Three connected topics for University of Tokyo mathematics practice

Original publication: Updated:

What this guide covers

The original article was written in January 2019, looking back at entrance papers and mock examinations available then. Its forecasts concerned the 2019 examination; they are not current predictions. The three topics remain useful examples of how a problem can combine several mathematical skills:

  1. Probability, especially probability recurrences.
  2. Complex-plane geometry.
  3. Areas and volumes of solids obtained from cross-sections.

The worked examples below come from the mock-exam problems discussed in that article. Use them to practice setting up a model, rather than assuming that a future paper must contain these topics.

1. Probability recurrences

A probability problem that involves repeated steps often also tests your understanding of sequences. Instead of enumerating every successful history, identify what information about the latest step determines which transitions remain possible. See the probability-recurrence guide for the basic method.

Problem: adjacent products divisible by three

Roll a fair six-sided die nn times, where n≥2n\ge2, obtaining a1,…,ana_1,\ldots,a_n. Find the probability pnp_n that akak+1a_ka_{k+1} is divisible by 33 for every 1≤k<n1\le k<n.

A product is divisible by 33 exactly when at least one factor is. Thus the forbidden event is a pair of consecutive outcomes neither of which is a multiple of 33.

Let qnq_n be the probability of a valid sequence ending in a multiple of 33, and rnr_n the probability of a valid sequence ending in a nonmultiple. A multiple occurs with probability 1/31/3.

Transitions between valid sequences ending in a multiple of three and those ending in a nonmultiple.

Appending one outcome gives

qn+1=13(qn+rn),rn+1=23qn,q1=13,r1=23.q_{n+1}=\frac13(q_n+r_n),\qquad r_{n+1}=\frac23q_n, \qquad q_1=\frac13,\quad r_1=\frac23.

We can solve the system by looking for linear combinations that form geometric sequences:

qn+1+12rn+1=23(qn+12rn),q_{n+1}+\frac12r_{n+1}=\frac23\left(q_n+\frac12r_n\right), qn+1−rn+1=−13(qn−rn).q_{n+1}-r_{n+1}=-\frac13(q_n-r_n).

The initial values therefore give

qn+12rn=(23)n,qn−rn=(−13)n.q_n+\frac12r_n=\left(\frac23\right)^n, \qquad q_n-r_n=\left(-\frac13\right)^n.

Solving,

qn=(23)n+1−(−13)n+1,q_n=\left(\frac23\right)^{n+1}-\left(-\frac13\right)^{n+1}, rn=(23)n+1+2(−13)n+1.r_n=\left(\frac23\right)^{n+1}+2\left(-\frac13\right)^{n+1}.

Consequently,

pn=2(23)n+1+(−13)n+1.\boxed{p_n=2\left(\frac23\right)^{n+1}+\left(-\frac13\right)^{n+1}.}

For instance, p2=5/9p_2=5/9: the only failure is that both outcomes are nonmultiples, with probability 4/94/9. This provides a quick check on the formula.

2. Complex numbers and geometric transformations

The historical guide highlighted the way complex numbers connect rotations, similarity, and loci. The following problem does not explicitly tell you to introduce a complex plane. Recognizing that option is the main strategic step.

Problem: similar isosceles triangles

Triangles OABOAB and OCDOCD are similar, noncongruent isosceles triangles with OA=OBOA=OB and OC=ODOC=OD. Their corresponding vertices are O,A,BO,A,B and O,C,DO,C,D, respectively, and each triple is ordered counterclockwise.

Choose EE so that BCEBCE is a right isosceles triangle with EB=ECEB=EC, and B,C,EB,C,E are ordered counterclockwise. Let MM be the midpoint of ADAD. If O,E,MO,E,M are collinear, find ∠AOB\angle AOB.

Normalize the configuration

A translation, rotation, and positive scaling preserve the required angle. Put O=0O=0, A=1A=1, and write B=bB=b, C=cC=c, where

∣b∣=1,Im⁡b>0,∣c∣≠1.|b|=1,\qquad \operatorname{Im}b>0,\qquad |c|\ne1.

Similarity then gives D=bcD=bc and M=(1+bc)/2M=(1+bc)/2. Because ∣bc∣≠1|bc|\ne1, the denominator 1+bc1+bc cannot vanish.

The right-angle and orientation conditions at EE give

c−eb−e=i,e=c−ib1−i.\frac{c-e}{b-e}=i, \qquad e=\frac{c-ib}{1-i}.

Collinearity means that e/(1+bc)e/(1+bc) is real. A complex number is real exactly when it equals its conjugate, so

e1+bc=eˉ1+bˉcˉ.\frac{e}{1+bc}=\frac{\bar e}{1+\bar b\bar c}.

Substitute the expression for ee and clear denominators:

(1+i)(c−ib)(1+bˉcˉ)=(1−i)(cˉ+ibˉ)(1+bc).(1+i)(c-ib)(1+\bar b\bar c) =(1-i)(\bar c+i\bar b)(1+bc).

Using bbˉ=1b\bar b=1, expansion and cancellation reduce this to

(∣c∣2−1)(b−ibˉ)=0.(|c|^2-1)(b-i\bar b)=0.

The triangles are not congruent, so ∣c∣≠1|c|\ne1. Hence b=ibˉb=i\bar b; multiplying by bb gives b2=ib^2=i. Of the two roots, only one lies above the real axis:

b=1+i2.b=\frac{1+i}{\sqrt2}.

Its argument is π/4\pi/4, so

∠AOB=45∘.\boxed{\angle AOB=45^\circ.}

The useful translations were: rotation becomes multiplication, a midpoint becomes an average, and collinearity becomes a real ratio. Fewer equations than unknown coordinates need not be a problem when the requested angle is invariant across a family of configurations.

3. Solid volumes from cross-sections

For a complicated solid, fixing one coordinate can turn a spatial problem into a plane region. Determine its area as a function of the fixed coordinate, then integrate. The main work is often identifying the values at which the shape of the section changes.

Problem: shortest axis-parallel paths in a cube

Let C=[−1,1]3C=[-1,1]^3, let DD be its surface, and let EE be the closed square pyramid with apex O=(0,0,0)O=(0,0,0) and base the face z=1z=1 of CC.

For points P,QP,Q, let d(P,Q)d(P,Q) be the shortest length of a polygonal path joining them using segments parallel to coordinate axes. For P∈CP\in C, define

m(P)=min⁡Q∈Dd(P,Q).m(P)=\min_{Q\in D}d(P,Q).

Let FF consist of the points P∈CP\in C satisfying m(P)≥d(O,P)m(P)\ge d(O,P).

  1. Express m(P)m(P) in coordinates for P=(x,y,z)∈EP=(x,y,z)\in E.
  2. For 0≤t≤10\le t\le1, find the area S(t)S(t) of the section of E∩FE\cap F by z=tz=t, taking an empty section to have area zero.
  3. Find the volume of FF.

Part 1: translate the distance definition

Any axis-parallel path must cover at least the absolute change in each coordinate, and moving along the three axes in succession attains that bound. Therefore

d(P,Q)=∣xP−xQ∣+∣yP−yQ∣+∣zP−zQ∣.d(P,Q)=|x_P-x_Q|+|y_P-y_Q|+|z_P-z_Q|.

To reach a specified face of the cube, the other two coordinates can remain unchanged. Thus, throughout CC,

m(P)=min⁡{1−∣x∣,1−∣y∣,1−∣z∣}.m(P)=\min\{1-|x|,1-|y|,1-|z|\}.

Inside EE we have 0≤z≤10\le z\le1, ∣x∣≤z|x|\le z, and ∣y∣≤z|y|\le z, giving

m(P)=1−z(P∈E).\boxed{m(P)=1-z\quad(P\in E).}

Part 2: intersect a square and a diamond

At height z=tz=t, the condition defining FF becomes

1−t≥∣x∣+∣y∣+t.1-t\ge|x|+|y|+t.

Together with the pyramid constraints, the section is

∣x∣≤t,∣y∣≤t,∣x∣+∣y∣≤1−2t.|x|\le t,\qquad |y|\le t,\qquad |x|+|y|\le1-2t.

The square expands as tt increases, while the diamond shrinks.

Animation of the changing square and diamond cross-section.

For 0≤t≤1/40\le t\le1/4, the square lies inside the diamond, since its corners satisfy 2t≤1−2t2t\le1-2t.

Square contained in the diamond, with the entire square shaded.

For 1/4≤t≤1/31/4\le t\le1/3, four corner triangles are cut from the square. Each triangle has perpendicular side length 4t−14t-1.

Square and diamond overlapping, with the four square corners removed.

For 1/3≤t≤1/21/3\le t\le1/2, the diamond lies inside the square, since 1−2t≤t1-2t\le t.

Diamond contained in the square, with the diamond shaded.

For t>1/2t>1/2 there is no section. At t=1/2t=1/2 it is a single point and still has area zero. Therefore

S(t)={4t20≤t≤14,4t2−2(4t−1)214≤t≤13,2(1−2t)213≤t≤12,012≤t≤1.\boxed{S(t)=\begin{cases} 4t^2&0\le t\le\frac14,\\ 4t^2-2(4t-1)^2&\frac14\le t\le\frac13,\\ 2(1-2t)^2&\frac13\le t\le\frac12,\\ 0&\frac12\le t\le1. \end{cases}}

The formulas agree at every shared endpoint.

Part 3: integrate and restore symmetry

The volume inside this one pyramid is

Vol⁡(E∩F)=∫01/44t2 dt+∫1/41/3{4t2−2(4t−1)2} dt+∫1/31/22(1−2t)2 dt=118.\begin{aligned} \operatorname{Vol}(E\cap F) &=\int_0^{1/4}4t^2\,dt +\int_{1/4}^{1/3}\{4t^2-2(4t-1)^2\}\,dt\\ &\quad+\int_{1/3}^{1/2}2(1-2t)^2\,dt =\frac1{18}. \end{aligned}

The cube decomposes into six congruent pyramids with apex OO, one for each face. Their common boundaries have volume zero, and FF is symmetric under the corresponding coordinate permutations and sign changes. Hence

Vol⁡(F)=6⋅118=13.\boxed{\operatorname{Vol}(F)=6\cdot\frac1{18}=\frac13.}

The factor six matters: integrating S(t)S(t) finds only the portion within EE.

What to carry into your practice

For repeated probability experiments, choose states that retain the information needed for the next step. For geometric similarity and rotation, try complex coordinates. For solids, work with cross-sections and locate every change of shape before integrating. These are ways of connecting familiar ideas, rather than a prediction of a particular future paper.

Related articles

About the author

Koki Saito

Physician and graduate of the University of Tokyo Faculty of Medicine. Drawing on his entrance-exam experience, he explains how to choose mathematical methods and construct complete solutions.

Full profile →

Have a question or spotted an error? Contact us.