Indian Statistical Institute, ISI BStat & BMath 2023 UGB Solutions & Discussions

ISI BMath & BSTAT 2023 Subjective Questions UGB: solutions and discussions

Past Year’s BMATH/BSTAT Papers & Solutions: Click Here


Indian Statistical Institute,ISI BMATH/BSTAT 2023 All UGA Objective Solutions: Click Here

Indian Statistical Institute,ISI BMATH/BSTAT 2021 All UGA Objective Solutions: Click Here

Indian Statistical Institute,ISI BMATH/BSTAT 2020 All UGA Objective Solutions: Click Here

Indian Statistical Institute,ISI BMATH/BSTAT 2020 All UGB Subjective Solutions: Click Here

Indian Statistical Institute,ISI BMATH/BSTAT 2020 UGA Answer Keys: Click Here

Indian Statistical Institute,ISI BMATH/BSTAT 2020 UGA Question Paper: Click Here

Indian Statistical Institute,ISI BMATH/BSTAT 2020 UGB Question Paper: Click Here

Problem 1.

Determine all integers n>1 such that every power of n has an odd number of digits.

Topic: Number Theory
Difficulty level: Medium

Indian Statistical Institute, ISI Subjective BStat & BMath 2023 Problem 1 UGB Solution

Solution:

We will prove that all powers of and integer n>1 have odd number of digits if and only if n=10^{2k} for some k\geq 1. We note that the number of digits of an integer n>1 is exactly

    \[1+\lfloor \log_{10}(n)\rfloor.\]

Indeed, if k=1+\lfloor \log_{10}(n)\rfloor, then 10^{k-1}\leq n<10^k and n has exactly k digits. If n=10^{2k}, then for any m\geq 1 the number of digits of n^m is odd:

    \[1+\lfloor \log_{10}(n^m)\rfloor =1+\lfloor 2km\rfloor =1+2km.\]

Assume that all powers of n have odd number of digits, but n is not of the form 10^{2k}. If n is a power of 10, then n is an odd power of 10, hence n^1=n has even number of digits. This is impossible.
So, n is not a power of 10 and \log_{10}(n) is not an integer number. Let us write \log_{10}(n)=l+r, where l\geq 0 is an integer and r\in (0,1). Since n^1=n has odd number of digits, we deduce that

    \[1+\lfloor \log_{10}(n)\rfloor=1+l\]

is odd. In particular, l is even.

If r=\frac{1}{m} for some integer m\geq 2, then

    \[1+\lfloor \log_{10}(n^m)\rfloor =1+\lfloor m(l+r)\rfloor=1+ml+1=ml+2\]

is even (since l is even). So, r\ne \frac{1}{m} for any m\geq 2, and r\in \left(\frac{1}{m+1},\frac{1}{m}\right) for some integer m\geq 1. But then n^{m+1} has even number of digits:

    \[1+\lfloor \log_{10}(n^{m+1})\rfloor=1+\lfloor (m+1)l+(m+1)r \rfloor=1+(m+1)l+1=(m+1)l+2.\]

The latter relation follows from the fact that (m+1)r\in\left(1,\frac{m+1}{m}\right)\subset (1,2).

Video Solution:

Problem 2.

Let a_0=\frac{1}{2} and a_n be defined inductively by

    \[a_n=\sqrt{\frac{1+a_{n-1}}{2}}, n \geq 1 .\]

(a) Show that for n=0,1,2, \ldots.

    \[a_n=\cos \theta_n \text { for some } 0<\theta_n<\frac{\pi}{2} \text {. }\]

and determine \theta_n.

(b) Using (a) or otherwise, calculate

    \[\lim _{n \rightarrow \infty} 4^n\left(1-a_n\right)\]

Topic: Trigonometric Sequences
Difficulty Level: Medium

Indian Statistical Institute, ISI Subjective BStat & BMath 2023 Problem 2 UGB Solution.

Solution:

(a) We note that a_0=\frac{1}{2}=\cos \frac{\pi}{3}. If a_{n-1}=\cos \theta_{n-1}, 0<\theta_{n-1}<\frac{\pi}{2}, then

    \[a_n=\sqrt{\frac{1+a_{n-1}}{2}}=\sqrt{\frac{1+\cos\theta_{n-1}}{2}}=\sqrt{\cos^2\left(\frac{\theta_{n-1}}{2}\right)}=\cos\left(\frac{\theta_{n-1}}{2}\right).\]

So, \theta_n=\frac{\theta_{n-1}}{2}. Since \theta_0=\frac{\pi}{3}, it follows that

    \[\theta_n=\frac{\pi}{3\times 2^n}.\]

(b)

    \[\lim_{n\to\infty} 4^n (1-a_n)=\lim_{n\to\infty}\frac{1-\cos \left(\frac{\pi}{3\times 2^{n}}\right)}{\frac{1}{4^{n}}}=\]

since 1-\cos x\sim \frac{x^2}{2}, x\to 0,

    \[=\lim_{n\to\infty}\frac{\frac{\pi^2}{18\times 4^n}}{\frac{1}{4^n}}=\frac{\pi^2}{18}\]

Video Solution:

Problem 3.

In a triangle A B C, consider points D and E on A C and A B. respectively, and assume that they do not coincide with any of the vertices A, B, C. If the segments B D and C E intersect at F, consider the areas w, x, y, z of the quadrilateral A E F D and the triangles B E F, B F C, C D F, respectively.

(a) Prove that y^2>xz.

(b) Determine w in terms of x, y, z.

Topic: Geometry
Difficulty Level: Medium

Indian Statistical Institute, ISI Subjective BStat & BMath 2023 Problem 3 UGB Solution.

Solution:

(a) Consider altitudes EG, FH, DI of length \alpha,\beta,\gamma respectively:

Denote length of BG, GH, HI, IC by a,b,c,d respectively, and let l be the length of BC. Then

    \[x+y=\frac{1}{2}\alpha l, \ z+y=\frac{1}{2}\gamma l, \ y=\frac{1}{2}\beta l.\]

The inequality y^2>xz is equivalent to

    \[y^2>\left(\frac{1}{2}\alpha l-y\right)\left(\frac{1}{2}\gamma l-y\right),\]

or to

    \[\frac{1}{4}\alpha \gamma l^2<\frac{1}{2} y l \left(\alpha+\gamma\right).\]

After simplifications we obtain, equivalently,

    \[y>\frac{\alpha \gamma l}{2(\alpha+\gamma)}.\]

Since y=\frac{1}{2}\beta l, the latter is equivalent to

    \[\beta>\frac{\alpha \gamma}{\alpha+\gamma},\]

or to

    \[\frac{\beta}{\alpha}+\frac{\beta}{\gamma}>1.\]

From triangles EGC and FHI we find that

    \[\frac{\beta}{\alpha}=\frac{c+d}{b+c+d}.\]

From triangles BDI and BFH we find that

    \[\frac{\beta}{\gamma}=\frac{a+b}{a+b+c}.\]

It remains to prove that

    \[\frac{a+b}{a+b+c}+\frac{c+d}{b+c+d}>1.\]

But it follows from simple inequalities \frac{a+b}{a+b+c}>\frac{b}{b+c}, \frac{c+d}{b+c+d}>\frac{c}{b+c}.

 

(b) Consider a segment AF and let u be the area of AEF, v be the area of AFD:

In particular, w=u+v. Triangles AEF and BFE share the same altitude on the line AB. Hence, \frac{u}{x}=\frac{AE}{EB}. Triangles AEC and BCE share the same altitude on the line AB. Hence, \frac{w+z}{x+y}=\frac{AE}{EB}. It follows that

    \[\frac{u}{x}=\frac{w+z}{x+y}.\]

Similarly, considering triangles AFD and CFD, and BAD and BCD, we deduce that

    \[\frac{v}{z}=\frac{w+x}{z+y}.\]

Hence,

    \[w=u+v=\frac{x(w+z)}{x+y}+\frac{z(w+x)}{z+y}.\]

Further,

    \[w\left((x+y)(z+y)-x(z+y)-z(x+y)\right)=xz(x+z+2y),\]

    \[w=\frac{xz(x+z+2y)}{y^2-xz}.\]

Video Solution:

Problem 4.

Let n_1,n_2,n_3,\cdots,n_{51}, be distinct natural numbers each of which has exactly 2023 positive integer factors. For instance, 2^{2022} has exactly 2023 positive integer factors 1,2,2^2, \cdots, 2^{2021}, 2^{2022}. Assume that no prime larger than 11 divides any of the n_i‘s. Show that there must be some perfect cube among the n_i‘s. You may use the fact that 2023=7 \times 17 \times 17

Topic: Combinatorics
Difficulty Level: Easy

Indian Statistical Institute, ISI Subjective BStat & BMath 2023 Problem 4 UGB Solution.

Solution:

Each number n_i has at most five distinct prime divisors: 2,3,5,7,11. If n_i=\prod^m_{j=1}p^{\alpha_j}_j, where \{p_1,\ldots,p_m\}\subset \{2,3,5,7,11\}, \alpha_1,\ldots,\alpha_m\geq 1, then the number of divisors of n_i equals \prod^m_{j=1}(\alpha_j+1). On the other hand,

    \[\prod^m_{j=1}(\alpha_j+1)=2023=7\times 17\times 17.\]

It follows that m\leq 3. Assume that there are no perfect cubes among n_1,\ldots,n_{51}. If n_i=p^\alpha for some i, then the number of divisors of n_i equals

    \[\alpha+1=2023.\]

Hence \alpha=2022=3\times 674 and n_i is a perfect cube. If n_i=p^{\alpha_1}_1p_2^{\alpha_2}, where p_1<p_2 are prime numbers among \{2,3,5,7,11\} and \alpha_1,\alpha_2\geq 1, then

    \[(\alpha_1+1)(\alpha_2+1)=2023=7\times 17\times 17.\]

If \alpha_1+1=7, \alpha_2+1=17\times 17, then \alpha_1=6=3\times 2 and \alpha_2=288=3\times 96. Hence n_i is a perfect cube. The same holds when \alpha_1+1=17\times 17, \alpha_2+1=7. It follows that either \alpha_1+1=7\times 17, \alpha_2+1=17, or \alpha_1+1=17, \alpha_2+1=7\times 17. Equivalently, either \alpha_1=118, \alpha_2=16, or \alpha_1=16, \alpha_2=118. The number of possible combinations for n_i is {5\choose 2}\times 2=20.

If n_i=p^{\alpha_1}_1p_2^{\alpha_2}p^{\alpha_3}_3, where p_1<p_2<p_3 are prime numbers among \{2,3,5,7,11\} and \alpha_1,\alpha_2,\alpha_3\geq 1, then

    \[(\alpha_1+1)(\alpha_2+1)(\alpha_3+1)=2023=7\times 17\times 17.\]

Hence, values of \alpha_1,\alpha_2,\alpha_3 are 6,16,16. The number of possible combinations for n_i is {5\choose 3}\times 3=30.

There are at most 50 natural numbers that are not perfect cubes, that have exactly 2023 divisors, and whose prime divisors are \leq 11. It follows that at least one of the numbers n_1,\ldots,n_{51} is a perfect cube.

Video Solution:

Problem 5.

There is a rectangular plot of size 1 \times n. This has to be covered by three types of tiles – red, blue and black. The red tiles are of size 1 \times 1, the blue tiles are of size 1 \times 1 and the black tiles are of size 1 \times 2. Let t_n denote the number of ways this can be done. For example, clearly t_1=2 because we can have either a red or a blue tile. Also, t_2=5 since we could have tiled the plot as: two red tiles, two blue tiles, a red tile on the left and a blue tile on the right, a blue tile on the left and a red tile on the right, or a single black tile.

(a) Prove that t_{2 n+1}=t_n\left(t_{n-1}+t_{n+1}\right) for all n>1.

(b) Prove that t_n=\sum_{d \geq 0}\left(\begin{array}{c}n-d \\ d\end{array}\right) 2^{n-2 d} for all n>0.
Here,

    \[\left(\begin{array}{c} m \\ r \end{array}\right)= \begin{cases}\frac{m !}{r !(m-r) !}, & \text { if } 0 \leq r \leq m \\ 0, & \text { otherwise }\end{cases}\]

for integers m, r.

Topic: Combinatorics Recurrence
Difficulty Level: Medium

Indian Statistical Institute, ISI Subjective BStat & BMath 2023 Problem 5 UGB Solution.

Solution:

(a) At first we note that t_{n+1}=2t_n+t_{n-1}. Indeed, if the first square is covered either by a red tile or by a blue tile, then there remains n squares more to cover. If the first two squares are covered by a black tile, then there remains n-1 squares to be covered.

Consider a plot of size 1\times (2n+1). If the middle square is covered either by a red tile or by a blue tile, then there remains n squares on the left and n squares on the right of it to be covered. This gives 2t^2_n possibilities. If the middle square is covered by a black tile, then either there remains n-1 squares on the left and n squares on the right of this tile to be covered, or there remains n squares on the left and n-1 squares on the right of this tile to be covered. This gives 2t_nt_{n-1} possibilities. Totally, we get

    \[t_{2n+1}=2t^2_n +2t_nt_{n-1}=t_n\left(t_{n-1}+2t_n+t_{n-1}\right)=t_n\left(t_{n-1}+t_{n+1}\right).\]

(b) Consider covering with exactly d black tiles (in particular, 2d\leq n). Assume that these tiles occupy squares n_1,n_1+1,n_2,n_2+1,\ldots,n_d,n_d+1. Remaining n-2d squares are partitioned by black tiles into d+1 groups (possibly empty):

    \[\{1,\ldots,n_1-1\}, \{n_1+2,\ldots,n_2-1\}, \ldots, \{n_d+2,\ldots,n\}.\]

Conversely, every such collection of d+1 groups define a configuration of black tiles. The number of such collections equals the number of non-negative integer solutions of the equation

    \[m_1+\ldots+m_{d+1}=n-2d,\]

which is known to be equal to {n-2d+d \choose d}={n-d \choose d}. When d black tiles are placed, there remains n-2d squares to be covered by red or blue tiles. There are 2^{n-2d} possibilities to do this. Totally, we get

    \[t_n=\sum_{d\geq 0}{n-d \choose d}2^{n-2d}.\]

Video Solution:

Problem 6.

Let \left\{u_n\right\}_{n \geq 1} be a sequence of real numbers defined as u_1=1 and

    \[u_{n+1}=u_n+\frac{1}{u_n} \text { for all } n \geq 1 \text {. }\]

Prove that u_n \leq \frac{3 \sqrt{n}}{2} for all n.

Topic: Sequence, Real Analysis
Difficulty Level: Medium

Indian Statistical Institute, ISI Subjective BStat & BMath 2023 Problem 6 UGB Solution.

Solution:

The function f(x)=x+\frac{1}{x} is strictly increasing on [1,\infty). Since u_{n+1}>u_n for all n\geq 1, and u_1=1, it follows that u_n\in [1,\infty) for all n\geq 1. We will verify the inequality u_n\leq \frac{3}{2}\sqrt{n} by induction on n. If n=1, the inequality is true. If n=2 the inequality is also true, as

    \[u_2=2<3\sqrt{2}=\frac{3}{2}\sqrt{2}.\]

Assume the inequality was verified for n\geq 2. Then

    \[u_{n+1}=f(u_{n})\leq f\left(\frac{3}{2}\sqrt{n}\right)=\frac{3}{2}\sqrt{n}+\frac{2}{3\sqrt{n}}=\frac{9n+4}{6\sqrt{n}}.\]

It is enough to prove that \frac{9n+4}{6\sqrt{n}}\leq \frac{3}{2}\sqrt{n+1}. Or, equivalently,

    \[9n+4\leq 9\sqrt{n(n+1)}.\]

Taking squares of both sides we obtain, equivalently,

    \[81n^2+72n+16\leq 81n^2+81n\]

or,

    \[9n\geq 16.\]

The inequality is true, since n\geq 2.

Video Solution:

Problem 7.

(a) Let n \geq 1 be an integer. Prove that X^n+Y^n+Z^n can be written as a polynomial with integer coefficients in the variables \alpha=X+Y+Z, \beta=X Y+Y Z+Z X and \gamma=X Y Z.

(b) Let G_n=x^n \sin (n A)+y^n \sin (n B)+z^n \sin (n C). where x, y, z, A, B, C are real numbers such that A+B+C is an integral multiple of \pi. Using (a) or otherwise, show that if G_1=G_2=0, then G_{\mathrm{n}}=0 for all positive integers n.

Topic: Polynomials, Mathematical Induction, Complex Numbers
Difficulty Level: Hard

Indian Statistical Institute, ISI Subjective BStat & BMath 2023 Problem 7 UGB Solution.

Solution:

(a) We prove the result by induction on n. If n=1, then X+Y+Z=\alpha. If n=2, then

    \[X^2+Y^2+Z^2=\left(X+Y+Z\right)^2-2XY-2YZ-2XZ=\alpha^2-2\beta.\]

If n=3, then

    \[X^3+Y^3+Z^3=\alpha^3-3\alpha\beta+3\gamma.\]

Assume that for every k\in \{1,\ldots,n\}, n\geq 3, there exists a polynomial with integer coefficients P_k in three variables, such that

    \[X^k+Y^k+Z^k=\alpha^k+P_k(\alpha,\beta,\gamma).\]

Then

    \[\begin{aligned} \alpha^{n+1}&=\alpha^n\times \alpha=\left(X^n+Y^n+Z^n-P_n(\alpha,\beta,\gamma)\right)\left(X+Y+Z\right)\\ &=X^{n+1}+Y^{n+1}+Z^{n+1}+X^n Y+X^nZ+XY^n+Y^nZ+XZ^n+YZ^n\\ &-\alpha P_n(\alpha,\beta,\gamma) \\ &=X^{n+1}+Y^{n+1}+Z^{n+1}+XY(X^{n-1}+Y^{n-1})+XZ(X^{n-1}+Z^{n-1})+YZ(Y^{n-1}+Z^{n-1})\\ &-\alpha P_n(\alpha,\beta,\gamma)\\ &=X^{n+1}+Y^{n+1}+Z^{n+1}+\left(XY+XZ+YZ\right) \left(X^{n-1}+Y^{n-1}+Z^{n-1}\right)\\ &-XYZ\left(X^{n-2}+Y^{n-2}+Z^{n-2}\right)-\alpha P_n(\alpha,\beta,\gamma)\\ &=X^{n+1}+Y^{n+1}+Z^{n+1}+\beta \left(\alpha^{n-1}+P_{n-1}(\alpha,\beta,\gamma)\right)-\gamma \left(\alpha^{n-2}+P_{n-2}(\alpha,\beta,\gamma)\right)-\alpha P_n(\alpha,\beta,\gamma) \end{aligned}\]

Hence,

    \[X^{n+1}+Y^{n+1}+Z^{n+1}=\alpha^{n+1}-\beta \alpha^{n-1}-\beta P_{n-1}(\alpha,\beta,\gamma)+\gamma \alpha^{n-2}+\gamma P_{n-2}(\alpha,\beta,\gamma)+\alpha P_{n}(\alpha,\beta,\gamma).\]

Hence, for every n\geq 1, X^n+Y^n+Z^n can be expressed as a polynomial in \alpha,\beta,\gamma with integer coefficients.

(b) According to part (a), for every n\geq1 there exists a polynomial Q_n in three variables with integer coefficients, such that

    \[X^n+Y^n+Z^n=Q_n\left(X+Y+Z,XY+YZ+XZ,XYZ\right).\]

Moreover, Q_1(\alpha,\beta,\gamma)=\alpha, Q_2(\alpha,\beta,\gamma)=\alpha^2-2\beta.

We note that

    \[G_n=\frac{x^n\left(e^{inA}-e^{-inA}\right)+y^n\left(e^{inB}-e^{-inB}\right)+z^n\left(e^{inC}-e^{-inC}\right)}{2i}=\]

    \[=\frac{\left((xe^{iA})^n+(ye^{iB})^n+(ze^{iC})^n\right)-\left((xe^{-iA})^n+(ye^{-iB})^n+(ze^{-iC})^n\right)}{2i}.\]

The first summand in the numerator is expressed as a polynomial Q_n applied to quantities

    \[xe^{iA}+ye^{iB}+ze^{iC}=:\alpha, xye^{i(A+B)}+xze^{i(A+C)}+yze^{i(B+C)}=:\beta,\]

    \[xyze^{i(A+B+C)}=:\gamma.\]

The second summand in the numerator is expressed as a polynomial Q_n applied to quantities

    \[xe^{-iA}+ye^{-iB}+ze^{-iC}=\overline{\alpha}, xye^{-i(A+B)}+xze^{-i(A+C)}+yze^{-i(B+C)}=\overline{\beta},\]

    \[xyze^{-i(A+B+C)}=\overline{\gamma},\]

where \overline{z} is a conjugate of a complex number z.

Hence, we get a representation

    \[G_n=\frac{Q_n(\alpha,\beta,\gamma)-Q_n(\overline{\alpha},\overline{\beta},\overline{\gamma})}{2i}.\]

It is enough to show that \alpha=\overline{\alpha}, \beta=\overline{\beta}, \gamma=\overline{\gamma} in order to deduce that all G_n=0.

 

Since G_1=0, and Q_1=\alpha, it follows that \alpha=\overline{\alpha}. Since G_2=0, it follows that

    \[Q_2(\alpha,\beta,\gamma)=Q_2(\overline{\alpha},\overline{\beta},\overline{\gamma}),\]

    \[\alpha^2-2\beta=\overline{\alpha}^2-2\overline{\beta}, \ \beta=\overline{\beta}.\]

Finally, since A+B+C is an integral multiple of \pi, A+B+C=m\pi, then

    \[\gamma=xyz e^{im\pi}=xyz e^{im\pi-2im\pi}=xyz e^{-im\pi}=\overline{\gamma}.\]

From relations \alpha=\overline{\alpha}, \beta=\overline{\beta}, \gamma=\overline{\gamma} it follows that G_n=0 for all n\geq 1.

Video Solution:

Problem 8.

Let f:[0,1] \rightarrow R be a continuous function which is differentiable on (0,1). Prove that either f is a linear function f(x)=a x+b or there exists t \in(0,1) such that |f(1)-f(0)|<\left|f^{\prime}(t)\right|.

Topic: Real Analysis, Differentiation
Difficulty Level: Medium

Indian Statistical Institute, ISI Subjective BStat & BMath 2023 Problem 8 UGB Solution.

Solution :

Assume that |f'(t)|\leq |f(1)-f(0)| for all t\in (0,1). If f(1)=f(0), then f is constant. Assume that f(1)>f(0). By the mean value theorem,

    \[|f(t)-f(0)|\leq (f(1)-f(0))t, \ t\in [0,1].\]

Assume that f(t)-f(0)<(f(1)-f(0))t for some t\in (0,1). Consider the function g(s)=f(s)-f(0), s\in [t,1]. Since |g'(s)|\leq f(1)-f(0), s\in [t,1), then by the mean value theorem

    \[|g(1)-g(t)|\leq (f(1)-f(0))(1-t).\]

Hence,

    \[|f(1)-f(t)|\leq (f(1)-f(0))(1-t)\]

and

    \[f(1)-f(0)=f(1)-f(t)+f(t)-f(0)<(f(1)-f(0))(1-t)+(f(1)-f(0))t=f(1)-f(0).\]

This is a contradiction. Hence, f(t)-f(0)\geq (f(1)-f(0))t for all t\in (0,1), and

    \[(f(1)-f(0))t\leq f(t)-f(0)\leq (f(1)-f(0))t, \ t\in [0,1].\]

It follows that f(t)=f(0)+(f(1)-f(0))t.

If f(1)<f(0), it is enough to apply the proved result to the fucntion -f(t).

Video Solution:

Past Year’s BMATH/BSTAT Papers & Solutions: Click Here


Indian Statistical Institute,ISI BMATH/BSTAT 2023 All UGA Objective Solutions: Click Here

Indian Statistical Institute,ISI BMATH/BSTAT 2021 All UGA Objective Solutions: Click Here

Indian Statistical Institute,ISI BMATH/BSTAT 2020 All UGA Objective Solutions: Click Here

Indian Statistical Institute,ISI BMATH/BSTAT 2020 All UGB Subjective Solutions: Click Here

Indian Statistical Institute,ISI BMATH/BSTAT 2020 UGA Answer Keys: Click Here

Indian Statistical Institute,ISI BMATH/BSTAT 2020 UGA Question Paper: Click Here

Indian Statistical Institute,ISI BMATH/BSTAT 2020 UGB Question Paper: Click Here