跳转到内容

STEP3 2002 -- Pure Mathematics

此内容尚不支持你的语言。

STEP3 2002 — Section A (Pure Mathematics)

Section titled “STEP3 2002 — Section A (Pure Mathematics)”

Exam: STEP3  |  Year: 2002  |  Questions: Q1—Q8  |  Total marks per question: 20

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

QTopicDifficultyKey Techniques
1微积分 CalculusChallenging换元积分,分部积分,旋转体体积公式,广义积分
2级数与数列 Sequences and SeriesHard数学归纳法,反正切加法公式,裂项相消,级数求和
3函数与方程 Functions and EquationsChallenging根式函数分析,函数图像描绘,不等式分析,代数方程求解
4数论 Number TheoryHard不等式放缩,反证法,因式分解,整数分析
5代数 AlgebraChallenging配方法,待定系数法,多项式恒等式,因式分解
6微分方程 Differential EquationsChallenging可分离变量法,分情况讨论,解曲线作图
7三角学与坐标几何 Trigonometry and Coordinate GeometryChallenging三角恒等式,坐标法,代数恒等变换,不等式放缩
8复数 Complex NumbersHard复数辐角计算,模的运算,几何解释,Ptolemy定理

Topic: 微积分 Calculus  |  Difficulty: Challenging  |  Marks: 20

1 Find the area of the region between the curve y=lnxxy = \frac{\ln x}{x} and the xx-axis, for 1xa1 \leqslant x \leqslant a. What happens to this area as aa tends to infinity?

Find the volume of the solid obtained when the region between the curve y=lnxxy = \frac{\ln x}{x} and the xx-axis, for 1xa1 \leqslant x \leqslant a, is rotated through 2π2\pi radians about the xx-axis. What happens to this volume as aa tends to infinity?

Model Solution

Part (i): Area

The area is

A=1alnxxdx=[(lnx)22]1a=(lna)22.A = \int_1^a \frac{\ln x}{x}\, dx = \left[\frac{(\ln x)^2}{2}\right]_1^a = \frac{(\ln a)^2}{2}.

As aa \to \infty, lna\ln a \to \infty, so AA \to \infty. The area diverges.

Part (ii): Volume of revolution

The volume obtained by rotating about the xx-axis is

V=π1a(lnxx)2dx=π1a(lnx)2x2dx.V = \pi \int_1^a \left(\frac{\ln x}{x}\right)^2 dx = \pi \int_1^a \frac{(\ln x)^2}{x^2}\, dx.

We evaluate this by parts. Let u=(lnx)2u = (\ln x)^2, dv=x2dxdv = x^{-2}\, dx, so du=2lnxxdxdu = \frac{2\ln x}{x}\, dx, v=1xv = -\frac{1}{x}. Then

1a(lnx)2x2dx=[(lnx)2x]1a+1a2lnxx2dx=(lna)2a+21alnxx2dx.\int_1^a \frac{(\ln x)^2}{x^2}\, dx = \left[-\frac{(\ln x)^2}{x}\right]_1^a + \int_1^a \frac{2\ln x}{x^2}\, dx = -\frac{(\ln a)^2}{a} + 2\int_1^a \frac{\ln x}{x^2}\, dx.

For the remaining integral, let u=lnxu = \ln x, dv=x2dxdv = x^{-2}\, dx, so du=1xdxdu = \frac{1}{x}\, dx, v=1xv = -\frac{1}{x}:

1alnxx2dx=[lnxx]1a+1a1x2dx=lnaa+[1x]1a=lnaa+11a.\int_1^a \frac{\ln x}{x^2}\, dx = \left[-\frac{\ln x}{x}\right]_1^a + \int_1^a \frac{1}{x^2}\, dx = -\frac{\ln a}{a} + \left[-\frac{1}{x}\right]_1^a = -\frac{\ln a}{a} + 1 - \frac{1}{a}.

Combining:

1a(lnx)2x2dx=(lna)2a2lnaa+22a=2(lna)2+2lna+2a.\int_1^a \frac{(\ln x)^2}{x^2}\, dx = -\frac{(\ln a)^2}{a} - \frac{2\ln a}{a} + 2 - \frac{2}{a} = 2 - \frac{(\ln a)^2 + 2\ln a + 2}{a}.

Therefore

V=π[2(lna)2+2lna+2a].V = \pi\left[2 - \frac{(\ln a)^2 + 2\ln a + 2}{a}\right].

As aa \to \infty, the fraction (lna)2+2lna+2a0\frac{(\ln a)^2 + 2\ln a + 2}{a} \to 0 (since polynomial growth of lna\ln a is slower than linear growth of aa), so V2πV \to 2\pi.


Topic: 级数与数列 Sequences and Series  |  Difficulty: Hard  |  Marks: 20

2 Prove that arctana+arctanb=arctan(a+b1ab)\arctan a + \arctan b = \arctan \left( \frac{a + b}{1 - ab} \right) when 0<a<10 < a < 1 and 0<b<10 < b < 1.

Prove by induction that, for n1n \geqslant 1,

r=1narctan(1r2+r+1)=arctan(nn+2)\sum_{r=1}^{n} \arctan \left( \frac{1}{r^2 + r + 1} \right) = \arctan \left( \frac{n}{n + 2} \right)

and hence find

r=1arctan(1r2+r+1).\sum_{r=1}^{\infty} \arctan \left( \frac{1}{r^2 + r + 1} \right) .

Hence prove that

r=1arctan(1r2r+1)=π2.\sum_{r=1}^{\infty} \arctan \left( \frac{1}{r^2 - r + 1} \right) = \frac{\pi}{2} .

Model Solution

Part (i): The addition formula

Let α=arctana\alpha = \arctan a and β=arctanb\beta = \arctan b, so tanα=a\tan\alpha = a and tanβ=b\tan\beta = b. Since 0<a<10 < a < 1 and 0<b<10 < b < 1, we have 0<α<π/40 < \alpha < \pi/4 and 0<β<π/40 < \beta < \pi/4, so 0<α+β<π/20 < \alpha + \beta < \pi/2.

Using the tangent addition formula:

tan(α+β)=tanα+tanβ1tanαtanβ=a+b1ab.\tan(\alpha + \beta) = \frac{\tan\alpha + \tan\beta}{1 - \tan\alpha\tan\beta} = \frac{a + b}{1 - ab}.

Since 0<a,b<10 < a, b < 1, we have ab<1ab < 1, so 1ab>01 - ab > 0 and a+b1ab>0\frac{a+b}{1-ab} > 0. Also α+β(0,π/2)\alpha + \beta \in (0, \pi/2), so

α+β=arctan(a+b1ab).\alpha + \beta = \arctan\left(\frac{a + b}{1 - ab}\right). \qquad \blacksquare

Part (ii): Induction proof

Base case (n=1n = 1): arctan ⁣(11+1+1)=arctan ⁣(13)=arctan ⁣(11+2)\arctan\!\left(\frac{1}{1+1+1}\right) = \arctan\!\left(\frac{1}{3}\right) = \arctan\!\left(\frac{1}{1+2}\right). \checkmark

Inductive step: Assume r=1narctan ⁣(1r2+r+1)=arctan ⁣(nn+2)\sum_{r=1}^{n} \arctan\!\left(\frac{1}{r^2+r+1}\right) = \arctan\!\left(\frac{n}{n+2}\right). Then

r=1n+1arctan ⁣(1r2+r+1)=arctan ⁣(nn+2)+arctan ⁣(1(n+1)2+(n+1)+1).\sum_{r=1}^{n+1} \arctan\!\left(\frac{1}{r^2+r+1}\right) = \arctan\!\left(\frac{n}{n+2}\right) + \arctan\!\left(\frac{1}{(n+1)^2+(n+1)+1}\right).

We have (n+1)2+(n+1)+1=n2+3n+3(n+1)^2 + (n+1) + 1 = n^2 + 3n + 3, so we need to compute

arctan ⁣(nn+2)+arctan ⁣(1n2+3n+3).\arctan\!\left(\frac{n}{n+2}\right) + \arctan\!\left(\frac{1}{n^2+3n+3}\right).

Since 0<nn+2<10 < \frac{n}{n+2} < 1 and 0<1n2+3n+3<10 < \frac{1}{n^2+3n+3} < 1, we apply the addition formula:

nn+2+1n2+3n+31n(n+2)(n2+3n+3)=n(n2+3n+3)+(n+2)(n+2)(n2+3n+3)n.\frac{\frac{n}{n+2} + \frac{1}{n^2+3n+3}}{1 - \frac{n}{(n+2)(n^2+3n+3)}} = \frac{n(n^2+3n+3) + (n+2)}{(n+2)(n^2+3n+3) - n}.

Numerator: n3+3n2+3n+n+2=n3+3n2+4n+2n^3 + 3n^2 + 3n + n + 2 = n^3 + 3n^2 + 4n + 2.

Denominator: n3+3n2+3n+2n2+6n+6n=n3+5n2+8n+6n^3 + 3n^2 + 3n + 2n^2 + 6n + 6 - n = n^3 + 5n^2 + 8n + 6.

We can verify that n3+3n2+4n+2n3+5n2+8n+6=n+1n+3\frac{n^3 + 3n^2 + 4n + 2}{n^3 + 5n^2 + 8n + 6} = \frac{n+1}{n+3} by cross-multiplying:

(n+1)(n3+5n2+8n+6)=n4+6n3+13n2+14n+6(n+1)(n^3 + 5n^2 + 8n + 6) = n^4 + 6n^3 + 13n^2 + 14n + 6.

(n+3)(n3+3n2+4n+2)=n4+6n3+13n2+14n+6(n+3)(n^3 + 3n^2 + 4n + 2) = n^4 + 6n^3 + 13n^2 + 14n + 6. \checkmark

Therefore r=1n+1arctan ⁣(1r2+r+1)=arctan ⁣(n+1n+3)\sum_{r=1}^{n+1} \arctan\!\left(\frac{1}{r^2+r+1}\right) = \arctan\!\left(\frac{n+1}{n+3}\right), completing the induction. \qquad \blacksquare

Part (iii): The infinite sum

r=1arctan ⁣(1r2+r+1)=limnarctan ⁣(nn+2)=arctan(1)=π4.\sum_{r=1}^{\infty} \arctan\!\left(\frac{1}{r^2+r+1}\right) = \lim_{n\to\infty} \arctan\!\left(\frac{n}{n+2}\right) = \arctan(1) = \frac{\pi}{4}.

Part (iv): The second identity

Note that 1r2r+1=1(r1)2+(r1)+1\frac{1}{r^2 - r + 1} = \frac{1}{(r-1)^2 + (r-1) + 1}. Substituting s=r1s = r - 1:

r=1arctan ⁣(1r2r+1)=s=0arctan ⁣(1s2+s+1)=arctan ⁣(10+0+1)+s=1arctan ⁣(1s2+s+1)\sum_{r=1}^{\infty} \arctan\!\left(\frac{1}{r^2-r+1}\right) = \sum_{s=0}^{\infty} \arctan\!\left(\frac{1}{s^2+s+1}\right) = \arctan\!\left(\frac{1}{0+0+1}\right) + \sum_{s=1}^{\infty} \arctan\!\left(\frac{1}{s^2+s+1}\right)

=arctan(1)+π4=π4+π4=π2.= \arctan(1) + \frac{\pi}{4} = \frac{\pi}{4} + \frac{\pi}{4} = \frac{\pi}{2}. \qquad \blacksquare


Topic: 函数与方程 Functions and Equations  |  Difficulty: Challenging  |  Marks: 20

3 Let

f(x)=axxb,f(x) = a\sqrt{x} - \sqrt{x - b} ,

where xb>0x \geqslant b > 0 and a>1a > 1. Sketch the graph of f(x)f(x). Hence show that the equation f(x)=cf(x) = c, where c>0c > 0, has no solution when c2<b(a21)c^2 < b \left( a^2 - 1 \right). Find conditions on c2c^2 in terms of aa and bb for the equation to have exactly one or exactly two solutions.

Solve the equations (i) 3xx2=43\sqrt{x} - \sqrt{x - 2} = 4 and (ii) 3xx3=53\sqrt{x} - \sqrt{x - 3} = 5.

Model Solution

Part (i): Sketch and analysis

We have f(x)=axxbf(x) = a\sqrt{x} - \sqrt{x-b} for xb>0x \geqslant b > 0, a>1a > 1.

At x=bx = b: f(b)=abf(b) = a\sqrt{b}.

As xx \to \infty: f(x)=x(a1b/x)f(x) = \sqrt{x}(a - \sqrt{1 - b/x}) \to \infty since a>1a > 1.

Derivative:

f(x)=a2x12xb=axbx2xxb.f'(x) = \frac{a}{2\sqrt{x}} - \frac{1}{2\sqrt{x-b}} = \frac{a\sqrt{x-b} - \sqrt{x}}{2\sqrt{x}\sqrt{x-b}}.

Setting f(x)=0f'(x) = 0: axb=xa\sqrt{x-b} = \sqrt{x}, so a2(xb)=xa^2(x-b) = x, giving x=a2ba21x = \frac{a^2 b}{a^2 - 1}.

Since a>1a > 1, we have a2a21>1\frac{a^2}{a^2-1} > 1, so xc=a2ba21>bx_c = \frac{a^2 b}{a^2-1} > b. This critical point lies in the domain.

For bx<xcb \leqslant x < x_c: f(x)<0f'(x) < 0 (decreasing). For x>xcx > x_c: f(x)>0f'(x) > 0 (increasing).

The minimum value is

f(xc)=aa2ba21a2ba21b=a2ba21ba21=(a21)ba21=b(a21).f(x_c) = a\sqrt{\frac{a^2 b}{a^2-1}} - \sqrt{\frac{a^2 b}{a^2-1} - b} = \frac{a^2\sqrt{b}}{\sqrt{a^2-1}} - \frac{\sqrt{b}}{\sqrt{a^2-1}} = \frac{(a^2-1)\sqrt{b}}{\sqrt{a^2-1}} = \sqrt{b(a^2-1)}.

The graph starts at (b,ab)(b, a\sqrt{b}), decreases to a minimum at xcx_c with value b(a21)\sqrt{b(a^2-1)}, then increases to infinity.

Part (ii): Conditions on cc

The equation f(x)=cf(x) = c with c>0c > 0:

  • No solution when c<b(a21)c < \sqrt{b(a^2-1)}, i.e. c2<b(a21)c^2 < b(a^2-1).
  • Exactly one solution when c=b(a21)c = \sqrt{b(a^2-1)} (at the minimum), or when c=abc = a\sqrt{b} (at x=bx = b, the function value from the right).
  • Exactly two solutions when b(a21)<c<ab\sqrt{b(a^2-1)} < c < a\sqrt{b}, i.e. b(a21)<c2<a2bb(a^2-1) < c^2 < a^2 b.
  • Exactly one solution when c>abc > a\sqrt{b}, i.e. c2>a2bc^2 > a^2 b.

Part (iii): Solving (i) 3xx2=43\sqrt{x} - \sqrt{x-2} = 4

Here a=3a = 3, b=2b = 2, c=4c = 4. We have f(2)=324.24f(2) = 3\sqrt{2} \approx 4.24 and b(a21)=16=4\sqrt{b(a^2-1)} = \sqrt{16} = 4.

Since c=4=b(a21)c = 4 = \sqrt{b(a^2-1)}, the equation has exactly one solution (at the minimum).

Rearranging: 3x=4+x23\sqrt{x} = 4 + \sqrt{x-2}. Squaring:

9x=16+8x2+x2    8x14=8x2    4x7=4x2.9x = 16 + 8\sqrt{x-2} + x - 2 \implies 8x - 14 = 8\sqrt{x-2} \implies 4x - 7 = 4\sqrt{x-2}.

Squaring again: 16x256x+49=16(x2)=16x3216x^2 - 56x + 49 = 16(x-2) = 16x - 32, so 16x272x+81=016x^2 - 72x + 81 = 0.

Discriminant: 7224(16)(81)=51845184=072^2 - 4(16)(81) = 5184 - 5184 = 0, so x=7232=94x = \frac{72}{32} = \frac{9}{4}.

Check: 39/49/42=33214=9212=43\sqrt{9/4} - \sqrt{9/4 - 2} = 3 \cdot \frac{3}{2} - \sqrt{\frac{1}{4}} = \frac{9}{2} - \frac{1}{2} = 4. \checkmark

The unique solution is x=94x = \dfrac{9}{4}.

Part (iv): Solving (ii) 3xx3=53\sqrt{x} - \sqrt{x-3} = 5

Here a=3a = 3, b=3b = 3, c=5c = 5. We have f(3)=335.20f(3) = 3\sqrt{3} \approx 5.20 and b(a21)=24=264.90\sqrt{b(a^2-1)} = \sqrt{24} = 2\sqrt{6} \approx 4.90.

Since 26<5<332\sqrt{6} < 5 < 3\sqrt{3}, the equation has exactly two solutions.

Rearranging: 3x=5+x33\sqrt{x} = 5 + \sqrt{x-3}. Squaring:

9x=25+10x3+x3    8x22=10x3    4x11=5x3.9x = 25 + 10\sqrt{x-3} + x - 3 \implies 8x - 22 = 10\sqrt{x-3} \implies 4x - 11 = 5\sqrt{x-3}.

Note: we need 4x1104x - 11 \geqslant 0, i.e. x11/4x \geqslant 11/4.

Squaring again: 16x288x+121=25(x3)=25x7516x^2 - 88x + 121 = 25(x-3) = 25x - 75, so 16x2113x+196=016x^2 - 113x + 196 = 0.

Discriminant: 11324(16)(196)=1276912544=225=152113^2 - 4(16)(196) = 12769 - 12544 = 225 = 15^2.

x=113±1532    x=4orx=4916.x = \frac{113 \pm 15}{32} \implies x = 4 \quad \text{or} \quad x = \frac{49}{16}.

Check x=4x = 4: 341=61=53\sqrt{4} - \sqrt{1} = 6 - 1 = 5. \checkmark

Check x=49/16x = 49/16: 349/1649/163=214116=21414=53\sqrt{49/16} - \sqrt{49/16 - 3} = \frac{21}{4} - \sqrt{\frac{1}{16}} = \frac{21}{4} - \frac{1}{4} = 5. \checkmark

The solutions are x=4916x = \dfrac{49}{16} and x=4x = 4.


Topic: 数论 Number Theory  |  Difficulty: Hard  |  Marks: 20

4 Show that if xx and yy are positive and x3+x2=y3y2x^3 + x^2 = y^3 - y^2 then x<yx < y.

Show further that if 0<xy10 < x \leqslant y - 1, then x3+x2<y3y2x^3 + x^2 < y^3 - y^2.

Prove that there does not exist a pair of positive integers such that the difference of their cubes is equal to the sum of their squares.

Find all the pairs of integers such that the difference of their cubes is equal to the sum of their squares.

Model Solution

Part (i): Show that x<yx < y

Suppose for contradiction that xyx \geqslant y. If x=yx = y, then x3+x2=x3x2x^3 + x^2 = x^3 - x^2, giving 2x2=02x^2 = 0, so x=0x = 0, contradicting x>0x > 0.

If x>y>0x > y > 0, then x3>y3x^3 > y^3 and x2>y2x^2 > y^2, so x3+x2>y3+y2>y3y2x^3 + x^2 > y^3 + y^2 > y^3 - y^2 (since y2>0y^2 > 0). This contradicts x3+x2=y3y2x^3 + x^2 = y^3 - y^2.

Therefore x<yx < y. \qquad \blacksquare

Part (ii): Show that 0<xy10 < x \leqslant y-1 implies x3+x2<y3y2x^3 + x^2 < y^3 - y^2

Since xy1x \leqslant y - 1, we have x+1yx + 1 \leqslant y, so

x3+x2=x2(x+1)x2y.x^3 + x^2 = x^2(x+1) \leqslant x^2 y.

Also y3y2=y2(y1)y2xy^3 - y^2 = y^2(y-1) \geqslant y^2 x (since y1xy - 1 \geqslant x).

So x3+x2x2yx^3 + x^2 \leqslant x^2 y and y3y2xy2y^3 - y^2 \geqslant xy^2. We need x2y<xy2x^2 y < xy^2, i.e. xy(yx)>0xy(y - x) > 0, which holds since x,y>0x, y > 0 and y>xy > x.

But we need strict inequality throughout. Since xy1x \leqslant y-1 and x>0x > 0, we have y>1y > 1, so y1>0y - 1 > 0. Also x<yx < y (from Part (i) or directly from xy1x \leqslant y-1).

More precisely: x2(x+1)x2yx^2(x+1) \leqslant x^2 y with equality iff x+1=yx + 1 = y. And y2(y1)xy2y^2(y-1) \geqslant xy^2 with equality iff x=y1x = y - 1. Even if x=y1x = y - 1, we get x2yx^2 y vs xy2xy^2, and x2y<xy2x^2 y < xy^2 since x<yx < y. Therefore x3+x2<y3y2x^3 + x^2 < y^3 - y^2. \qquad \blacksquare

Part (iii): No pair of positive integers

Suppose m,nm, n are positive integers with m3n3=m2+n2m^3 - n^3 = m^2 + n^2. Since m3n3=(mn)(m2+mn+n2)>0m^3 - n^3 = (m-n)(m^2 + mn + n^2) > 0 and m2+n2>0m^2 + n^2 > 0, we need m>nm > n, so mn+1m \geqslant n + 1.

Setting x=nx = n and y=my = m in Part (ii): since 0<nm10 < n \leqslant m - 1, we have n3+n2<m3m2n^3 + n^2 < m^3 - m^2.

But we need m3n3=m2+n2m^3 - n^3 = m^2 + n^2, i.e. n3+n2=m3m2n^3 + n^2 = m^3 - m^2. This contradicts n3+n2<m3m2n^3 + n^2 < m^3 - m^2.

Therefore no such pair exists. \qquad \blacksquare

Part (iv): Find all integer pairs

The equation is y3x3=y2+x2y^3 - x^3 = y^2 + x^2, i.e. y3y2=x3+x2y^3 - y^2 = x^3 + x^2, which gives

(yx)(y2+xy+x2)=x2+y2.()(y - x)(y^2 + xy + x^2) = x^2 + y^2. \qquad (\star)

Case 1: x=0x = 0. Then y3=y2y^3 = y^2, so y=0y = 0 or y=1y = 1. Solutions: (0,0)(0, 0) and (0,1)(0, 1).

Case 2: y=0y = 0. Then x3=x2-x^3 = x^2, so x=0x = 0 or x=1x = -1. Solution: (1,0)(-1, 0).

Case 3: x>0x > 0. From Part (iii), no positive integer pair exists.

Case 4: x<0x < 0, y>0y > 0. Let x=px = -p with p1p \geqslant 1. Then ()(\star) becomes

(y+p)(y2py+p2)=p2+y2.(y + p)(y^2 - py + p^2) = p^2 + y^2.

The left side equals y3+p31+1=2y^3 + p^3 \geqslant 1 + 1 = 2, while p2+y22p^2 + y^2 \geqslant 2. For p2,y2p \geqslant 2, y \geqslant 2: y3+p316>8p2+y2y^3 + p^3 \geqslant 16 > 8 \geqslant p^2 + y^2. So we only need to check small values.

p=1,y=1p = 1, y = 1: (1+1)(11+1)=2(1+1)(1-1+1) = 2 and 1+1=21 + 1 = 2. \checkmark Solution: (1,1)(-1, 1).

p=1,y2p = 1, y \geqslant 2: (y+1)(y2y+1)=y3+1>y2+1(y+1)(y^2 - y + 1) = y^3 + 1 > y^2 + 1. No solution.

p=2,y=1p = 2, y = 1: (3)(42+1)=9(3)(4-2+1) = 9 and 4+1=54 + 1 = 5. No solution.

p2,y=1p \geqslant 2, y = 1: (1+p)(1p+p2)=1+p39>p2+1(1+p)(1-p+p^2) = 1 + p^3 \geqslant 9 > p^2 + 1. No solution.

Case 5: x<0x < 0, y<0y < 0. Let x=p,y=qx = -p, y = -q with p,q1p, q \geqslant 1. Then

(q+p)(q2pq+p2)=p2+q2,i.e.(pq)(p2pq+q2)=p2+q2.(-q+p)(q^2 - pq + p^2) = p^2 + q^2, \quad \text{i.e.} \quad (p-q)(p^2 - pq + q^2) = p^2 + q^2.

This is p3q3=p2+q2p^3 - q^3 = p^2 + q^2. If p=qp = q: 0=2p20 = 2p^2, impossible. If p>qp > q: by Part (iii) applied to positive integers q,pq, p with p3q3=p2+q2p^3 - q^3 = p^2 + q^2, no solution exists (same argument). If p<qp < q: left side is negative, right side is positive. No solution.

Case 6: x<0x < 0, y=0y = 0. Already covered in Case 2.

Therefore the complete set of integer solutions is

(x,y){(0,0), (0,1), (1,0), (1,1)}.(x, y) \in \{(0, 0),\ (0, 1),\ (-1, 0),\ (-1, 1)\}.


Topic: 代数 Algebra  |  Difficulty: Challenging  |  Marks: 20

5 Give a condition that must be satisfied by pp, qq and rr for it to be possible to write the quadratic polynomial px2+qx+rpx^2 + qx + r in the form p(x+h)2p(x + h)^2, for some hh.

Obtain an equation, which you need not simplify, that must be satisfied by tt if it is possible to write (x2+12bx+t)2(x4+bx3+cx2+dx+e)\left(x^2 + \frac{1}{2}bx + t\right)^2 - \left(x^4 + bx^3 + cx^2 + dx + e\right) in the form k(x+h)2k(x + h)^2, for some kk and hh.

Hence, or otherwise, write x4+6x3+9x22x7x^4 + 6x^3 + 9x^2 - 2x - 7 as a product of two quadratic factors.

Model Solution

Part (i): Condition for px2+qx+r=p(x+h)2px^2 + qx + r = p(x+h)^2

Expanding: p(x+h)2=px2+2phx+ph2p(x+h)^2 = px^2 + 2phx + ph^2. Comparing coefficients with px2+qx+rpx^2 + qx + r:

q=2ph,r=ph2.q = 2ph, \qquad r = ph^2.

From the first equation h=q/(2p)h = q/(2p), so r=pq2/(4p2)=q2/(4p)r = p \cdot q^2/(4p^2) = q^2/(4p).

The necessary and sufficient condition is q2=4prq^2 = 4pr, i.e. the discriminant is zero. \qquad \blacksquare

Part (ii): Equation for tt

We want

(x2+12bx+t)2(x4+bx3+cx2+dx+e)=k(x+h)2.\left(x^2 + \tfrac{1}{2}bx + t\right)^2 - \left(x^4 + bx^3 + cx^2 + dx + e\right) = k(x+h)^2.

The left side must be a perfect square (as a quadratic in xx). Expanding the square:

(x2+12bx+t)2=x4+bx3+(14b2+2t)x2+btx+t2.\left(x^2 + \tfrac{1}{2}bx + t\right)^2 = x^4 + bx^3 + \left(\tfrac{1}{4}b^2 + 2t\right)x^2 + btx + t^2.

Subtracting:

(14b2+2tc)x2+(btd)x+(t2e)=k(x+h)2.\left(\tfrac{1}{4}b^2 + 2t - c\right)x^2 + (bt - d)x + (t^2 - e) = k(x+h)^2.

From Part (i), the condition for this quadratic Ax2+Bx+CAx^2 + Bx + C to be a perfect square is B2=4ACB^2 = 4AC:

(btd)2=4(14b2+2tc)(t2e).(bt - d)^2 = 4\left(\tfrac{1}{4}b^2 + 2t - c\right)(t^2 - e). \qquad \blacksquare

Part (iii): Factorizing x4+6x3+9x22x7x^4 + 6x^3 + 9x^2 - 2x - 7

Here b=6b = 6, c=9c = 9, d=2d = -2, e=7e = -7. The condition becomes

(6t+2)2=4(9+2t9)(t2+7)=42t(t2+7).(6t + 2)^2 = 4(9 + 2t - 9)(t^2 + 7) = 4 \cdot 2t \cdot (t^2 + 7).

Simplifying: (3t+1)2=2t(t2+7)(3t+1)^2 = 2t(t^2 + 7), i.e. 9t2+6t+1=2t3+14t9t^2 + 6t + 1 = 2t^3 + 14t.

2t39t2+8t1=0.2t^3 - 9t^2 + 8t - 1 = 0.

Testing t=1t = -1: 29810-2 - 9 - 8 - 1 \neq 0. Wait, let me recheck: 2(1)39(1)2+8(1)1=2981=202(-1)^3 - 9(-1)^2 + 8(-1) - 1 = -2 - 9 - 8 - 1 = -20. That’s wrong.

Let me recompute: (6t+2)2=36t2+24t+4(6t+2)^2 = 36t^2 + 24t + 4 and 8t(t2+7)=8t3+56t8t(t^2+7) = 8t^3 + 56t. So 8t3+56t=36t2+24t+48t^3 + 56t = 36t^2 + 24t + 4, giving 8t336t2+32t4=08t^3 - 36t^2 + 32t - 4 = 0, i.e. 2t39t2+8t1=02t^3 - 9t^2 + 8t - 1 = 0.

Testing t=12t = \frac{1}{2}: 218914+8121=1494+41=2+3=102 \cdot \frac{1}{8} - 9 \cdot \frac{1}{4} + 8 \cdot \frac{1}{2} - 1 = \frac{1}{4} - \frac{9}{4} + 4 - 1 = -2 + 3 = 1 \neq 0.

Let me try the factorization directly. We seek x4+6x3+9x22x7=(x2+ax+p)(x2+cx+q)x^4 + 6x^3 + 9x^2 - 2x - 7 = (x^2 + ax + p)(x^2 + cx + q) with a+c=6a + c = 6. Taking a=c=3a = c = 3:

(x2+3x+p)(x2+3x+q)=x4+6x3+(9+p+q)x2+3(p+q)x+pq.(x^2 + 3x + p)(x^2 + 3x + q) = x^4 + 6x^3 + (9 + p + q)x^2 + 3(p+q)x + pq.

Matching: 9+p+q=99 + p + q = 9 so p+q=0p + q = 0, giving q=pq = -p. Then 3(p+q)=03(p+q) = 0, but we need 2-2. So a=c=3a = c = 3 doesn’t work directly.

Let a=3+δa = 3 + \delta, c=3δc = 3 - \delta. Then the x2x^2 coefficient is ac+p+q=9δ2+p+q=9ac + p + q = 9 - \delta^2 + p + q = 9, so p+q=δ2p + q = \delta^2. The xx coefficient is aq+cp=(3+δ)q+(3δ)p=3(p+q)+δ(qp)=3δ2+δ(qp)=2aq + cp = (3+\delta)q + (3-\delta)p = 3(p+q) + \delta(q-p) = 3\delta^2 + \delta(q-p) = -2.

With p+q=δ2p + q = \delta^2 and pq=7pq = -7: pp and qq are roots of z2δ2z7=0z^2 - \delta^2 z - 7 = 0, so qp=δ4+28q - p = \sqrt{\delta^4 + 28} (taking q>pq > p).

From 3δ2+δδ4+28=23\delta^2 + \delta\sqrt{\delta^4 + 28} = -2, we need δ<0\delta < 0. Let δ=ϵ\delta = -\epsilon with ϵ>0\epsilon > 0: 3ϵ2ϵϵ4+28=23\epsilon^2 - \epsilon\sqrt{\epsilon^4 + 28} = -2, so ϵϵ4+28=3ϵ2+2\epsilon\sqrt{\epsilon^4 + 28} = 3\epsilon^2 + 2.

Squaring: ϵ2(ϵ4+28)=9ϵ4+12ϵ2+4\epsilon^2(\epsilon^4 + 28) = 9\epsilon^4 + 12\epsilon^2 + 4, giving ϵ6+28ϵ2=9ϵ4+12ϵ2+4\epsilon^6 + 28\epsilon^2 = 9\epsilon^4 + 12\epsilon^2 + 4, i.e. ϵ69ϵ4+16ϵ24=0\epsilon^6 - 9\epsilon^4 + 16\epsilon^2 - 4 = 0.

Let u=ϵ2u = \epsilon^2: u39u2+16u4=0u^3 - 9u^2 + 16u - 4 = 0. Testing u=1u = 1: 19+164=401 - 9 + 16 - 4 = 4 \neq 0. Testing u=2u = 2: 836+324=08 - 36 + 32 - 4 = 0. \checkmark

So u=2u = 2, ϵ=2\epsilon = \sqrt{2}, δ=2\delta = -\sqrt{2}.

Then p+q=2p + q = 2, pq=7pq = -7, so p,qp, q are roots of z22z7=0z^2 - 2z - 7 = 0: z=1±22z = 1 \pm 2\sqrt{2}.

Taking p=122p = 1 - 2\sqrt{2}, q=1+22q = 1 + 2\sqrt{2}, a=32a = 3 - \sqrt{2}, c=3+2c = 3 + \sqrt{2}:

x4+6x3+9x22x7=(x2+(32)x+122)(x2+(3+2)x+1+22).x^4 + 6x^3 + 9x^2 - 2x - 7 = \left(x^2 + (3-\sqrt{2})x + 1 - 2\sqrt{2}\right)\left(x^2 + (3+\sqrt{2})x + 1 + 2\sqrt{2}\right).

Verification: The xx coefficient is (32)(1+22)+(3+2)(122)(3-\sqrt{2})(1+2\sqrt{2}) + (3+\sqrt{2})(1-2\sqrt{2}).

(32)(1+22)=3+6224=1+52(3-\sqrt{2})(1+2\sqrt{2}) = 3 + 6\sqrt{2} - \sqrt{2} - 4 = -1 + 5\sqrt{2}.

(3+2)(122)=362+24=152(3+\sqrt{2})(1-2\sqrt{2}) = 3 - 6\sqrt{2} + \sqrt{2} - 4 = -1 - 5\sqrt{2}.

Sum =2= -2. \checkmark

The constant term: (122)(1+22)=18=7(1-2\sqrt{2})(1+2\sqrt{2}) = 1 - 8 = -7. \checkmark


Topic: 微分方程 Differential Equations  |  Difficulty: Challenging  |  Marks: 20

6 Find all the solution curves of the differential equation y4(dydx)4=(y21)2y^4 \left(\frac{dy}{dx}\right)^4 = \left(y^2 - 1\right)^2 that pass through either of the points (i) (0,32)\left(0, \frac{\sqrt{3}}{2}\right), (ii) (0,52)\left(0, \frac{\sqrt{5}}{2}\right).

Show also that y=1y = 1 and y=1y = -1 are solutions of the differential equation. Sketch all these solution curves on a single set of axes.

Model Solution

Part (i): Solving the differential equation

The equation y4(dydx)4=(y21)2y^4\left(\frac{dy}{dx}\right)^4 = (y^2-1)^2 can be written as (ydydx)4=(y21)2\left(y\frac{dy}{dx}\right)^4 = (y^2-1)^2.

Taking square roots (both sides are non-negative): (ydydx)2=y21\left(y\frac{dy}{dx}\right)^2 = |y^2 - 1|.

So ydydx=±y21y\frac{dy}{dx} = \pm\sqrt{|y^2-1|}, which gives four ODEs depending on the signs.

Case A: y>1|y| > 1 (so y21>0y^2 - 1 > 0)

ydydx=±y21y\frac{dy}{dx} = \pm\sqrt{y^2-1}, i.e. yy21dy=±dx\frac{y}{\sqrt{y^2-1}}\, dy = \pm\, dx.

Integrating: y21=±x+C\sqrt{y^2-1} = \pm x + C, so

y2=(xC)2+1(choosing appropriate sign).()y^2 = (x - C)^2 + 1 \qquad \text{(choosing appropriate sign)}. \qquad (\star)

Case B: y<1|y| < 1 (so 1y2>01 - y^2 > 0)

ydydx=±1y2y\frac{dy}{dx} = \pm\sqrt{1-y^2}, i.e. y1y2dy=±dx\frac{y}{\sqrt{1-y^2}}\, dy = \pm\, dx.

Integrating: 1y2=±x+C-\sqrt{1-y^2} = \pm x + C, so 1y2=x+C\sqrt{1-y^2} = \mp x + C', giving

y2=1(xC)2.()y^2 = 1 - (x - C)^2. \qquad (\star\star)

Solution curves through (0,3/2)(0, \sqrt{3}/2)

Since y=3/2<1y = \sqrt{3}/2 < 1, we use Case B. From ()(\star\star): 34=1C2\frac{3}{4} = 1 - C^2, so C=±12C = \pm\frac{1}{2}.

This gives two circles:

(x12)2+y2=1and(x+12)2+y2=1.\left(x - \tfrac{1}{2}\right)^2 + y^2 = 1 \qquad \text{and} \qquad \left(x + \tfrac{1}{2}\right)^2 + y^2 = 1.

Both are valid solution curves (the upper semicircles satisfy the ODE with y>0y > 0).

Solution curves through (0,5/2)(0, \sqrt{5}/2)

Since y=5/2>1y = \sqrt{5}/2 > 1, we use Case A. From ()(\star): 54=C2+1\frac{5}{4} = C^2 + 1, so C=±12C = \pm\frac{1}{2}.

This gives two hyperbolas:

y2(x12)2=1andy2(x+12)2=1.y^2 - \left(x - \tfrac{1}{2}\right)^2 = 1 \qquad \text{and} \qquad y^2 - \left(x + \tfrac{1}{2}\right)^2 = 1.

Part (ii): Constant solutions

If y=1y = 1: LHS =10=0= 1 \cdot 0 = 0, RHS =0= 0. \checkmark

If y=1y = -1: LHS =10=0= 1 \cdot 0 = 0, RHS =0= 0. \checkmark

Part (iii): Sketch

The solution curves are:

  • Two circles: (x±12)2+y2=1\left(x \pm \frac{1}{2}\right)^2 + y^2 = 1 (valid for y<1|y| < 1)
  • Two hyperbolas: y2(x±12)2=1y^2 - \left(x \pm \frac{1}{2}\right)^2 = 1 (valid for y>1|y| > 1)
  • Two horizontal lines: y=1y = 1 and y=1y = -1

The circles are centred at (±1/2,0)(\pm 1/2, 0) with radius 1, fitting within the strip y<1|y| < 1. The hyperbolas have vertices at (±1/2,±1)(\pm 1/2, \pm 1) and open vertically. The lines y=±1y = \pm 1 are asymptotes that the hyperbolas approach and that separate the circular from the hyperbolic regions.

The curve through (0,3/2)(0, \sqrt{3}/2) consists of the upper arcs of both circles. The curve through (0,5/2)(0, \sqrt{5}/2) consists of the upper branches of both hyperbolas.


Topic: 三角学与坐标几何 Trigonometry and Coordinate Geometry  |  Difficulty: Challenging  |  Marks: 20

7 Given that α\alpha and β\beta are acute angles, show that α+β=π/2\alpha + \beta = \pi/2 if and only if cos2α+cos2β=1\cos^2 \alpha + \cos^2 \beta = 1.

In the xxyy plane, the point AA has coordinates (0,s)(0, s) and the point CC has coordinates (s,0)(s, 0), where s>0s > 0. The point BB lies in the first quadrant (x>0,y>0x > 0, y > 0). The lengths of ABAB, OBOB and CBCB are respectively a,ba, b and cc.

Show that

(s2+b2a2)2+(s2+b2c2)2=4s2b2(s^2 + b^2 - a^2)^2 + (s^2 + b^2 - c^2)^2 = 4s^2b^2

and hence that

(2s2a2c2)2+(2b2a2c2)2=4a2c2.(2s^2 - a^2 - c^2)^2 + (2b^2 - a^2 - c^2)^2 = 4a^2c^2 .

Deduce that

(ac)22b2(a+c)2.(a - c)^2 \leqslant 2b^2 \leqslant (a + c)^2 .

Model Solution

Part (i): α+β=π/2\alpha + \beta = \pi/2 if and only if cos2α+cos2β=1\cos^2\alpha + \cos^2\beta = 1

Forward direction. Suppose α+β=π/2\alpha + \beta = \pi/2. Then β=π/2α\beta = \pi/2 - \alpha, so

cosβ=cos ⁣(π2α)=sinα.\cos\beta = \cos\!\left(\frac{\pi}{2} - \alpha\right) = \sin\alpha.

Since α\alpha is acute, both sinα\sin\alpha and cosα\cos\alpha are non-negative, so cos2β=sin2α\cos^2\beta = \sin^2\alpha. Therefore

cos2α+cos2β=cos2α+sin2α=1.\cos^2\alpha + \cos^2\beta = \cos^2\alpha + \sin^2\alpha = 1.

Reverse direction. Suppose cos2α+cos2β=1\cos^2\alpha + \cos^2\beta = 1. Using cos2α=1sin2α\cos^2\alpha = 1 - \sin^2\alpha, we get sin2α+cos2β=1\sin^2\alpha + \cos^2\beta = 1, so cos2β=sin2α\cos^2\beta = \sin^2\alpha.

Since α\alpha and β\beta are acute, cosβ>0\cos\beta > 0 and sinα>0\sin\alpha > 0, so cosβ=sinα=cos(π/2α)\cos\beta = \sin\alpha = \cos(\pi/2 - \alpha). Since cos\cos is strictly decreasing on [0,π][0, \pi] and both β\beta and π/2α\pi/2 - \alpha lie in (0,π/2)(0, \pi/2), we conclude β=π/2α\beta = \pi/2 - \alpha, i.e. α+β=π/2\alpha + \beta = \pi/2. \qquad \blacksquare

Part (ii): The first identity

Let B=(x,y)B = (x, y) with x>0,y>0x > 0, y > 0. Let θ=BOx\theta = \angle BOx, so x=bcosθx = b\cos\theta and y=bsinθy = b\sin\theta.

Using A=(0,s)A = (0, s), C=(s,0)C = (s, 0), O=(0,0)O = (0, 0):

a2=AB2=x2+(ys)2=x2+y22sy+s2=b2+s22sbsinθa^2 = |AB|^2 = x^2 + (y - s)^2 = x^2 + y^2 - 2sy + s^2 = b^2 + s^2 - 2sb\sin\theta

c2=CB2=(xs)2+y2=x22sx+s2+y2=b2+s22sbcosθc^2 = |CB|^2 = (x - s)^2 + y^2 = x^2 - 2sx + s^2 + y^2 = b^2 + s^2 - 2sb\cos\theta

Therefore

s2+b2a2=2sbsinθ,s2+b2c2=2sbcosθ.s^2 + b^2 - a^2 = 2sb\sin\theta, \qquad s^2 + b^2 - c^2 = 2sb\cos\theta.

Squaring and adding:

(s2+b2a2)2+(s2+b2c2)2=4s2b2sin2θ+4s2b2cos2θ=4s2b2.(s^2 + b^2 - a^2)^2 + (s^2 + b^2 - c^2)^2 = 4s^2b^2\sin^2\theta + 4s^2b^2\cos^2\theta = 4s^2b^2. \qquad \blacksquare

Part (iii): The second identity

From Part (ii), a2=s2+b22sbsinθa^2 = s^2 + b^2 - 2sb\sin\theta and c2=s2+b22sbcosθc^2 = s^2 + b^2 - 2sb\cos\theta, so

a2+c2=2(s2+b2)2sb(sinθ+cosθ).a^2 + c^2 = 2(s^2 + b^2) - 2sb(\sin\theta + \cos\theta).

Hence

2s2a2c2=2s22(s2+b2)+2sb(sinθ+cosθ)=2sb(sinθ+cosθ)2b22s^2 - a^2 - c^2 = 2s^2 - 2(s^2 + b^2) + 2sb(\sin\theta + \cos\theta) = 2sb(\sin\theta + \cos\theta) - 2b^2

2b2a2c2=2b22(s2+b2)+2sb(sinθ+cosθ)=2sb(sinθ+cosθ)2s2.2b^2 - a^2 - c^2 = 2b^2 - 2(s^2 + b^2) + 2sb(\sin\theta + \cos\theta) = 2sb(\sin\theta + \cos\theta) - 2s^2.

Wait — let me redo this more cleanly. Define u=a2+c2u = a^2 + c^2 and v=a2c2v = a^2 - c^2. Then

u=2(s2+b2)2sb(sinθ+cosθ),v=2sb(cosθsinθ).u = 2(s^2 + b^2) - 2sb(\sin\theta + \cos\theta), \qquad v = 2sb(\cos\theta - \sin\theta).

So

2s2u=2sb(sinθ+cosθ)2b2,2b2u=2sb(sinθ+cosθ)2s2.2s^2 - u = 2sb(\sin\theta + \cos\theta) - 2b^2, \qquad 2b^2 - u = 2sb(\sin\theta + \cos\theta) - 2s^2.

Let w=sinθ+cosθw = \sin\theta + \cos\theta. Then 2s2u=2sbw2b2=2b(swb)2s^2 - u = 2sbw - 2b^2 = 2b(sw - b) and 2b2u=2sbw2s2=2s(bws)2b^2 - u = 2sbw - 2s^2 = 2s(bw - s).

Computing the left side:

(2s2u)2+(2b2u)2=4b2(swb)2+4s2(bws)2.(2s^2 - u)^2 + (2b^2 - u)^2 = 4b^2(sw - b)^2 + 4s^2(bw - s)^2.

Expanding:

=4b2(s2w22sbw+b2)+4s2(b2w22sbw+s2)= 4b^2(s^2w^2 - 2sbw + b^2) + 4s^2(b^2w^2 - 2sbw + s^2)

=4s2b2w28s2b2w+4b4+4s2b2w28s2b2w+4s4= 4s^2b^2w^2 - 8s^2b^2w + 4b^4 + 4s^2b^2w^2 - 8s^2b^2w + 4s^4

Wait, that has w2w^2 and ww terms that don’t simplify neatly. Let me try a direct computation instead.

(2s2u)2+(2b2u)2=4s44s2u+u2+4b44b2u+u2(2s^2 - u)^2 + (2b^2 - u)^2 = 4s^4 - 4s^2u + u^2 + 4b^4 - 4b^2u + u^2

=2u24u(s2+b2)+4(s4+b4)= 2u^2 - 4u(s^2 + b^2) + 4(s^4 + b^4)

=2[u22u(s2+b2)+2(s4+b4)]= 2[u^2 - 2u(s^2 + b^2) + 2(s^4 + b^4)]

=2[(s2+b2u)2+(s4+b4)(s2+b2)2+2(s4+b4)2(s4+b4)]= 2[(s^2 + b^2 - u)^2 + (s^4 + b^4) - (s^2 + b^2)^2 + 2(s^4 + b^4) - 2(s^4 + b^4)]

This is getting circular. Let me use the identity from Part (ii) directly.

From Part (ii): (s2+b2a2)2+(s2+b2c2)2=4s2b2(s^2 + b^2 - a^2)^2 + (s^2 + b^2 - c^2)^2 = 4s^2b^2.

Now compute 4a2c24a^2c^2. We have a2c2=(s2+b22sbsinθ)(s2+b22sbcosθ)a^2c^2 = (s^2 + b^2 - 2sb\sin\theta)(s^2 + b^2 - 2sb\cos\theta).

Let m=s2+b2m = s^2 + b^2. Then

a2c2=(m2sbsinθ)(m2sbcosθ)=m22sbm(sinθ+cosθ)+4s2b2sinθcosθ.a^2c^2 = (m - 2sb\sin\theta)(m - 2sb\cos\theta) = m^2 - 2sbm(\sin\theta + \cos\theta) + 4s^2b^2\sin\theta\cos\theta.

Also, u=2m2sb(sinθ+cosθ)u = 2m - 2sb(\sin\theta + \cos\theta), so u2=4m28sbm(sinθ+cosθ)+4s2b2(sinθ+cosθ)2u^2 = 4m^2 - 8sbm(\sin\theta + \cos\theta) + 4s^2b^2(\sin\theta + \cos\theta)^2.

And v2=4s2b2(cosθsinθ)2v^2 = 4s^2b^2(\cos\theta - \sin\theta)^2.

Therefore

u2v2=4m28sbm(sinθ+cosθ)+4s2b2[(sinθ+cosθ)2(cosθsinθ)2].u^2 - v^2 = 4m^2 - 8sbm(\sin\theta + \cos\theta) + 4s^2b^2[(\sin\theta + \cos\theta)^2 - (\cos\theta - \sin\theta)^2].

Since (sinθ+cosθ)2(cosθsinθ)2=4sinθcosθ(\sin\theta + \cos\theta)^2 - (\cos\theta - \sin\theta)^2 = 4\sin\theta\cos\theta:

u2v2=4[m22sbm(sinθ+cosθ)+4s2b2sinθcosθ]=4a2c2.u^2 - v^2 = 4[m^2 - 2sbm(\sin\theta + \cos\theta) + 4s^2b^2\sin\theta\cos\theta] = 4a^2c^2.

Since u2v2=(a2+c2)2(a2c2)2=4a2c2u^2 - v^2 = (a^2 + c^2)^2 - (a^2 - c^2)^2 = 4a^2c^2, this is just the identity (u+v)(uv)=4a2c2(u+v)(u-v) = 4a^2c^2, which is always true. So I need a different approach to connect to the left side.

Let me instead directly compute (2s2u)2+(2b2u)2(2s^2 - u)^2 + (2b^2 - u)^2 and compare with 4a2c24a^2c^2.

(2s2u)2+(2b2u)2=(2s2a2c2)2+(2b2a2c2)2(2s^2 - u)^2 + (2b^2 - u)^2 = (2s^2 - a^2 - c^2)^2 + (2b^2 - a^2 - c^2)^2

=[2s22m+2sbw]2+[2b22m+2sbw]2= [2s^2 - 2m + 2sbw]^2 + [2b^2 - 2m + 2sbw]^2 where w=sinθ+cosθw = \sin\theta + \cos\theta

=[2b2+2sbw]2+[2s2+2sbw]2= [-2b^2 + 2sbw]^2 + [-2s^2 + 2sbw]^2

=4b2(swb)2+4s2(bws)2= 4b^2(sw - b)^2 + 4s^2(bw - s)^2

=4[b2(s2w22bsw+b2)+s2(b2w22bsw+s2)]= 4[b^2(s^2w^2 - 2bsw + b^2) + s^2(b^2w^2 - 2bsw + s^2)]

=4[s2b2w22b3sw+b4+s2b2w22s3bw+s4]= 4[s^2b^2w^2 - 2b^3sw + b^4 + s^2b^2w^2 - 2s^3bw + s^4]

=4[2s2b2w22sbw(b2+s2)+s4+b4]= 4[2s^2b^2w^2 - 2sbw(b^2 + s^2) + s^4 + b^4]

=4[2s2b2w22sbwm+m22s2b2]= 4[2s^2b^2w^2 - 2sbwm + m^2 - 2s^2b^2]

Now 4a2c2=4[m22sbmw+4s2b2sinθcosθ]4a^2c^2 = 4[m^2 - 2sbmw + 4s^2b^2\sin\theta\cos\theta].

So we need: 2s2b2w22sbwm+m22s2b2=m22sbmw+4s2b2sinθcosθ2s^2b^2w^2 - 2sbwm + m^2 - 2s^2b^2 = m^2 - 2sbmw + 4s^2b^2\sin\theta\cos\theta.

Simplifying: 2s2b2w22s2b2=4s2b2sinθcosθ2s^2b^2w^2 - 2s^2b^2 = 4s^2b^2\sin\theta\cos\theta.

w21=(sinθ+cosθ)21=sin2θ+2sinθcosθ+cos2θ1=2sinθcosθw^2 - 1 = (\sin\theta + \cos\theta)^2 - 1 = \sin^2\theta + 2\sin\theta\cos\theta + \cos^2\theta - 1 = 2\sin\theta\cos\theta.

So 2s2b2(w21)=4s2b2sinθcosθ2s^2b^2(w^2 - 1) = 4s^2b^2\sin\theta\cos\theta. This is true! \qquad \blacksquare

OK so the approach works but the write-up is messy. Let me restructure it cleanly.

Part (iv): The deduction

From Part (ii), rearranging: (2b2a2c2)2=4a2c2(2s2a2c2)2(2b^2 - a^2 - c^2)^2 = 4a^2c^2 - (2s^2 - a^2 - c^2)^2.

Since (2s2a2c2)20(2s^2 - a^2 - c^2)^2 \geqslant 0, we have

(2b2a2c2)24a2c2,(2b^2 - a^2 - c^2)^2 \leqslant 4a^2c^2,

so 2b2a2c22ac|2b^2 - a^2 - c^2| \leqslant 2ac. This gives

2ac2b2a2c22ac.-2ac \leqslant 2b^2 - a^2 - c^2 \leqslant 2ac.

Adding a2+c2a^2 + c^2 throughout: a2+c22ac2b2a2+c2+2aca^2 + c^2 - 2ac \leqslant 2b^2 \leqslant a^2 + c^2 + 2ac, i.e.

(ac)22b2(a+c)2.(a - c)^2 \leqslant 2b^2 \leqslant (a + c)^2. \qquad \blacksquare


Topic: 复数 Complex Numbers  |  Difficulty: Hard  |  Marks: 20

8 Four complex numbers u1,u2,u3u_1, u_2, u_3 and u4u_4 have unit modulus, and arguments θ1,θ2,θ3\theta_1, \theta_2, \theta_3 and θ4\theta_4, respectively, with π<θ1<θ2<θ3<θ4<π-\pi < \theta_1 < \theta_2 < \theta_3 < \theta_4 < \pi.

Show that

arg(u1u2)=12(θ1+θ2π)+2nπ\arg (u_1 - u_2) = \frac{1}{2} (\theta_1 + \theta_2 - \pi) + 2n\pi

where n=0n = 0 or 11. Deduce that

arg((u1u2)(u4u3))=arg((u1u4)(u3u2))+2nπ\arg ((u_1 - u_2) (u_4 - u_3)) = \arg ((u_1 - u_4) (u_3 - u_2)) + 2n\pi

for some integer nn.

Prove that

(u1u2)(u4u3)+(u1u4)(u3u2)=(u1u3)(u4u2).| (u_1 - u_2) (u_4 - u_3) | + | (u_1 - u_4) (u_3 - u_2) | = | (u_1 - u_3) (u_4 - u_2) | .

Model Solution

Part (i): Argument of u1u2u_1 - u_2

Write uk=eiθku_k = e^{i\theta_k}. Then

u1u2=eiθ1eiθ2=ei(θ1+θ2)/2(ei(θ1θ2)/2ei(θ1θ2)/2)=2isin ⁣(θ1θ22)ei(θ1+θ2)/2.u_1 - u_2 = e^{i\theta_1} - e^{i\theta_2} = e^{i(\theta_1+\theta_2)/2}\left(e^{i(\theta_1-\theta_2)/2} - e^{-i(\theta_1-\theta_2)/2}\right) = 2i\sin\!\left(\frac{\theta_1-\theta_2}{2}\right)e^{i(\theta_1+\theta_2)/2}.

Since θ1<θ2\theta_1 < \theta_2, we have θ1θ22<0\frac{\theta_1-\theta_2}{2} < 0, so sin ⁣(θ1θ22)<0\sin\!\left(\frac{\theta_1-\theta_2}{2}\right) < 0. Thus

u1u2=2sin ⁣(θ1θ22)i(1)ei(θ1+θ2)/2=2sin ⁣(θ1θ22)ei(θ1+θ2)/2eiπ/2eiπ.u_1 - u_2 = 2\left|\sin\!\left(\frac{\theta_1-\theta_2}{2}\right)\right| \cdot i \cdot (-1) \cdot e^{i(\theta_1+\theta_2)/2} = 2\left|\sin\!\left(\frac{\theta_1-\theta_2}{2}\right)\right| e^{i(\theta_1+\theta_2)/2} \cdot e^{-i\pi/2} \cdot e^{i\pi}.

Therefore

arg(u1u2)=θ1+θ22+π2+π(mod2π)=θ1+θ22+3π2(mod2π).\arg(u_1 - u_2) = \frac{\theta_1+\theta_2}{2} + \frac{\pi}{2} + \pi \pmod{2\pi} = \frac{\theta_1+\theta_2}{2} + \frac{3\pi}{2} \pmod{2\pi}.

Wait — let me redo this more carefully. We have isinαi \cdot \sin\alpha where α<0\alpha < 0 and sinα<0\sin\alpha < 0. So isinα=sinαi(1)=sinαeiπeiπ/2=sinαei3π/2i\sin\alpha = |\sin\alpha| \cdot i \cdot (-1) = |\sin\alpha| \cdot e^{i\pi} \cdot e^{i\pi/2} = |\sin\alpha| \cdot e^{i3\pi/2}.

Thus u1u2=2sinαei(θ1+θ2)/2ei3π/2u_1 - u_2 = 2|\sin\alpha| \cdot e^{i(\theta_1+\theta_2)/2} \cdot e^{i3\pi/2}, giving

arg(u1u2)=θ1+θ22+3π2=θ1+θ22π2(mod2π).\arg(u_1 - u_2) = \frac{\theta_1+\theta_2}{2} + \frac{3\pi}{2} = \frac{\theta_1+\theta_2}{2} - \frac{\pi}{2} \pmod{2\pi}.

Since θ1+θ2π2\frac{\theta_1+\theta_2-\pi}{2} is equivalent to θ1+θ22π2\frac{\theta_1+\theta_2}{2} - \frac{\pi}{2}, we have

arg(u1u2)=12(θ1+θ2π)+2nπ,n=0 or 1.\arg(u_1 - u_2) = \frac{1}{2}(\theta_1 + \theta_2 - \pi) + 2n\pi, \qquad n = 0 \text{ or } 1. \qquad \blacksquare

Part (ii): Argument of products

Using the result from Part (i):

arg(u1u2)=θ1+θ2π2,arg(u4u3)=θ3+θ4π2,\arg(u_1 - u_2) = \frac{\theta_1+\theta_2-\pi}{2}, \qquad \arg(u_4 - u_3) = \frac{\theta_3+\theta_4-\pi}{2},

arg(u1u4)=θ1+θ4π2,arg(u3u2)=θ2+θ3π2.\arg(u_1 - u_4) = \frac{\theta_1+\theta_4-\pi}{2}, \qquad \arg(u_3 - u_2) = \frac{\theta_2+\theta_3-\pi}{2}.

Therefore

arg((u1u2)(u4u3))=θ1+θ2+θ3+θ42π2(mod2π),\arg((u_1-u_2)(u_4-u_3)) = \frac{\theta_1+\theta_2+\theta_3+\theta_4 - 2\pi}{2} \pmod{2\pi},

arg((u1u4)(u3u2))=θ1+θ2+θ3+θ42π2(mod2π).\arg((u_1-u_4)(u_3-u_2)) = \frac{\theta_1+\theta_2+\theta_3+\theta_4 - 2\pi}{2} \pmod{2\pi}.

These are equal modulo 2π2\pi, so

arg((u1u2)(u4u3))=arg((u1u4)(u3u2))+2nπ.\arg((u_1-u_2)(u_4-u_3)) = \arg((u_1-u_4)(u_3-u_2)) + 2n\pi. \qquad \blacksquare

Part (iii): The magnitude identity

Since all uku_k lie on the unit circle, uiuj=2sinθiθj2|u_i - u_j| = 2\left|\sin\frac{\theta_i-\theta_j}{2}\right|. With the ordering π<θ1<θ2<θ3<θ4<π-\pi < \theta_1 < \theta_2 < \theta_3 < \theta_4 < \pi, all differences θjθi\theta_j - \theta_i for i<ji < j are positive and less than 2π2\pi, so uiuj=2sinθjθi2|u_i - u_j| = 2\sin\frac{\theta_j-\theta_i}{2}.

Let α=θ2θ1>0\alpha = \theta_2 - \theta_1 > 0, β=θ3θ2>0\beta = \theta_3 - \theta_2 > 0, γ=θ4θ3>0\gamma = \theta_4 - \theta_3 > 0. Then

(u1u2)(u4u3)=4sinα2sinγ2,|(u_1-u_2)(u_4-u_3)| = 4\sin\frac{\alpha}{2}\sin\frac{\gamma}{2},

(u1u4)(u3u2)=4sinα+β+γ2sinβ2,|(u_1-u_4)(u_3-u_2)| = 4\sin\frac{\alpha+\beta+\gamma}{2}\sin\frac{\beta}{2},

(u1u3)(u4u2)=4sinα+β2sinβ+γ2.|(u_1-u_3)(u_4-u_2)| = 4\sin\frac{\alpha+\beta}{2}\sin\frac{\beta+\gamma}{2}.

We need to show:

sinα2sinγ2+sinα+β+γ2sinβ2=sinα+β2sinβ+γ2.()\sin\frac{\alpha}{2}\sin\frac{\gamma}{2} + \sin\frac{\alpha+\beta+\gamma}{2}\sin\frac{\beta}{2} = \sin\frac{\alpha+\beta}{2}\sin\frac{\beta+\gamma}{2}. \qquad (\dagger)

Let A=α/2A = \alpha/2, B=β/2B = \beta/2, C=γ/2C = \gamma/2. Then ()(\dagger) becomes

sinAsinC+sin(A+B+C)sinB=sin(A+B)sin(B+C).\sin A\sin C + \sin(A+B+C)\sin B = \sin(A+B)\sin(B+C).

Using the product-to-sum formula sinXsinY=12[cos(XY)cos(X+Y)]\sin X \sin Y = \frac{1}{2}[\cos(X-Y) - \cos(X+Y)]:

RHS: sin(A+B)sin(B+C)=12[cos(AC)cos(A+2B+C)]\sin(A+B)\sin(B+C) = \frac{1}{2}[\cos(A-C) - \cos(A+2B+C)].

LHS: sinAsinC=12[cos(AC)cos(A+C)]\sin A\sin C = \frac{1}{2}[\cos(A-C) - \cos(A+C)] and

sin(A+B+C)sinB=12[cos(A+C)cos(A+2B+C)]\sin(A+B+C)\sin B = \frac{1}{2}[\cos(A+C) - \cos(A+2B+C)].

Adding: LHS=12[cos(AC)cos(A+2B+C)]=RHS\text{LHS} = \frac{1}{2}[\cos(A-C) - \cos(A+2B+C)] = \text{RHS}. \qquad \blacksquare

Geometric interpretation: The four points u1,u2,u3,u4u_1, u_2, u_3, u_4 lie in order on the unit circle, forming a cyclic quadrilateral. The identity (u1u2)(u4u3)+(u1u4)(u3u2)=(u1u3)(u4u2)|(u_1-u_2)(u_4-u_3)| + |(u_1-u_4)(u_3-u_2)| = |(u_1-u_3)(u_4-u_2)| is precisely Ptolemy’s theorem: for a cyclic quadrilateral, the sum of the products of opposite sides equals the product of the diagonals.