2)(-1, 2), and horizontal tangents at the four summits. Symmetry reduced all the work to a quarter-period.
The astroid ( 3 t, 3 t): four arcs meeting at four cusps. It is the curve traced by a point of a circle of radius 1/4 rolling inside the unit circle.
The astroid (cos3t,sin3t)(\cos^3 t, \sin^3 t): four arcs meeting at four cusps. It is the curve traced by a point of a circle of radius 14\frac14 rolling inside the unit circle.

24.2 Polar curves

Definition 24.6

A polar curve is given by r=r(θ)r = r(\theta): the point of parameter θ\theta is

M(θ)=r(θ)u(θ),u(θ)=(cosθ,sinθ).M(\theta) = r(\theta)\,\vec u(\theta), \qquad \vec u(\theta) = (\cos\theta, \sin\theta).

With v(θ)=(sinθ,cosθ)=u(θ)\vec v(\theta) = (-\sin\theta, \cos\theta) = \vec u\,'(\theta), the velocity is

M(θ)=r(θ)u(θ)+r(θ)v(θ).M'(\theta) = r'(\theta)\, \vec u(\theta) + r(\theta)\, \vec v(\theta) .

Consequences: where r0r \neq 0, the point is regular, and the tangent makes with the ray the angle VV given by tanV=rr\tan V = \frac{r}{r'} (angle between MM' and u\vec u); where r(θ0)=0r(\theta_0) = 0, the curve passes through the origin with tangent the ray θ=θ0\theta = \theta_0 (direction u(θ0)\vec u(\theta_0), read off M(θ0)=r(θ0)u(θ0)M'(\theta_0) = r'(\theta_0)\vec u(\theta_0), or from the limit chord: the chord from OO to M(θ)M(\theta) is carried by u(θ)\vec u(\theta) itself, which tends to u(θ0)\vec u(\theta_0) — so the rule holds even when r(θ0)=0r'(\theta_0) = 0 and the point is singular, which is why polar cusps at the origin, like the cardioid’s, get their tangent for free).

Example 24.7 (The cardioid)

r(θ)=1+cosθr(\theta) = 1 + \cos\theta. Symmetry: r(θ)=r(θ)r(-\theta) = r(\theta): reflection in the xx-axis; study θ[0,π]\theta \in \intcc{0}{\pi}. rr decreases from 22 to 00; at θ=π\theta = \pi, r=0r = 0: the curve reaches the origin tangentially to the ray θ=π\theta = \pi (the xx-axis), forming a cusp there — the heart’s point. Tangent at θ=0\theta = 0: r=0r' = 0, so tanV=\tan V = \infty: perpendicular to the axis.

The angle VV deserves one more reading. At θ=π2\theta = \frac\pi2: r=1r = 1 and r=1r' = -1, so tanV=rr=1\tan V = \frac{r}{r'} = -1: the tangent makes three-quarters of a right angle with the outgoing ray — the curve is already bending back toward its cusp. The formula tanV=r/r\tan V = r/r' delivers tangent directions along the whole curve with no computation of M(θ)M'(\theta) whatsoever: it is the polar analogue of reading a slope.

Example 24.8 (Polar to cartesian: a hidden circle)

What is the polar curve r=2cosθr = 2\cos\theta? Multiply by rr: r2=2rcosθr^2 = 2r\cos\theta, i.e. x2+y2=2xx^2 + y^2 = 2x, i.e.

(x1)2+y2=1:(x - 1)^2 + y^2 = 1 :

the circle of center (1,0)(1, 0) and radius 11, passing through the origin. Bookkeeping matters: as θ\theta runs over (π2,π2]\intoc{-\frac\pi2}{\frac\pi2} the whole circle is swept exactly once (rr vanishes at both ends), and at θ=±π2\theta = \pm\frac\pi2 the rule “tangent at the origin along the ray θ=θ0\theta = \theta_0” gives a vertical tangent there — matching the geometry, since the vertical axis is indeed tangent to this circle at OO. For θ\theta beyond that range, r<0r < 0 retraces the same circle: a reminder that a polar curve is a parametrized object, allowed to pass over itself.

The four-petaled rose r = 2 (). The petals along the y-axis are traced with r < 0 (the point plots on the ray opposite to ); the dashed diagonals = ± π4 are the tangents at the origin, where r vanishes.
The four-petaled rose r=cos2θr = \cos 2\theta (Exercise 24.4). The petals along the yy-axis are traced with r<0r < 0 (the point plots on the ray opposite to θ\theta); the dashed diagonals θ=±π4\theta = \pm\frac\pi4 are the tangents at the origin, where rr vanishes.
The cardioid r = 1 +. Polar curves are read by sweeping the angle: the radius swells and shrinks as  turns.
The cardioid r=1+cosθr = 1 + \cos\theta. Polar curves are read by sweeping the angle: the radius swells and shrinks as θ\theta turns.

24.3 Arc length

Definition 24.9 (Length of an arc)

The length of a C1C^1 arc f ⁣:[a,b]R2f \colon \intcc{a}{b} \to \R^2 is the integral of the speed:

L=abf(t) ⁣dt=abx(t)2+y(t)2   ⁣dt.L = \int_a^b \norm{f'(t)}\,\dd t = \int_a^b \sqrt{x'(t)^2 + y'(t)^2}\;\dd t .

(Motivation: on a small interval, M(t+h)M(t)+hf(t)M(t + h) \approx M(t) + h f'(t), so the arc is close to a polygon whose segment lengths sum to a Riemann sum of f\norm{f'}, Theorem 15.20.) For a polar curve r=r(θ)r = r(\theta), the velocity ru+rvr'\vec u + r\vec v has orthogonal components, so

L=θ1θ2r(θ)2+r(θ)2   ⁣dθ.L = \int_{\theta_1}^{\theta_2} \sqrt{r'(\theta)^2 + r(\theta)^2}\;\dd\theta .

The length does not depend on the (monotone, C1C^1) parametrization chosen: substituting t=φ(s)t = \varphi(s) in the integral (Theorem 15.15) multiplies ff' by φ\varphi' and  ⁣dt\dd t by φ1\varphi'^{-1}.

Example 24.10 (Sanity check: the circle)

For f(t)=(Rcost,Rsint)f(t) = (R\cos t, R\sin t) on [0,2π]\intcc{0}{2\pi}: f=R\norm{f'} = R, so L=2πRL = 2\pi R — the definition returns the circumference. Reparametrization test: the map g(t)=(Rcos2t,Rsin2t)g(t) = (R\cos 2t, R\sin 2t) on [0,π]\intcc{0}{\pi} draws the same circle at doubled speed g=2R\norm{g'} = 2R, and

0π2R ⁣dt=2πR\int_0^{\pi} 2R\,\dd t = 2\pi R

again: half the time, twice the speed, same length — the invariance promised in the definition, watched once on numbers. (Running gg on all of [0,2π]\intcc{0}{2\pi} would give 4πR4\pi R: a curve traversed twice is twice as long as a journey; length measures the parametrized path, and honest bookkeeping of the interval is part of the computation.) For the astroid (cos3t,sin3t)(\cos^3t, \sin^3t): f=3sintcost=32sin2t\norm{f'} = 3\abs{\sin t\cos t} = \tfrac32\abs{\sin 2t}, and by symmetry L=40π/232sin2t ⁣dt=6L = 4\int_0^{\pi/2} \tfrac32\sin 2t\,\dd t = 6: a curve drawn inside the unit circle, of length 6<2π6 < 2\pi. The weekend problem measures the most famous arch of all.

One arch of the cycloid (radius R = 1), the rolling circle at t = 2, and the two guide lines of the weekend problem: the chord MC to the contact point is normal to the curve, the chord MT to the top of the circle is tangent.
One arch of the cycloid (radius R=1R = 1), the rolling circle at t=2t = 2, and the two guide lines of the weekend problem: the chord MCMC to the contact point is normal to the curve, the chord MTMT to the top of the circle is tangent.

24.4 Conics

Definition 24.11 (Focus–directrix definition)

Fix a point FF (focus), a line DD not through FF (directrix) and e>0e > 0 (eccentricity). The conic of these data is

C={M:d(M,F)=e  d(M,D)}:\mathcal{C} = \{M : d(M, F) = e\; d(M, D)\}:

an ellipse for e<1e < 1, a parabola for e=1e = 1, a hyperbola for e>1e > 1. (The circle appears as a degenerate limit e0e \to 0.)

Example 24.12 (The definition, checked on a parabola)

Take the parabola y2=4xy^2 = 4x, i.e. 2p=42p = 4: focus F=(1,0)F = (1, 0) and directrix D:x=1D : x = -1 (the reduced form Theorem 24.13 puts them at ±p2\pm\frac p2). At the point M=(1,2)M = (1, 2) of the curve:

MF=(11)2+22=2,d(M,D)=1(1)=2:MF = \sqrt{(1-1)^2 + 2^2} = 2, \qquad d(M, D) = 1 - (-1) = 2 :

equal, as e=1e = 1 demands. At M=(4,4)M' = (4, 4): MF=9+16=5MF' = \sqrt{9 + 16} = 5 and d(M,D)=5d(M', D) = 5 again. The focus–directrix definition is not an abstraction: it is a pair of distances one can measure on any point, and the algebra of the reduced equations is nothing but this measurement done once and for all.

Theorem 24.13 (Reduced equations)

In a well-chosen orthonormal frame:

ellipse: x2a2+y2b2=1(ab>0),hyperbola: x2a2y2b2=1,parabola: y2=2px,\text{ellipse: } \frac{x^2}{a^2} + \frac{y^2}{b^2} = 1 \quad (a \geq b > 0), \qquad \text{hyperbola: } \frac{x^2}{a^2} - \frac{y^2}{b^2} = 1, \qquad \text{parabola: } y^2 = 2px,

with, for the ellipse: foci at (±c,0)(\pm c, 0), c=a2b2c = \sqrt{a^2 - b^2}, e=cae = \frac ca, and the bifocal characterization MF+MF=2aMF + MF' = 2a; for the hyperbola: c=a2+b2c = \sqrt{a^2 + b^2}, e=cae = \frac ca, MFMF=2a\abs{MF - MF'} = 2a, asymptotes y=±baxy = \pm\frac ba x.

Proof. Take the focus at the origin and the directrix vertical, x=hx = -h (h>0h > 0): the condition MF2=e2d(M,D)2MF^2 = e^2 d(M, D)^2 reads x2+y2=e2(x+h)2x^2 + y^2 = e^2 (x + h)^2. For e=1e = 1: y2=2hx+h2y^2 = 2hx + h^2, a parabola after the shift xxh2x \mapsto x - \frac h2 (so p=hp = h). For e1e \neq 1: completing the square in xx,

(1e2)(xe2h1e2) ⁣2+y2=e2h2+e4h21e2=e2h21e2.(1 - e^2)\Bigl(x - \frac{e^2 h}{1 - e^2}\Bigr)^{\!2} + y^2 = e^2h^2 + \frac{e^4 h^2}{1 - e^2} = \frac{e^2 h^2}{1 - e^2} .

Set X=xe2h1e2X = x - \frac{e^2h}{1-e^2} (a shift of frame). When e<1e < 1, divide by the positive right-hand side: X2a2+y2b2=1\frac{X^2}{a^2} + \frac{y^2}{b^2} = 1 with a=eh1e2a = \frac{eh}{1 - e^2}, b=eh1e2b = \frac{eh}{\sqrt{1-e^2}}; when e>1e > 1, both sides of the display flip sign appropriately and the same division gives X2a2y2b2=1\frac{X^2}{a^2} - \frac{y^2}{b^2} = 1 with a=ehe21a = \frac{eh}{e^2 - 1}, b=ehe21b = \frac{eh}{\sqrt{e^2 - 1}}. The stated values of cc follow (c2=a2b2c^2 = a^2 - b^2 or a2+b2a^2 + b^2 gives c=eac = ea in both cases, placing the focus correctly), and the bifocal properties are direct verifications on the reduced equations. In detail for the ellipse, with F=(c,0)F' = (c, 0) and M=(X,y)M = (X, y) on the curve:

MF2=(Xc)2+y2=X22cX+c2+b2(1X2a2)=c2a2X22cX+a2=(aeX)2,MF'^2 = (X - c)^2 + y^2 = X^2 - 2cX + c^2 + b^2\Bigl(1 - \frac{X^2}{a^2}\Bigr) = \frac{c^2}{a^2}X^2 - 2cX + a^2 = (a - eX)^2 ,

using c2+b2=a2c^2 + b^2 = a^2 and e=cae = \frac ca; since Xa\abs X \leq a and e<1e < 1, aeX>0a - eX > 0, so MF=aeXMF' = a - eX with no square root ever extracted. The mirror computation gives MF=a+eXMF = a + eX, whence MF+MF=2aMF + MF' = 2a, constant. For the hyperbola the same algebra yields MF=a+eXMF = \abs{a + eX} and MF=eXaMF' = \abs{eX - a}, with difference ±2a\pm2a according to the branch.

Example 24.14

x225+y29=1\frac{x^2}{25} + \frac{y^2}{9} = 1: ellipse, a=5a = 5, b=3b = 3, c=4c = 4: foci (±4,0)(\pm4, 0), eccentricity 45\frac45. Its parametrization: (5cost,3sint)(5\cos t, 3\sin t) — a circle stretched anisotropically; the sum of distances to the foci of any of its points is 1010.

Example 24.15 (Reading an orbit from its polar equation)

The polar conic r=11+12cosθr = \dfrac{1}{1 + \frac12\cos\theta} (Exercise 24.7 with p=1p = 1, e=12e = \frac12) is an ellipse with a focus at the origin — the geometry of a planetary orbit with the sun at OO. Extract everything from pp and ee: from the reduction of Theorem 24.13, a=p1e2=13/4=43a = \dfrac{p}{1 - e^2} = \dfrac{1}{3/4} = \dfrac43 and c=ea=23c = ea = \dfrac23. The two apsides check it with no theory at all:

r(0)=13/2=23=ac(perihelion),r(π)=11/2=2=a+c(aphelion),r(0) = \frac{1}{3/2} = \frac23 = a - c \quad (\text{perihelion}), \qquad r(\pi) = \frac{1}{1/2} = 2 = a + c \quad (\text{aphelion}),

and their sum 23+2=83=2a\frac23 + 2 = \frac83 = 2a recovers the major axis. The polar form is the natural one whenever a focus is physically distinguished; the reduced cartesian form, whenever the symmetry axes are. Converting between the two is exactly what the completed-square computation of the theorem does.

Eccentricity as a dial, finally: keep p=1p = 1 and turn ee in r=11+ecosθr = \frac{1}{1 + e\cos\theta}. At e=0e = 0: the circle r=1r = 1. At e=12e = \frac12: the ellipse just studied, rr oscillating between 23\frac23 and 22. At e=1e = 1: r(θ)r(\theta) \to \infty as θπ\theta \to \pi — the curve no longer closes: a parabola, its farthest point pushed to infinity. At e=2e = 2: the denominator vanishes at cosθ=12\cos\theta = -\frac12, and only θ(2π3,2π3)\theta \in \intoo{-\frac{2\pi}3}{\frac{2\pi}3} survives: one branch of a hyperbola, escaping along two asymptotic directions. One formula, the entire conic family, and the transition points e=1e = 1 visible as the moment the denominator first reaches zero.

Remark 24.16 (Common pitfalls)

Double points are not singular points: at a self-crossing, each branch is regular; “singular” refers to f(t0)=0f'(t_0) = 0 for one value of the parameter (Example 24.5 vs the astroid’s cusps). Vertical tangent versus cusp: x(t0)=0y(t0)x'(t_0) = 0 \neq y'(t_0) is a regular point with vertical tangent; only x=y=0x' = y' = 0 demands the higher-order expansion. Negative radii plot on the opposite ray: for r(θ)<0r(\theta) < 0 the point is ru(θ)-\abs{r}\,\vec u(\theta), at angle θ+π\theta + \pi (Exercise 24.5); forgetting this loses inner loops or doubles curves. Arc length integrates f\norm{f'}, not ff': split the integral at the zeros of the speed — for the astroid, integrating 32sin2t\frac32\sin 2t over a full period without absolute values gives 00, not 66. Conics must be reduced before being read: in 9x2+25y236x50y164=09x^2 + 25y^2 - 36x - 50y - 164 = 0, neither the axes, nor the center, nor the eccentricity is visible until the squares are completed (Exercise 24.6); and a vanishing quadratic part on one variable means parabola, not “degenerate ellipse”.

Remark 24.17 (Where these curves go)

Parametrized curves are the language of mechanics: trajectories are curves, velocity vectors are tangent vectors, and the arc length of Definition 24.9 is the distance run. The conics reappear wherever an inverse-square law acts — planetary orbits are ellipses with the sun at a focus. In Chapter 25, curves become the level sets of functions of two variables, and the tangent of this chapter meets the gradient of the next. The Year 2 volume adds curvature and the local canonical form of a curve; the weekend problem below already extracts, with this year’s tools only, everything the seventeenth century knew about its most celebrated curve.

Remark 24.18 (Perspectives inside Book 3)

This chapter consumes the whole first half of the volume and feeds the last chapter. Consumed: tangent vectors are derivatives (Chapter 14), arc length and areas are integrals (Chapter 15), cusps are settled by Taylor expansions (Chapter 16), the tautochrone is a linear differential equation (Chapter 5), and every distance and angle is Euclidean (Chapter 23). Fed: in Chapter 25, a curve t(x(t),y(t))t \mapsto (x(t), y(t)) drawn inside a level set {f=c}\{f = c\} is differentiated by the chain rule, and the resulting identity f, f(t)=0\langle \nabla f,\ f'(t)\rangle = 0 marries this chapter’s tangent vectors to the next one’s gradients — the tangent line of the curve and the normal direction of the surface are the same computation seen from both banks.

24.5 Exercises

Exercise 24.1

Study and sketch the curve x(t)=t2x(t) = t^2, y(t)=t3y(t) = t^3 (symmetries, variations, behavior at the singular point t=0t = 0).

Solution

Solution of Exercise 24.1.

M(t)=(t2,t3)M(-t) = (t^2, -t^3): reflection in the xx-axis; study t0t \geq 0. Both x=2tx' = 2t and y=3t2y' = 3t^2 are 0\geq 0: the branch moves right and up, from (0,0)(0,0) to infinity. At t=0t = 0 the velocity vanishes; expansions x=t2x = t^2, y=t3y = t^3 show the tangent is the xx-axis (y/x=t0y/x = t \to 0) with yy changing sign while x0x \geq 0: a cusp pointing left. The curve is the semi-cubical parabola y2=x3y^2 = x^3.

Exercise 24.2

For the cycloid x(t)=tsintx(t) = t - \sin t, y(t)=1costy(t) = 1 - \cos t (the trajectory of a point of a rolling wheel of radius 11): identify the period-translation symmetry, the singular points, and the tangent direction at t=0t = 0 (expand xx and yy to the first nonzero orders).

Solution

Solution of Exercise 24.2.

M(t+2π)=M(t)+(2π,0)M(t + 2\pi) = M(t) + (2\pi, 0): the curve repeats, translated by one wheel circumference; study one period. x(t)=1cost0x'(t) = 1 - \cos t \geq 0, y(t)=sinty'(t) = \sin t: the arch rises on (0,π)\intoo{0}{\pi}, falls on (π,2π)\intoo{\pi}{2\pi}, culminating at (π,2)(\pi, 2). Singular points where x=y=0x' = y' = 0: t2πZt \in 2\pi\Z, on the ground. Near t=0t = 0:

x(t)=t36+o(t3),y(t)=t22+o(t2):x(t) = \frac{t^3}{6} + o(t^3), \qquad y(t) = \frac{t^2}{2} + o(t^2):

xx changes sign, y0y \geq 0, and xy3/2\frac{x}{y^{3/2}} bounded: the tangent is vertical (direction (0,1)(0,1)), a cusp where the tracked point momentarily has zero speed — the physical signature of rolling without slipping.

Exercise 24.3

Give the tangent line to the curve (cos3t,sin3t)(\cos^3 t, \sin^3 t) at t=π4t = \frac\pi4, and check that the segment of this tangent cut by the axes has length 11 — a famous property of the astroid (true at every regular point).

Solution

Solution of Exercise 24.3.

At t=π4t = \frac\pi4: M=(24,24)M = \bigl(\frac{\sqrt2}{4}, \frac{\sqrt2}{4}\bigr), tangent direction (cost,sint)=12(1,1)(-\cos t, \sin t) = \frac{1}{\sqrt2}(-1, 1) (Example 24.4). Tangent line: y24=(x24)y - \frac{\sqrt2}{4} = -(x - \frac{\sqrt2}{4}), i.e. x+y=22x + y = \frac{\sqrt2}{2}. Intercepts: (22,0)\bigl(\frac{\sqrt2}{2}, 0\bigr) and (0,22)\bigl(0, \frac{\sqrt2}{2}\bigr); the segment between them has length 12+12=1\sqrt{\frac12 + \frac12} = 1.

(General point tt: the tangent at (cos3t,sin3t)(\cos^3 t, \sin^3 t) cuts the axes at (cost,0)(\cos t, 0) and (0,sint)(0, \sin t) — check that the line through these points has direction (cost,sint)(-\cos t, \sin t) and passes through M(t)M(t) — and the cut segment has length cos2t+sin2t=1\sqrt{\cos^2 t + \sin^2 t} = 1.)

Exercise 24.4

Sketch the polar curves r=cos2θr = \cos 2\theta (four-petaled rose) and r=1cosθr = \frac{1}{\cos\theta} on (π2,π2)\intoo{-\frac\pi2}{\frac\pi2} (recognize a line).

Solution

Solution of Exercise 24.4.

r=cos2θr = \cos 2\theta: period π\pi in θ\theta, symmetric in both axes; rr vanishes at θ=±π4\theta = \pm\frac\pi4 (tangents at the origin along the diagonals) and is negative for θ(π4,3π4)\theta \in \intoo{\frac\pi4}{\frac{3\pi}{4}}, where the points plot on the opposite ray — producing four petals along the axes directions θ=0,π2,π,3π2\theta = 0, \frac\pi2, \pi, \frac{3\pi}{2}, each of maximal radius 11.

rcosθ=1r\cos\theta = 1 is the equation x=1x = 1: the polar curve r=1cosθr = \frac{1}{\cos\theta} is the vertical line x=1x = 1 (swept once for θ(π2,π2)\theta \in \intoo{-\frac\pi2}{\frac\pi2}).

Exercise 24.5 ★★

Study the polar curve r=1+2cosθr = 1 + 2\cos\theta (a limaçon): domain where r0r \geq 0 vs r<0r < 0 (points plotted with negative radius sit on the opposite ray), passage through the origin, inner loop, sketch.

Solution

Solution of Exercise 24.5.

r(θ)=1+2cosθr(\theta) = 1 + 2\cos\theta vanishes for cosθ=12\cos\theta = -\frac12: θ=±2π3\theta = \pm\frac{2\pi}{3}. Symmetry in the xx-axis; study θ[0,π]\theta \in \intcc{0}{\pi}. For θ[0,2π3)\theta \in \intco{0}{\frac{2\pi}{3}}: r>0r > 0, decreasing from 33 to 00: outer arc, entering the origin tangent to the ray θ=2π3\theta = \frac{2\pi}{3}. For θ(2π3,π]\theta \in \intoc{\frac{2\pi}{3}}{\pi}: r<0r < 0: the points ru(θ)r\vec u(\theta) sit on the opposite ray (angle θπ(π3,0]\theta - \pi \in \intoc{-\frac\pi3}{0}), with distance r\abs r growing from 00 to 11: this draws a small inner loop through the origin. At θ=π\theta = \pi, r=1r = -1 and the point is u(π)=(1,0)-\vec u(\pi) = (1, 0): the loop closes on the positive xx-axis. Sketch: a big heart-like outer curve with maximum reach 33 at θ=0\theta = 0, plus a loop inside it through OO and (1,0)(1,0).

Exercise 24.6 ★★

Identify the conic 9x2+25y236x50y164=09x^2 + 25y^2 - 36x - 50y - 164 = 0: reduce by completing squares, give center, semi-axes, foci, eccentricity.

Solution

Solution of Exercise 24.6.

Complete squares:

9(x24x)+25(y22y)=164    9(x2)2+25(y1)2=164+36+25=225.9(x^2 - 4x) + 25(y^2 - 2y) = 164 \iff 9(x-2)^2 + 25(y-1)^2 = 164 + 36 + 25 = 225 .

Dividing by 225225: (x2)225+(y1)29=1\frac{(x-2)^2}{25} + \frac{(y-1)^2}{9} = 1: an ellipse of center (2,1)(2, 1), semi-axes a=5a = 5 (horizontal), b=3b = 3; c=259=4c = \sqrt{25 - 9} = 4: foci (2±4,1)=(2,1)(2 \pm 4,\, 1) = (-2, 1) and (6,1)(6, 1); eccentricity e=45e = \frac45.

Exercise 24.7 ★★

Prove that the polar equation of a conic with focus at the origin is

r=p1+ecosθr = \frac{p}{1 + e\cos\theta}

(directrix vertical at distance pe\frac pe right of the focus). Which values of θ\theta are allowed when e>1e > 1?

Solution

Solution of Exercise 24.7.

Focus at the origin, directrix D:x=dD: x = d with d=pe>0d = \frac pe > 0. For a point M=(rcosθ,rsinθ)M = (r\cos\theta, r\sin\theta) with r>0r > 0:

MF=r,d(M,D)=drcosθ,MF = r, \qquad d(M, D) = \abs{d - r\cos\theta},

and the conic condition MF=ed(M,D)MF = e\,d(M, D), in the regime where MM is on the focus side of the directrix (rcosθ<dr\cos\theta < d), reads r=e(drcosθ)r = e(d - r\cos\theta), i.e.

r(1+ecosθ)=ed=p,r=p1+ecosθ.r(1 + e\cos\theta) = ed = p, \qquad r = \frac{p}{1 + e\cos\theta} .

For e<1e < 1 the denominator never vanishes: all θ\theta allowed (ellipse). For e=1e = 1: θπ\theta \neq \pi (parabola, open towards the directrix’s far side). For e>1e > 1: need 1+ecosθ>01 + e\cos\theta > 0, i.e. θ(θ0,θ0)\theta \in \intoo{-\theta_0}{\theta_0} with θ0=arccos(1e)\theta_0 = \arccos\bigl(-\frac1e\bigr): one branch of the hyperbola (the other branch corresponds to the sign choice r<0r < 0, or to the second focus).

Exercise 24.8 ★★

From the bifocal property MF+MF=2aMF + MF' = 2a, deduce the “gardener’s construction” of the ellipse, and prove that the tangent at MM makes equal angles with MFMF and MFMF' (reflection property; differentiate M(t)F+M(t)F=2a\norm{M(t) - F} + \norm{M(t) - F'} = 2a and interpret the vanishing sum of unit-vector products).

Solution

Solution of Exercise 24.8.

Gardener: attach a string of length 2a2a to two stakes F,FF, F' and keep it taut with the tracing point MM: the constraint is exactly MF+MF=2aMF + MF' = 2a, so the traced curve is the ellipse.

Reflection property: let tM(t)t \mapsto M(t) be a regular parametrization and u(t)=M(t)FM(t)Fu(t) = \frac{M(t) - F}{\norm{M(t) - F}}, u(t)u'(t) the analogous unit vector to FF'. Differentiating MF+MF=2a\norm{M - F} + \norm{M - F'} = 2a, using  ⁣d ⁣dtMF=MFMF,M\frac{\dd}{\dd t}\norm{M - F} = \bigl\langle \frac{M - F}{\norm{M-F}},\, M'\bigr\rangle (chain rule on ,\sqrt{\langle\cdot,\cdot\rangle}):

u+u,M=0.\bigl\langle u + u',\, M'\bigr\rangle = 0 .

So the tangent direction MM' is orthogonal to the bisector direction u+uu + u' of the two focal rays: the tangent makes equal angles with MFMF and MFMF'. (A light ray from one focus reflects off the ellipse to the other focus.)

Exercise 24.9 ★★★

The lemniscate of Bernoulli is the polar curve r2=cos2θr^2 = \cos 2\theta (take r=cos2θr = \sqrt{\cos2\theta} where defined).

  1. Give its domain, symmetries, tangents at the origin, and sketch it.
  2. Prove that it is the locus of points MM with MFMF=12MF \cdot MF' = \frac12 where F,F=(±12,0)F, F' = \bigl(\pm\frac{1}{\sqrt2}, 0\bigr). (Compute MF2MF2MF^2\,MF'^2 in polar coordinates.)
Solution

Solution of Exercise 24.9.

  1. Domain: cos2θ0\cos 2\theta \geq 0: θ[π4,π4][3π4,5π4]\theta \in \intcc{-\frac\pi4}{\frac\pi4} \cup \intcc{\frac{3\pi}{4}}{\frac{5\pi}{4}}. Symmetries: θθ\theta \mapsto -\theta (xx-axis) and θπθ\theta \mapsto \pi - \theta (yy-axis): study [0,π4]\intcc{0}{\frac\pi4}; rr decreases from 11 to 00. At θ=±π4\theta = \pm\frac\pi4: r=0r = 0, tangents at the origin along the diagonals. The curve is the \infty symbol: two symmetric loops meeting at OO, reaching (±1,0)(\pm 1, 0).
  2. With F,F=(±c,0)F, F' = (\pm c, 0), c=12c = \frac{1}{\sqrt 2}, and M=(rcosθ,rsinθ)M = (r\cos\theta, r\sin\theta):

    MF2MF2=((rcosθc)2+r2sin2θ)((rcosθ+c)2+r2sin2θ)=(r2+c2)2(2rccosθ)2,MF^2\, MF'^2 = \bigl((r\cos\theta - c)^2 + r^2\sin^2\theta\bigr) \bigl((r\cos\theta + c)^2 + r^2\sin^2\theta\bigr) = (r^2 + c^2)^2 - (2rc\cos\theta)^2 ,

    by the identity (AB)(A+B)=A2B2(A - B)(A + B) = A^2 - B^2 with A=r2+c2A = r^2 + c^2, B=2rccosθB = 2rc\cos\theta. With c2=12c^2 = \frac12:

    MF2MF2=(r2+12)22r2cos2θ=r4+r2(12cos2θ)+14=r4r2cos2θ+14.MF^2 MF'^2 = \Bigl(r^2 + \frac12\Bigr)^2 - 2r^2\cos^2\theta = r^4 + r^2\bigl(1 - 2\cos^2\theta\bigr) + \frac14 = r^4 - r^2\cos 2\theta + \frac14 .

    On the lemniscate, r2=cos2θr^2 = \cos 2\theta: the first two terms cancel, leaving MF2MF2=14MF^2MF'^2 = \frac14, i.e. MFMF=12MF \cdot MF' = \frac12. Conversely, the computation read backwards shows the locus equation MFMF=12MF\,MF' = \frac12 is r2=cos2θr^2 = \cos 2\theta (for r0r \neq 0; and OO satisfies both).

Exercise 24.10 ★★

Compute the total length of the cardioid r=1+cosθr = 1 + \cos\theta using the polar formula of Definition 24.9 (the identity 1+cosθ=2cos2θ21 + \cos\theta = 2\cos^2\frac\theta2 turns the square root into 2cosθ22\abs{\cos\frac\theta2}).

Solution

Solution of Exercise 24.10.

r=1+cosθr = 1 + \cos\theta, r=sinθr' = -\sin\theta:

r2+r2=1+2cosθ+cos2θ+sin2θ=2+2cosθ=4cos2θ2,r^2 + r'^2 = 1 + 2\cos\theta + \cos^2\theta + \sin^2\theta = 2 + 2\cos\theta = 4\cos^2\frac\theta2 ,

so r2+r2=2cosθ2\sqrt{r^2 + r'^2} = 2\abs{\cos\frac\theta2} and, by the symmetry in the xx-axis,

L=02π2cosθ2 ⁣dθ=20π2cosθ2 ⁣dθ=2[4sinθ2]0π=8.L = \int_0^{2\pi} 2\Bigl|\cos\frac\theta2\Bigr|\,\dd\theta = 2\int_0^{\pi} 2\cos\frac\theta2\,\dd\theta = 2\Bigl[4\sin\frac\theta2\Bigr]_0^{\pi} = 8 .

Another algebraic, π\pi-free perimeter.

Exercise 24.11 ★★

Show that the tangent to the ellipse (acost,bsint)(a\cos t, b\sin t) at the point of parameter tt has equation

xcosta+ysintb=1,\frac{x\cos t}{a} + \frac{y\sin t}{b} = 1 ,

and deduce the tangent to x2a2+y2b2=1\frac{x^2}{a^2} + \frac{y^2}{b^2} = 1 at a point (x0,y0)(x_0, y_0) of the ellipse: xx0a2+yy0b2=1\frac{x\,x_0}{a^2} + \frac{y\,y_0}{b^2} = 1 (the “splitting of the squares” rule).

Solution

Solution of Exercise 24.11.

The point (acost,bsint)(a\cos t, b\sin t) satisfies the equation: cos2t+sin2t=1\cos^2 t + \sin^2 t = 1. The line’s normal vector is (costa,sintb)\bigl(\frac{\cos t}a, \frac{\sin t}b\bigr), and its product with the velocity (asint,bcost)(-a\sin t, b\cos t) is sintcost+sintcost=0-\sin t\cos t + \sin t\cos t = 0: the line passes through the point with the tangent direction — it is the tangent. For (x0,y0)(x_0, y_0) on the ellipse, write cost=x0a\cos t = \frac{x_0}a, sint=y0b\sin t = \frac{y_0}b and substitute:

xx0a2+yy0b2=1,\frac{x\,x_0}{a^2} + \frac{y\,y_0}{b^2} = 1 ,

obtained from the ellipse equation by “splitting” x2xx0x^2 \mapsto x\,x_0 and y2yy0y^2 \mapsto y\,y_0.

Exercise 24.12 ★★★

The logarithmic spiral is the polar curve r=ekθr = \eu^{k\theta} (k>0k > 0 fixed, θR\theta \in \R).

  1. Show that the angle VV between the radius and the tangent is constant (tanV=1k\tan V = \frac1k) — the spiral crosses every ray at the same angle.
  2. Show that rotating the spiral by an angle cc maps it onto its scaling by the factor ekc\eu^{-kc}: every rotation of the spiral is a magnification of it (self-similarity).
  3. Compute the length of the arc θ(,θ0]\theta \in \intoc{-\infty}{\theta_0} (as a limit of lengths on [A,θ0]\intcc{A}{\theta_0}, AA \to -\infty) and observe that it is finite: a curve spiraling infinitely many times around the origin, of finite length.
Solution

Solution of Exercise 24.12.

  1. r=kekθr' = k\eu^{k\theta}, so tanV=rr=1k\tan V = \frac{r}{r'} = \frac1k: constant. The spiral cuts every ray from the origin at the same angle V=arctan1kV = \arctan\frac1k.
  2. In complex notation the spiral is {ekθeiθ:θR}\{\eu^{k\theta} \eu^{\iu\theta} : \theta \in \R\}. Rotating by cc multiplies by eic\eu^{\iu c}:

    ekθei(θ+c)=ekcek(θ+c)ei(θ+c),\eu^{k\theta}\eu^{\iu(\theta + c)} = \eu^{-kc}\,\eu^{k(\theta + c)}\eu^{\iu(\theta+c)} ,

    and as θ+c\theta + c runs over R\R this describes ekc\eu^{-kc} times the spiral: rotation == scaling. No other smooth curve but lines and circles has this property.

  3. r2+r2=1+k2ekθ\sqrt{r^2 + r'^2} = \sqrt{1 + k^2}\,\eu^{k\theta}, so on [A,θ0]\intcc{A}{\theta_0} the length is 1+k2k(ekθ0ekA)\frac{\sqrt{1+k^2}}{k}\bigl(\eu^{k\theta_0} - \eu^{kA}\bigr), and as AA \to -\infty:

    L=1+k2kekθ0<:L = \frac{\sqrt{1 + k^2}}{k}\,\eu^{k\theta_0} < \infty :

    infinitely many turns around the origin, finite total length (the turns shrink geometrically).

24.6 Problem: the cycloid, queen of curves

Problem 24.1

A wheel of radius RR rolls without slipping along the xx-axis; the point of the rim initially at the origin traces the cycloid. The seventeenth century fought over this curve — Galileo weighed paper cutouts of it, Wren measured it, Roberval computed its area, Huygens built clocks on it — and every one of their results is within reach of this chapter. We prove the four classics: the tangent construction, Wren’s length 8R8R, the area 3πR23\pi R^2, and Huygens’ tautochrone property.

Part I — Rolling and the tangent.

  1. After the wheel has turned by an angle tt, its center sits at Ω(t)=(Rt,R)\Omega(t) = (Rt, R) (rolling without slipping: contact distance == arc rolled). Show that the marked point is at

    M(t)=(R(tsint),  R(1cost)).M(t) = \bigl(R(t - \sin t),\; R(1 - \cos t)\bigr).
  2. Compute f(t)f'(t) and show f(t)=2Rsint2\norm{f'(t)} = 2R\,\abs{\sin\frac t2}; locate the singular points (cusps — cf. Exercise 24.2) and the top of each arch.
  3. Let C(t)=(Rt,0)C(t) = (Rt, 0) be the contact point and T(t)=(Rt,2R)T(t) = (Rt, 2R) the top of the wheel. Prove that for 0<t<2π0 < t < 2\pi the vector MCM - C is normal to the curve at M(t)M(t) and the vector TMT - M is tangent: to draw the tangent to a cycloid, join the point to the top of its rolling circle. (Factor everything through sint2\sin\frac t2 and cost2\cos\frac t2.)
  4. Interpret question 3 kinematically: the contact point is the instantaneous center of rotation, and the speed of MM equals its distance to CC (for unit angular velocity). Verify MC=2Rsint2\norm{M - C} = 2R\abs{\sin\frac t2}.
  5. Prove the height–speed relation

    f(t)2=2Ry(t):\norm{f'(t)}^2 = 2R\,y(t) :

    on a cycloid traversed at unit angular velocity, the speed at each point is exactly the free-fall speed for a drop equal to the current height. (Keep this for Part IV.)

Part II — Wren’s theorem: the arch has length 8R8R.

  1. Using Definition 24.9, compute the length of one arch:

    L=02π2Rsint2 ⁣dt=8RL = \int_0^{2\pi} 2R\sin\frac t2\,\dd t = 8R

    (Wren’s theorem, 1658). Four wheel diameters, and no π\pi anywhere.

  2. Compute the arc length from the cusp: s(t)=4R(1cost2)s(t) = 4R\bigl(1 - \cos\frac t2\bigr), and check s(2π)=8Rs(2\pi) = 8R.
  3. Now measure the arc from the apex t=πt = \pi: σ(t)=4Rcost2\sigma(t) = \abs{4R\cos\frac t2}. Prove the intrinsic relation

    σ2=8R(2Ry):\sigma^2 = 8R\,\bigl(2R - y\bigr) :

    the squared arc distance from the top is proportional to the height drop below the top.

  4. Sanity checks: recover from question 8 that the half-arch from apex to cusp has length 4R4R, and compare with the astroid computation of Example 24.10 — both curves have algebraic, π\pi-free lengths; explain what makes this possible even though both are built from circles. (Look at the form of f\norm{f'}.)

Part III — Roberval’s area: 3πR23\pi R^2.

  1. Justify that the area between one arch and the ground is A=02πy(t)x(t) ⁣dtA = \int_0^{2\pi} y(t)\,x'(t)\,\dd t (the substitution x=x(t)x = x(t) in y ⁣dx\int y\,\dd x, Theorem 15.15; xx is increasing).
  2. Compute

    A=R202π(1cost)2 ⁣dt=3πR2:A = R^2\int_0^{2\pi}(1 - \cos t)^2\,\dd t = 3\pi R^2 :

    exactly three times the area of the wheel — the ratio Galileo had guessed by weighing.

  3. Same method for the astroid (cos3t,sin3t)(\cos^3t, \sin^3t): show that the enclosed area is 3π8\frac{3\pi}8 (linearize sin4tcos2t\sin^4 t\cos^2 t; only the constant term survives over a full period).
  4. Sanity check the formula of question 10 on the upper unit semicircle (cost,sint)(\cos t, \sin t), tt from π\pi to 00: does it return π2\frac\pi2?

Part IV — Huygens’ tautochrone. Flip the arch: a frictionless bead slides, under gravity gg, inside the cycloidal bowl

x(t)=R(t+sint),y(t)=R(1cost)(t[π,π]),x(t) = R(t + \sin t), \qquad y(t) = R(1 - \cos t) \qquad (t \in \intcc{-\pi}{\pi}),

whose lowest point is the origin (yy measured upward).

  1. Compute the speed f(t)=2Rcost2\norm{f'(t)} = 2R\cos\frac t2, the arc length from the bottom s(t)=4Rsint2s(t) = 4R\sin\frac t2, and prove the key identity

    y=s28R.y = \frac{s^2}{8R} .
  2. The bead released at rest from the point of parameter t0>0t_0 > 0 obeys energy conservation: if s(τ)s(\tau) denotes its arc position at time τ\tau, then 12( ⁣ds ⁣dτ)2+gy=gy0\frac12\bigl( \frac{\dd s}{\dd\tau}\bigr)^2 + g\,y = g\,y_0. Rewrite this, using question 14, as

    ( ⁣ds ⁣dτ) ⁣2=g4R(s02s2),s0=s(t0).\Bigl(\frac{\dd s}{\dd\tau}\Bigr)^{\!2} = \frac{g}{4R}\,\bigl(s_0^2 - s^2\bigr), \qquad s_0 = s(t_0).
  3. Differentiate with respect to τ\tau and obtain the harmonic oscillator

     ⁣d2s ⁣dτ2=g4Rs;\frac{\dd^2 s}{\dd\tau^2} = -\frac{g}{4R}\,s ;

    solve it with Theorem 5.10 and the initial conditions: s(τ)=s0cos(ωτ)s(\tau) = s_0\cos(\omega\tau), ω=g/4R\omega = \sqrt{g/4R}.

  4. Deduce the tautochrone property (Huygens, 1659): the time to reach the bottom,

    T=π2ω=πRg,T_{\downarrow} = \frac{\pi}{2\omega} = \pi\sqrt{\frac Rg}\,,

    does not depend on the release point — beads released together from any two heights of the bowl arrive together.

  5. Verify by substitution that s(τ)=s0cosωτs(\tau) = s_0\cos\omega\tau satisfies the first-order energy equation of question 15 exactly (not only the differentiated one), and explain in one sentence why a circular pendulum is only approximately isochronous while the cycloid is exactly so.
  6. Numerically: what radius RR makes the descent time exactly one second (g=9.81g = 9.81)? Note how close the answer is to one meter, and how the full oscillation period 2π4R/g2\pi\sqrt{4R/g} compares with the small-angle pendulum formula 2π/g2\pi\sqrt{\ell/g} for =4R\ell = 4R.

Part V — Dividends, and synthesis.

  1. Apply the tangent rule of question 3 at t=π2t = \frac\pi2 (take R=1R = 1): compute MM, TT, the direction of MTMT, and check it against f(π2)f'(\frac\pi2).
  2. (Trochoids) Mark instead a point at distance dd from the center (dRd \neq R): the curve is x=Rtdsintx = Rt - d\sin t, y=Rdcosty = R - d\cos t. Show that for d<Rd < R the curve is regular everywhere and is the graph of no singular behavior (x>0x' > 0: it advances), while for d>Rd > R the abscissa xx' changes sign and the curve makes loops — the flanged railway wheel whose rim points travel backwards.
  3. (Brachistochrone teaser) From the cusp (πR,2R)(\pi R, 2R) of the bowl to the bottom, compare the cycloid descent time πR/g\pi\sqrt{R/g} with the time along the straight chute joining the same points (constant acceleration gsinαg\sin\alpha along the chord): show the chord takes π2+4R/g3.72R/g\sqrt{\pi^2 + 4}\,\sqrt{R/g} \approx 3.72\sqrt{R/g}. The curve beats the line — it is, in fact, the fastest of all curves, a result of the calculus of variations.
  4. Show that on the arch (0<t<2π0 < t < 2\pi),

     ⁣dy ⁣dx=cott2, ⁣d2y ⁣dx2=14Rsin4t2<0:\frac{\dd y}{\dd x} = \cot\frac t2, \qquad \frac{\dd^2 y}{\dd x^2} = -\frac{1}{4R\sin^4\frac t2} < 0 :

    the arch is concave, with vertical tangents exactly at the cusps.

  5. Recover the vertical cusp tangent of Exercise 24.2 geometrically: compute the limit direction of the chord MTMT of question 3 as t0+t \to 0^{+}, with no expansions at all.
  6. Synthesis, in four sentences: which three named theorems this problem proved (with their numbers 8R8R, 3πR23\pi R^2, πR/g\pi\sqrt{R/g} and their authors); which single computational device (sint2\sin\frac t2, cost2\cos\frac t2 factorizations) powered all of Parts I, II and IV; how the intrinsic relation y=s2/8Ry = s^2/8R converted geometry into a linear differential equation; and which chapter of this book each Part leaned on.
Solution

Solution of Problem 24.1.

1. Rolling without slipping means the contact point has traveled a distance equal to the arc of wheel unrolled: after turning by tt, the center is at (Rt,R)(Rt, R). The marked point sits on the rim at angle tt behind the downward vertical (the wheel turns clockwise while advancing):

M(t)=Ω(t)+R(sint,cost)=(R(tsint), R(1cost)),M(t) = \Omega(t) + R(-\sin t, -\cos t) = \bigl(R(t - \sin t),\ R(1 - \cos t)\bigr),

which is correct at t=0t = 0 (M=(0,0)M = (0,0)) and at t=πt = \pi (M=(πR,2R)M = (\pi R, 2R), the top).

2. f(t)=R(1cost, sint)f'(t) = R(1 - \cos t,\ \sin t) and

f2=R2((1cost)2+sin2t)=2R2(1cost)=4R2sin2t2,\norm{f'}^2 = R^2\bigl((1-\cos t)^2 + \sin^2 t\bigr) = 2R^2(1 - \cos t) = 4R^2\sin^2\frac t2 ,

so f=2Rsint2\norm{f'} = 2R\abs{\sin\frac t2}. Singular points at t2πZt \in 2\pi\Z: the cusps on the ground (Exercise 24.2); the top of the arch is t=πt = \pi, where the speed 2R2R is maximal.

3. Half-angle factorizations:

f(t)=2Rsint2(sint2, cost2),MC=2Rsint2(cost2, sint2),f'(t) = 2R\sin\frac t2\,\Bigl(\sin\frac t2,\ \cos\frac t2\Bigr), \quad M - C = 2R\sin\frac t2\,\Bigl(-\cos\frac t2,\ \sin\frac t2\Bigr),
TM=(Rsint, R(1+cost))=2Rcost2(sint2, cost2).T - M = \bigl(R\sin t,\ R(1 + \cos t)\bigr) = 2R\cos\frac t2\,\Bigl(\sin\frac t2,\ \cos\frac t2\Bigr).

For 0<t<2π0 < t < 2\pi, sint20\sin\frac t2 \neq 0: MCM - C is orthogonal to the tangent direction (sint2,cost2)\bigl(\sin\frac t2, \cos\frac t2\bigr) (their product is sint2cost2+sint2cost2=0-\sin\frac t2\cos\frac t2 + \sin\frac t2\cos\frac t2 = 0), and TMT - M is parallel to it. The chord to the top of the wheel is the tangent; the chord to the contact point is the normal.

4. At each instant the wheel pivots about its contact point (that point has zero velocity: rolling without slipping), so every rigid point of the wheel moves orthogonally to the line joining it to CC, with speed (angular velocity 11) equal to that distance. Check: MC=2Rsint2=f(t)\norm{M - C} = 2R\abs{\sin\frac t2} = \norm{f'(t)}.

5. 2Ry(t)=2R2Rsin2t2=4R2sin2t2=f(t)22R\,y(t) = 2R\cdot 2R\sin^2\frac t2 = 4R^2\sin^2\frac t2 = \norm{f'(t)}^2. The speed at height yy is 2Ry\sqrt{2R\,y} — formally the law v=2ghv = \sqrt{2gh} of free fall, with the ground playing the ceiling; Part IV turns this observation into clockwork.

6. By Definition 24.9 and question 2 (sint20\sin\frac t2 \geq 0 on [0,2π]\intcc{0}{2\pi}):

L=02π2Rsint2 ⁣dt=2R[2cost2]02π=2R(2+2)=8R.L = \int_0^{2\pi} 2R\sin\frac t2\,\dd t = 2R\Bigl[-2\cos\frac t2\Bigr]_0^{2\pi} = 2R(2 + 2) = 8R .

Wren’s theorem: four diameters exactly.

7. s(t)=0t2Rsinu2 ⁣du=4R(1cost2)s(t) = \int_0^t 2R\sin\frac u2\,\dd u = 4R\bigl(1 - \cos\frac t2\bigr); s(2π)=4R(1+1)=8Rs(2\pi) = 4R(1+1) = 8R, consistent.

8. σ(t)=s(t)s(π)=4R(1cost2)4R=4Rcost2\sigma(t) = \abs{s(t) - s(\pi)} = \abs{4R(1 - \cos\frac t2) - 4R} = 4R\abs{\cos\frac t2}, so σ2=16R2cos2t2\sigma^2 = 16R^2\cos^2\frac t2; and 2Ry=R(1+cost)=2Rcos2t22R - y = R(1 + \cos t) = 2R\cos^2\frac t2, whence 8R(2Ry)=16R2cos2t2=σ28R(2R - y) = 16R^2\cos^2\frac t2 = \sigma^2.

9. At the cusp t=0t = 0 (or 2π2\pi): σ=4R\sigma = 4R, half of 8R8R: the apex halves the arch. In both computations the speed is trigonometric polynomial in t/2\abs{\text{trigonometric polynomial in } t/2}, whose antiderivative is again trigonometric: the length is a difference of values of cosines — rational numbers times RR — with no arc of circle to measure, hence no π\pi. The circle itself has constant speed, so its length integral produces the full interval length 2π2\pi; the cycloid’s speed vanishes at the ends and integrates algebraically.

10. The arch is swept with xx increasing from 00 to 2πR2\pi R (x=R(1cost)0x' = R(1 - \cos t) \geq 0, vanishing only at isolated points). Substituting x=x(t)x = x(t) in the area integral 02πRy ⁣dx\int_0^{2\pi R} y\,\dd x (Theorem 15.15) gives A=02πy(t)x(t) ⁣dtA = \int_0^{2\pi} y(t)\,x'(t)\,\dd t.

11.

A=02πR(1cost)R(1cost) ⁣dt=R202π(12cost+cos2t) ⁣dt=R2(2π0+π)=3πR2,A = \int_0^{2\pi} R(1 - \cos t)\cdot R(1 - \cos t)\,\dd t = R^2\int_0^{2\pi}\bigl(1 - 2\cos t + \cos^2 t\bigr)\dd t = R^2\Bigl(2\pi - 0 + \pi\Bigr) = 3\pi R^2 ,

using 02πcos2=π\int_0^{2\pi}\cos^2 = \pi. Exactly three wheel areas: Galileo’s balance said “about 33”; Roberval’s computation says “exactly”.

12. With x=cos3tx = \cos^3 t, y=sin3ty = \sin^3 t: yx=3sin4tcos2ty\,x' = -3\sin^4 t\cos^2 t. Linearize:

sin4tcos2t=1cos2t2sin22t4=sin22t8sin22tcos2t8,\sin^4 t\cos^2 t = \frac{1 - \cos 2t}{2}\cdot\frac{\sin^2 2t}{4} = \frac{\sin^2 2t}{8} - \frac{\sin^2 2t\cos 2t}{8} ,

and over [0,2π]\intcc{0}{2\pi}: sin22t=π\int\sin^2 2t = \pi, sin22tcos2t=[sin32t6]=0\int \sin^2 2t\cos 2t = \bigl[\frac{\sin^3 2t}{6}\bigr] = 0. So y ⁣dx=3π8\oint y\,\dd x = -3\cdot\frac\pi8, and the enclosed area is 3π8\frac{3\pi}8 (the sign records the counterclockwise orientation).

13. For (cost,sint)(\cos t, \sin t) with tt from π\pi to 00, xx increases from 1-1 to 11 and

π0sint(sint) ⁣dt=0πsin2t ⁣dt=π2:\int_\pi^0 \sin t\cdot(-\sin t)\,\dd t = \int_0^\pi \sin^2 t\,\dd t = \frac\pi2 :

the formula returns the area of the upper half-disc, as it must.

14. x=R(1+cost)=2Rcos2t2x' = R(1 + \cos t) = 2R\cos^2\frac t2, y=Rsint=2Rsint2cost2y' = R\sin t = 2R\sin\frac t2\cos\frac t2, so f=2Rcost2\norm{f'} = 2R\cos\frac t2 (nonnegative on [π,π]\intcc{-\pi}{\pi}). Arc from the bottom: s(t)=0t2Rcosu2 ⁣du=4Rsint2s(t) = \int_0^t 2R\cos\frac u2\,\dd u = 4R\sin\frac t2. Then

y=R(1cost)=2Rsin2t2=2R(s4R) ⁣2=s28R.y = R(1 - \cos t) = 2R\sin^2\frac t2 = 2R\Bigl(\frac{s}{4R}\Bigr)^{\!2} = \frac{s^2}{8R} .

15. Energy conservation with v= ⁣ds ⁣dτv = \frac{\dd s}{\dd\tau} and y=s28Ry = \frac{s^2}{8R}, y0=s028Ry_0 = \frac{s_0^2}{8R}:

( ⁣ds ⁣dτ) ⁣2=2g(y0y)=2g8R(s02s2)=g4R(s02s2).\Bigl(\frac{\dd s}{\dd\tau}\Bigr)^{\!2} = 2g(y_0 - y) = \frac{2g}{8R}\bigl(s_0^2 - s^2\bigr) = \frac{g}{4R}\bigl(s_0^2 - s^2\bigr).

16. Differentiating in τ\tau: 2ss=g4R2ss2s's'' = -\frac{g}{4R}\,2ss', so wherever s0s' \neq 0 (hence everywhere by continuity), s=g4Rss'' = -\frac{g}{4R}s: the harmonic oscillator. By Theorem 5.10, s(τ)=Acosωτ+Bsinωτs(\tau) = A\cos\omega\tau + B\sin\omega\tau with ω=g/(4R)\omega = \sqrt{g/(4R)}; the initial conditions s(0)=s0s(0) = s_0, s(0)=0s'(0) = 0 give s(τ)=s0cosωτs(\tau) = s_0\cos\omega\tau.

17. The bead reaches the bottom when s=0s = 0, i.e. at ωτ=π2\omega\tau = \frac\pi2:

T=π2ω=π24Rg=πRg,T_{\downarrow} = \frac{\pi}{2\omega} = \frac\pi2\sqrt{\frac{4R}{g}} = \pi\sqrt{\frac Rg}\,,

independent of s0s_0: released from anywhere in the bowl, all beads arrive at the same instant — the tautochrone.

18. Substituting s=s0cosωτs = s_0\cos\omega\tau: the left side is s02ω2sin2ωτs_0^2\omega^2\sin^2\omega\tau and the right side is g4Rs02(1cos2ωτ)=s02ω2sin2ωτ\frac{g}{4R}s_0^2(1 - \cos^2\omega\tau) = s_0^2\omega^2\sin^2\omega\tau: exact equality, so no spurious solution was introduced. For a circular arc, yy is not proportional to s2s^2 (y=(1coss)=s22s4243+y = \ell(1 - \cos\frac s\ell) = \frac{s^2}{2\ell} - \frac{s^4}{24\ell^3} + \dots): the restoring term is only approximately linear, so the period of a circular pendulum drifts with amplitude, while the cycloid’s is rigorously constant.

19. πR/g=1\pi\sqrt{R/g} = 1 gives R=gπ29.819.870.994R = \frac{g}{\pi^2} \approx \frac{9.81}{9.87} \approx 0.994 m — almost exactly one meter. The full oscillation takes 2πω=2π4R/g\frac{2\pi}\omega = 2\pi\sqrt{4R/g}, which is precisely the small-angle formula 2π/g2\pi\sqrt{\ell/g} for a pendulum of length =4R\ell = 4R: Huygens suspended his pendulum from cycloidal cheeks of exactly that proportion.

20. At t=π2t = \frac\pi2, R=1R = 1: M=(π21, 1)M = \bigl(\frac\pi2 - 1,\ 1\bigr), T=(π2, 2)T = \bigl(\frac\pi2,\ 2\bigr), so TM=(1,1)T - M = (1, 1). And f(π2)=(10, 1)=(1,1)f'(\frac\pi2) = (1 - 0,\ 1) = (1, 1): the chord to the top is the velocity, as promised.

21. For the trochoid, x(t)=Rdcostx'(t) = R - d\cos t and y(t)=dsinty'(t) = d\sin t. If d<Rd < R: xRd>0x' \geq R - d > 0, the point always advances; the velocity never vanishes (its first component is positive): no singular point, a regular wave. If d>Rd > R: x(0)=Rd<0<R+d=x(π)x'(0) = R - d < 0 < R + d = x'(\pi), so the point moves backwards near the contact instants and forwards elsewhere: the curve crosses itself in loops. A point on the flange of a railway wheel (below the rail head, d>Rd > R) travels backwards at every turn.

22. The chord from the cusp (πR,2R)(\pi R, 2R) to the origin has length L=Rπ2+4L = R\sqrt{\pi^2 + 4} and slope angle α\alpha with sinα=2RL\sin\alpha = \frac{2R}{L}. Sliding from rest with constant acceleration gsinαg\sin\alpha: L=12gsinαT2L = \frac12 g\sin\alpha\,T^2, so

T=2Lgsinα=2L22Rg=LRg=π2+4Rg3.72Rg,T = \sqrt{\frac{2L}{g\sin\alpha}} = \sqrt{\frac{2L^2}{2Rg}} = \frac{L}{\sqrt{Rg}} = \sqrt{\pi^2 + 4}\,\sqrt{\frac Rg} \approx 3.72\sqrt{\frac Rg}\,,

against πR/g3.14R/g\pi\sqrt{R/g} \approx 3.14\sqrt{R/g} for the cycloid: the curved path is faster. It is in fact the fastest possible — the brachistochrone — a theorem of the calculus of variations, beyond this volume.

23.  ⁣dy ⁣dx=yx=sint1cost=cott2\dfrac{\dd y}{\dd x} = \dfrac{y'}{x'} = \dfrac{\sin t}{1 - \cos t} = \cot\dfrac t2 (half-angle formulas). Then

 ⁣d2y ⁣dx2= ⁣d ⁣dt(cott2)1x(t)=12sin2t212Rsin2t2=14Rsin4t2<0:\frac{\dd^2y}{\dd x^2} = \frac{\dd}{\dd t}\Bigl(\cot\frac t2\Bigr)\cdot\frac{1}{x'(t)} = -\frac{1}{2\sin^2\frac t2}\cdot\frac{1}{2R\sin^2\frac t2} = -\frac{1}{4R\sin^4\frac t2} < 0 :

concave throughout the arch; as t0+t \to 0^+ or 2π2\pi^- the slope cott2±\cot\frac t2 \to \pm\infty: vertical tangents at the cusps.

24. The chord direction is TM(sint2,cost2)T - M \parallel \bigl(\sin\frac t2, \cos\frac t2\bigr), which tends to (0,1)(0, 1) as t0+t \to 0^{+}: the tangent at the cusp is vertical — recovered from pure geometry, with no Taylor expansion.

25. (i) Wren’s theorem, L=8RL = 8R; Roberval’s area, A=3πR2A = 3\pi R^2 (Galileo’s conjectured ratio 33); Huygens’ tautochrone, T=πR/gT_\downarrow = \pi\sqrt{R/g}. (ii) Every computation ran on the half-angle factorizations f=2Rsint2(sint2,cost2)f' = 2R\sin\frac t2(\sin\frac t2, \cos\frac t2) and their bowl analogue — one identity powering tangent, length and clock alike. (iii) The intrinsic relation y=s2/8Ry = s^2/8R converted the energy equation into s=g4Rss'' = -\frac{g}{4R}s, a linear equation with constant coefficients whose solutions are exactly isochronous. (iv) Part I used the differential calculus of Chapter 14, Part II–III the integral of Chapter 15, Part IV the differential equations of Chapter 5 — the cycloid is this book’s curriculum rolled into one curve.