跳转到内容

STEP3 2005 -- Pure Mathematics

此内容尚不支持你的语言。

STEP3 2005 — Section A (Pure Mathematics)

Section titled “STEP3 2005 — Section A (Pure Mathematics)”

Exam: STEP3  |  Year: 2005  |  Questions: Q1—Q8  |  Total marks per question: 20

All questions are pure mathematics. Solutions and examiner commentary are included below.

QTopicDifficultyKey Techniques
1三角函数 TrigonometryStandard三角恒等式变换,辅助角公式,不等式估计,函数图像分析
2微分方程与曲线 Differential Equations and CurvesChallenging分离变量法,隐函数求导,二阶导数判别法,参数分类讨论
3多项式与方程 Polynomials and EquationsStandard多项式复合,待定系数法,配方法,四次方程因式分解
4数列与递推关系 Sequences and Recurrence RelationsChallenging数学归纳法,递推关系化简,等比数列条件,周期性分析
5二次曲线与切线 Quadratic Curves and TangentsChallenging切线方程推导,判别式分析,必要充分条件论证,线性方程解的存在性
6三次方程与双曲函数Cubic Equations and Hyperbolic FunctionsHard
7积分技巧Integration TechniquesChallenging
8复数与几何Complex Numbers and GeometryChallenging

Topic: 三角函数 Trigonometry  |  Difficulty: Standard  |  Marks: 20

1 Show that sinA=cosB\sin A = \cos B if and only if A=(4n+1)π2±BA = (4n + 1) \frac{\pi}{2} \pm B for some integer nn.

Show also that sinx±cosx2|\sin x \pm \cos x| \leqslant \sqrt{2} for all values of xx and deduce that there are no solutions to the equation sin(sinx)=cos(cosx)\sin (\sin x) = \cos (\cos x).

Sketch, on the same axes, the graphs of y=sin(sinx)y = \sin (\sin x) and y=cos(cosx)y = \cos (\cos x). Sketch, not on the previous axes, the graph of y=sin(2sinx)y = \sin (2 \sin x).

Hint

1 To prove the first part, use the results: cosB=sin(π2B)\cos B = \sin \left( \frac{\pi}{2} - B \right), whatever the value of BB; and

sinA=sinYA=Y+2nπ or A=πY+2nπ;\sin A = \sin Y \Leftrightarrow A = Y + 2n\pi \text{ or } A = \pi - Y + 2n\pi;

thus here, replacing YY by π2B\frac{\pi}{2} - B, A=2nπ+π2±BA = 2n\pi + \frac{\pi}{2} \pm B.

For the next part, it is probably easiest to use the fact that asinx±bcosxa \sin x \pm b \cos x can be written in the form Rsin(x±α)R \sin(x \pm \alpha); here,

sinx±cosx=2(sinxcosπ4±cosxsinπ4)=2sin(x±π4)\sin x \pm \cos x = \sqrt{2} \left( \sin x \cos \frac{\pi}{4} \pm \cos x \sin \frac{\pi}{4} \right) = \sqrt{2} \sin \left( x \pm \frac{\pi}{4} \right)

so sinx±cosx2|\sin x \pm \cos x| \leqslant \sqrt{2}.

Now, from the first part,

sin(sinx)=cos(cosx)sinx=2nπ+π2±cosx\sin (\sin x) = \cos (\cos x) \Leftrightarrow \sin x = 2n\pi + \frac{\pi}{2} \pm \cos x

so

sinx±cosx2nπ+π2π2>2,|\sin x \pm \cos x| \geqslant \left| 2n\pi + \frac{\pi}{2} \right| \geqslant \frac{\pi}{2} > \sqrt{2},

which is a contradiction.

All the curves asked for have period 2π2\pi, so they will be sketched for xx in this range only.

For y=sin(sinx)y = \sin (\sin x), y=0y = 0 when sinx=0\sin x = 0 only (since sinx<π|\sin x| < \pi), so at 00, π\pi and 2π2\pi; the turning points are at cosxcos(sinx)=0\cos x \cos (\sin x) = 0, so when cosx=0\cos x = 0, that is at x=π2,3π2x = \frac{\pi}{2}, \frac{3\pi}{2}, or when cos(sinx)=0\cos (\sin x) = 0, which is impossible since sinx<π2|\sin x| < \frac{\pi}{2}; the turning points are a maximum at (π2,sin(1))\left( \frac{\pi}{2}, \sin(1) \right) and a minimum at (3π2,sin(1))\left( \frac{3\pi}{2}, -\sin(1) \right), where sin(1)0.84\sin(1) \approx 0.84.

For y=cos(cosx)y = \cos (\cos x), y>0y > 0 for all xx, since cosx1<π2|\cos x| \leqslant 1 < \frac{\pi}{2}; the turning points are at sinxsin(cosx)=0\sin x \sin (\cos x) = 0, so when either sinx=0\sin x = 0 or cosx=0\cos x = 0, that is at x=0,π2,π,3π2,2πx = 0, \frac{\pi}{2}, \pi, \frac{3\pi}{2}, 2\pi; the turning points are maxima at (π2,1)\left( \frac{\pi}{2}, 1 \right) and (3π2,1)\left( \frac{3\pi}{2}, 1 \right), and minima at (0,cos(1))(0, \cos(1)), (π,cos(1))(\pi, \cos(1)), (2π,cos(1))(2\pi, \cos(1)), where cos(1)0.54\cos(1) \approx 0.54.

x y = cos(cos x) y = sin(sin x)
0 0.54 0
π/2 1 0.84
π 0.54 0
3π/2 1 -0.84
0.54 0

For the curve y=sin(2sinx)y = \sin(2 \sin x), y=0y = 0 if 2sinx2 \sin x is a multiple of π\pi, which is only possible if sinx=0\sin x = 0 (since 2sinx<π|2 \sin x| < \pi), so when xx is 00, π\pi and 2π2\pi; the turning points are at 2cosxcos(2sinx)=02 \cos x \cos(2 \sin x) = 0; so when cosx=0\cos x = 0, that is at x=π2,3π2x = \frac{\pi}{2}, \frac{3\pi}{2}, or when 2sinx2 \sin x is an odd multiple of π2\frac{\pi}{2}, which is only possible if 2sinx=±π22 \sin x = \pm \frac{\pi}{2}, so when sinx=π4±0.8\sin x = \frac{\pi}{4} \approx \pm 0.8; the turning points are a minimum at (π2,sin2)\left(\frac{\pi}{2}, \sin 2\right), where sin20.91\sin 2 \approx 0.91 and maxima either side of this, with yy-coordinates 11 and a maximum at (3π2,sin2)\left(\frac{3\pi}{2}, -\sin 2\right) with minima either side with yy-coordinates 1-1.

x y
0 0
π 0
0
Model Solution

Part 1: Show that sinA=cosB\sin A = \cos B if and only if A=(4n+1)π2±BA = (4n+1)\frac{\pi}{2} \pm B

We start with the identity cosB=sin(π2B)\cos B = \sin\left(\frac{\pi}{2} - B\right), which holds for all BB.

Therefore sinA=cosB\sin A = \cos B if and only if sinA=sin(π2B)\sin A = \sin\left(\frac{\pi}{2} - B\right).

The general solution of sinA=sinY\sin A = \sin Y is A=2nπ+YA = 2n\pi + Y or A=2nπ+(πY)A = 2n\pi + (\pi - Y) for integer nn.

Substituting Y=π2BY = \frac{\pi}{2} - B:

A=2nπ+π2BorA=2nπ+π2+BA = 2n\pi + \frac{\pi}{2} - B \qquad \text{or} \qquad A = 2n\pi + \frac{\pi}{2} + B

which can be combined as

A=2nπ+π2±B=(4n+1)π2±B.A = 2n\pi + \frac{\pi}{2} \pm B = (4n + 1)\frac{\pi}{2} \pm B. \qquad \blacksquare


Part 2: Show that sinx±cosx2|\sin x \pm \cos x| \leqslant \sqrt{2}

We write sinx±cosx\sin x \pm \cos x in the form Rsin(x±α)R\sin(x \pm \alpha).

sinx+cosx=2(12sinx+12cosx)=2(sinxcosπ4+cosxsinπ4)=2sin(x+π4)\sin x + \cos x = \sqrt{2}\left(\frac{1}{\sqrt{2}}\sin x + \frac{1}{\sqrt{2}}\cos x\right) = \sqrt{2}\left(\sin x \cos\frac{\pi}{4} + \cos x \sin\frac{\pi}{4}\right) = \sqrt{2}\sin\left(x + \frac{\pi}{4}\right)

Similarly:

sinxcosx=2(sinxcosπ4cosxsinπ4)=2sin(xπ4)\sin x - \cos x = \sqrt{2}\left(\sin x \cos\frac{\pi}{4} - \cos x \sin\frac{\pi}{4}\right) = \sqrt{2}\sin\left(x - \frac{\pi}{4}\right)

Since sinθ1|\sin\theta| \leqslant 1 for all θ\theta, we have

sinx±cosx=2sin(x±π4)2.|\sin x \pm \cos x| = \sqrt{2}\left|\sin\left(x \pm \frac{\pi}{4}\right)\right| \leqslant \sqrt{2}. \qquad \blacksquare


Part 3: Deduce that sin(sinx)=cos(cosx)\sin(\sin x) = \cos(\cos x) has no solutions

From Part 1, sin(sinx)=cos(cosx)\sin(\sin x) = \cos(\cos x) holds if and only if

sinx=2nπ+π2±cosx\sin x = 2n\pi + \frac{\pi}{2} \pm \cos x

for some integer nn. Rearranging:

sinxcosx=2nπ+π2\sin x \mp \cos x = 2n\pi + \frac{\pi}{2}

Taking absolute values and applying Part 2:

2sinxcosx=2nπ+π2\sqrt{2} \geqslant |\sin x \mp \cos x| = \left|2n\pi + \frac{\pi}{2}\right|

If n1n \geqslant 1: 2nπ+π22π+π2=5π2>2\left|2n\pi + \frac{\pi}{2}\right| \geqslant 2\pi + \frac{\pi}{2} = \frac{5\pi}{2} > \sqrt{2}.

If n1n \leqslant -1: 2nπ+π22ππ2=3π2>2\left|2n\pi + \frac{\pi}{2}\right| \geqslant 2\pi - \frac{\pi}{2} = \frac{3\pi}{2} > \sqrt{2}.

If n=0n = 0: sinxcosx=π2|\sin x \mp \cos x| = \frac{\pi}{2}. But π21.571>21.414\frac{\pi}{2} \approx 1.571 > \sqrt{2} \approx 1.414, so π2>2\frac{\pi}{2} > \sqrt{2}, which contradicts sinxcosx2|\sin x \mp \cos x| \leqslant \sqrt{2}.

In all cases we reach a contradiction, so sin(sinx)=cos(cosx)\sin(\sin x) = \cos(\cos x) has no solutions. \qquad \blacksquare


Part 4: Sketches

Graph of y=sin(sinx)y = \sin(\sin x):

This has period 2π2\pi.

  • Zeros: sin(sinx)=0\sin(\sin x) = 0 when sinx=0\sin x = 0, i.e. x=0,π,2πx = 0, \pi, 2\pi (within [0,2π][0, 2\pi]).
  • Turning points: From dydx=cosxcos(sinx)=0\frac{dy}{dx} = \cos x \cdot \cos(\sin x) = 0, either cosx=0\cos x = 0 or cos(sinx)=0\cos(\sin x) = 0. The second requires sinx=π2|\sin x| = \frac{\pi}{2}, which is impossible since sinx1<π2|\sin x| \leqslant 1 < \frac{\pi}{2}. So turning points occur at x=π2,3π2x = \frac{\pi}{2}, \frac{3\pi}{2}.
    • At x=π2x = \frac{\pi}{2}: y=sin(1)0.84y = \sin(1) \approx 0.84 (maximum).
    • At x=3π2x = \frac{3\pi}{2}: y=sin(1)=sin(1)0.84y = \sin(-1) = -\sin(1) \approx -0.84 (minimum).

The curve is a smooth wave oscillating between ±sin(1)\pm\sin(1), resembling a sinusoid but slightly flattened.

Graph of y=cos(cosx)y = \cos(\cos x):

This has period 2π2\pi.

  • Since cosx1<π2|\cos x| \leqslant 1 < \frac{\pi}{2}, we have cos(cosx)>0\cos(\cos x) > 0 for all xx. The curve never crosses the xx-axis.
  • Turning points: From dydx=sinxsin(cosx)=0\frac{dy}{dx} = \sin x \cdot \sin(\cos x) = 0, either sinx=0\sin x = 0 or sin(cosx)=0\sin(\cos x) = 0.
    • sinx=0\sin x = 0 at x=0,π,2πx = 0, \pi, 2\pi: y=cos(1)0.54y = \cos(1) \approx 0.54 (minima).
    • sin(cosx)=0\sin(\cos x) = 0 when cosx=0\cos x = 0, i.e. x=π2,3π2x = \frac{\pi}{2}, \frac{3\pi}{2}: y=cos(0)=1y = \cos(0) = 1 (maxima).

The curve oscillates between cos(1)0.54\cos(1) \approx 0.54 and 11, always positive, with a “double-hump” shape per period.

Graph of y=sin(2sinx)y = \sin(2\sin x):

This has period 2π2\pi.

  • Zeros: sin(2sinx)=0\sin(2\sin x) = 0 when 2sinx=kπ2\sin x = k\pi for integer kk. Since 2sinx2<π|2\sin x| \leqslant 2 < \pi, the only possibility is k=0k = 0, i.e. sinx=0\sin x = 0, so x=0,π,2πx = 0, \pi, 2\pi.
  • Turning points: From dydx=2cosxcos(2sinx)=0\frac{dy}{dx} = 2\cos x \cdot \cos(2\sin x) = 0:
    • cosx=0\cos x = 0 at x=π2,3π2x = \frac{\pi}{2}, \frac{3\pi}{2}:
      • x=π2x = \frac{\pi}{2}: y=sin20.91y = \sin 2 \approx 0.91 (local minimum — the curve dips here from the surrounding peaks).
      • x=3π2x = \frac{3\pi}{2}: y=sin(2)=sin20.91y = \sin(-2) = -\sin 2 \approx -0.91 (local maximum).
    • cos(2sinx)=0\cos(2\sin x) = 0 when 2sinx=±π22\sin x = \pm\frac{\pi}{2}, i.e. sinx=±π4±0.785\sin x = \pm\frac{\pi}{4} \approx \pm 0.785. This gives four xx-values in [0,2π][0, 2\pi] where y=±1y = \pm 1 (the global maxima and minima).

The curve oscillates between 1-1 and 11 with a characteristic shape: two “side peaks” reaching ±1\pm 1 on either side of each main turning point at x=π2x = \frac{\pi}{2} and x=3π2x = \frac{3\pi}{2}, where y±0.91y \approx \pm 0.91.


Topic: 微分方程与曲线 Differential Equations and Curves  |  Difficulty: Challenging  |  Marks: 20

2 Find the general solution of the differential equation dydx=xyx2+a2\frac{\mathrm{d}y}{\mathrm{d}x} = -\frac{xy}{x^2 + a^2}, where a0a \neq 0, and show that it can be written in the form y2(x2+a2)=c2y^2(x^2 + a^2) = c^2, where cc is an arbitrary constant. Sketch this curve.

Find an expression for ddx(x2+y2)\frac{\mathrm{d}}{\mathrm{d}x}(x^2 + y^2) and show that

d2dx2(x2+y2)=2(1c2(x2+a2)2)+8c2x2(x2+a2)3 .\frac{\mathrm{d}^2}{\mathrm{d}x^2}(x^2 + y^2) = 2 \left( 1 - \frac{c^2}{(x^2 + a^2)^2} \right) + \frac{8c^2x^2}{(x^2 + a^2)^3} \text{ .}

(i) Show that, if 0<c<a20 < c < a^2, the points on the curve whose distance from the origin is least are (0,±ca)(0, \pm \frac{c}{a}).

(ii) If c>a2c > a^2, determine the points on the curve whose distance from the origin is least.

Hint

2 This equation can be solved by separating the variables:

2dyy=2xdxx2+a2soln(y2)=ln(x2+a2)+kory2(x2+a2)=c2.\int \frac{2 \, dy}{y} = - \int \frac{2x \, dx}{x^2 + a^2} \quad \text{so} \quad \ln(y^2) = -\ln(x^2 + a^2) + k \quad \text{or} \quad y^2(x^2 + a^2) = c^2.

The curve has two branches: one has y>0y > 0, reflection symmetry about the yy-axis, a maximum at (0,ca)(0, \frac{c}{a}) and y0y \to 0 as x|x| \to \infty; the other has y<0y < 0 and is a reflection of the first branch in the xx-axis.

ddx(x2+y2)=2x2ydydx=2x2xy2x2+a2=2x2xc2(x2+a2)2\frac{d}{dx}(x^2 + y^2) = 2x - 2y \frac{dy}{dx} = 2x - \frac{2xy^2}{x^2 + a^2} = 2x - \frac{2xc^2}{(x^2 + a^2)^2}

d2dx2(x2+y2)=22c2(x2+a2)2+4xc22x(x2+a2)3=2(1c2(x2+a2)2)+8c2x2(x2+a2)3\frac{d^2}{dx^2}(x^2 + y^2) = 2 - \frac{2c^2}{(x^2 + a^2)^2} + \frac{4xc^2 \cdot 2x}{(x^2 + a^2)^3} = 2 \left( 1 - \frac{c^2}{(x^2 + a^2)^2} \right) + \frac{8c^2x^2}{(x^2 + a^2)^3}

(i) ddx(x2+y2)=0\frac{\text{d}}{\text{d}x}(x^2 + y^2) = 0 when x=0x = 0 and when c2=(x2+a2)2c^2 = (x^2 + a^2)^2, but the latter is not possible if 0<c<a20 < c < a^2. If x=0x = 0, y=±cay = \pm \frac{c}{a} and d2dx2(x2+y2)=1c2a4\frac{\text{d}^2}{\text{d}x^2}(x^2 + y^2) = 1 - \frac{c^2}{a^4} which is positive if 0<c<a20 < c < a^2, indicating a local minimum. Hence the points on the curve whose distance from the origin is least are (0,±ca)(0, \pm \frac{c}{a}).

(ii) If c>a2c > a^2 then d2dx2(x2+y2)\frac{\text{d}^2}{\text{d}x^2}(x^2 + y^2) is negative at x=0x = 0, indicating a local maximum there; but in this case there are further stationary points at x2=ca2x^2 = c - a^2, y=±cy = \pm \sqrt{c} and at these points d2dx2(x2+y2)=8x2c>0\frac{\text{d}^2}{\text{d}x^2}(x^2 + y^2) = \frac{8x^2}{c} > 0. Hence the points on the curve whose distance from the origin is least are (±ca2,±c)(\pm \sqrt{c - a^2}, \pm \sqrt{c}).

Model Solution

Part 1: Solve the differential equation and show y2(x2+a2)=c2y^2(x^2 + a^2) = c^2

We have dydx=xyx2+a2\frac{dy}{dx} = -\frac{xy}{x^2 + a^2}. This is separable.

For y0y \neq 0:

dyy=xx2+a2dx\frac{dy}{y} = -\frac{x}{x^2 + a^2}\, dx

Integrating both sides:

lny=12ln(x2+a2)+k\ln|y| = -\frac{1}{2}\ln(x^2 + a^2) + k

where kk is an arbitrary constant. Exponentiating:

y=ekx2+a2|y| = \frac{e^k}{\sqrt{x^2 + a^2}}

so y=Ax2+a2y = \frac{A}{\sqrt{x^2 + a^2}} where AA is an arbitrary constant (which can be positive, negative, or zero — the zero case y=0y = 0 also satisfies the original ODE). Squaring:

y2=A2x2+a2y^2 = \frac{A^2}{x^2 + a^2}

y2(x2+a2)=c2where c2=A2.y^2(x^2 + a^2) = c^2 \qquad \text{where } c^2 = A^2. \qquad \blacksquare

Sketch of the curve:

The curve y2(x2+a2)=c2y^2(x^2 + a^2) = c^2 has two branches:

  • Upper branch (y>0y > 0): y=cx2+a2y = \frac{c}{\sqrt{x^2 + a^2}}, which is symmetric about the yy-axis, has a maximum of y=cay = \frac{c}{a} at x=0x = 0, and y0y \to 0 as x|x| \to \infty.
  • Lower branch (y<0y < 0): y=cx2+a2y = -\frac{c}{\sqrt{x^2 + a^2}}, which is the reflection of the upper branch in the xx-axis.

The curve looks like a symmetric pair of bell-shaped curves, one above and one below the xx-axis, never touching the xx-axis.


Part 2: Find ddx(x2+y2)\frac{d}{dx}(x^2 + y^2) and verify the second derivative

Since y2(x2+a2)=c2y^2(x^2 + a^2) = c^2, we have y2=c2x2+a2y^2 = \frac{c^2}{x^2 + a^2}.

First derivative. Using implicit differentiation of x2+y2x^2 + y^2:

ddx(x2+y2)=2x+2ydydx\frac{d}{dx}(x^2 + y^2) = 2x + 2y\frac{dy}{dx}

Substituting dydx=xyx2+a2\frac{dy}{dx} = -\frac{xy}{x^2 + a^2}:

ddx(x2+y2)=2x2xy2x2+a2=2x2xc2x2+a2x2+a2=2x2xc2(x2+a2)2\frac{d}{dx}(x^2 + y^2) = 2x - \frac{2xy^2}{x^2 + a^2} = 2x - \frac{2x \cdot \frac{c^2}{x^2 + a^2}}{x^2 + a^2} = 2x - \frac{2xc^2}{(x^2 + a^2)^2}

=2x(1c2(x2+a2)2).(*)= 2x\left(1 - \frac{c^2}{(x^2 + a^2)^2}\right). \qquad \text{(*)}

Second derivative. Differentiating ()(*) with respect to xx:

d2dx2(x2+y2)=ddx[2x2xc2(x2+a2)2]\frac{d^2}{dx^2}(x^2 + y^2) = \frac{d}{dx}\left[2x - \frac{2xc^2}{(x^2 + a^2)^2}\right]

=2c2ddx[2x(x2+a2)2]= 2 - c^2 \cdot \frac{d}{dx}\left[\frac{2x}{(x^2 + a^2)^2}\right]

Computing the derivative of 2x(x2+a2)2\frac{2x}{(x^2 + a^2)^2} using the quotient rule:

ddx[2x(x2+a2)2]=2(x2+a2)22x2(x2+a2)2x(x2+a2)4=2(x2+a2)8x2(x2+a2)3=2a26x2(x2+a2)3\frac{d}{dx}\left[\frac{2x}{(x^2 + a^2)^2}\right] = \frac{2(x^2 + a^2)^2 - 2x \cdot 2(x^2 + a^2) \cdot 2x}{(x^2 + a^2)^4} = \frac{2(x^2 + a^2) - 8x^2}{(x^2 + a^2)^3} = \frac{2a^2 - 6x^2}{(x^2 + a^2)^3}

Therefore:

d2dx2(x2+y2)=2c22a26x2(x2+a2)3=22a2c26c2x2(x2+a2)3\frac{d^2}{dx^2}(x^2 + y^2) = 2 - c^2 \cdot \frac{2a^2 - 6x^2}{(x^2 + a^2)^3} = 2 - \frac{2a^2c^2 - 6c^2x^2}{(x^2 + a^2)^3}

=22a2c2(x2+a2)3+6c2x2(x2+a2)3= 2 - \frac{2a^2c^2}{(x^2 + a^2)^3} + \frac{6c^2x^2}{(x^2 + a^2)^3}

Now we show this equals the required form. Observe that

2a2c2(x2+a2)3=2c2(x2+a2)2c2x2(x2+a2)3=2c2(x2+a2)22c2x2(x2+a2)3\frac{2a^2c^2}{(x^2 + a^2)^3} = \frac{2c^2(x^2 + a^2) - 2c^2x^2}{(x^2 + a^2)^3} = \frac{2c^2}{(x^2 + a^2)^2} - \frac{2c^2x^2}{(x^2 + a^2)^3}

Substituting:

d2dx2(x2+y2)=22c2(x2+a2)2+2c2x2(x2+a2)3+6c2x2(x2+a2)3\frac{d^2}{dx^2}(x^2 + y^2) = 2 - \frac{2c^2}{(x^2 + a^2)^2} + \frac{2c^2x^2}{(x^2 + a^2)^3} + \frac{6c^2x^2}{(x^2 + a^2)^3}

=2(1c2(x2+a2)2)+8c2x2(x2+a2)3.= 2\left(1 - \frac{c^2}{(x^2 + a^2)^2}\right) + \frac{8c^2x^2}{(x^2 + a^2)^3}. \qquad \blacksquare


Part (i): 0<c<a20 < c < a^2 — points closest to the origin

The distance from the origin squared is x2+y2x^2 + y^2. Stationary points satisfy ddx(x2+y2)=0\frac{d}{dx}(x^2 + y^2) = 0.

From ()(*):

2x(1c2(x2+a2)2)=02x\left(1 - \frac{c^2}{(x^2 + a^2)^2}\right) = 0

Either x=0x = 0, or (x2+a2)2=c2(x^2 + a^2)^2 = c^2, i.e. x2+a2=cx^2 + a^2 = c (taking the positive root since x2+a2>0x^2 + a^2 > 0), giving x2=ca2x^2 = c - a^2.

If 0<c<a20 < c < a^2, then ca2<0c - a^2 < 0, so x2=ca2x^2 = c - a^2 has no real solutions. The only stationary point is at x=0x = 0.

At x=0x = 0: y2=c2a2y^2 = \frac{c^2}{a^2}, so y=±cay = \pm\frac{c}{a}.

To confirm this is a minimum, we check the second derivative at x=0x = 0:

d2dx2(x2+y2)x=0=2(1c2a4)+0=2(1c2a4)\frac{d^2}{dx^2}(x^2 + y^2)\bigg|_{x=0} = 2\left(1 - \frac{c^2}{a^4}\right) + 0 = 2\left(1 - \frac{c^2}{a^4}\right)

Since 0<c<a20 < c < a^2, we have c2<a4c^2 < a^4, so 1c2a4>01 - \frac{c^2}{a^4} > 0. The second derivative is positive, confirming a local minimum.

Therefore, when 0<c<a20 < c < a^2, the points on the curve closest to the origin are (0,±ca)\left(0, \pm\frac{c}{a}\right). \qquad \blacksquare


Part (ii): c>a2c > a^2 — points closest to the origin

When c>a2c > a^2, we have ca2>0c - a^2 > 0, so x2=ca2x^2 = c - a^2 gives x=±ca2x = \pm\sqrt{c - a^2}.

Check x=0x = 0: The second derivative is 2(1c2a4)2\left(1 - \frac{c^2}{a^4}\right). Since c>a2c > a^2, we have c2>a4c^2 > a^4, so the second derivative is negative — this is a local maximum, not a minimum.

Check x2=ca2x^2 = c - a^2: At these points, x2+a2=cx^2 + a^2 = c, so (x2+a2)2=c2(x^2 + a^2)^2 = c^2. The second derivative becomes:

d2dx2(x2+y2)=2(1c2c2)+8c2x2c3=0+8x2c=8(ca2)c>0\frac{d^2}{dx^2}(x^2 + y^2) = 2\left(1 - \frac{c^2}{c^2}\right) + \frac{8c^2x^2}{c^3} = 0 + \frac{8x^2}{c} = \frac{8(c - a^2)}{c} > 0

since c>a2>0c > a^2 > 0. This confirms a local minimum.

At these points:

y2=c2x2+a2=c2c=cy^2 = \frac{c^2}{x^2 + a^2} = \frac{c^2}{c} = c

so y=±cy = \pm\sqrt{c}.

Therefore, when c>a2c > a^2, the points on the curve closest to the origin are (±ca2,±c)\left(\pm\sqrt{c - a^2}, \pm\sqrt{c}\right). \qquad \blacksquare


Topic: 多项式与方程 Polynomials and Equations  |  Difficulty: Standard  |  Marks: 20

3 Let f(x)=x2+px+qf(x) = x^2 + px + q and g(x)=x2+rx+sg(x) = x^2 + rx + s. Find an expression for f(g(x))f(g(x)) and hence find a necessary and sufficient condition on aa, bb and cc for it to be possible to write the quartic expression x4+ax3+bx2+cx+dx^4 + ax^3 + bx^2 + cx + d in the form f(g(x))f(g(x)), for some choice of values of p,q,rp, q, r and ss.

Show further that this condition holds if and only if it is possible to write the quartic expression x4+ax3+bx2+cx+dx^4 + ax^3 + bx^2 + cx + d in the form (x2+vx+w)2k(x^2 + vx + w)^2 - k, for some choice of values of v,wv, w and kk.

Find the roots of the quartic equation x44x3+10x212x+4=0x^4 - 4x^3 + 10x^2 - 12x + 4 = 0.

Wednesday 29 JUNE 2005 Afternoon 3 hours

Before the start of the examination please give each candidate a copy of this erratum notice.

There is an error in the sentence starting ‘Show that the equation …’ . This sentence should read:

‘Show that the equation x33a2x=2a3coshTx^3 - 3a^2x = 2a^3 \cosh T is satisfied by 2acosh(13T)2a \cosh (\frac{1}{3}T) and hence that, if c2b3>0c^2 \ge b^3 > 0, one of the roots of the equation x33bx=2cx^3 - 3bx = 2c is u+buu + \frac{b}{u}, where

u=(c+c2b3)13.u = \left( c + \sqrt{c^2 - b^3} \right)^{\frac{1}{3}}.

There is an error in this part question. The question should read:

‘Show that, if the discs collide at least once, their total kinetic energy just before the first collision is 43mgr(23μ)\frac{4}{3}mgr(2 - 3\mu).’

Any enquiry about this notice should be referred to the Information Bureau on 01223 553 998 or helpdesk@ocr.org.uk

Jun05/erratum09

Hint

3 Direct substitution gives

f(g(x))=(x2+rx+s)2+p(x2+rx+s)+qf(g(x)) = (x^2 + rx + s)^2 + p(x^2 + rx + s) + q =x4+2rx3+(r2+2s+p)x2+(2rs+rp)x+s2+ps+q.= x^4 + 2rx^3 + (r^2 + 2s + p)x^2 + (2rs + rp)x + s^2 + ps + q.

If x4+ax3+bx2+cx+dx^4 + ax^3 + bx^2 + cx + d is to have this form then it is necessary to choose r=a2r = \frac{a}{2} and to choose ss and pp to satisfy 2s+p=br2=ba242s + p = b - r^2 = b - \frac{a^2}{4} and r(2s+p)=cr(2s + p) = c or a(2s+p)=2ca(2s + p) = 2c. Thus a(ba24)=2ca\left(b - \frac{a^2}{4}\right) = 2c is a necessary condition for this to be possible.

It is also sufficient: in fact, pick p=0p = 0; then s=4ba28s = \frac{4b - a^2}{8} and q=ds2q = d - s^2 will do.

Expanding the second form gives

(x2+vx+w)2k=x4+2vx3+(v2+2w)x2+2vwx+w2k,(x^2 + vx + w)^2 - k = x^4 + 2vx^3 + (v^2 + 2w)x^2 + 2vwx + w^2 - k,

but this is identical to x4+2rx3+(r2+2s+p)x2+(2rs+rp)x+s2+ps+qx^4 + 2rx^3 + (r^2 + 2s + p)x^2 + (2rs + rp)x + s^2 + ps + q with p=0p = 0, v=rv = r, w=sw = s and k=qk = -q, and so, since the sufficiency demonstrated above allowed the choice p=0p = 0, the condition is the same.

To solve the final equation, write the quartic in the second form:

x44x3+10x212x+4=(x22x+3)25=0x^4 - 4x^3 + 10x^2 - 12x + 4 = (x^2 - 2x + 3)^2 - 5 = 0 so x22x+35=0 or x22x+3+5=0x^2 - 2x + 3 - \sqrt{5} = 0 \text{ or } x^2 - 2x + 3 + \sqrt{5} = 0 so x=1±52 or 1±i5+2.x = 1 \pm \sqrt{\sqrt{5} - 2} \text{ or } 1 \pm i\sqrt{\sqrt{5} + 2}.

Model Solution

Part 1: Find f(g(x))f(g(x)) and the necessary and sufficient condition

We compute f(g(x))f(g(x)) directly:

f(g(x))=(g(x))2+pg(x)+q=(x2+rx+s)2+p(x2+rx+s)+qf(g(x)) = (g(x))^2 + p \cdot g(x) + q = (x^2 + rx + s)^2 + p(x^2 + rx + s) + q

Expanding (x2+rx+s)2(x^2 + rx + s)^2:

=x4+2rx3+(r2+2s)x2+2rsx+s2= x^4 + 2rx^3 + (r^2 + 2s)x^2 + 2rsx + s^2

So:

f(g(x))=x4+2rx3+(r2+2s+p)x2+(2rs+rp)x+s2+ps+qf(g(x)) = x^4 + 2rx^3 + (r^2 + 2s + p)x^2 + (2rs + rp)x + s^2 + ps + q

For x4+ax3+bx2+cx+d=f(g(x))x^4 + ax^3 + bx^2 + cx + d = f(g(x)), we equate coefficients:

2r=a(coefficient of x3)2r = a \qquad \text{(coefficient of } x^3\text{)} r2+2s+p=b(coefficient of x2)r^2 + 2s + p = b \qquad \text{(coefficient of } x^2\text{)} 2rs+rp=c(coefficient of x)2rs + rp = c \qquad \text{(coefficient of } x\text{)} s2+ps+q=d(constant term)s^2 + ps + q = d \qquad \text{(constant term)}

From the first equation: r=a2r = \frac{a}{2}.

From the third equation: r(2s+p)=cr(2s + p) = c, so a2(2s+p)=c\frac{a}{2}(2s + p) = c, giving 2s+p=2ca2s + p = \frac{2c}{a} (provided a0a \neq 0).

From the second equation: 2s+p=br2=ba242s + p = b - r^2 = b - \frac{a^2}{4}.

For consistency we need:

ba24=2cab - \frac{a^2}{4} = \frac{2c}{a}

a(ba24)=2ca\left(b - \frac{a^2}{4}\right) = 2c

aba34=2cab - \frac{a^3}{4} = 2c

4aba3=8c4ab - a^3 = 8c

This is the necessary and sufficient condition. It is necessary because the coefficient equations force it. It is sufficient because, given this condition, we can choose p=0p = 0, r=a2r = \frac{a}{2}, s=4ba28s = \frac{4b - a^2}{8}, and q=ds2q = d - s^2, and all four coefficient equations are satisfied.


Part 2: Equivalence with the form (x2+vx+w)2k(x^2 + vx + w)^2 - k

Expanding (x2+vx+w)2k(x^2 + vx + w)^2 - k:

=x4+2vx3+(v2+2w)x2+2vwx+w2k= x^4 + 2vx^3 + (v^2 + 2w)x^2 + 2vwx + w^2 - k

This is exactly the same as f(g(x))f(g(x)) with p=0p = 0, v=rv = r, w=sw = s, and k=q=s2dk = -q = s^2 - d.

Since the sufficiency argument in Part 1 allowed us to choose p=0p = 0, the condition 4aba3=8c4ab - a^3 = 8c is both necessary and sufficient for writing the quartic in the form (x2+vx+w)2k(x^2 + vx + w)^2 - k. \qquad \blacksquare


Part 3: Find the roots of x44x3+10x212x+4=0x^4 - 4x^3 + 10x^2 - 12x + 4 = 0

We check the condition: a=4a = -4, b=10b = 10, c=12c = -12.

4aba3=4(4)(10)(4)3=160+64=96=8(12)=8c4ab - a^3 = 4(-4)(10) - (-4)^3 = -160 + 64 = -96 = 8(-12) = 8c \qquad \checkmark

So we can write this quartic as (x2+vx+w)2k(x^2 + vx + w)^2 - k.

We need v=a2=2v = \frac{a}{2} = -2 and w=4ba28=40168=3w = \frac{4b - a^2}{8} = \frac{40 - 16}{8} = 3.

Then k=w2d=94=5k = w^2 - d = 9 - 4 = 5.

x44x3+10x212x+4=(x22x+3)25x^4 - 4x^3 + 10x^2 - 12x + 4 = (x^2 - 2x + 3)^2 - 5

Verification: (x22x+3)2=x44x3+10x212x+9(x^2 - 2x + 3)^2 = x^4 - 4x^3 + 10x^2 - 12x + 9, so (x22x+3)25=x44x3+10x212x+4(x^2 - 2x + 3)^2 - 5 = x^4 - 4x^3 + 10x^2 - 12x + 4. \checkmark

Setting (x22x+3)25=0(x^2 - 2x + 3)^2 - 5 = 0:

(x22x+3)2=5(x^2 - 2x + 3)^2 = 5

x22x+3=5orx22x+3=5x^2 - 2x + 3 = \sqrt{5} \qquad \text{or} \qquad x^2 - 2x + 3 = -\sqrt{5}

Case 1: x22x+3=5x^2 - 2x + 3 = \sqrt{5}

x22x+(35)=0x^2 - 2x + (3 - \sqrt{5}) = 0

x=2±44(35)2=1±13+5=1±52x = \frac{2 \pm \sqrt{4 - 4(3 - \sqrt{5})}}{2} = 1 \pm \sqrt{1 - 3 + \sqrt{5}} = 1 \pm \sqrt{\sqrt{5} - 2}

Since 52.236>2\sqrt{5} \approx 2.236 > 2, these are real roots.

Case 2: x22x+3=5x^2 - 2x + 3 = -\sqrt{5}

x22x+(3+5)=0x^2 - 2x + (3 + \sqrt{5}) = 0

x=2±44(3+5)2=1±135=1±(2+5)=1±i2+5x = \frac{2 \pm \sqrt{4 - 4(3 + \sqrt{5})}}{2} = 1 \pm \sqrt{1 - 3 - \sqrt{5}} = 1 \pm \sqrt{-(2 + \sqrt{5})} = 1 \pm i\sqrt{2 + \sqrt{5}}

The four roots are:

x=1±52andx=1±i5+2x = 1 \pm \sqrt{\sqrt{5} - 2} \qquad \text{and} \qquad x = 1 \pm i\sqrt{\sqrt{5} + 2} \qquad \blacksquare


Topic: 数列与递推关系 Sequences and Recurrence Relations  |  Difficulty: Challenging  |  Marks: 20

4 The sequence unu_n (n=1,2,n = 1, 2, \dots) satisfies the recurrence relation

un+2=un+1un(kunun+1)u_{n+2} = \frac{u_{n+1}}{u_n}(ku_n - u_{n+1})

where kk is a constant.

If u1=au_1 = a and u2=bu_2 = b, where aa and bb are non-zero and bkab \neq ka, prove by induction that

u2n=(ba)u2n1u_{2n} = \left(\frac{b}{a}\right)u_{2n-1}

u2n+1=cu2nu_{2n+1} = cu_{2n}

for n1n \geq 1, where cc is a constant to be found in terms of kk, aa and bb. Hence express u2nu_{2n} and u2n1u_{2n-1} in terms of a,b,ca, b, c and nn.

Find conditions on a,ba, b and kk in the three cases:

(i) the sequence unu_n is geometric;

(ii) the sequence unu_n has period 2;

(iii) the sequence unu_n has period 4.

Hint

4 For the base case you need to verify that u2n=bau2n1u_{2n} = \frac{b}{a}u_{2n-1} and u2n+1=cu2nu_{2n+1} = cu_{2n} when n=1n = 1:

u1=a,u2=b so u2n=bau2n1 when n=1;u_1 = a, u_2 = b \text{ so } u_{2n} = \frac{b}{a}u_{2n-1} \text{ when } n = 1;

u3=u2u1(ku1u2)=u2kaba so u2n+1=cu2n when n=1, provided c=kba.u_3 = \frac{u_2}{u_1}(ku_1 - u_2) = u_2 \frac{ka - b}{a} \text{ so } u_{2n+1} = cu_{2n} \text{ when } n = 1, \text{ provided } c = k - \frac{b}{a}.

For the induction step, assume that u2n=bau2n1u_{2n} = \frac{b}{a}u_{2n-1} and u2n+1=cu2nu_{2n+1} = cu_{2n} when n=Nn = N then

u2N+2=u2N+1u2N(ku2Nu2N+1)u_{2N+2} = \frac{u_{2N+1}}{u_{2N}}(ku_{2N} - u_{2N+1}) =u2N+1(kc) (by the induction hypothesis)= u_{2N+1}(k - c) \text{ (by the induction hypothesis)} =bau2N+1 (by the definition of c)= \frac{b}{a}u_{2N+1} \text{ (by the definition of } c)

and

u2N+3=u2N+2u2N+1(ku2N+1u2N+2)u_{2N+3} = \frac{u_{2N+2}}{u_{2N+1}}(ku_{2N+1} - u_{2N+2}) =u2N+2(kba) (by what has just been shown)= u_{2N+2} \left( k - \frac{b}{a} \right) \text{ (by what has just been shown)} =cu2N+2.= cu_{2N+2}.

which completes the induction.

Hence u2n=bcau2n2==(bca)n1u2=b(bca)n1u_{2n} = \frac{bc}{a}u_{2n-2} = \dots = \left( \frac{bc}{a} \right)^{n-1} u_2 = b \left( \frac{bc}{a} \right)^{n-1} and u2n1=bcau2n3==(bca)n1u1=a(bca)n1u_{2n-1} = \frac{bc}{a}u_{2n-3} = \dots = \left( \frac{bc}{a} \right)^{n-1} u_1 = a \left( \frac{bc}{a} \right)^{n-1}

(i) For unu_n to be geometric requires u2nu2n1=u2n+1u2n\frac{u_{2n}}{u_{2n-1}} = \frac{u_{2n+1}}{u_{2n}}; that is, ba=c=kba\frac{b}{a} = c = k - \frac{b}{a} or ak=2bak = 2b;

(ii) For unu_n to have period 2 requires u2n+1=u2n1u_{2n+1} = u_{2n-1}, but u2n+1=cu2n=cbau2n1u_{2n+1} = cu_{2n} = \frac{cb}{a}u_{2n-1}, so it is necessary that cba=1\frac{cb}{a} = 1 or a2+b2=kaba^2 + b^2 = kab;

(iii) For unu_n to have period 4 requires u2n+3=u2n1u_{2n+3} = u_{2n-1} so, by the previous part, it is necessary that (cba)2=1\left( \frac{cb}{a} \right)^2 = 1 but cba1\frac{cb}{a} \neq 1 (to avoid period 2) so cba=1\frac{cb}{a} = -1 or b2a2=kabb^2 - a^2 = kab.

Model Solution

Part 1: Prove by induction that u2n=bau2n1u_{2n} = \frac{b}{a}u_{2n-1} and u2n+1=cu2nu_{2n+1} = cu_{2n}

Finding cc: From the recurrence with n=1n = 1:

u3=u2u1(ku1u2)=ba(kab)=b(kba)u_3 = \frac{u_2}{u_1}(ku_1 - u_2) = \frac{b}{a}(ka - b) = b\left(k - \frac{b}{a}\right)

So u3=cu2u_3 = cu_2 requires c=kbac = k - \frac{b}{a}.

Base case (n=1n = 1):

u2=b=baa=bau1u_2 = b = \frac{b}{a} \cdot a = \frac{b}{a}u_1 \checkmark

u3=cu2u_3 = cu_2 by definition of cc. \checkmark

Induction step: Assume u2N=bau2N1u_{2N} = \frac{b}{a}u_{2N-1} and u2N+1=cu2Nu_{2N+1} = cu_{2N} for some N1N \geq 1.

We need to show u2N+2=bau2N+1u_{2N+2} = \frac{b}{a}u_{2N+1} and u2N+3=cu2N+2u_{2N+3} = cu_{2N+2}.

From the recurrence:

u2N+2=u2N+1u2N(ku2Nu2N+1)u_{2N+2} = \frac{u_{2N+1}}{u_{2N}}(ku_{2N} - u_{2N+1})

Using the induction hypothesis u2N+1=cu2Nu_{2N+1} = cu_{2N}:

u2N+2=cu2Nu2N(ku2Ncu2N)=c(kc)u2Nu_{2N+2} = \frac{cu_{2N}}{u_{2N}}(ku_{2N} - cu_{2N}) = c(k - c)u_{2N}

Now c=kbac = k - \frac{b}{a}, so kc=bak - c = \frac{b}{a}. Therefore:

u2N+2=cbau2N=bacu2N=bau2N+1u_{2N+2} = c \cdot \frac{b}{a} \cdot u_{2N} = \frac{b}{a} \cdot cu_{2N} = \frac{b}{a}u_{2N+1} \qquad \checkmark

For the next term:

u2N+3=u2N+2u2N+1(ku2N+1u2N+2)u_{2N+3} = \frac{u_{2N+2}}{u_{2N+1}}(ku_{2N+1} - u_{2N+2})

Using u2N+2=bau2N+1u_{2N+2} = \frac{b}{a}u_{2N+1}:

u2N+3=bau2N+1u2N+1(ku2N+1bau2N+1)=ba(kba)u2N+1=bacu2N+1u_{2N+3} = \frac{\frac{b}{a}u_{2N+1}}{u_{2N+1}}\left(ku_{2N+1} - \frac{b}{a}u_{2N+1}\right) = \frac{b}{a}\left(k - \frac{b}{a}\right)u_{2N+1} = \frac{b}{a} \cdot c \cdot u_{2N+1}

But also u2N+3=cu2N+2u_{2N+3} = cu_{2N+2}, so we need cu2N+2=bcau2N+1cu_{2N+2} = \frac{bc}{a}u_{2N+1}. Since u2N+2=bau2N+1u_{2N+2} = \frac{b}{a}u_{2N+1}, this gives cbau2N+1=bcau2N+1c \cdot \frac{b}{a}u_{2N+1} = \frac{bc}{a}u_{2N+1}. \checkmark

By induction, the result holds for all n1n \geq 1, with c=kbac = k - \frac{b}{a}. \qquad \blacksquare


Part 2: Express u2nu_{2n} and u2n1u_{2n-1} in terms of a,b,c,na, b, c, n

From the relations u2n=bau2n1u_{2n} = \frac{b}{a}u_{2n-1} and u2n+1=cu2nu_{2n+1} = cu_{2n}, we get:

u2n+1=cu2n=bcau2n1u_{2n+1} = cu_{2n} = \frac{bc}{a}u_{2n-1}

So the odd-indexed terms satisfy u2n+1=bcau2n1u_{2n+1} = \frac{bc}{a}u_{2n-1}, which is a geometric progression with ratio bca\frac{bc}{a}.

u2n1=(bca)n1u1=a(bca)n1u_{2n-1} = \left(\frac{bc}{a}\right)^{n-1} u_1 = a\left(\frac{bc}{a}\right)^{n-1}

u2n=bau2n1=baa(bca)n1=b(bca)n1u_{2n} = \frac{b}{a}u_{2n-1} = \frac{b}{a} \cdot a\left(\frac{bc}{a}\right)^{n-1} = b\left(\frac{bc}{a}\right)^{n-1}


Part (i): unu_n is geometric

For unu_n to be geometric, we need the common ratio un+1un\frac{u_{n+1}}{u_n} to be constant for all nn. In particular:

u2u1=u3u2\frac{u_2}{u_1} = \frac{u_3}{u_2}

ba=u3b=cbb=c\frac{b}{a} = \frac{u_3}{b} = \frac{cb}{b} = c

So c=bac = \frac{b}{a}, which means kba=bak - \frac{b}{a} = \frac{b}{a}, giving:

k=2bai.e.ak=2bk = \frac{2b}{a} \qquad \text{i.e.} \qquad ak = 2b \qquad \blacksquare


Part (ii): unu_n has period 2

Period 2 means un+2=unu_{n+2} = u_n for all nn. In particular u3=u1u_3 = u_1, i.e. cu2=u1cu_2 = u_1, so cb=acb = a.

cb=a    (kba)b=a    kbb2a=a    kabb2=a2cb = a \implies \left(k - \frac{b}{a}\right)b = a \implies kb - \frac{b^2}{a} = a \implies kab - b^2 = a^2

a2+b2=kaba^2 + b^2 = kab \qquad \blacksquare

We can verify this gives period 2 for all terms: u2n+1=bcau2n1=bcau2n1u_{2n+1} = \frac{bc}{a}u_{2n-1} = \frac{bc}{a}u_{2n-1}. With cb=acb = a, this ratio is bca=1\frac{bc}{a} = 1, so u2n+1=u2n1u_{2n+1} = u_{2n-1} for all nn. Similarly u2n=u2n2u_{2n} = u_{2n-2}.


Part (iii): unu_n has period 4

Period 4 means un+4=unu_{n+4} = u_n but un+2unu_{n+2} \neq u_n (otherwise it would be period 2).

From Part 2, u2n1=a(bca)n1u_{2n-1} = a\left(\frac{bc}{a}\right)^{n-1}. For period 4, we need u2n+3=u2n1u_{2n+3} = u_{2n-1}, so (bca)2=1\left(\frac{bc}{a}\right)^2 = 1, giving bca=±1\frac{bc}{a} = \pm 1.

If bca=1\frac{bc}{a} = 1, this is the period 2 case (excluded). So we need bca=1\frac{bc}{a} = -1:

bc=a    (kba)b=a    kbb2a=a    kabb2=a2bc = -a \implies \left(k - \frac{b}{a}\right)b = -a \implies kb - \frac{b^2}{a} = -a \implies kab - b^2 = -a^2

b2a2=kabb^2 - a^2 = kab \qquad \blacksquare


Topic: 二次曲线与切线 Quadratic Curves and Tangents  |  Difficulty: Challenging  |  Marks: 20

5 Let PP be the point on the curve y=ax2+bx+cy = ax^2 + bx + c (where aa is non-zero) at which the gradient is mm. Show that the equation of the tangent at PP is

ymx=c(mb)24a.y - mx = c - \frac{(m - b)^2}{4a} .

Show that the curves y=a1x2+b1x+c1y = a_1x^2 + b_1x + c_1 and y=a2x2+b2x+c2y = a_2x^2 + b_2x + c_2 (where a1a_1 and a2a_2 are non-zero) have a common tangent with gradient mm if and only if

(a2a1)m2+2(a1b2a2b1)m+4a1a2(c2c1)+a2b12a1b22=0.(a_2 - a_1)m^2 + 2(a_1b_2 - a_2b_1)m + 4a_1a_2(c_2 - c_1) + a_2b_1^2 - a_1b_2^2 = 0 .

Show that, in the case a1a2a_1 \neq a_2, the two curves have exactly one common tangent if and only if they touch each other. In the case a1=a2a_1 = a_2, find a necessary and sufficient condition for the two curves to have exactly one common tangent.

Hint

5 The point on the curve with the required gradient is given by

dydx=2ax+b=m or x=mb2a,\frac{dy}{dx} = 2ax + b = m \text{ or } x = \frac{m - b}{2a},

with

y=a(mb2a)2+b(mb2a)+c=m2b24a+c.y = a \left( \frac{m - b}{2a} \right)^2 + b \left( \frac{m - b}{2a} \right) + c = \frac{m^2 - b^2}{4a} + c.

The equation of the tangent is therefore:

ymx=a(mb2a)2+b(mb2a)+cm(mb2a)y - mx = a \left( \frac{m - b}{2a} \right)^2 + b \left( \frac{m - b}{2a} \right) + c - m \left( \frac{m - b}{2a} \right)

=c(mb)2a(mb(mb)2)=c(mb)24a.= c - \frac{(m - b)}{2a} \left( m - b - \frac{(m - b)}{2} \right) = c - \frac{(m - b)^2}{4a}.

The curves have a common tangent with gradient mm if and only if the equations of the tangents to the two curves with gradient mm are identical; that is, have the same intercept, so if and only if

c1(mb1)24a1=c2(mb2)24a2c_1 - \frac{(m - b_1)^2}{4a_1} = c_2 - \frac{(m - b_2)^2}{4a_2}

that is,

4a1a2c1a2m2+2a2b1ma2b12=4a1a2c2a1m2+2a1b2ma1b224a_1a_2c_1 - a_2m^2 + 2a_2b_1m - a_2b_1^2 = 4a_1a_2c_2 - a_1m^2 + 2a_1b_2m - a_1b_2^2

which gives the quoted result.

There is exactly one common tangent when a1a2a_1 \neq a_2 when the quadratic equation for mm has exactly one root, which occurs if and only if the discriminant of the equation is zero; that is

4(a1b2a2b1)2=4(a2a1)(4a1a2(c2c1)+a2b12a1b22)4(a_1b_2 - a_2b_1)^2 = 4(a_2 - a_1)(4a_1a_2(c_2 - c_1) + a_2b_1^2 - a_1b_2^2)

4a12b228a1a2b1b2+4a22b12=16a1a2(a2a1)(c2c1)+4a22b12+4a12b224a1a2(b12+b22)\Leftrightarrow \quad 4a_1^2b_2^2 - 8a_1a_2b_1b_2 + 4a_2^2b_1^2 = 16a_1a_2(a_2 - a_1)(c_2 - c_1) + 4a_2^2b_1^2 + 4a_1^2b_2^2 - 4a_1a_2(b_1^2 + b_2^2)

4a1a2(b12+b22)8a1a2b1b2=16a1a2(a2a1)(c2c1)\Leftrightarrow \quad 4a_1a_2(b_1^2 + b_2^2) - 8a_1a_2b_1b_2 = 16a_1a_2(a_2 - a_1)(c_2 - c_1)

b12+b222b1b2=4(a2a1)(c2c1)(dividing by 4a1a2 which is non-zero).\Leftrightarrow \quad b_1^2 + b_2^2 - 2b_1b_2 = 4(a_2 - a_1)(c_2 - c_1) \quad \text{(dividing by } 4a_1a_2 \text{ which is non-zero).}

The curves touch if there is exactly one solution to the simultaneous equations

y=a1x2+b1x+c1 and y=a2x2+b2x+c2;y = a_1x^2 + b_1x + c_1 \text{ and } y = a_2x^2 + b_2x + c_2;

that is, if the equation (a2a1)x2+(b2b1)x+(c2c1)=0(a_2 - a_1)x^2 + (b_2 - b_1)x + (c_2 - c_1) = 0 has exactly one root so, again using the discriminant condition, if and only if (b2b1)2=4(a2a1)(c2c1)(b_2 - b_1)^2 = 4(a_2 - a_1)(c_2 - c_1), which is the same condition.

If a1=a2a_1 = a_2 the curves have exactly one common tangent if there is exactly one solution to

2m(b2b1)+4a(c2c1)+(b12b22)=0;2m(b_2 - b_1) + 4a(c_2 - c_1) + (b_1^2 - b_2^2) = 0;

since this is just a linear equation, the only condition is that b2b10b_2 - b_1 \neq 0.

Model Solution

Part 1: Equation of the tangent at PP

The gradient of y=ax2+bx+cy = ax^2 + bx + c is dydx=2ax+b\frac{dy}{dx} = 2ax + b. Setting this equal to mm:

2ax+b=m    x=mb2a2ax + b = m \implies x = \frac{m - b}{2a}

The yy-coordinate at PP is:

y=a(mb2a)2+b(mb2a)+c=(mb)24a+b(mb)2a+cy = a\left(\frac{m - b}{2a}\right)^2 + b\left(\frac{m - b}{2a}\right) + c = \frac{(m - b)^2}{4a} + \frac{b(m - b)}{2a} + c

=(mb)2+2b(mb)4a+c=(mb)(mb+2b)4a+c=(mb)(m+b)4a+c=m2b24a+c= \frac{(m - b)^2 + 2b(m - b)}{4a} + c = \frac{(m - b)(m - b + 2b)}{4a} + c = \frac{(m - b)(m + b)}{4a} + c = \frac{m^2 - b^2}{4a} + c

The tangent at PP has equation yyP=m(xxP)y - y_P = m(x - x_P), i.e. y=mx+(yPmxP)y = mx + (y_P - mx_P):

ymx=m2b24a+cmmb2a=c+m2b22m(mb)4ay - mx = \frac{m^2 - b^2}{4a} + c - m \cdot \frac{m - b}{2a} = c + \frac{m^2 - b^2 - 2m(m - b)}{4a}

=c+m2b22m2+2mb4a=c+m2+2mbb24a=c(mb)24a= c + \frac{m^2 - b^2 - 2m^2 + 2mb}{4a} = c + \frac{-m^2 + 2mb - b^2}{4a} = c - \frac{(m - b)^2}{4a} \qquad \blacksquare


Part 2: Condition for a common tangent with gradient mm

The tangent to the first curve with gradient mm is ymx=c1(mb1)24a1y - mx = c_1 - \frac{(m - b_1)^2}{4a_1}.

The tangent to the second curve with gradient mm is ymx=c2(mb2)24a2y - mx = c_2 - \frac{(m - b_2)^2}{4a_2}.

These are the same line if and only if their intercepts are equal:

c1(mb1)24a1=c2(mb2)24a2c_1 - \frac{(m - b_1)^2}{4a_1} = c_2 - \frac{(m - b_2)^2}{4a_2}

Multiplying through by 4a1a24a_1a_2:

4a1a2c1a2(mb1)2=4a1a2c2a1(mb2)24a_1a_2c_1 - a_2(m - b_1)^2 = 4a_1a_2c_2 - a_1(m - b_2)^2

4a1a2c1a2(m22b1m+b12)=4a1a2c2a1(m22b2m+b22)4a_1a_2c_1 - a_2(m^2 - 2b_1m + b_1^2) = 4a_1a_2c_2 - a_1(m^2 - 2b_2m + b_2^2)

Rearranging:

a1m2a2m2+2a2b1m2a1b2m+4a1a2c14a1a2c2a2b12+a1b22=0a_1m^2 - a_2m^2 + 2a_2b_1m - 2a_1b_2m + 4a_1a_2c_1 - 4a_1a_2c_2 - a_2b_1^2 + a_1b_2^2 = 0

(a1a2)m2+2(a2b1a1b2)m+4a1a2(c1c2)+a1b22a2b12=0(a_1 - a_2)m^2 + 2(a_2b_1 - a_1b_2)m + 4a_1a_2(c_1 - c_2) + a_1b_2^2 - a_2b_1^2 = 0

Multiplying by 1-1:

(a2a1)m2+2(a1b2a2b1)m+4a1a2(c2c1)+a2b12a1b22=0(a_2 - a_1)m^2 + 2(a_1b_2 - a_2b_1)m + 4a_1a_2(c_2 - c_1) + a_2b_1^2 - a_1b_2^2 = 0 \qquad \blacksquare


Part 3: Case a1a2a_1 \neq a_2 — exactly one common tangent iff the curves touch

When a1a2a_1 \neq a_2, the condition from Part 2 is a quadratic in mm. There is exactly one common tangent if and only if this quadratic has exactly one root, i.e. its discriminant is zero.

The discriminant is:

Δ=4(a1b2a2b1)24(a2a1)[4a1a2(c2c1)+a2b12a1b22]\Delta = 4(a_1b_2 - a_2b_1)^2 - 4(a_2 - a_1)\left[4a_1a_2(c_2 - c_1) + a_2b_1^2 - a_1b_2^2\right]

Setting Δ=0\Delta = 0 and dividing by 4:

(a1b2a2b1)2=(a2a1)[4a1a2(c2c1)+a2b12a1b22](a_1b_2 - a_2b_1)^2 = (a_2 - a_1)\left[4a_1a_2(c_2 - c_1) + a_2b_1^2 - a_1b_2^2\right]

Expanding the left side: a12b222a1a2b1b2+a22b12a_1^2b_2^2 - 2a_1a_2b_1b_2 + a_2^2b_1^2.

Expanding the right side: 4a1a2(a2a1)(c2c1)+a22b12a1a2b22a1a2b12+a12b224a_1a_2(a_2 - a_1)(c_2 - c_1) + a_2^2b_1^2 - a_1a_2b_2^2 - a_1a_2b_1^2 + a_1^2b_2^2.

Subtracting the common terms a12b22+a22b12a_1^2b_2^2 + a_2^2b_1^2 from both sides:

2a1a2b1b2=4a1a2(a2a1)(c2c1)a1a2(b12+b22)-2a_1a_2b_1b_2 = 4a_1a_2(a_2 - a_1)(c_2 - c_1) - a_1a_2(b_1^2 + b_2^2)

Dividing by a1a2a_1a_2 (both non-zero):

2b1b2=4(a2a1)(c2c1)b12b22-2b_1b_2 = 4(a_2 - a_1)(c_2 - c_1) - b_1^2 - b_2^2

b12+b222b1b2=4(a2a1)(c2c1)b_1^2 + b_2^2 - 2b_1b_2 = 4(a_2 - a_1)(c_2 - c_1)

(b2b1)2=4(a2a1)(c2c1)(*)(b_2 - b_1)^2 = 4(a_2 - a_1)(c_2 - c_1) \qquad \text{(*)}

Now the curves touch when the equation a1x2+b1x+c1=a2x2+b2x+c2a_1x^2 + b_1x + c_1 = a_2x^2 + b_2x + c_2 has exactly one solution, i.e.

(a2a1)x2+(b2b1)x+(c2c1)=0(a_2 - a_1)x^2 + (b_2 - b_1)x + (c_2 - c_1) = 0

has discriminant zero:

(b2b1)2=4(a2a1)(c2c1)(b_2 - b_1)^2 = 4(a_2 - a_1)(c_2 - c_1)

This is exactly condition ()(*). Therefore, when a1a2a_1 \neq a_2, the curves have exactly one common tangent if and only if they touch each other. \qquad \blacksquare


Part 4: Case a1=a2a_1 = a_2 — condition for exactly one common tangent

When a1=a2=aa_1 = a_2 = a, the quadratic from Part 2 becomes linear (the m2m^2 term vanishes):

2(ab2ab1)m+4a2(c2c1)+ab12ab22=02(ab_2 - ab_1)m + 4a^2(c_2 - c_1) + ab_1^2 - ab_2^2 = 0

2a(b2b1)m+4a2(c2c1)+a(b12b22)=02a(b_2 - b_1)m + 4a^2(c_2 - c_1) + a(b_1^2 - b_2^2) = 0

Dividing by aa (non-zero):

2(b2b1)m+4a(c2c1)+(b12b22)=02(b_2 - b_1)m + 4a(c_2 - c_1) + (b_1^2 - b_2^2) = 0

This is a linear equation in mm. It has exactly one solution if and only if the coefficient of mm is non-zero, i.e.

b2b1b_2 \neq b_1 \qquad \blacksquare

(If b2=b1b_2 = b_1, the equation becomes 4a(c2c1)=04a(c_2 - c_1) = 0. If also c2=c1c_2 = c_1, the curves are identical and every tangent is common; if c2c1c_2 \neq c_1, there are no common tangents.)


Topic: 三次方程与双曲函数 | Cubic Equations and Hyperbolic Functions  |  Difficulty: Hard  |  Marks: 20

6 In this question, you may use without proof the results

4cosh3y3coshy=cosh(3y)andarcosh y=ln(y+y21).4 \cosh^3 y - 3 \cosh y = \cosh(3y) \quad \text{and} \quad \text{arcosh } y = \ln (y + \sqrt{y^2 - 1}) .

[ Note: arcosh y\text{arcosh } y is another notation for cosh1y\cosh^{-1} y ]

Show that the equation x33a2x=2a3coshTx^3 - 3a^2x = 2a^3 \cosh T is satisfied by 2acosh(13T)2a \cosh (\frac{1}{3}T) and hence that, if c2b3c^2 \geqslant b^3, one of the roots of the equation x33bx=2cx^3 - 3bx = 2c is u+buu + \frac{b}{u}, where u=(c+c2b3)13u = \left( c + \sqrt{c^2 - b^3} \right)^{\frac{1}{3}}.

Show that the other two roots of the equation x33bx=2cx^3 - 3bx = 2c are the roots of the quadratic equation x2+(u+bu)x+u2+b2u2b=0x^2 + (u + \frac{b}{u})x + u^2 + \frac{b^2}{u^2} - b = 0, and find these roots in terms of uu, bb and ω\omega, where ω=12(1+i3)\omega = \frac{1}{2}(-1 + i\sqrt{3}).

Solve completely the equation x36x=6x^3 - 6x = 6.

Hint

6 Direct substitution of x = 2a cosh(T/3) into the left hand side of the equation gives

(2a cosh(T/3))^3 - 6a^3 cosh(T/3) = 2a^3(4(cosh(T/3))^3 - 3 cosh(T/3)) = 2a^3 cosh T

(by the first result given at the start of the question).

Let a^2 = b, which is possible since b > 0, and cosh T = c/a^3, which requires c/a^3 >= 1; but this holds if you choose a to have the same sign as c, since then c/a^3 > 0 and c^2 > b^3 = a^6.

Then, by the second result given at the start of the question,

T = ln(c/a^3 + sqrt(c^2/a^6 - 1)) = ln((c + sqrt(c^2 - b^3))/a^3) = 3 ln(u/a),

so one of the roots of the equation x^3 - 3bx = 2c is

2a cosh(ln(u/a)) = 2a * (u/a + a/u)/2 = u + b/u.

Note that, since u + b/u is a root of the equation x^3 - 3bx = 2c,

2c = (u + b/u)^3 - 3b(u + b/u) = (u + b/u)(u^2 + b^2/u^2 - b)

and that

u^2 + b^2/u^2 - b - (u + b/u)^2 = -3b,

so

x^3 - 3bx - 2c = (x - (u + b/u))(x^2 + (u + b/u)x + u^2 + b^2/u^2 - b),

so the other two roots of x^3 - 3bx = 2c are the roots of x^2 + (u + b/u)x + u^2 + b^2/u^2 - b = 0, which are

1/2 * (-(u + b/u) +/- sqrt((u + b/u)^2 - 4(u^2 + b^2/u^2 - b))) = 1/2 * (-(u + b/u) +/- sqrt(-3(u^2 + b^2/u^2 - 2b))) = 1/2 * (-(u + b/u) +/- sqrt(3)j(u - b/u)),

that is omegau + omega^2(b/u) and omega^2u + omega(b/u).

In x^3 - 6x = 6, b = 2, c = 3, so a = sqrt(2) and so u = cbrt(3 + 1) = 2^(2/3) and b/u = 2^(1/3), so the solutions are 2^(1/3) + 2^(2/3), omega2^(1/3) + omega^22^(2/3) and omega^22^(1/3) + omega2^(2/3).

Model Solution

Part 1: Show that x=2acosh(T3)x = 2a\cosh\left(\frac{T}{3}\right) satisfies x33a2x=2a3coshTx^3 - 3a^2x = 2a^3\cosh T

Substituting x=2acosh(T3)x = 2a\cosh\left(\frac{T}{3}\right):

x33a2x=8a3cosh3(T3)6a3cosh(T3)x^3 - 3a^2x = 8a^3\cosh^3\left(\frac{T}{3}\right) - 6a^3\cosh\left(\frac{T}{3}\right)

=2a3(4cosh3(T3)3cosh(T3))= 2a^3\left(4\cosh^3\left(\frac{T}{3}\right) - 3\cosh\left(\frac{T}{3}\right)\right)

Using the identity 4cosh3y3coshy=cosh(3y)4\cosh^3 y - 3\cosh y = \cosh(3y) with y=T3y = \frac{T}{3}:

=2a3coshT= 2a^3\cosh T \qquad \blacksquare


Part 2: One root of x33bx=2cx^3 - 3bx = 2c is u+buu + \frac{b}{u}

We need to reduce x33bx=2cx^3 - 3bx = 2c to the form x33a2x=2a3coshTx^3 - 3a^2x = 2a^3\cosh T.

Setting a2=ba^2 = b (so a=ba = \sqrt{b}, which is valid since c2b3>0c^2 \geqslant b^3 > 0 implies b>0b > 0) and a3coshT=ca^3\cosh T = c, i.e. coshT=ca3=cb3/2\cosh T = \frac{c}{a^3} = \frac{c}{b^{3/2}}.

This requires cb3/21\frac{c}{b^{3/2}} \geqslant 1, i.e. cb3/2c \geqslant b^{3/2}, which holds since c2b3c^2 \geqslant b^3 and we can choose aa to have the same sign as cc (so c/a3>0c/a^3 > 0).

From Part 1, one root is x=2acosh(T3)x = 2a\cosh\left(\frac{T}{3}\right).

Now using arcosh y=ln(y+y21)\text{arcosh } y = \ln(y + \sqrt{y^2 - 1}):

T=arcosh(cb3/2)=ln(cb3/2+c2b31)=ln(c+c2b3b3/2)T = \text{arcosh}\left(\frac{c}{b^{3/2}}\right) = \ln\left(\frac{c}{b^{3/2}} + \sqrt{\frac{c^2}{b^3} - 1}\right) = \ln\left(\frac{c + \sqrt{c^2 - b^3}}{b^{3/2}}\right)

Let u=(c+c2b3)1/3u = \left(c + \sqrt{c^2 - b^3}\right)^{1/3}. Then:

T3=13ln(c+c2b3b3/2)=ln(ub1/2)=ln(ua)\frac{T}{3} = \frac{1}{3}\ln\left(\frac{c + \sqrt{c^2 - b^3}}{b^{3/2}}\right) = \ln\left(\frac{u}{b^{1/2}}\right) = \ln\left(\frac{u}{a}\right)

Therefore:

cosh(T3)=cosh(lnua)=12(ua+au)\cosh\left(\frac{T}{3}\right) = \cosh\left(\ln\frac{u}{a}\right) = \frac{1}{2}\left(\frac{u}{a} + \frac{a}{u}\right)

and the root is:

x=2a12(ua+au)=u+a2u=u+bux = 2a \cdot \frac{1}{2}\left(\frac{u}{a} + \frac{a}{u}\right) = u + \frac{a^2}{u} = u + \frac{b}{u} \qquad \blacksquare


Part 3: The other two roots are roots of x2+(u+bu)x+u2+b2u2b=0x^2 + \left(u + \frac{b}{u}\right)x + u^2 + \frac{b^2}{u^2} - b = 0

Since u+buu + \frac{b}{u} is a root of x33bx2c=0x^3 - 3bx - 2c = 0, we can factor:

x33bx2c=(xubu)(x2+px+q)x^3 - 3bx - 2c = \left(x - u - \frac{b}{u}\right)(x^2 + px + q)

for some p,qp, q. Expanding and comparing with x33bx2cx^3 - 3bx - 2c:

The coefficient of x2x^2: pubu=0p - u - \frac{b}{u} = 0, so p=u+bup = u + \frac{b}{u}.

The coefficient of xx: qp(u+bu)=3bq - p\left(u + \frac{b}{u}\right) = -3b, so q=p23b=(u+bu)23bq = p^2 - 3b = \left(u + \frac{b}{u}\right)^2 - 3b.

q=u2+2b+b2u23b=u2+b2u2bq = u^2 + 2b + \frac{b^2}{u^2} - 3b = u^2 + \frac{b^2}{u^2} - b

The constant term: q(u+bu)=2c-q\left(u + \frac{b}{u}\right) = -2c. We verify: since u3=c+c2b3u^3 = c + \sqrt{c^2 - b^3} and b3u3=cc2b3\frac{b^3}{u^3} = c - \sqrt{c^2 - b^3} (as b3u3=b3c+c2b3=b3(cc2b3)b3=cc2b3\frac{b^3}{u^3} = \frac{b^3}{c + \sqrt{c^2 - b^3}} = \frac{b^3(c - \sqrt{c^2 - b^3})}{b^3} = c - \sqrt{c^2 - b^3}), we have u3+b3u3=2cu^3 + \frac{b^3}{u^3} = 2c.

Also u3+b3u3=(u+bu)(u2b+b2u2)=(u+bu)qu^3 + \frac{b^3}{u^3} = \left(u + \frac{b}{u}\right)\left(u^2 - b + \frac{b^2}{u^2}\right) = \left(u + \frac{b}{u}\right)q, confirming q(u+bu)=2cq\left(u + \frac{b}{u}\right) = 2c. \checkmark

Therefore the other two roots satisfy:

x2+(u+bu)x+u2+b2u2b=0x^2 + \left(u + \frac{b}{u}\right)x + u^2 + \frac{b^2}{u^2} - b = 0 \qquad \blacksquare


Part 4: Find these roots in terms of u,b,ωu, b, \omega

Using the quadratic formula:

x=(u+bu)±(u+bu)24(u2+b2u2b)2x = \frac{-\left(u + \frac{b}{u}\right) \pm \sqrt{\left(u + \frac{b}{u}\right)^2 - 4\left(u^2 + \frac{b^2}{u^2} - b\right)}}{2}

The discriminant:

(u+bu)24(u2+b2u2b)=u2+2b+b2u24u24b2u2+4b\left(u + \frac{b}{u}\right)^2 - 4\left(u^2 + \frac{b^2}{u^2} - b\right) = u^2 + 2b + \frac{b^2}{u^2} - 4u^2 - \frac{4b^2}{u^2} + 4b

=3u2+6b3b2u2=3(u22b+b2u2)=3(ubu)2= -3u^2 + 6b - \frac{3b^2}{u^2} = -3\left(u^2 - 2b + \frac{b^2}{u^2}\right) = -3\left(u - \frac{b}{u}\right)^2

So:

x=(u+bu)±3(ubu)22=(u+bu)±i3(ubu)2x = \frac{-\left(u + \frac{b}{u}\right) \pm \sqrt{-3\left(u - \frac{b}{u}\right)^2}}{2} = \frac{-\left(u + \frac{b}{u}\right) \pm i\sqrt{3}\left(u - \frac{b}{u}\right)}{2}

=12(u+bu)±i32(ubu)= -\frac{1}{2}\left(u + \frac{b}{u}\right) \pm \frac{i\sqrt{3}}{2}\left(u - \frac{b}{u}\right)

With ω=12(1+i3)\omega = \frac{1}{2}(-1 + i\sqrt{3}) and ω2=12(1i3)\omega^2 = \frac{1}{2}(-1 - i\sqrt{3}), we can write:

Taking the ++ sign:

x=12ub2u+i32ui3b2u=u(12+i32)+bu(12i32)=ωu+ω2bux = -\frac{1}{2}u - \frac{b}{2u} + \frac{i\sqrt{3}}{2}u - \frac{i\sqrt{3}b}{2u} = u\left(-\frac{1}{2} + \frac{i\sqrt{3}}{2}\right) + \frac{b}{u}\left(-\frac{1}{2} - \frac{i\sqrt{3}}{2}\right) = \omega u + \omega^2 \cdot \frac{b}{u}

Wait, let me recheck. ω=12(1+i3)\omega = \frac{1}{2}(-1 + i\sqrt{3}), so 12+i32=ω-\frac{1}{2} + \frac{i\sqrt{3}}{2} = \omega and 12i32=ω2-\frac{1}{2} - \frac{i\sqrt{3}}{2} = \omega^2. So:

x+=ωu+ω2bux_+ = \omega u + \omega^2 \cdot \frac{b}{u}

Taking the - sign:

x=ω2u+ωbux_- = \omega^2 u + \omega \cdot \frac{b}{u}

The other two roots are:

ωu+ω2buandω2u+ωbu\omega u + \frac{\omega^2 b}{u} \qquad \text{and} \qquad \omega^2 u + \frac{\omega b}{u} \qquad \blacksquare


Part 5: Solve x36x=6x^3 - 6x = 6 completely

Here b=2b = 2 and c=3c = 3. Check: c2=9b3=8c^2 = 9 \geqslant b^3 = 8. \checkmark

u=(3+98)1/3=(3+1)1/3=41/3=22/3u = \left(3 + \sqrt{9 - 8}\right)^{1/3} = (3 + 1)^{1/3} = 4^{1/3} = 2^{2/3}

bu=222/3=21/3\frac{b}{u} = \frac{2}{2^{2/3}} = 2^{1/3}

The three roots are:

x1=u+bu=22/3+21/3x_1 = u + \frac{b}{u} = 2^{2/3} + 2^{1/3}

x2=ω22/3+ω221/3x_2 = \omega \cdot 2^{2/3} + \omega^2 \cdot 2^{1/3}

x3=ω222/3+ω21/3x_3 = \omega^2 \cdot 2^{2/3} + \omega \cdot 2^{1/3}

where ω=12(1+i3)\omega = \frac{1}{2}(-1 + i\sqrt{3}) and ω2=12(1i3)\omega^2 = \frac{1}{2}(-1 - i\sqrt{3}).

Numerically: x11.587+1.260=2.847x_1 \approx 1.587 + 1.260 = 2.847.

The complex roots can also be written as:

x2,3=12(22/3+21/3)±i32(22/321/3)x_{2,3} = -\frac{1}{2}(2^{2/3} + 2^{1/3}) \pm \frac{i\sqrt{3}}{2}(2^{2/3} - 2^{1/3})

\qquad \blacksquare


Topic: 积分技巧 | Integration Techniques  |  Difficulty: Challenging  |  Marks: 20

7 Show that if 1uf(u)du=F(u)+c\int \frac{1}{u f(u)} \, du = F(u) + c, then mxf(xm)dx=F(xm)+c\int \frac{m}{x f(x^m)} \, dx = F(x^m) + c, where m0m \neq 0.

Find:

(i) 1xnxdx\int \frac{1}{x^n - x} \, dx;

(ii) 1xn+x2dx\int \frac{1}{\sqrt{x^n + x^2}} \, dx.

Hint

7 Substituting u = x^m gives

integral m dx/(x f(x^m)) = integral m x^(m-1) dx/(x^m f(x^m)) = integral du/(u f(u)) = F(u) + c = F(x^m) + c

(i) integral dx/(x^n - x) = integral dx/(x(x^(n-1) - 1)), so letting u = x^(n-1) and f(u) = u - 1, integral (n-1) dx/(x^n - x) = integral du/(u(u-1)) = integral (1/(u-1) - 1/u) du = ln|(u-1)/u| so integral dx/(x^n - x) = (1/(n-1)) ln|(x^(n-1) - 1)/x^(n-1)| + c.

(ii) integral dx/sqrt(x^n + x^2) = integral dx/(x sqrt(x^(n-2) + 1)) (for x > 0) so letting u = x^(n-2) and f(u) = sqrt(u+1) (and assuming n != 2) integral (n-2) dx/sqrt(x^n + x^2) = integral du/(u sqrt(u+1)). Substituting u = v^2 - 1 with v > 0, integral du/(u sqrt(u+1)) = integral 2v dv/((v^2 - 1)v) = integral (1/(v-1) - 1/(v+1)) dv = ln|(v-1)/(v+1)| so integral dx/sqrt(x^n + x^2) = (1/(n-2)) ln|(sqrt(x^(n-2) + 1) - 1)/(sqrt(x^(n-2) + 1) + 1)| + c.

Model Solution

Part 1: Show the general result

Given that 1uf(u)du=F(u)+c\int \frac{1}{uf(u)}\,du = F(u) + c, we substitute u=xmu = x^m, so du=mxm1dxdu = mx^{m-1}\,dx.

mxf(xm)dx=mxm1xmf(xm)dx=1uf(u)du=F(u)+c=F(xm)+c\int \frac{m}{xf(x^m)}\,dx = \int \frac{mx^{m-1}}{x^m f(x^m)}\,dx = \int \frac{1}{uf(u)}\,du = F(u) + c = F(x^m) + c \qquad \blacksquare

(Here we used the substitution u=xmu = x^m to convert the integral, noting that mxm1xm=mx\frac{mx^{m-1}}{x^m} = \frac{m}{x}.)


Part (i): Find 1xnxdx\int \frac{1}{x^n - x}\,dx

1xnxdx=1x(xn11)dx\int \frac{1}{x^n - x}\,dx = \int \frac{1}{x(x^{n-1} - 1)}\,dx

We want to use the general result with m=n1m = n - 1 and f(u)=u1f(u) = u - 1. Then f(xm)=f(xn1)=xn11f(x^m) = f(x^{n-1}) = x^{n-1} - 1, and xf(xm)=x(xn11)=xnxxf(x^m) = x(x^{n-1} - 1) = x^n - x.

We need 1uf(u)du=1u(u1)du\int \frac{1}{uf(u)}\,du = \int \frac{1}{u(u - 1)}\,du.

Using partial fractions:

1u(u1)=1u11u\frac{1}{u(u - 1)} = \frac{1}{u - 1} - \frac{1}{u}

1u(u1)du=lnu1lnu+c=lnu1u+c\int \frac{1}{u(u - 1)}\,du = \ln|u - 1| - \ln|u| + c = \ln\left|\frac{u - 1}{u}\right| + c

So F(u)=lnu1uF(u) = \ln\left|\frac{u - 1}{u}\right|. By the general result with m=n1m = n - 1:

n1x(xn11)dx=F(xn1)+c\int \frac{n - 1}{x(x^{n-1} - 1)}\,dx = F(x^{n-1}) + c

Therefore:

1xnxdx=1n1lnxn11xn1+c\int \frac{1}{x^n - x}\,dx = \frac{1}{n - 1}\ln\left|\frac{x^{n-1} - 1}{x^{n-1}}\right| + c \qquad \blacksquare

(valid for n1n \neq 1.)


Part (ii): Find 1xn+x2dx\int \frac{1}{\sqrt{x^n + x^2}}\,dx

1xn+x2dx=1xxn2+1dx(for x>0)\int \frac{1}{\sqrt{x^n + x^2}}\,dx = \int \frac{1}{x\sqrt{x^{n-2} + 1}}\,dx \qquad \text{(for } x > 0\text{)}

We use the general result with m=n2m = n - 2 and f(u)=u+1f(u) = \sqrt{u + 1}. Then f(xm)=xn2+1f(x^m) = \sqrt{x^{n-2} + 1}, and xf(xm)=xxn2+1=xn+x2xf(x^m) = x\sqrt{x^{n-2} + 1} = \sqrt{x^n + x^2}.

We need 1uf(u)du=1uu+1du\int \frac{1}{uf(u)}\,du = \int \frac{1}{u\sqrt{u + 1}}\,du.

Substitute u=v21u = v^2 - 1 (with v>0v > 0), so du=2vdvdu = 2v\,dv and u+1=v\sqrt{u + 1} = v:

1uu+1du=2v(v21)vdv=2v21dv\int \frac{1}{u\sqrt{u + 1}}\,du = \int \frac{2v}{(v^2 - 1)v}\,dv = \int \frac{2}{v^2 - 1}\,dv

Using partial fractions:

2v21=2(v1)(v+1)=1v11v+1\frac{2}{v^2 - 1} = \frac{2}{(v - 1)(v + 1)} = \frac{1}{v - 1} - \frac{1}{v + 1}

2v21dv=lnv1lnv+1+c=lnv1v+1+c\int \frac{2}{v^2 - 1}\,dv = \ln|v - 1| - \ln|v + 1| + c = \ln\left|\frac{v - 1}{v + 1}\right| + c

So F(u)=lnu+11u+1+1F(u) = \ln\left|\frac{\sqrt{u + 1} - 1}{\sqrt{u + 1} + 1}\right|. By the general result with m=n2m = n - 2:

n2xxn2+1dx=F(xn2)+c\int \frac{n - 2}{x\sqrt{x^{n-2} + 1}}\,dx = F(x^{n-2}) + c

Therefore:

1xn+x2dx=1n2lnxn2+11xn2+1+1+c\int \frac{1}{\sqrt{x^n + x^2}}\,dx = \frac{1}{n - 2}\ln\left|\frac{\sqrt{x^{n-2} + 1} - 1}{\sqrt{x^{n-2} + 1} + 1}\right| + c \qquad \blacksquare

(valid for n2n \neq 2.)


Topic: 复数与几何 | Complex Numbers and Geometry  |  Difficulty: Challenging  |  Marks: 20

8 In this question, aa and cc are distinct non-zero complex numbers. The complex conjugate of any complex number zz is denoted by zz^*.

Show that

ac2=aa+ccacca|a - c|^2 = aa^* + cc^* - ac^* - ca^*

and hence prove that the triangle OACOAC in the Argand diagram, whose vertices are represented by 00, aa and cc respectively, is right angled at AA if and only if 2aa=ac+ca2aa^* = ac^* + ca^*.

Points PP and PP' in the Argand diagram are represented by the complex numbers abab and ab\frac{a}{b^*}, where bb is a non-zero complex number. A circle in the Argand diagram has centre CC and passes through the point AA, and is such that OAOA is tangent to the circle. Show that the point PP lies on the circle if and only if point PP' lies on the circle.

Conversely, show that if the points represented by the complex numbers abab and ab\frac{a}{b^*}, for some non-zero complex number bb with bb1bb^* \neq 1, both lie on a circle centre CC in the Argand diagram which passes through AA, then OAOA is a tangent to the circle.

Hint

8 Direct use of the important result |z|^2 = zz* gives

|a - c|^2 = (a - c)(a* - c*) = aa* + cc* - ac* - ca*.

OAC is a right angle if and only if |AC|^2 + |OA|^2 = |OC|^2; that is, |a - c|^2 + |a|^2 = |c|^2 or, using the result above, 2aa* - ac* - ca* = 0.

The circle has centre C and radius AC, so complex numbers representing points on the circle satisfy |z - c|^2 = |a - c|^2 or zz* - zc* - cz* = aa* - ac* - ca*.

Because OA is a tangent to the circle, angle OAC is a right angle and so 2aa* - ac* - ca* = 0 as above; thus the condition for points to lie on the circle becomes zz* - zc* - cz* + aa* = 0.

P lies on this circle if and only if

abab - abc* - cab + aa* = 0

and P’ lies on the circle if and only if

aa*/(bb*) - a/b* * c* - c * a*/b + aa* = 0

but multiplying this by bb* (which is not equal to zero) gives the same condition.

Conversely, if the points lie on the circle represented by |z - c|^2 = |a - c|^2,

abab - abc* - cab + aa* = 2aa* - ac* - ca*

and

aa*/(bb*) - a/b* * c* - c * a*/b = 2aa* - ac* - ca*,

so that

abab - abc* - cab + aa* = bb*(2aa* - ac* - ca*)

and so, provided bb* != 1, it must be the case that 2aa* - ac* - ca* = 0, and this shows that OAC is a right angle and hence that OA is a tangent to the circle.

Model Solution

Part 1: Show that ac2=aa+ccacca|a - c|^2 = aa^* + cc^* - ac^* - ca^*

Using the result z2=zz|z|^2 = zz^*:

ac2=(ac)(ac)=(ac)(ac)|a - c|^2 = (a - c)(a - c)^* = (a - c)(a^* - c^*)

=aaacca+cc=aa+ccacca= aa^* - ac^* - ca^* + cc^* = aa^* + cc^* - ac^* - ca^* \qquad \blacksquare


Part 2: Triangle OACOAC is right-angled at AA iff 2aa=ac+ca2aa^* = ac^* + ca^*

Triangle OACOAC is right-angled at AA if and only if OA2+AC2=OC2|OA|^2 + |AC|^2 = |OC|^2 (Pythagoras).

OA2=a2=aa|OA|^2 = |a|^2 = aa^* AC2=ac2=aa+ccacca|AC|^2 = |a - c|^2 = aa^* + cc^* - ac^* - ca^* OC2=c2=cc|OC|^2 = |c|^2 = cc^*

The condition OA2+AC2=OC2|OA|^2 + |AC|^2 = |OC|^2 becomes:

aa+aa+ccacca=ccaa^* + aa^* + cc^* - ac^* - ca^* = cc^*

2aaacca=02aa^* - ac^* - ca^* = 0

2aa=ac+ca2aa^* = ac^* + ca^* \qquad \blacksquare


Part 3: PP lies on the circle iff PP' lies on the circle

The circle has centre CC and passes through AA, so a point zz lies on the circle if and only if:

zc2=ac2|z - c|^2 = |a - c|^2

Expanding using z2=zz|z|^2 = zz^*:

zzzccz+cc=aaacca+cczz^* - zc^* - cz^* + cc^* = aa^* - ac^* - ca^* + cc^*

zzzccz=aaacca(*)zz^* - zc^* - cz^* = aa^* - ac^* - ca^* \qquad \text{(*)}

Since OAOA is tangent to the circle, angle OACOAC is a right angle, so by Part 2: 2aa=ac+ca2aa^* = ac^* + ca^*, i.e. ac+ca=2aaac^* + ca^* = 2aa^*.

Substituting into ()(*):

zzzccz=aa2aa=aazz^* - zc^* - cz^* = aa^* - 2aa* = -aa*

zzzccz+aa=0(**)zz^* - zc^* - cz^* + aa^* = 0 \qquad \text{(**)}

PP lies on the circle: z=abz = ab, so z=abz^* = a^*b^*.

(ab)(ab)(ab)cc(ab)+aa=0(ab)(a^*b^*) - (ab)c^* - c(a^*b^*) + aa^* = 0

aabbabcabc+aa=0(P)aa^*bb^* - abc^* - a^*b^*c + aa^* = 0 \qquad \text{(P)}

PP' lies on the circle: z=abz = \frac{a}{b^*}, so z=abz^* = \frac{a^*}{b}.

abababccab+aa=0\frac{a}{b^*} \cdot \frac{a^*}{b} - \frac{a}{b^*}c^* - c\frac{a^*}{b} + aa^* = 0

aabbacbcab+aa=0\frac{aa^*}{bb^*} - \frac{ac^*}{b^*} - \frac{ca^*}{b} + aa^* = 0

Multiplying through by bbbb^* (which is non-zero):

aaabcabc+aabb=0(P’)aa^* - abc^* - a^*b^*c + aa^*bb^* = 0 \qquad \text{(P')}

Conditions (P) and (P’) are identical. Therefore PP lies on the circle if and only if PP' lies on the circle. \qquad \blacksquare


Part 4: Converse — if both abab and ab\frac{a}{b^*} lie on a circle through AA with centre CC, then OAOA is tangent

Suppose both z=abz = ab and z=abz = \frac{a}{b^*} lie on the circle zc2=ac2|z - c|^2 = |a - c|^2.

From the general circle condition ()(*):

For z=abz = ab: aabbabcabc=aaacca(I)\quad aa^*bb^* - abc^* - a^*b^*c = aa^* - ac^* - ca^* \qquad \text{(I)}

For z=abz = \frac{a}{b^*}: aabbacbcab=aaacca(II)\quad \frac{aa^*}{bb^*} - \frac{ac^*}{b^*} - \frac{ca^*}{b} = aa^* - ac^* - ca^* \qquad \text{(II)}

Let K=aaaccaK = aa^* - ac^* - ca^* (the right-hand side of both equations).

From (II), multiplying by bbbb^*:

aaabcabc=bbK(II’)aa^* - abc^* - a^*b^*c = bb^* K \qquad \text{(II')}

From (I): the left-hand side equals KK, so:

K=aabbabcabcK = aa^*bb^* - abc^* - a^*b^*c

But from (II’): abc+abc=aabbKabc^* + a^*b^*c = aa^* - bb^*K, substituting into (I):

K=aabb(aabbK)=aabbaa+bbKK = aa^*bb^* - (aa^* - bb^*K) = aa^*bb^* - aa^* + bb^*K

KbbK=aa(bb1)K - bb^*K = aa^*(bb^* - 1)

K(1bb)=aa(bb1)K(1 - bb^*) = aa^*(bb^* - 1)

K=aaK = -aa^*

(since bb1bb^* \neq 1, we can divide by (1bb)=(bb1)(1 - bb^*) = -(bb^* - 1).)

So K=aaacca=aaK = aa^* - ac^* - ca^* = -aa^*, which gives:

2aa=ac+ca2aa^* = ac^* + ca^*

By Part 2, this means triangle OACOAC is right-angled at AA, so OAOA is perpendicular to ACAC. Since ACAC is a radius of the circle and OAOA passes through AA, this means OAOA is tangent to the circle. \qquad \blacksquare