Skip to content

STEP3 2016 -- Pure Mathematics

STEP3 2016 — Section A (Pure Mathematics)

Section titled “STEP3 2016 — Section A (Pure Mathematics)”

Exam: STEP3  |  Year: 2016  |  Questions: Q1—Q8  |  Total marks per question: 20

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

QTopicDifficultyKey Techniques
1积分 (Integration)Standard三角换元 x+a=ba2tanux+a=\sqrt{b-a^2}\tan u, 完全平方配方, 递推关系推导, 数学归纳法
2坐标几何 (Coordinate Geometry)Challenging参数化表示 (at2,2at)(at^2, 2at), 隐函数求导求法线方程, 韦达定理 (q+r=p,qr=2q+r=-p, qr=2), 联立方程求交点
3微分方程 (Differential Equations)Challenging对含指数函数的表达式求导, 因式定理/余式定理, 多项式次数比较, 反证法
4级数与数列 (Series and Sequences)Challenging裂项法(望远镜求和), 双曲函数定义 sech(ry)=2ery1+e2ry\text{sech}(ry) = \frac{2e^{-ry}}{1+e^{-2ry}}, 变量替换 x=e2yx=e^{-2y}, 利用偶函数性质处理负下标
5组合与数论 (Combinatorics and Number Theory)Hard二项式展开对称性, 素数整除性论证, 乘积分解 P1,2m+1=P1,m+1Pm+1,2m+1P_{1,2m+1} = P_{1,m+1} \cdot P_{m+1,2m+1}, 强归纳法(分奇偶讨论)
6双曲函数 (Hyperbolic Functions)Challenging双曲函数加法公式,反双曲函数,分类讨论,必要充分条件论证
7复数 (Complex Numbers)Hard单位根因式分解,因子定理,复数模的乘积,正多边形几何性质
8函数方程 (Functional Equations)Standard代入-x建立方程组,函数迭代构造循环,代数消元法,验证解

Topic: 积分 (Integration)  |  Difficulty: Standard  |  Marks: 20

1 Let In=1(x2+2ax+b)ndx,I_n = \int_{-\infty}^{\infty} \frac{1}{(x^2 + 2ax + b)^n} \, dx \, , where aa and bb are constants with b>a2b > a^2, and nn is a positive integer.

(i) By using the substitution x+a=ba2tanux + a = \sqrt{b - a^2} \tan u , or otherwise, show that I1=πba2.I_1 = \frac{\pi}{\sqrt{b - a^2}} \, .

(ii) Show that 2n(ba2)In+1=(2n1)In2n(b - a^2) \, I_{n+1} = (2n - 1) \, I_n .

(iii) Hence prove by induction that In=π22n2(ba2)n12(2n2n1).I_n = \frac{\pi}{2^{2n-2}(b - a^2)^{n-\frac{1}{2}}} \binom{2n-2}{n-1} \, .

Hint

Part (i) is most simply dealt with by the suggested method, change of variable, and it is worth completing the square in the denominator to simplify the algebra leading to a trivial integral. Part (ii) can either be attempted immediately using integration by parts, starting from InI_n and obtaining 2nx(x+a)(x2+2ax+b)n+1dx\int_{-\infty}^{\infty} \frac{2nx(x+a)}{(x^2+2ax+b)^{n+1}} dx and then writing the numerator as 2n(x2+2ax+b)na(2x+2a)2n(ba2)2n(x^2 + 2ax + b) - na(2x + 2a) - 2n(b - a^2). Alternatively, use of the same substitution in In+1I_{n+1} as in part (i) leads to the need to integrate cos2nu\cos^{2n} u, which in turn can be written as cos2n2u(1sin2u)=cos2n2ucos2n2usinusinu\cos^{2n-2} u(1 - \sin^2 u) = \cos^{2n-2} u - \cos^{2n-2} u \sin u \cdot \sin u, with the second term being susceptible to integration by parts. Part (iii) follows from the previous parts by induction using part (ii) to achieve the inductive step and (i) the base case.

Model Solution

Part (i)

Complete the square in the denominator:

x2+2ax+b=(x+a)2+(ba2)x^2 + 2ax + b = (x + a)^2 + (b - a^2)

Let x+a=ba2tanux + a = \sqrt{b - a^2} \tan u, so dx=ba2sec2ududx = \sqrt{b - a^2} \sec^2 u \, du. As xx \to \infty, uπ2u \to \frac{\pi}{2}; as xx \to -\infty, uπ2u \to -\frac{\pi}{2}.

The denominator becomes:

(x+a)2+(ba2)=(ba2)tan2u+(ba2)=(ba2)sec2u(x + a)^2 + (b - a^2) = (b - a^2)\tan^2 u + (b - a^2) = (b - a^2)\sec^2 u

Therefore:

I1=π/2π/2ba2sec2u(ba2)sec2udu=1ba2π/2π/2du=πba2I_1 = \int_{-\pi/2}^{\pi/2} \frac{\sqrt{b - a^2} \sec^2 u}{(b - a^2)\sec^2 u} \, du = \frac{1}{\sqrt{b - a^2}} \int_{-\pi/2}^{\pi/2} du = \frac{\pi}{\sqrt{b - a^2}}

Part (ii)

Let c2=ba2c^2 = b - a^2 and substitute t=x+at = x + a, so dt=dxdt = dx and x2+2ax+b=t2+c2x^2 + 2ax + b = t^2 + c^2:

In=dt(t2+c2)nI_n = \int_{-\infty}^{\infty} \frac{dt}{(t^2 + c^2)^n}

Write 1=(t2+c2)t2t2+c21 = \frac{(t^2 + c^2) - t^2}{t^2 + c^2} in the numerator:

In=t2+c2(t2+c2)n+1dtt2(t2+c2)n+1dtI_n = \int_{-\infty}^{\infty} \frac{t^2 + c^2}{(t^2 + c^2)^{n+1}} \, dt - \int_{-\infty}^{\infty} \frac{t^2}{(t^2 + c^2)^{n+1}} \, dt

In=In+1+c2In+1t2(t2+c2)n+1dt(*)I_n = I_{n+1} + c^2 I_{n+1} - \int_{-\infty}^{\infty} \frac{t^2}{(t^2 + c^2)^{n+1}} \, dt \qquad \text{(*)}

Evaluate the remaining integral by parts. Let u=tu = t and dv=t(t2+c2)n+1dtdv = \frac{t}{(t^2 + c^2)^{n+1}} \, dt, so du=dtdu = dt and v=12n(t2+c2)nv = \frac{-1}{2n(t^2 + c^2)^n}:

t2(t2+c2)n+1dt=[t2n(t2+c2)n]+12ndt(t2+c2)n\int_{-\infty}^{\infty} \frac{t^2}{(t^2 + c^2)^{n+1}} \, dt = \left[ \frac{-t}{2n(t^2 + c^2)^n} \right]_{-\infty}^{\infty} + \frac{1}{2n} \int_{-\infty}^{\infty} \frac{dt}{(t^2 + c^2)^n}

The boundary term vanishes (the numerator is O(t)O(t) while the denominator is O(t2n)O(t^{2n}) with n1n \geq 1), so:

t2(t2+c2)n+1dt=12nIn\int_{-\infty}^{\infty} \frac{t^2}{(t^2 + c^2)^{n+1}} \, dt = \frac{1}{2n} I_n

Substituting back into ()(*):

In=(1+c2)In+112nIn=c2In+1+2n12nInI_n = (1 + c^2) I_{n+1} - \frac{1}{2n} I_n = c^2 I_{n+1} + \frac{2n - 1}{2n} I_n

Rearranging:

In2n12nIn=c2In+1I_n - \frac{2n - 1}{2n} I_n = c^2 I_{n+1}

12nIn=c2In+1\frac{1}{2n} I_n = c^2 I_{n+1}

(2n1)In=2nc2In+1(2n - 1) I_n = 2n c^2 I_{n+1}

That is, 2n(ba2)In+1=(2n1)In2n(b - a^2) I_{n+1} = (2n - 1) I_n.

Part (iii)

We prove by induction on nn that

In=π22n2(ba2)n1/2(2n2n1)I_n = \frac{\pi}{2^{2n-2}(b - a^2)^{n - 1/2}} \binom{2n - 2}{n - 1}

Base case (n=1n = 1): The formula gives π20(ba2)1/2(00)=πba2\frac{\pi}{2^0 (b - a^2)^{1/2}} \binom{0}{0} = \frac{\pi}{\sqrt{b - a^2}}, which matches part (i).

Inductive step: Assume the formula holds for some n1n \geq 1. By part (ii):

In+1=2n12n(ba2)In=2n12n(ba2)π22n2(ba2)n1/2(2n2n1)I_{n+1} = \frac{2n - 1}{2n(b - a^2)} I_n = \frac{2n - 1}{2n(b - a^2)} \cdot \frac{\pi}{2^{2n-2}(b - a^2)^{n - 1/2}} \binom{2n - 2}{n - 1}

Combine the powers of (ba2)(b - a^2):

(ba2)(n1/2)1=(ba2)(n+1/2)(b - a^2)^{-(n - 1/2) - 1} = (b - a^2)^{-(n + 1/2)}

Combine the powers of 22:

12n22n2=122n1n\frac{1}{2n \cdot 2^{2n-2}} = \frac{1}{2^{2n-1} \cdot n}

So:

In+1=π(2n1)22n1n(ba2)n+1/2(2n2n1)I_{n+1} = \frac{\pi (2n - 1)}{2^{2n-1} \cdot n \cdot (b - a^2)^{n + 1/2}} \binom{2n - 2}{n - 1}

We need to show this equals π22n(ba2)n+1/2(2nn)\frac{\pi}{2^{2n}(b - a^2)^{n+1/2}} \binom{2n}{n}, i.e., we need:

(2n1)22n1n(2n2n1)=122n(2nn)\frac{(2n - 1)}{2^{2n-1} \cdot n} \binom{2n - 2}{n - 1} = \frac{1}{2^{2n}} \binom{2n}{n}

Multiply both sides by 22n2^{2n}:

2(2n1)n(2n2n1)=(2nn)\frac{2(2n - 1)}{n} \binom{2n - 2}{n - 1} = \binom{2n}{n}

Expand the right side:

(2nn)=(2n)!n!n!=2n(2n1)n2(2n2)!(n1)!(n1)!=2(2n1)n(2n2n1)\binom{2n}{n} = \frac{(2n)!}{n! \, n!} = \frac{2n(2n - 1)}{n^2} \cdot \frac{(2n - 2)!}{(n - 1)! \, (n - 1)!} = \frac{2(2n - 1)}{n} \binom{2n - 2}{n - 1}

This confirms the identity. By induction, the formula holds for all positive integers nn.

Examiner Notes

这是最常被尝试的题目,超过93%的考生作答,成功率也较高(第二高),平均得分约三分之二。多数考生顺利完成(i)和(iii),但有各种代数错误,部分考生在(i)中忘记代入积分限。许多考生在(ii)中使用(i)的换元后开始,但面对所得积分时卡住。


Topic: 坐标几何 (Coordinate Geometry)  |  Difficulty: Challenging  |  Marks: 20

2 The distinct points P(ap2,2ap)P(ap^2, 2ap), Q(aq2,2aq)Q(aq^2, 2aq) and R(ar2,2ar)R(ar^2, 2ar) lie on the parabola y2=4axy^2 = 4ax, where a>0a > 0. The points are such that the normal to the parabola at QQ and the normal to the parabola at RR both pass through PP.

(i) Show that q2+qp+2=0q^2 + qp + 2 = 0 .

(ii) Show that QRQR passes through a certain point that is independent of the choice of PP.

(iii) Let TT be the point of intersection of OPOP and QRQR, where OO is the coordinate origin. Show that TT lies on a line that is independent of the choice of PP.

Show further that the distance from the xx-axis to TT is less than a2\frac{a}{\sqrt{2}} .

Hint

Working parametrically with x=at2,y=2atx = at^2, y = 2at gives a normal as tx+y=at3+2attx + y = at^3 + 2at and imposing that this passes through (ap2,2ap)(ap^2, 2ap) yields t2+tp+2=0t^2 + tp + 2 = 0 () which has roots qq and rr, the former giving (i). As a consequence, r+q=pr + q = -p and rq=2rq = 2, so that QR, 2x(r+q)y+2aqr=02x - (r + q)y + 2aqr = 0, simplifies to 2x+py+4a=02x + py + 4a = 0, and thus passes through (2a,0)(-2a, 0) for (ii). T can be shown to be (a,2ap)(-a, \frac{-2a}{p}), which, of course, lies on x=ax = -a, and as () had two real distinct roots, qq and rr, p28>0p^2 - 8 > 0, which yields 2ap<a2|\frac{-2a}{p}| < \frac{a}{\sqrt{2}}.

Model Solution

Part (i)

The parabola is y2=4axy^2 = 4ax. Differentiating implicitly: 2ydydx=4a2y \frac{dy}{dx} = 4a, so dydx=2ay\frac{dy}{dx} = \frac{2a}{y}.

At the point (at2,2at)(at^2, 2at) on the parabola, the slope of the tangent is 2a2at=1t\frac{2a}{2at} = \frac{1}{t}, so the slope of the normal is t-t.

The normal at (at2,2at)(at^2, 2at) is:

y2at=t(xat2)y - 2at = -t(x - at^2)

For the normal at Q(aq2,2aq)Q(aq^2, 2aq) to pass through P(ap2,2ap)P(ap^2, 2ap), substitute x=ap2x = ap^2, y=2apy = 2ap:

2ap2aq=q(ap2aq2)2ap - 2aq = -q(ap^2 - aq^2)

2a(pq)=qa(pq)(p+q)2a(p - q) = -qa(p - q)(p + q)

Since PP and QQ are distinct points, pqp \neq q, so we may divide both sides by a(pq)a(p - q):

2=q(p+q)2 = -q(p + q)

q2+qp+2=0q^2 + qp + 2 = 0

Part (ii)

Since both qq and rr satisfy t2+pt+2=0t^2 + pt + 2 = 0, by Vieta’s formulas:

q+r=p,qr=2q + r = -p, \qquad qr = 2

The line QRQR passes through Q(aq2,2aq)Q(aq^2, 2aq) and R(ar2,2ar)R(ar^2, 2ar). Its slope is:

2aq2araq2ar2=2a(qr)a(qr)(q+r)=2q+r\frac{2aq - 2ar}{aq^2 - ar^2} = \frac{2a(q - r)}{a(q - r)(q + r)} = \frac{2}{q + r}

The equation of QRQR is:

y2aq=2q+r(xaq2)y - 2aq = \frac{2}{q + r}(x - aq^2)

Substituting q+r=pq + r = -p and qr=2qr = 2:

y2aq=2p(xaq2)y - 2aq = \frac{2}{-p}(x - aq^2)

p(y2aq)=2(xaq2)-p(y - 2aq) = 2(x - aq^2)

py+2apq=2x2aq2-py + 2apq = 2x - 2aq^2

2x+py=2aq2+2apq=2aq(q+p)2x + py = 2aq^2 + 2apq = 2aq(q + p)

From q2+pq+2=0q^2 + pq + 2 = 0, we get q(q+p)=2q(q + p) = -2, so:

2x+py=2a(2)=4a2x + py = 2a(-2) = -4a

2x+py+4a=02x + py + 4a = 0

This line passes through (2a,0)(-2a, 0) regardless of PP, since 2(2a)+p(0)+4a=02(-2a) + p(0) + 4a = 0.

Part (iii)

The line OPOP passes through O(0,0)O(0, 0) and P(ap2,2ap)P(ap^2, 2ap), so its equation is y=2pxy = \frac{2}{p}x, i.e., py=2xpy = 2x.

To find the intersection TT with QRQR: 2x+py+4a=02x + py + 4a = 0. Substitute 2x=py2x = py:

py+py+4a=0    2py=4a    y=2appy + py + 4a = 0 \implies 2py = -4a \implies y = \frac{-2a}{p}

Then x=py2=p(2a/p)2=ax = \frac{py}{2} = \frac{p \cdot (-2a/p)}{2} = -a.

So T=(a,2ap)T = \left(-a, \frac{-2a}{p}\right), which lies on the line x=ax = -a. This line is independent of PP.

Showing the distance bound: The equation t2+pt+2=0t^2 + pt + 2 = 0 must have two distinct real roots qq and rr (since QQ and RR are distinct points). The discriminant must be positive:

p28>0    p2>8    p>22p^2 - 8 > 0 \implies p^2 > 8 \implies |p| > 2\sqrt{2}

The distance from TT to the xx-axis is:

2ap=2ap\left| \frac{-2a}{p} \right| = \frac{2a}{|p|}

Since p>22|p| > 2\sqrt{2}:

2ap<2a22=a2\frac{2a}{|p|} < \frac{2a}{2\sqrt{2}} = \frac{a}{\sqrt{2}}

Therefore the distance from TT to the xx-axis is less than a2\frac{a}{\sqrt{2}}.

Examiner Notes

约四分之三的考生尝试此题,成功率中等。多数人使用隐函数求导而非参数方法,能找到切线、法线和弦的方程但不总是因式分解简化。那些能因式分解的考生进展良好,而不能的则在(i)上挣扎。其他弱点包括未能利用 r+qr+qrqrq 的关系,从而找不到(ii)中的固定点。在最后一部分,对不等式开方后只考虑正数情况并不罕见。


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

3 (i) Given that x32(x+1)2exdx=P(x)Q(x)ex+constant,\int \frac{x^3 - 2}{(x + 1)^2} \text{e}^x \text{d}x = \frac{P(x)}{Q(x)} \text{e}^x + \text{constant} , where P(x)P(x) and Q(x)Q(x) are polynomials, show that Q(x)Q(x) has a factor of x+1x + 1.

Show also that the degree of P(x)P(x) is exactly one more than the degree of Q(x)Q(x), and find P(x)P(x) in the case Q(x)=x+1Q(x) = x + 1.

(ii) Show that there are no polynomials P(x)P(x) and Q(x)Q(x) such that 1x+1exdx=P(x)Q(x)ex+constant.\int \frac{1}{x + 1} \text{e}^x \text{d}x = \frac{P(x)}{Q(x)} \text{e}^x + \text{constant} . You need consider only the case when P(x)P(x) and Q(x)Q(x) have no common factors.

Hint

Differentiating, multiplying by denominators and dividing by the exponential function, gives Q(P+P)PQ2=(x32)Q2Q(P + P') - PQ'^2 = (x^3 - 2)Q^2 which, invoking the factor theorem, gives the first required result. Denoting the degree of P by pp and that of Q by qq in this expression yields p+q+2=2q+3p + q + 2 = 2q + 3 and hence the desired result in (i). Furthermore, in the given case, substitution in the same result and postulating P(x)=ax2+bx+cP(x) = ax^2 + bx + c yields consistent equations for a,ba, b and cc and thus P(x)=x22xP(x) = x^2 - 2x.

For part (ii), commencing as in part (i) demonstrates again that Q has a factor (x+1)(x + 1) as Q(P+P)PQ=Q2Q(P + P') - PQ' = Q^2. Supposing Q(x)=(x+1)nS(x)Q(x) = (x + 1)^n S(x), where n2n \ge 2 and S(1)0S(-1) \ne 0, with P(1)0P(-1) \ne 0 and substituting in the expression already derived leads to a contradiction.

Model Solution

Part (i)

We start by noting that if P(x)Q(x)ex\frac{P(x)}{Q(x)}e^x differentiates to x32(x+1)2ex\frac{x^3 - 2}{(x+1)^2}e^x, then by the product rule:

ddx[PQex]=(PQ+PQPQQ2)ex=PQ+PQPQQ2ex\frac{d}{dx}\left[\frac{P}{Q}e^x\right] = \left(\frac{P}{Q} + \frac{P'Q - PQ'}{Q^2}\right)e^x = \frac{PQ + P'Q - PQ'}{Q^2}\,e^x

Setting this equal to the integrand:

PQ+PQPQQ2=x32(x+1)2(...)\frac{PQ + P'Q - PQ'}{Q^2} = \frac{x^3 - 2}{(x+1)^2} \qquad \text{(...)}

Cross-multiplying:

(x32)Q2=(x+1)2[PQ+PQPQ]=(x+1)2Q[P+P](x+1)2PQ(...)(x^3 - 2)\,Q^2 = (x+1)^2\bigl[PQ + P'Q - PQ'\bigr] = (x+1)^2\,Q\bigl[P + P'\bigr] - (x+1)^2\,PQ' \qquad \text{(...)}

To show QQ has a factor of x+1x + 1: substitute x=1x = -1 into the equation (x32)Q2=Q2(P+P)QQP(x^3 - 2)Q^2 = Q^2(P + P') - QQ'P (which follows from cross-multiplying the original identity differently — multiply both sides by Q2(x+1)2Q^2(x+1)^2 and divide by (x+1)2(x+1)^2):

(x32)Q=Q(P+P)QP(...)(x^3 - 2)Q = Q(P + P') - Q'P \qquad \text{(...)}

(This comes from dividing both sides of (x32)Q2=Q2(P+P)QQP(x^3-2)Q^2 = Q^2(P+P') - QQ'P by QQ, noting Q≢0Q \not\equiv 0.)

At x=1x = -1:

(12)Q(1)=Q(1)[P(1)+P(1)]Q(1)P(1)(-1 - 2)\,Q(-1) = Q(-1)\bigl[P(-1) + P'(-1)\bigr] - Q'(-1)\,P(-1)

3Q(1)=Q(1)[P(1)+P(1)]Q(1)P(1)-3\,Q(-1) = Q(-1)\bigl[P(-1) + P'(-1)\bigr] - Q'(-1)\,P(-1)

This is a relation between values at x=1x = -1 and does not immediately force Q(1)=0Q(-1) = 0. Instead, we use equation (**) more carefully.

From (x32)Q=Q(P+P)QP(x^3 - 2)Q = Q(P + P') - Q'P, rearrange:

Q[(x32)PP]=QP(...*)Q\bigl[(x^3 - 2) - P - P'\bigr] = -Q'P \qquad \text{(...*)}

If Q(1)0Q(-1) \neq 0, then at x=1x = -1:

Q(1)[3P(1)P(1)]=Q(1)P(1)Q(-1)\bigl[-3 - P(-1) - P'(-1)\bigr] = -Q'(-1)\,P(-1)

This is just a constraint on the values — it does not lead to a contradiction on its own. However, we can establish the result by a different route. Differentiate the given identity directly.

Suppose P(x)Q(x)ex\frac{P(x)}{Q(x)}e^x has derivative x32(x+1)2ex\frac{x^3 - 2}{(x+1)^2}e^x. Write Q(x)=(x+1)nS(x)Q(x) = (x+1)^n S(x) where S(1)0S(-1) \neq 0 and n0n \geq 0. We want to show n1n \geq 1.

From the equation (x32)Q=Q(P+P)QP(x^3 - 2)Q = Q(P + P') - Q'P:

The left side has a zero of order nn at x=1x = -1 (from the factor QQ).

On the right side:

  • Q(P+P)Q(P + P') has a zero of order n\geq n at x=1x = -1.
  • QPQ'P: since Q=(x+1)nSQ = (x+1)^n S, we have Q=n(x+1)n1S+(x+1)nSQ' = n(x+1)^{n-1}S + (x+1)^n S', so QQ' has a zero of order n1n - 1 at x=1x = -1 (assuming n1n \geq 1; if n=0n = 0, Q=SQ' = S' has no guaranteed zero).

If n=0n = 0: the left side is (x32)S(x)(x^3 - 2)S(x), which at x=1x = -1 gives 3S(1)0-3S(-1) \neq 0. The right side is S(P+P)SPS(P + P') - S'P. At x=1x = -1: S(1)[P(1)+P(1)]S(1)P(1)S(-1)[P(-1) + P'(-1)] - S'(-1)P(-1). Setting equal: 3S(1)=S(1)[P(1)+P(1)]S(1)P(1)-3S(-1) = S(-1)[P(-1) + P'(-1)] - S'(-1)P(-1). This can be satisfied, so we need a different argument.

We use the degree argument instead. From (x32)Q=Q(P+P)QP(x^3 - 2)Q = Q(P + P') - Q'P, let degP=p\deg P = p, degQ=q\deg Q = q. The left side has degree q+3q + 3. On the right side, Q(P+P)Q(P + P') has degree q+pq + p (the leading terms of PP and PP' do not cancel since they have different degrees), and QPQ'P has degree (q1)+p=q+p1(q - 1) + p = q + p - 1. So the right side has degree q+pq + p.

Therefore q+3=q+pq + 3 = q + p, giving p=3p = 3.

Now, from equation (*), cross-multiplying the original identity PQ+PQPQQ2=x32(x+1)2\frac{PQ + P'Q - PQ'}{Q^2} = \frac{x^3 - 2}{(x+1)^2}:

(PQ+PQPQ)(x+1)2=(x32)Q2(PQ + P'Q - PQ')(x+1)^2 = (x^3 - 2)Q^2

At x=1x = -1: 0=(3)Q(1)20 = (-3)Q(-1)^2, so Q(1)=0Q(-1) = 0.

This means Q(1)=0Q(-1) = 0, so (x+1)(x + 1) divides Q(x)Q(x).

Showing degP=degQ+1\deg P = \deg Q + 1: From the degree analysis above, p=3p = 3 regardless of qq. Wait — let me redo this more carefully from the cross-multiplied form.

From (PQ+PQPQ)(x+1)2=(x32)Q2(PQ + P'Q - PQ')(x+1)^2 = (x^3 - 2)Q^2:

  • LHS degree: max(deg(PQ),deg(PQ),deg(PQ))+2=max(p+q,p+q1,p+q1)+2=p+q+2\max(\deg(PQ), \deg(P'Q), \deg(PQ')) + 2 = \max(p+q, p+q-1, p+q-1) + 2 = p + q + 2.
  • RHS degree: 3+2q3 + 2q.

So p+q+2=2q+3p + q + 2 = 2q + 3, giving p=q+1p = q + 1.

Finding P(x)P(x) when Q(x)=x+1Q(x) = x + 1: Here q=1q = 1, so p=2p = 2. Write P(x)=ax2+bx+cP(x) = ax^2 + bx + c.

From equation (**): (x32)(x+1)=(x+1)(P+P)P(x^3 - 2)(x+1) = (x+1)(P + P') - P

where P=2ax+bP' = 2ax + b, so P+P=ax2+(2a+b)x+(b+c)P + P' = ax^2 + (2a + b)x + (b + c).

Left side: x4+x32x2x^4 + x^3 - 2x - 2.

Right side: (x+1)[ax2+(2a+b)x+(b+c)](ax2+bx+c)(x+1)[ax^2 + (2a+b)x + (b+c)] - (ax^2 + bx + c).

Expanding (x+1)[ax2+(2a+b)x+(b+c)](x+1)[ax^2 + (2a+b)x + (b+c)]:

=ax3+(2a+b)x2+(b+c)x+ax2+(2a+b)x+(b+c)= ax^3 + (2a+b)x^2 + (b+c)x + ax^2 + (2a+b)x + (b+c) =ax3+(3a+b)x2+(2a+2b+c)x+(b+c)= ax^3 + (3a+b)x^2 + (2a+2b+c)x + (b+c)

Subtracting P=ax2+bx+cP = ax^2 + bx + c:

ax3+(3a+b)x2+(2a+2b+c)x+(b+c)ax2bxcax^3 + (3a+b)x^2 + (2a+2b+c)x + (b+c) - ax^2 - bx - c =ax3+(2a+b)x2+(2a+b+c)x+b= ax^3 + (2a+b)x^2 + (2a+b+c)x + b

Matching coefficients with x4+x32x2x^4 + x^3 - 2x - 2:

  • x4x^4: 0=10 = 1??

This is a contradiction, so let me re-examine. The issue is that the equation (**) came from dividing by QQ, but I need to use the original cross-multiplied form.

From (PQ+PQPQ)(x+1)2=(x32)Q2(PQ + P'Q - PQ')(x+1)^2 = (x^3 - 2)Q^2 with Q=x+1Q = x + 1:

LHS: [(x+1)P+(x+1)PP](x+1)2[(x+1)P + (x+1)P' - P](x+1)^2

=[xP+P+xP+PP](x+1)2= [xP + P + xP' + P' - P](x+1)^2

=[xP+xP+P](x+1)2= [xP + xP' + P'](x+1)^2

=[x(P+P)+P](x+1)2= [x(P + P') + P'](x+1)^2

With P=ax2+bx+cP = ax^2 + bx + c:

P+P=ax2+(2a+b)x+(b+c)P + P' = ax^2 + (2a+b)x + (b+c)

x(P+P)=ax3+(2a+b)x2+(b+c)xx(P + P') = ax^3 + (2a+b)x^2 + (b+c)x

P=2ax+bP' = 2ax + b

x(P+P)+P=ax3+(2a+b)x2+(b+c+2a)x+bx(P+P') + P' = ax^3 + (2a+b)x^2 + (b+c+2a)x + b

Multiply by (x+1)2=x2+2x+1(x+1)^2 = x^2 + 2x + 1:

[ax3+(2a+b)x2+(2a+b+c)x+b](x2+2x+1)[ax^3 + (2a+b)x^2 + (2a+b+c)x + b](x^2 + 2x + 1)

This is getting unwieldy. Let me use a cleaner approach.

Cleaner approach for finding PP: Since ddx[PQex]=x32(x+1)2ex\frac{d}{dx}\left[\frac{P}{Q}e^x\right] = \frac{x^3-2}{(x+1)^2}e^x with Q=x+1Q = x+1, we need:

Px+1+P(x+1)P(x+1)2=x32(x+1)2\frac{P}{x+1} + \frac{P'(x+1) - P}{(x+1)^2} = \frac{x^3 - 2}{(x+1)^2}

Multiply through by (x+1)2(x+1)^2:

P(x+1)+P(x+1)P=x32P(x+1) + P'(x+1) - P = x^3 - 2

xP+Px+P=x32xP + P'x + P' = x^3 - 2

x(P+P)+P=x32(***)x(P + P') + P' = x^3 - 2 \qquad \text{(***)}

With P=ax2+bx+cP = ax^2 + bx + c, P=2ax+bP' = 2ax + b:

P+P=ax2+(2a+b)x+(b+c)P + P' = ax^2 + (2a+b)x + (b+c)

x(P+P)=ax3+(2a+b)x2+(b+c)xx(P+P') = ax^3 + (2a+b)x^2 + (b+c)x

x(P+P)+P=ax3+(2a+b)x2+(b+c+2a)x+bx(P+P') + P' = ax^3 + (2a+b)x^2 + (b+c+2a)x + b

Matching with x32x^3 - 2:

  • x3x^3: a=1a = 1
  • x2x^2: 2a+b=0b=22a + b = 0 \Rightarrow b = -2
  • x1x^1: b+c+2a=02+c+2=0c=0b + c + 2a = 0 \Rightarrow -2 + c + 2 = 0 \Rightarrow c = 0
  • x0x^0: b=2b = -2

So P(x)=x22xP(x) = x^2 - 2x.

Verification: ddx[x22xx+1ex]\frac{d}{dx}\left[\frac{x^2 - 2x}{x+1}e^x\right].

Write x22xx+1=x3+3x+1\frac{x^2 - 2x}{x+1} = x - 3 + \frac{3}{x+1} (by polynomial division). Then:

ddx[(x3+3x+1)ex]=(13(x+1)2+x3+3x+1)ex\frac{d}{dx}\left[\left(x - 3 + \frac{3}{x+1}\right)e^x\right] = \left(1 - \frac{3}{(x+1)^2} + x - 3 + \frac{3}{x+1}\right)e^x

=(x2+3x+13(x+1)2)ex= \left(x - 2 + \frac{3}{x+1} - \frac{3}{(x+1)^2}\right)e^x

=(x2)(x+1)2+3(x+1)3(x+1)2ex= \frac{(x-2)(x+1)^2 + 3(x+1) - 3}{(x+1)^2}\,e^x

Numerator: (x2)(x2+2x+1)+3x+33=x3+2x2+x2x24x2+3x=x32(x-2)(x^2+2x+1) + 3x + 3 - 3 = x^3 + 2x^2 + x - 2x^2 - 4x - 2 + 3x = x^3 - 2. ✓


Part (ii)

Suppose for contradiction that 1x+1exdx=PQex+C\int \frac{1}{x+1}e^x\,dx = \frac{P}{Q}e^x + C where P,QP, Q are polynomials with no common factors.

Differentiating and clearing denominators as before:

P(x+1)+P(x+1)P=Qfrom PQ+PQPQQ2=1x+1P(x+1) + P'(x+1) - P = Q \qquad \text{from } \frac{P}{Q} + \frac{P'Q - PQ'}{Q^2} = \frac{1}{x+1}

More precisely, PQ+PQPQQ2=1x+1\frac{PQ + P'Q - PQ'}{Q^2} = \frac{1}{x+1}, so:

(PQ+PQPQ)(x+1)=Q2(...)(PQ + P'Q - PQ')(x+1) = Q^2 \qquad \text{(...)}

At x=1x = -1: 0=Q(1)20 = Q(-1)^2, so Q(1)=0Q(-1) = 0. Write Q(x)=(x+1)nS(x)Q(x) = (x+1)^n S(x) with S(1)0S(-1) \neq 0, n1n \geq 1.

Since PP and QQ share no common factors, P(1)0P(-1) \neq 0.

We compute the order of vanishing of each term in PQ+PQPQPQ + P'Q - PQ' at x=1x = -1:

  • PQ=P(x+1)nSPQ = P \cdot (x+1)^n S vanishes to order nn.
  • PQ=P(x+1)nSP'Q = P' \cdot (x+1)^n S vanishes to order n\geq n.
  • PQ=P[n(x+1)n1S+(x+1)nS]PQ' = P\bigl[n(x+1)^{n-1}S + (x+1)^n S'\bigr]: since P(1)0P(-1) \neq 0, this vanishes to order exactly n1n - 1.

Therefore PQ+PQPQPQ + P'Q - PQ' vanishes to order exactly n1n - 1 (the PQPQ' term dominates).

The left side of (*), (PQ+PQPQ)(x+1)(PQ + P'Q - PQ')(x+1), vanishes to order nn.

The right side, Q2=(x+1)2nS2Q^2 = (x+1)^{2n}S^2, vanishes to order 2n2n.

For the equation to hold: n=2nn = 2n, giving n=0n = 0. But n1n \geq 1, a contradiction.

Therefore no such polynomials PP and QQ exist.

Examiner Notes

比 Q2 稍少人尝试但成功率略高。两部分中,考生成功将右侧表达式的导数等于被积函数(消去指数函数)。第一部分中,许多人求出了给定 Q(x)Q(x) 情况下的 P(x)P(x),但证明 degP\deg PdegQ\deg Q 大一以及(ii)中不存在这样的多项式时出现许多不合逻辑的步骤。许多人如果乘开分母并使用余式/因式定理会更好,而非尝试基于有理表达式次数的论证。


Topic: 级数与数列 (Series and Sequences)  |  Difficulty: Challenging  |  Marks: 20

4 (i) By considering 11+xr11+xr+1\frac{1}{1 + x^r} - \frac{1}{1 + x^{r+1}} for x1|x| \neq 1, simplify r=1Nxr(1+xr)(1+xr+1).\sum_{r=1}^{N} \frac{x^r}{(1 + x^r)(1 + x^{r+1})} . Show that, for x<1|x| < 1, r=1xr(1+xr)(1+xr+1)=x1x2.\sum_{r=1}^{\infty} \frac{x^r}{(1 + x^r)(1 + x^{r+1})} = \frac{x}{1 - x^2} . (ii) Deduce that r=1sech(ry)sech((r+1)y)=2eycosech(2y)\sum_{r=1}^{\infty} \text{sech}(ry) \text{sech}((r + 1)y) = 2\text{e}^{-y} \text{cosech}(2y) for y>0y > 0.

Hence simplify r=sech(ry)sech((r+1)y),\sum_{r=-\infty}^{\infty} \text{sech}(ry) \text{sech}((r + 1)y) , for y>0y > 0.

Hint

The considered expression equates to (x1)xr(1+xr)(1+xr+1)\frac{(x-1)x^r}{(1+x^r)(1+x^{r+1})} and so, by the method of differences, r=1Nxr(1+xr)(1+xr+1)=1(x1)[11+x11+xN+1]\sum_{r=1}^N \frac{x^r}{(1+x^r)(1+x^{r+1})} = \frac{1}{(x-1)} \left[ \frac{1}{1+x} - \frac{1}{1+x^{N+1}} \right], and letting NN \to \infty, the desired result is obtained. Writing sech(ry)\text{sech}(ry) as 2ery1+e2ry\frac{2e^{-ry}}{1+e^{-2ry}} and similarly sech((r+1)y)\text{sech}((r + 1)y), the result of part (i) with x=e2yx = e^{-2y} can be used to obtain the result. Care needs to be taken to write r=sech(ry)sech((r+1)y)\sum_{r=-\infty}^{\infty} \text{sech}(ry) \text{sech}((r + 1)y) as 2[r=1sech(ry)sech((r+1)y)+sech y]2[\sum_{r=1}^{\infty} \text{sech}(ry) \text{sech}((r + 1)y) + \text{sech } y] which with the previous deduction of (ii) can be simplified to 2 cosech y2 \text{ cosech } y.

Model Solution

Part (i)

Consider the expression suggested in the problem:

11+xr11+xr+1\frac{1}{1 + x^r} - \frac{1}{1 + x^{r+1}}

Combining over a common denominator:

=(1+xr+1)(1+xr)(1+xr)(1+xr+1)=xr+1xr(1+xr)(1+xr+1)=xr(x1)(1+xr)(1+xr+1)= \frac{(1 + x^{r+1}) - (1 + x^r)}{(1 + x^r)(1 + x^{r+1})} = \frac{x^{r+1} - x^r}{(1 + x^r)(1 + x^{r+1})} = \frac{x^r(x - 1)}{(1 + x^r)(1 + x^{r+1})}

Therefore:

xr(1+xr)(1+xr+1)=1x1(11+xr11+xr+1)\frac{x^r}{(1 + x^r)(1 + x^{r+1})} = \frac{1}{x - 1}\left(\frac{1}{1 + x^r} - \frac{1}{1 + x^{r+1}}\right)

Summing from r=1r = 1 to NN (telescoping):

r=1Nxr(1+xr)(1+xr+1)=1x1r=1N(11+xr11+xr+1)\sum_{r=1}^{N} \frac{x^r}{(1 + x^r)(1 + x^{r+1})} = \frac{1}{x - 1}\sum_{r=1}^{N}\left(\frac{1}{1 + x^r} - \frac{1}{1 + x^{r+1}}\right)

=1x1(11+x11+xN+1)= \frac{1}{x - 1}\left(\frac{1}{1 + x} - \frac{1}{1 + x^{N+1}}\right)

For x<1|x| < 1: as NN \to \infty, xN+10x^{N+1} \to 0, so 11+xN+11\frac{1}{1 + x^{N+1}} \to 1.

r=1xr(1+xr)(1+xr+1)=1x1(11+x1)=1x11(1+x)1+x=1x1x1+x\sum_{r=1}^{\infty} \frac{x^r}{(1 + x^r)(1 + x^{r+1})} = \frac{1}{x - 1}\left(\frac{1}{1 + x} - 1\right) = \frac{1}{x - 1} \cdot \frac{1 - (1 + x)}{1 + x} = \frac{1}{x - 1} \cdot \frac{-x}{1 + x}

=x(x1)(1+x)=x(1x)(1+x)=x1x2= \frac{-x}{(x - 1)(1 + x)} = \frac{x}{(1 - x)(1 + x)} = \frac{x}{1 - x^2}


Part (ii)

Using the definition coshθ=eθ+eθ2\cosh \theta = \frac{e^\theta + e^{-\theta}}{2}, we write:

sech(ry)=2ery+ery=2ery1+e2ry\text{sech}(ry) = \frac{2}{e^{ry} + e^{-ry}} = \frac{2e^{-ry}}{1 + e^{-2ry}}

sech((r+1)y)=2e(r+1)y1+e2(r+1)y\text{sech}((r+1)y) = \frac{2e^{-(r+1)y}}{1 + e^{-2(r+1)y}}

Therefore:

sech(ry)sech((r+1)y)=4e(2r+1)y(1+e2ry)(1+e2(r+1)y)\text{sech}(ry)\,\text{sech}((r+1)y) = \frac{4e^{-(2r+1)y}}{(1 + e^{-2ry})(1 + e^{-2(r+1)y})}

Set x=e2yx = e^{-2y}. Since y>0y > 0, we have 0<x<10 < x < 1, so x<1|x| < 1. Then:

e2ry=xr,e(2r+1)y=eyxre^{-2ry} = x^r, \qquad e^{-(2r+1)y} = e^{-y} \cdot x^r

sech(ry)sech((r+1)y)=4eyxr(1+xr)(1+xr+1)\text{sech}(ry)\,\text{sech}((r+1)y) = \frac{4e^{-y}\,x^r}{(1 + x^r)(1 + x^{r+1})}

Summing from r=1r = 1 to \infty and using part (i):

r=1sech(ry)sech((r+1)y)=4eyr=1xr(1+xr)(1+xr+1)=4eyx1x2\sum_{r=1}^{\infty} \text{sech}(ry)\,\text{sech}((r+1)y) = 4e^{-y}\sum_{r=1}^{\infty} \frac{x^r}{(1 + x^r)(1 + x^{r+1})} = 4e^{-y} \cdot \frac{x}{1 - x^2}

Substituting x=e2yx = e^{-2y}:

=4eye2y1e4y=4e3y1e4y= \frac{4e^{-y} \cdot e^{-2y}}{1 - e^{-4y}} = \frac{4e^{-3y}}{1 - e^{-4y}}

Multiply numerator and denominator by e4ye^{4y}:

=4eye4y1=4ey(e2y1)(e2y+1)= \frac{4e^{y}}{e^{4y} - 1} = \frac{4e^{y}}{(e^{2y} - 1)(e^{2y} + 1)}

Now we express the right side 2eycosech(2y)2e^{-y}\,\text{cosech}(2y) in the same form:

2eycosech(2y)=2eysinh(2y)=2eye2ye2y2=4eye2ye2y2e^{-y}\,\text{cosech}(2y) = \frac{2e^{-y}}{\sinh(2y)} = \frac{2e^{-y}}{\frac{e^{2y} - e^{-2y}}{2}} = \frac{4e^{-y}}{e^{2y} - e^{-2y}}

Multiply numerator and denominator by e2ye^{2y}:

=4eye4y1= \frac{4e^{y}}{e^{4y} - 1}

These are identical, so:

r=1sech(ry)sech((r+1)y)=2eycosech(2y)\sum_{r=1}^{\infty} \text{sech}(ry)\,\text{sech}((r+1)y) = 2e^{-y}\,\text{cosech}(2y)


Hence result:

We evaluate r=sech(ry)sech((r+1)y)\sum_{r=-\infty}^{\infty} \text{sech}(ry)\,\text{sech}((r+1)y).

Since sech\text{sech} is an even function (sech(θ)=sech(θ)\text{sech}(-\theta) = \text{sech}(\theta)):

For r2r \leq -2, set k=(r+1)1k = -(r+1) \geq 1:

sech(ry)sech((r+1)y)=sech((k+1)y)sech(ky)\text{sech}(ry)\,\text{sech}((r+1)y) = \text{sech}((k+1)y)\,\text{sech}(ky)

So the terms for r=2,3,4,r = -2, -3, -4, \ldots match the terms for r=1,2,3,r = 1, 2, 3, \ldots.

The remaining terms are r=1r = -1 and r=0r = 0:

  • r=1r = -1: sech(y)sech(0)=sech(y)1=sech(y)\text{sech}(-y)\,\text{sech}(0) = \text{sech}(y) \cdot 1 = \text{sech}(y)
  • r=0r = 0: sech(0)sech(y)=1sech(y)=sech(y)\text{sech}(0)\,\text{sech}(y) = 1 \cdot \text{sech}(y) = \text{sech}(y)

Therefore:

r=sech(ry)sech((r+1)y)=2r=1sech(ry)sech((r+1)y)+2sech(y)\sum_{r=-\infty}^{\infty} \text{sech}(ry)\,\text{sech}((r+1)y) = 2\sum_{r=1}^{\infty} \text{sech}(ry)\,\text{sech}((r+1)y) + 2\,\text{sech}(y)

=22eycosech(2y)+2sech(y)= 2 \cdot 2e^{-y}\,\text{cosech}(2y) + 2\,\text{sech}(y)

=4eycosech(2y)+2sech(y)= 4e^{-y}\,\text{cosech}(2y) + 2\,\text{sech}(y)

We simplify. Using cosech(2y)=1sinh(2y)=12sinhycoshy\text{cosech}(2y) = \frac{1}{\sinh(2y)} = \frac{1}{2\sinh y\,\cosh y}:

4eycosech(2y)=4ey2sinhycoshy=2eysinhycoshy4e^{-y}\,\text{cosech}(2y) = \frac{4e^{-y}}{2\sinh y\,\cosh y} = \frac{2e^{-y}}{\sinh y\,\cosh y}

And 2sech(y)=2coshy2\,\text{sech}(y) = \frac{2}{\cosh y}.

=2eysinhycoshy+2coshy=2ey+2sinhysinhycoshy\sum = \frac{2e^{-y}}{\sinh y\,\cosh y} + \frac{2}{\cosh y} = \frac{2e^{-y} + 2\sinh y}{\sinh y\,\cosh y}

Now simplify 2ey+2sinhy2e^{-y} + 2\sinh y:

2ey+2eyey2=2ey+eyey=ey+ey=2coshy2e^{-y} + 2 \cdot \frac{e^y - e^{-y}}{2} = 2e^{-y} + e^y - e^{-y} = e^y + e^{-y} = 2\cosh y

Therefore:

r=sech(ry)sech((r+1)y)=2coshysinhycoshy=2sinhy=2cosech(y)\sum_{r=-\infty}^{\infty} \text{sech}(ry)\,\text{sech}((r+1)y) = \frac{2\cosh y}{\sinh y\,\cosh y} = \frac{2}{\sinh y} = 2\,\text{cosech}(y)

Examiner Notes

比 Q2 稍多人尝试(约五分之四),但成功率略低。第一部分(有明确提示)完成得不错,但 NN 趋于无穷的极限论证中记号很差。在(ii)中,相当多考生用正指数幂写出 sech(ry) 的表达式,导致无法应用(i)的结果。很少有考生完全简化最终结果,且许多人未能正确处理正负部分的求和,有些人只是简单地将从1到无穷的求和乘以2。


Topic: 组合与数论 (Combinatorics and Number Theory)  |  Difficulty: Hard  |  Marks: 20

5 (i) By considering the binomial expansion of (1+x)2m+1(1 + x)^{2m+1}, prove that

(2m+1m)<22m,\binom{2m+1}{m} < 2^{2m},

for any positive integer mm.

(ii) For any positive integers rr and ss with r<sr < s, Pr,sP_{r,s} is defined as follows: Pr,sP_{r,s} is the product of all the prime numbers greater than rr and less than or equal to ss, if there are any such primes numbers; if there are no such primes numbers, then Pr,s=1P_{r,s} = 1.

For example, P3,7=35P_{3,7} = 35, P7,10=1P_{7,10} = 1 and P14,18=17P_{14,18} = 17.

Show that, for any positive integer mm, Pm+1,2m+1P_{m+1, 2m+1} divides (2m+1m)\binom{2m+1}{m}, and deduce that

Pm+1,2m+1<22m.P_{m+1, 2m+1} < 2^{2m}.

(iii) Show that, if P1,k<4kP_{1,k} < 4^k for k=2,3,,2mk = 2, 3, \dots, 2m, then P1,2m+1<42m+1P_{1,2m+1} < 4^{2m+1}.

(iv) Prove that P1,n<4nP_{1,n} < 4^n for n2n \geqslant 2.

Hint

The binomial expansion, evaluated for x=1x = 1, appreciating that terms are symmetrical contains two terms equal to the LHS of the inequality, and so truncating to them gives double the required result in (i). Appreciating that (2m+1)!(m+1)!m!\frac{(2m+1)!}{(m+1)!m!} is an integer and that if m+1<p2m+1m + 1 < p \le 2m + 1, with p a prime, implies p divides the numerator and not the denominator of this expression and hence divides the integer then can be extended for all such primes yielding the result, with the deduction following from (i). For (iii), it can be shown that m+12mm + 1 \le 2m and writing P1,2m+1P_{1,2m+1} as P1,m+1Pm+1,2m+1P_{1,m+1} P_{m+1,2m+1}, combining the given result and (ii), the desired result is obtained. Part (iv) is obtained by use of strong induction with the supposition, P1,m<4mP_{1,m} < 4^m for all mkm \le k for some particular k2k \ge 2, and considering the cases k even and odd separately and making use of (iii).

Model Solution

Part (i)

Consider the binomial expansion of (1+x)2m+1(1+x)^{2m+1}:

(1+x)2m+1=k=02m+1(2m+1k)xk.(1+x)^{2m+1} = \sum_{k=0}^{2m+1} \binom{2m+1}{k} x^k.

Setting x=1x = 1:

22m+1=k=02m+1(2m+1k).2^{2m+1} = \sum_{k=0}^{2m+1} \binom{2m+1}{k}.

The binomial coefficients satisfy the symmetry (2m+1k)=(2m+12m+1k)\binom{2m+1}{k} = \binom{2m+1}{2m+1-k}. We pair the terms: the kk-th term with the (2m+1k)(2m+1-k)-th term. Since 2m+12m+1 is odd, every term has a distinct partner, and there are m+1m+1 pairs:

22m+1=2[(2m+10)+(2m+11)++(2m+1m)].2^{2m+1} = 2\left[\binom{2m+1}{0} + \binom{2m+1}{1} + \cdots + \binom{2m+1}{m}\right].

Since all binomial coefficients are strictly positive:

(2m+10)+(2m+11)++(2m+1m)>(2m+1m).\binom{2m+1}{0} + \binom{2m+1}{1} + \cdots + \binom{2m+1}{m} > \binom{2m+1}{m}.

Therefore:

22m+1=2[(2m+10)++(2m+1m)]>2(2m+1m),2^{2m+1} = 2\left[\binom{2m+1}{0} + \cdots + \binom{2m+1}{m}\right] > 2\binom{2m+1}{m},

which gives (2m+1m)<22m\binom{2m+1}{m} < 2^{2m}. (proved)\qquad \text{(proved)}

Part (ii)

Recall that (2m+1m)=(2m+1)!m!(m+1)!\binom{2m+1}{m} = \frac{(2m+1)!}{m!\,(m+1)!}.

Let pp be any prime with m+1<p2m+1m + 1 < p \leq 2m + 1. We show that p(2m+1m)p \mid \binom{2m+1}{m} by counting the multiplicity of pp in the numerator and denominator.

  • Since m+1<p2m+1m + 1 < p \leq 2m+1 and 2p>2(m+1)>2m+12p > 2(m+1) > 2m+1, the number pp appears exactly once in 1,2,,2m+11, 2, \ldots, 2m+1. So pp has multiplicity exactly 1 in (2m+1)!(2m+1)!.
  • Since p>m+1>mp > m + 1 > m, the prime pp does not divide any of 1,2,,m1, 2, \ldots, m, so pp has multiplicity 0 in m!m!.
  • Since p>m+1p > m + 1, the prime pp does not divide any of 1,2,,m+11, 2, \ldots, m+1, so pp has multiplicity 0 in (m+1)!(m+1)!.

Therefore pp has multiplicity 100=11 - 0 - 0 = 1 in (2m+1m)\binom{2m+1}{m}, so p(2m+1m)p \mid \binom{2m+1}{m}.

Since this holds for every prime pp with m+1<p2m+1m + 1 < p \leq 2m+1, and these primes are distinct, their product Pm+1,2m+1P_{m+1, 2m+1} divides (2m+1m)\binom{2m+1}{m}.

From part (i), (2m+1m)<22m\binom{2m+1}{m} < 2^{2m}. Since Pm+1,2m+1P_{m+1, 2m+1} is a positive divisor of (2m+1m)\binom{2m+1}{m}:

Pm+1,2m+1(2m+1m)<22m.(deduced)P_{m+1, 2m+1} \leq \binom{2m+1}{m} < 2^{2m}. \qquad \text{(deduced)}

Part (iii)

We decompose the product of all primes up to 2m+12m+1:

P1,2m+1=P1,m+1Pm+1,2m+1.P_{1,2m+1} = P_{1,m+1} \cdot P_{m+1,2m+1}.

This is valid because every prime 2m+1\leq 2m+1 is either m+1\leq m+1 or lies in (m+1,2m+1](m+1, 2m+1], and the two sets are disjoint.

Since m1m \geq 1, we have m+12m + 1 \geq 2 and m+12mm + 1 \leq 2m, so m+1m+1 lies in the range {2,3,,2m}\{2, 3, \ldots, 2m\}. By hypothesis:

P1,m+1<4m+1.P_{1,m+1} < 4^{m+1}.

By part (ii):

Pm+1,2m+1<22m=4m.P_{m+1,2m+1} < 2^{2m} = 4^m.

Multiplying:

P1,2m+1=P1,m+1Pm+1,2m+1<4m+14m=42m+1.(proved)P_{1,2m+1} = P_{1,m+1} \cdot P_{m+1,2m+1} < 4^{m+1} \cdot 4^m = 4^{2m+1}. \qquad \text{(proved)}

Part (iv)

We prove P1,n<4nP_{1,n} < 4^n for all n2n \geq 2 by strong induction.

Base cases.

P1,2=2<16=42P_{1,2} = 2 < 16 = 4^2.

P1,3=2×3=6<64=43P_{1,3} = 2 \times 3 = 6 < 64 = 4^3.

Inductive step. Suppose P1,k<4kP_{1,k} < 4^k for all kk with 2kn2 \leq k \leq n, where n3n \geq 3. We show P1,n+1<4n+1P_{1,n+1} < 4^{n+1}.

Case 1: n+1n + 1 is even. Write n+1=2mn + 1 = 2m with m2m \geq 2 (since n+14n + 1 \geq 4).

Since P1,sP_{1,s} is the product of primes s\leq s, increasing the upper bound from 2m12m - 1 to 2m2m either multiplies by a new prime (if 2m2m is prime) or leaves the product unchanged (if 2m2m is composite). In either case:

P1,2mP1,2m1.P_{1,2m} \leq P_{1,2m-1}.

Since 2m1322m - 1 \geq 3 \geq 2, the inductive hypothesis gives P1,2m1<42m1P_{1,2m-1} < 4^{2m-1}. Therefore:

P1,2mP1,2m1<42m1<42m.P_{1,2m} \leq P_{1,2m-1} < 4^{2m-1} < 4^{2m}.

Case 2: n+1n + 1 is odd. Write n+1=2m+1n + 1 = 2m + 1 with m2m \geq 2 (since n+15n + 1 \geq 5).

We decompose as in part (iii):

P1,2m+1=P1,m+1Pm+1,2m+1.P_{1,2m+1} = P_{1,m+1} \cdot P_{m+1,2m+1}.

Since m2m \geq 2, we have 3m+12mn3 \leq m + 1 \leq 2m \leq n, so m+1m + 1 lies in the range {2,3,,n}\{2, 3, \ldots, n\}. By the inductive hypothesis:

P1,m+1<4m+1.P_{1,m+1} < 4^{m+1}.

By part (ii):

Pm+1,2m+1<22m=4m.P_{m+1,2m+1} < 2^{2m} = 4^m.

Therefore:

P1,2m+1<4m+14m=42m+1.P_{1,2m+1} < 4^{m+1} \cdot 4^m = 4^{2m+1}.

In both cases, P1,n+1<4n+1P_{1,n+1} < 4^{n+1}. By strong induction, P1,n<4nP_{1,n} < 4^n for all n2n \geq 2. (proved)\qquad \text{(proved)}

Examiner Notes

约一半考生尝试此题,得分与 Q2 相似。二项展开及其对称性在第一部分处理得不错,(iii)的结果在(iv)中的应用也可以。但(ii)回答得不好,因为关于整除性的论证往往过于随意;在(iii)中,很少有人注意到需要 m+12mm+1 \le 2m 的条件才能使用(ii)的结果。


Topic: 双曲函数 (Hyperbolic Functions)  |  Difficulty: Challenging  |  Marks: 20

6 Show, by finding RR and γ\gamma, that Asinhx+BcoshxA \sinh x + B \cosh x can be written in the form Rcosh(x+γ)R \cosh(x + \gamma) if B>A>0B > A > 0. Determine the corresponding forms in the other cases that arise, for A>0A > 0, according to the value of BB.

Two curves have equations y=sechxy = \operatorname{sech} x and y=atanhx+by = a \tanh x + b, where a>0a > 0.

(i) In the case b>ab > a, show that if the curves intersect then the xx-coordinates of the points of intersection can be written in the form

±arcosh(1b2a2)artanhab.\pm \operatorname{arcosh} \left( \frac{1}{\sqrt{b^2 - a^2}} \right) - \operatorname{artanh} \frac{a}{b}.

(ii) Find the corresponding result in the case a>b>0a > b > 0.

(iii) Find necessary and sufficient conditions on aa and bb for the curves to intersect at two distinct points.

(iv) Find necessary and sufficient conditions on aa and bb for the curves to touch and, given that they touch, express the yy-coordinate of the point of contact in terms of aa.

Hint

Using Rcosh(x+γ)=R(coshxcoshγ+sinhxsinhγ)R \cosh(x + \gamma) = R(\cosh x \cosh \gamma + \sinh x \sinh \gamma), R=B2A2R = \sqrt{B^2 - A^2} and γ=tanh1AB\gamma = \tanh^{-1} \frac{A}{B} if B>A>0B > A > 0. If B=AB = A, then Asinhx+Bcoshx=AexA \sinh x + B \cosh x = Ae^x. If A<B<A-A < B < A, the expression can be written as Rsinh(x+γ)R \sinh(x + \gamma) with R=A2B2R = \sqrt{A^2 - B^2} and γ=tanh1BA\gamma = \tanh^{-1} \frac{B}{A}. If B=AB = -A, then Asinhx+Bcoshx=AexA \sinh x + B \cosh x = -Ae^{-x}, and if B<AB < -A, the expression can be written as Rcosh(x+γ)R \cosh(x + \gamma) with R=B2A2R = -\sqrt{B^2 - A^2} and γ=tanh1AB\gamma = \tanh^{-1} \frac{A}{B}.

For part (i), solving simultaneously gives asinhx+bcoshx=1a \sinh x + b \cosh x = 1, which gives the desired solutions using the first result of the question.

Similarly for part (ii) using the appropriate result, x=sinh1(1a2b2)tanh1bax = \sinh^{-1} \left( \frac{1}{\sqrt{a^2-b^2}} \right) - \tanh^{-1} \frac{b}{a}.

For (iii), we require that the conditions for (i) give two solutions, i.e. that b>ab > a and (1b2a2)>1\left( \frac{1}{\sqrt{b^2-a^2}} \right) > 1, and so a<b<a2+1a < b < \sqrt{a^2 + 1}, and vice versa, if this applies there are indeed two solutions.

For (iv), we require case (i) to give coincident solutions, i.e. b=a2+1b = \sqrt{a^2 + 1} and hence x=tanh1aa2+1x = -\tanh^{-1} \frac{a}{\sqrt{a^2+1}}, and so y=1a2+1y = \frac{1}{\sqrt{a^2+1}}. The reverse argument also applies.

Model Solution

Preamble: Combining Asinhx+BcoshxA \sinh x + B \cosh x

We expand Rcosh(x+γ)=Rcoshxcoshγ+RsinhxsinhγR \cosh(x + \gamma) = R \cosh x \cosh \gamma + R \sinh x \sinh \gamma and match coefficients:

A=Rsinhγ,B=Rcoshγ.A = R \sinh \gamma, \qquad B = R \cosh \gamma.

Dividing: tanhγ=AB\tanh \gamma = \frac{A}{B}, so γ=artanhAB\gamma = \operatorname{artanh} \frac{A}{B} (valid when A/B<1|A/B| < 1).

Using cosh2γsinh2γ=1\cosh^2 \gamma - \sinh^2 \gamma = 1:

B2R2A2R2=1    R2=B2A2.\frac{B^2}{R^2} - \frac{A^2}{R^2} = 1 \implies R^2 = B^2 - A^2.

Case 1: B>A>0B > A > 0. Then B2A2>0B^2 - A^2 > 0, so R=B2A2>0R = \sqrt{B^2 - A^2} > 0 and γ=artanhAB\gamma = \operatorname{artanh} \frac{A}{B}.

Asinhx+Bcoshx=B2A2cosh ⁣(x+artanhAB).A \sinh x + B \cosh x = \sqrt{B^2 - A^2} \, \cosh\!\left(x + \operatorname{artanh} \frac{A}{B}\right).

Case 2: B=A>0B = A > 0. Then Asinhx+Acoshx=A(sinhx+coshx)=AexA \sinh x + A \cosh x = A(\sinh x + \cosh x) = Ae^x.

Case 3: 0<B<A0 < B < A (i.e., A<B<A-A < B < A with B>0B > 0). We try Rsinh(x+γ)=Rsinhxcoshγ+RcoshxsinhγR \sinh(x + \gamma) = R \sinh x \cosh \gamma + R \cosh x \sinh \gamma, matching A=RcoshγA = R \cosh \gamma and B=RsinhγB = R \sinh \gamma. Then tanhγ=B/A\tanh \gamma = B/A, so γ=artanhBA\gamma = \operatorname{artanh} \frac{B}{A}, and R2=A2B2>0R^2 = A^2 - B^2 > 0, giving R=A2B2R = \sqrt{A^2 - B^2}.

Asinhx+Bcoshx=A2B2sinh ⁣(x+artanhBA).A \sinh x + B \cosh x = \sqrt{A^2 - B^2} \, \sinh\!\left(x + \operatorname{artanh} \frac{B}{A}\right).

Case 4: B=A<0B = -A < 0. Then AsinhxAcoshx=A(sinhxcoshx)=AexA \sinh x - A \cosh x = A(\sinh x - \cosh x) = -Ae^{-x}.

Case 5: B<A<0B < -A < 0. We return to the cosh\cosh form. R2=B2A2>0R^2 = B^2 - A^2 > 0, but Rcoshγ=B<0R \cosh \gamma = B < 0, so we take R=B2A2<0R = -\sqrt{B^2 - A^2} < 0. Then γ=artanhAB\gamma = \operatorname{artanh} \frac{A}{B} (note A/B<1|A/B| < 1 since B>A|B| > A).

Asinhx+Bcoshx=B2A2cosh ⁣(x+artanhAB).A \sinh x + B \cosh x = -\sqrt{B^2 - A^2} \, \cosh\!\left(x + \operatorname{artanh} \frac{A}{B}\right).


Part (i)

Setting sechx=atanhx+b\operatorname{sech} x = a \tanh x + b and multiplying both sides by coshx\cosh x:

1=asinhx+bcoshx.1 = a \sinh x + b \cosh x.

In the case b>a>0b > a > 0, we apply Case 1 with A=aA = a, B=bB = b:

b2a2cosh ⁣(x+artanhab)=1\sqrt{b^2 - a^2} \, \cosh\!\left(x + \operatorname{artanh} \frac{a}{b}\right) = 1

cosh ⁣(x+artanhab)=1b2a2.\cosh\!\left(x + \operatorname{artanh} \frac{a}{b}\right) = \frac{1}{\sqrt{b^2 - a^2}}.

For this to have a solution, we need 1b2a21\frac{1}{\sqrt{b^2 - a^2}} \geq 1, i.e., b2a21b^2 - a^2 \leq 1.

Since cosh\cosh is an even function, coshu=α\cosh u = \alpha gives u=±arcoshαu = \pm \operatorname{arcosh} \alpha for α1\alpha \geq 1. Therefore:

x+artanhab=±arcosh ⁣(1b2a2)x + \operatorname{artanh} \frac{a}{b} = \pm \operatorname{arcosh}\!\left(\frac{1}{\sqrt{b^2 - a^2}}\right)

x=±arcosh ⁣(1b2a2)artanhab.(shown)x = \pm \operatorname{arcosh}\!\left(\frac{1}{\sqrt{b^2 - a^2}}\right) - \operatorname{artanh} \frac{a}{b}. \qquad \text{(shown)}

Part (ii)

Again the intersection condition is asinhx+bcoshx=1a \sinh x + b \cosh x = 1. In the case a>b>0a > b > 0, we apply Case 3 with A=aA = a, B=bB = b:

a2b2sinh ⁣(x+artanhba)=1\sqrt{a^2 - b^2} \, \sinh\!\left(x + \operatorname{artanh} \frac{b}{a}\right) = 1

sinh ⁣(x+artanhba)=1a2b2.\sinh\!\left(x + \operatorname{artanh} \frac{b}{a}\right) = \frac{1}{\sqrt{a^2 - b^2}}.

Since sinh\sinh is a strictly increasing function with range (,)(-\infty, \infty), the equation sinhu=α\sinh u = \alpha has the unique solution u=arsinhαu = \operatorname{arsinh} \alpha for any real α\alpha. Therefore:

x+artanhba=arsinh ⁣(1a2b2)x + \operatorname{artanh} \frac{b}{a} = \operatorname{arsinh}\!\left(\frac{1}{\sqrt{a^2 - b^2}}\right)

x=arsinh ⁣(1a2b2)artanhba.(shown)x = \operatorname{arsinh}\!\left(\frac{1}{\sqrt{a^2 - b^2}}\right) - \operatorname{artanh} \frac{b}{a}. \qquad \text{(shown)}

Part (iii)

From part (i), when b>ab > a, the intersection xx-coordinates satisfy cosh(x+γ)=1b2a2\cosh(x + \gamma) = \frac{1}{\sqrt{b^2 - a^2}}. This equation has two distinct solutions (the ±\pm branch) if and only if 1b2a2>1\frac{1}{\sqrt{b^2 - a^2}} > 1, i.e., b2a2<1b^2 - a^2 < 1, i.e., b<a2+1b < \sqrt{a^2 + 1}.

Combining with b>ab > a: the curves intersect at two distinct points when a<b<a2+1a < b < \sqrt{a^2 + 1}.

From part (ii), when a>b>0a > b > 0, the equation sinh(x+γ)=1a2b2\sinh(x + \gamma) = \frac{1}{\sqrt{a^2 - b^2}} has exactly one solution (since sinh\sinh is strictly monotone), so there is only one intersection point.

Therefore the necessary and sufficient condition for two distinct intersection points is:

a<b<a2+1.(shown)a < b < \sqrt{a^2 + 1}. \qquad \text{(shown)}

Part (iv)

“Touching” means the curves meet at exactly one point with a common tangent, i.e., both the function values and derivatives agree.

From y=sechxy = \operatorname{sech} x and y=atanhx+by = a \tanh x + b:

y=sechxtanhxandy=asech2x.y' = -\operatorname{sech} x \tanh x \qquad \text{and} \qquad y' = a \operatorname{sech}^2 x.

At a touching point, sechxtanhx=asech2x-\operatorname{sech} x \tanh x = a \operatorname{sech}^2 x. Since sechx>0\operatorname{sech} x > 0, divide by sechx\operatorname{sech} x:

tanhx=asechx-\tanh x = a \operatorname{sech} x

sinhxcoshx=acoshx-\frac{\sinh x}{\cosh x} = \frac{a}{\cosh x}

sinhx=a.\sinh x = -a.

Substituting into the intersection condition asinhx+bcoshx=1a \sinh x + b \cosh x = 1:

a(a)+bcoshx=1    bcoshx=1+a2    coshx=1+a2b.a(-a) + b \cosh x = 1 \implies b \cosh x = 1 + a^2 \implies \cosh x = \frac{1 + a^2}{b}.

Using cosh2xsinh2x=1\cosh^2 x - \sinh^2 x = 1:

(1+a2b)2a2=1\left(\frac{1 + a^2}{b}\right)^2 - a^2 = 1

(1+a2)2b2=1+a2\frac{(1 + a^2)^2}{b^2} = 1 + a^2

1+a2b2=1\frac{1 + a^2}{b^2} = 1

b2=1+a2b^2 = 1 + a^2

b=a2+1(since b>0).b = \sqrt{a^2 + 1} \quad (\text{since } b > 0).

Conversely, if b=a2+1b = \sqrt{a^2 + 1}, then from part (i) (since b=a2+1>ab = \sqrt{a^2+1} > a), the cosh\cosh equation becomes cosh(x+γ)=1b2a2=11=1\cosh(x + \gamma) = \frac{1}{\sqrt{b^2 - a^2}} = \frac{1}{1} = 1, which has the unique solution x+γ=0x + \gamma = 0, confirming exactly one intersection point.

The necessary and sufficient condition for the curves to touch is:

b=a2+1.(shown)b = \sqrt{a^2 + 1}. \qquad \text{(shown)}

yy-coordinate of the point of contact. At the touching point, coshx=1+a2b=1+a2a2+1=a2+1\cosh x = \frac{1+a^2}{b} = \frac{1+a^2}{\sqrt{a^2+1}} = \sqrt{a^2+1}, so:

y=sechx=1coshx=1a2+1.y = \operatorname{sech} x = \frac{1}{\cosh x} = \frac{1}{\sqrt{a^2 + 1}}.

Examiner Notes

约40%考生尝试此题,得分约三分之一。多数知道 cosh 加法公式,但 A=B 的情况处理很弱。(i) 部分解释不清,± 号理解有误,导致 (ii) 出现虚假的 ±。(iii)(iv) 必要充分条件论证薄弱。仅少数优秀考生完整解答。


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

7 Let ω=e2πi/n\omega = e^{2\pi i/n}, where nn is a positive integer. Show that, for any complex number zz,

(z1)(zω)(zωn1)=zn1.(z - 1)(z - \omega) \cdots (z - \omega^{n-1}) = z^n - 1 .

The points X0,X1,,Xn1X_0, X_1, \dots, X_{n-1} lie on a circle with centre OO and radius 1, and are the vertices of a regular polygon.

(i) The point PP is equidistant from X0X_0 and X1X_1. Show that, if nn is even,

PX0×PX1××PXn1=OPn+1,|PX_0| \times |PX_1| \times \cdots \times |PX_{n-1}| = |OP|^n + 1 ,

where PXk|PX_k| denotes the distance from PP to XkX_k.

Give the corresponding result when nn is odd. (There are two cases to consider.)

(ii) Show that

X0X1×X0X2××X0Xn1=n.|X_0X_1| \times |X_0X_2| \times \cdots \times |X_0X_{n-1}| = n .

Hint

Considering (ωr)n(\omega^r)^n establishes by the factor theorem that each factor on the LHS is a factor of the RHS, and comparing coefficients of znz^n between the two sides establishes that no numerical factor is required.

For part (i), representing XrX_r by ωr\omega^r, then there are two cases to consider, PP will be represented either by reπinre^{\frac{\pi i}{n}}, or re(πn+π)ire^{(\frac{\pi}{n}+\pi)i}. The product of moduli is the moduli of the product of factors, and the product of the factors can be simplified using the stem and choosing zz in turn as the representations of PP to give the required result in both cases.

Proceeding similarly for nn odd, the first case yields OPn+1|OP|^n + 1, and the second, OPn1|OP|^n - 1, if OP1|OP| \ge 1, and 1OPn1 - |OP|^n if OP<1|OP| < 1.

Using the same representations for the XrX_r in part (ii), and the same technique with the moduli, the stem can be divided by (z1)(z - 1) to give (zω)(zω2)(zωn1)=zn1+zn2++1(z - \omega)(z - \omega^2) \dots (z - \omega^{n-1}) = z^{n-1} + z^{n-2} + \dots + 1 which then gives the desired result when z=1z = 1.

Model Solution

Proof of the stem result.

The equation zn=1z^n = 1 has exactly nn roots: z=1,ω,ω2,,ωn1z = 1, \omega, \omega^2, \ldots, \omega^{n-1}, where ω=e2πi/n\omega = e^{2\pi i/n}.

To verify these are distinct: if ωj=ωk\omega^j = \omega^k for 0j,kn10 \le j, k \le n-1, then e2πi(jk)/n=1e^{2\pi i(j-k)/n} = 1, requiring n(jk)n \mid (j-k). Since jk<n|j-k| < n, we must have j=kj = k.

By the Factor Theorem, each root ωk\omega^k means (zωk)(z - \omega^k) divides zn1z^n - 1. Since ω0,ω1,,ωn1\omega^0, \omega^1, \ldots, \omega^{n-1} are nn distinct roots, the product

(z1)(zω)(zω2)(zωn1)(z - 1)(z - \omega)(z - \omega^2) \cdots (z - \omega^{n-1})

divides zn1z^n - 1. Both sides are monic polynomials of degree nn, so they must be equal:

(z1)(zω)(zωn1)=zn1.(*)(z - 1)(z - \omega) \cdots (z - \omega^{n-1}) = z^n - 1. \qquad \text{(*)}

Part (i)

We represent each vertex XkX_k by ωk\omega^k, so X0=1X_0 = 1 and X1=ω=e2πi/nX_1 = \omega = e^{2\pi i/n}.

Since PP is equidistant from X0X_0 and X1X_1, it lies on the perpendicular bisector of the chord X0X1X_0 X_1. This bisector passes through the center OO and makes an angle π/n\pi/n with the positive real axis. We can write

P=reiπ/nP = r e^{i\pi/n}

for some real rr, where OP=r|OP| = |r|.

Taking the modulus of both sides of ()(*) with z=Pz = P:

k=0n1Pωk=Pn1.\prod_{k=0}^{n-1} |P - \omega^k| = |P^n - 1|.

Computing Pn=rneiπ=rnP^n = r^n e^{i\pi} = -r^n, so

Pn1=rn1=rn+1.|P^n - 1| = |-r^n - 1| = |r^n + 1|.

When nn is even: rn=rn0r^n = |r|^n \ge 0, so rn+1>0r^n + 1 > 0, giving

k=0n1PXk=rn+1=OPn+1.\prod_{k=0}^{n-1} |PX_k| = r^n + 1 = |OP|^n + 1.

This holds for all real rr (both sides of OO), since rn=rnr^n = |r|^n when nn is even.

When nn is odd, there are two cases depending on the sign of rr (i.e., which side of OO the point PP lies):

Case 1: r>0r > 0 (same side as the arc midpoint eiπ/ne^{i\pi/n}). Then rn>0r^n > 0, so

rn+1=rn+1=OPn+1.|r^n + 1| = r^n + 1 = |OP|^n + 1.

Case 2: r<0r < 0 (opposite side of OO). Write r=sr = -s where s=OP>0s = |OP| > 0. Since nn is odd, rn=snr^n = -s^n, so

rn+1=1sn=1OPn.|r^n + 1| = |1 - s^n| = |1 - |OP|^n|.

Concretely:

  • If OP>1|OP| > 1: the product equals OPn1|OP|^n - 1.
  • If OP<1|OP| < 1: the product equals 1OPn1 - |OP|^n.
  • If OP=1|OP| = 1: the product equals 00 (here PP coincides with the vertex X(n+1)/2X_{(n+1)/2}).

Part (ii)

The distance X0Xk=1ωk|X_0 X_k| = |1 - \omega^k|, so the required product is

k=1n11ωk=k=1n1(1ωk).\prod_{k=1}^{n-1} |1 - \omega^k| = \left| \prod_{k=1}^{n-1} (1 - \omega^k) \right|.

From ()(*), dividing both sides by (z1)(z - 1):

(zω)(zω2)(zωn1)=zn1z1=zn1+zn2++z+1.(z - \omega)(z - \omega^2) \cdots (z - \omega^{n-1}) = \frac{z^n - 1}{z - 1} = z^{n-1} + z^{n-2} + \cdots + z + 1.

Setting z=1z = 1:

(1ω)(1ω2)(1ωn1)=1+1++1=n.(1 - \omega)(1 - \omega^2) \cdots (1 - \omega^{n-1}) = 1 + 1 + \cdots + 1 = n.

Taking moduli:

k=1n11ωk=n=n.\prod_{k=1}^{n-1} |1 - \omega^k| = |n| = n.

Examiner Notes

纯数部分最不受欢迎的题目,不到30%尝试,整体表现不佳。多数能完成 stem 部分,但经常忽略数值因子为1的验证。尽管有 stem,多数不知道如何处理 (i),因此进展甚少。


Topic: 函数方程 (Functional Equations)  |  Difficulty: Standard  |  Marks: 20

8 (i) The function ff satisfies, for all xx, the equation

f(x)+(1x)f(x)=x2.f(x) + (1 - x)f(-x) = x^2 .

Show that f(x)+(1+x)f(x)=x2f(-x) + (1 + x)f(x) = x^2. Hence find f(x)f(x) in terms of xx. You should verify that your function satisfies the original equation.

(ii) The function KK is defined, for x1x \neq 1, by

K(x)=x+1x1.K(x) = \frac{x + 1}{x - 1} .

Show that, for x1x \neq 1, K(K(x))=xK(K(x)) = x.

The function gg satisfies the equation

g(x)+xg(x+1x1)=x(x1).g(x) + x g \left( \frac{x + 1}{x - 1} \right) = x \quad (x \neq 1) .

Show that, for x1x \neq 1, g(x)=2xx2+1g(x) = \frac{2x}{x^2 + 1}.

(iii) Find h(x)h(x), for x0,x1x \neq 0, x \neq 1, given that

h(x)+h(11x)=1x11x(x0,x1).h(x) + h \left( \frac{1}{1 - x} \right) = 1 - x - \frac{1}{1 - x} \quad (x \neq 0, \enspace x \neq 1) .

Hint

The first result in (i) is obtained by the substitution x=ux = -u (followed by a second u=xu = x!). Substituting for f(x)f(-x) in the initial statement using the result obtained readily leads to f(x)=xf(x) = x which is simply verified. Alternatively, subtracting the result from the initial equation leads to f(x)=f(x)f(x) = f(-x) which substituting gives the required result again.

In part (ii), substituting K(x)K(x) for xx in the equation for g(x)g(x) gives an equation for g(x+1x1)g\left(\frac{x+1}{x-1}\right) which can be substituted in the equation to be solved to give the desired result.

Similarly, in part (iii), substituting 11x\frac{1}{1-x} for xx gives an equation for h(11x)h\left(\frac{1}{1-x}\right) and h(x1x)h\left(\frac{x-1}{x}\right), and then repeating this substitution in the equation just obtained gives an equation for h(x1x)h\left(\frac{x-1}{x}\right) and h(x)h(x). Adding the given and last equations and subtracting that first found leads to h(x)=12xh(x) = \frac{1}{2} - x.

Model Solution

Part (i)

The given equation is

f(x)+(1x)f(x)=x2.(1)f(x) + (1 - x)f(-x) = x^2. \qquad \text{(1)}

Replace xx by x-x throughout:

f(x)+(1+x)f(x)=(x)2=x2.(2)f(-x) + (1 + x)f(x) = (-x)^2 = x^2. \qquad \text{(2)}

This is the first required result.

From (2), solve for f(x)f(-x):

f(x)=x2(1+x)f(x).f(-x) = x^2 - (1 + x)f(x).

Substitute into (1):

f(x)+(1x)[x2(1+x)f(x)]=x2.f(x) + (1 - x)\bigl[x^2 - (1 + x)f(x)\bigr] = x^2.

Expand:

f(x)+x2(1x)(1x)(1+x)f(x)=x2.f(x) + x^2(1 - x) - (1 - x)(1 + x)f(x) = x^2.

f(x)+x2x3(1x2)f(x)=x2.f(x) + x^2 - x^3 - (1 - x^2)f(x) = x^2.

Collect the f(x)f(x) terms:

f(x)[1(1x2)]=x2x2+x3.f(x)\bigl[1 - (1 - x^2)\bigr] = x^2 - x^2 + x^3.

f(x)x2=x3.f(x) \cdot x^2 = x^3.

f(x)=x.(3)f(x) = x. \qquad \text{(3)}

Verification: Substitute f(x)=xf(x) = x into (1):

x+(1x)(x)=xx+x2=x2.x + (1 - x)(-x) = x - x + x^2 = x^2. \checkmark

Part (ii)

Showing K(K(x))=xK(K(x)) = x:

K(K(x))=K ⁣(x+1x1)=x+1x1+1x+1x11=x+1+x1x1x+1x+1x1=2xx12x1=x.(4)K(K(x)) = K\!\left(\frac{x+1}{x-1}\right) = \frac{\dfrac{x+1}{x-1} + 1}{\dfrac{x+1}{x-1} - 1} = \frac{\dfrac{x+1+x-1}{x-1}}{\dfrac{x+1-x+1}{x-1}} = \frac{\dfrac{2x}{x-1}}{\dfrac{2}{x-1}} = x. \qquad \text{(4)}

Finding g(x)g(x):

The equation for gg is

g(x)+xg ⁣(x+1x1)=x.(5)g(x) + x \, g\!\left(\frac{x+1}{x-1}\right) = x. \qquad \text{(5)}

Replace xx by x+1x1\frac{x+1}{x-1} in (5). Using (4)(4), K(K(x))=xK(K(x)) = x, so

g ⁣(x+1x1)+x+1x1g(x)=x+1x1.(6)g\!\left(\frac{x+1}{x-1}\right) + \frac{x+1}{x-1} \, g(x) = \frac{x+1}{x-1}. \qquad \text{(6)}

From (6), solve for g ⁣(x+1x1)g\!\bigl(\frac{x+1}{x-1}\bigr):

g ⁣(x+1x1)=x+1x1x+1x1g(x)=x+1x1(1g(x)).(7)g\!\left(\frac{x+1}{x-1}\right) = \frac{x+1}{x-1} - \frac{x+1}{x-1} \, g(x) = \frac{x+1}{x-1}\bigl(1 - g(x)\bigr). \qquad \text{(7)}

Substitute (7) into (5):

g(x)+xx+1x1(1g(x))=x.g(x) + x \cdot \frac{x+1}{x-1}\bigl(1 - g(x)\bigr) = x.

Multiply through by (x1)(x - 1):

(x1)g(x)+x(x+1)(1g(x))=x(x1).(x - 1)g(x) + x(x + 1)(1 - g(x)) = x(x - 1).

Expand the left side:

(x1)g(x)+x(x+1)x(x+1)g(x)=x2x.(x - 1)g(x) + x(x + 1) - x(x + 1)g(x) = x^2 - x.

Collect the g(x)g(x) terms:

g(x)[(x1)x(x+1)]=x2xx(x+1).g(x)\bigl[(x - 1) - x(x + 1)\bigr] = x^2 - x - x(x + 1).

g(x)[x1x2x]=x2xx2x.g(x)\bigl[x - 1 - x^2 - x\bigr] = x^2 - x - x^2 - x.

g(x)(1x2)=2x.g(x)(-1 - x^2) = -2x.

g(x)=2xx2+1.(8)g(x) = \frac{2x}{x^2 + 1}. \qquad \text{(8)}

Verification: We need to check that (8)(8) satisfies (5). First compute g ⁣(x+1x1)g\!\bigl(\frac{x+1}{x-1}\bigr):

g ⁣(x+1x1)=2x+1x1(x+1x1)2+1=2(x+1)x1(x+1)2+(x1)2(x1)2=2(x+1)x1(x1)2(x+1)2+(x1)2.g\!\left(\frac{x+1}{x-1}\right) = \frac{2 \cdot \frac{x+1}{x-1}}{\left(\frac{x+1}{x-1}\right)^2 + 1} = \frac{\frac{2(x+1)}{x-1}}{\frac{(x+1)^2 + (x-1)^2}{(x-1)^2}} = \frac{2(x+1)}{x-1} \cdot \frac{(x-1)^2}{(x+1)^2 + (x-1)^2}.

The denominator simplifies: (x+1)2+(x1)2=x2+2x+1+x22x+1=2x2+2=2(x2+1)(x+1)^2 + (x-1)^2 = x^2 + 2x + 1 + x^2 - 2x + 1 = 2x^2 + 2 = 2(x^2 + 1). So

g ⁣(x+1x1)=2(x+1)(x1)2(x2+1)=x21x2+1.g\!\left(\frac{x+1}{x-1}\right) = \frac{2(x+1)(x-1)}{2(x^2+1)} = \frac{x^2 - 1}{x^2 + 1}.

Now check (5):

g(x)+xg ⁣(x+1x1)=2xx2+1+xx21x2+1=2x+x3xx2+1=x3+xx2+1=x(x2+1)x2+1=x.g(x) + x \, g\!\left(\frac{x+1}{x-1}\right) = \frac{2x}{x^2+1} + x \cdot \frac{x^2-1}{x^2+1} = \frac{2x + x^3 - x}{x^2+1} = \frac{x^3 + x}{x^2+1} = \frac{x(x^2+1)}{x^2+1} = x. \checkmark

Part (iii)

Define T(x)=11xT(x) = \frac{1}{1 - x}. We first check that TT has order 3 (i.e., applying TT three times returns to xx):

T2(x)=T ⁣(11x)=1111x=11x11x=1xx=x1x.T^2(x) = T\!\left(\frac{1}{1-x}\right) = \frac{1}{1 - \frac{1}{1-x}} = \frac{1}{\frac{1-x-1}{1-x}} = \frac{1-x}{-x} = \frac{x-1}{x}.

T3(x)=T ⁣(x1x)=11x1x=1x(x1)x=11x=x.(9)T^3(x) = T\!\left(\frac{x-1}{x}\right) = \frac{1}{1 - \frac{x-1}{x}} = \frac{1}{\frac{x - (x-1)}{x}} = \frac{1}{\frac{1}{x}} = x. \qquad \text{(9)}

The given equation is

h(x)+h ⁣(11x)=1x11x.(I)h(x) + h\!\left(\frac{1}{1-x}\right) = 1 - x - \frac{1}{1-x}. \qquad \text{(I)}

Replace xx by T(x)=11xT(x) = \frac{1}{1-x} in (I):

h ⁣(11x)+h ⁣(x1x)=111x1111x.h\!\left(\frac{1}{1-x}\right) + h\!\left(\frac{x-1}{x}\right) = 1 - \frac{1}{1-x} - \frac{1}{1 - \frac{1}{1-x}}.

The right side simplifies:

111x1xx=111x+1xx.1 - \frac{1}{1-x} - \frac{1-x}{-x} = 1 - \frac{1}{1-x} + \frac{1-x}{x}.

So:

h ⁣(11x)+h ⁣(x1x)=111x+1xx.(II)h\!\left(\frac{1}{1-x}\right) + h\!\left(\frac{x-1}{x}\right) = 1 - \frac{1}{1-x} + \frac{1-x}{x}. \qquad \text{(II)}

Replace xx by T2(x)=x1xT^2(x) = \frac{x-1}{x} in (I). Using (9)(9), T3(x)=xT^3(x) = x:

h ⁣(x1x)+h(x)=1x1x11x1x.h\!\left(\frac{x-1}{x}\right) + h(x) = 1 - \frac{x-1}{x} - \frac{1}{1 - \frac{x-1}{x}}.

The right side simplifies:

1x1x11x=1x1xx=11+1xx=1xx.1 - \frac{x-1}{x} - \frac{1}{\frac{1}{x}} = 1 - \frac{x-1}{x} - x = 1 - 1 + \frac{1}{x} - x = \frac{1}{x} - x.

So:

h ⁣(x1x)+h(x)=1xx.(III)h\!\left(\frac{x-1}{x}\right) + h(x) = \frac{1}{x} - x. \qquad \text{(III)}

Now compute (I)+(III)(II)\text{(I)} + \text{(III)} - \text{(II)}:

Left side: [h(x)+h(T(x))]+[h(T2(x))+h(x)][h(T(x))+h(T2(x))]=2h(x)[h(x) + h(T(x))] + [h(T^2(x)) + h(x)] - [h(T(x)) + h(T^2(x))] = 2h(x).

Right side:

(1x11x)+(1xx)(111x+1xx).\left(1 - x - \frac{1}{1-x}\right) + \left(\frac{1}{x} - x\right) - \left(1 - \frac{1}{1-x} + \frac{1-x}{x}\right).

Group terms:

=1x11x+1xx1+11x1xx.= 1 - x - \frac{1}{1-x} + \frac{1}{x} - x - 1 + \frac{1}{1-x} - \frac{1-x}{x}.

The 11x\frac{1}{1-x} terms cancel:

=2x+1x1xx=2x+1(1x)x=2x+xx=2x+1.= -2x + \frac{1}{x} - \frac{1-x}{x} = -2x + \frac{1 - (1-x)}{x} = -2x + \frac{x}{x} = -2x + 1.

Therefore 2h(x)=12x2h(x) = 1 - 2x, giving

h(x)=12x.h(x) = \frac{1}{2} - x.

Verification:

h(x)+h ⁣(11x)=(12x)+(1211x)=1x11x.h(x) + h\!\left(\frac{1}{1-x}\right) = \left(\frac{1}{2} - x\right) + \left(\frac{1}{2} - \frac{1}{1-x}\right) = 1 - x - \frac{1}{1-x}. \checkmark

Examiner Notes

尝试人数接近 Q1(最多),且略为更成功,不少人得满分。函数迭代构造循环的想法普遍被发现。(ii) 中找 g(x) 时有困难,部分考生代入已知的 g(x) 而非求解它。(iii) 中部分考生只做了第一次代换就停止,无法找到解答。也有部分猜出答案但未完整论证。