Unofficial archive — problems, solutions & results © IMC, reproduced with permission.

IMC / 2025 / Problems / Day 1, P5

IMC 2025 · Day 1 · P5

killer

For a positive integer nn, let [n]={1,2,,n}[n] = \{1, 2, \dots, n\}. Denote by SnS_n the set of all bijections from [n][n] to [n][n], and let TnT_n be the set of all maps from [n][n] to [n][n]. Define the order ord(τ)\operatorname{ord}(\tau) of a map τTn\tau \in T_n as the number of distinct maps in the set {τ,ττ,τττ,}\{\tau, \tau \circ \tau, \tau \circ \tau \circ \tau, \dots\} where \circ denotes composition. Finally, let f(n)=maxτSnord(τ)andg(n)=maxτTnord(τ).f(n) = \max_{\tau \in S_n} \operatorname{ord}(\tau) \quad \text{and} \quad g(n) = \max_{\tau \in T_n} \operatorname{ord}(\tau). Prove that g(n)<f(n)+n0.501g(n) < f(n) + n^{0.501} for sufficiently large nn.

(proposed by Fedor Petrov, St Petersburg State University)

Solution (official)

For every τTn\tau \in T_n we need to prove that ord(τ)f(n)+n0.501\operatorname{ord}(\tau) \leqslant f(n) + n^{0.501} (if nn is large enough). Denote by C(τ)C(\tau) the set of elements x[n]x \in [n] for which τk(x)=τ(τk(x))=x\tau^k(x) = \underbrace{\tau(\dots\tau}_{k}(x)\dots) = x for some k>0k > 0. It is immediate that C(τ)C(\tau) is a τ\tau-invariant set; let τc=τC(τ)\tau_c = \tau|_{C(\tau)}, that is a permutation on C(τ)C(\tau).

Let NN be the order of this permutation τc\tau_c, clearly Ng(n)N \leqslant g(n). Consider an arbitrary element x[n]C(τ)x \in [n] \setminus C(\tau). The sequence x,τ(x),τ2(x),x, \tau(x), \tau^2(x), \dots is eventually periodic, but not from the beginning, because xC(τ)x \notin C(\tau).

Let h(x)>0h(x) > 0 be the minimal number for which τh(x)(x)\tau^{h(x)}(x) is in the period; equivalently, this is the minimal number for which τh(x)(x)C(τ)\tau^{h(x)}(x) \in C(\tau). Let R=maxx[n]C(τ)h(x)R = \max\limits_{x \in [n] \setminus C(\tau)} h(x). Note that τR(x)=τR+N(x)\tau^R(x) = \tau^{R+N}(x) for all x[n]x \in [n], since τR(x)C(τ)\tau^R(x) \in C(\tau) for all x[n]x \in [n]. Therefore, ord(τ)N+R\operatorname{ord}(\tau) \leqslant N + R. Thus, if R<n0.501R < n^{0.501}, we are done.

Now assume that Rn0.501R \geqslant n^{0.501}, that is, h(x)n0.501h(x) \geqslant n^{0.501} for some xC(τ)x \notin C(\tau). It yields C(τ)nn0.501|C(\tau)| \leqslant n - n^{0.501}. Consider the cycle lengths of the permutation τc\tau_c. We claim that there exists a prime number p<n0.501p < n^{0.501} which does not divide any of these lengths. Indeed, otherwise the sum of cycle lengths of τc\tau_c is not less than the sum of all prime numbers not exceeding n0.501n^{0.501} (because each positive integer is not less than the sum of its prime divisors, that in turn follows from aba+bab \geqslant a + b for a,b2a, b \geqslant 2). But for large nn, the number of prime numbers less than n0.501n^{0.501} is at least n0.5009n^{0.5009} by some weak version of Prime Number Theorem (that also has many short proofs). Therefore, their sum exceeds nn, contradiction.

Now we may consider the permutation τ0Sn\tau_0 \in S_n which acts as τc\tau_c on C(τ)C(\tau) and has a cycle of length pp on [n]C(τ)[n] \setminus C(\tau). The order of τ0\tau_0 is not less than pNp \cdot N, therefore Nf(n)/pf(n)/2N \leqslant f(n)/p \leqslant f(n)/2, and by the above argument we get ord(τ)N+Rf(n)/2+n<f(n)\operatorname{ord}(\tau) \leqslant N + R \leqslant f(n)/2 + n < f(n) for large nn.

How the field did

contestants scored
425
average (of 10)
1.02
solved (≥ 80%)
4.7%
near-0 (≤ 10%)
79.1%
discrimination
0.49

Score distribution (field cohort)

Computed on contestants with a meaningful total (field cohort); discrimination is the corrected item–total correlation.

Similar problems

IMC 1999 · Day 2 · P12killeravg 0.4/10 · solved 2% · near-0 91% · disc 0.30
IMC 2016 · Day 1 · P5killeravg 0.3/10 · solved 2% · near-0 96% · disc 0.29
IMC 2016 · Day 2 · P8very hardavg 2.1/10 · solved 10% · near-0 62% · disc 0.38
IMC 2022 · Day 1 · P3very hardavg 2.3/10 · solved 12% · near-0 55% · disc 0.64