Maths Olympiad Prep

Library / /18 of 36

Combinatorics Difficulty 6.6 National Olympiad Prove it Italy

Problem:

Dedalus buys a finite number of binary strings (each of finite length) made up of the binary digits 0 and 1. For each of the strings he buys, Dedalus pays (1/2)L(1 / 2)^L drachmas, where LL denotes the length of the string. The Minotaur escapes if he manages to find an infinite sequence of binary digits that does not contain any of the strings bought by Dedalus. Dedalus must trap the Minotaur.
For example, if Dedalus were to buy the strings 00 and 11, paying half a drachma, the Minotaur could escape with the infinite sequence 0101010101010101\ldots.
On the other hand, Dedalus can trap the Minotaur with an expense of 75 cents of a drachma: it is enough, for example, to buy the strings 0 and 11, or the strings 00, 11, 01.
Determine all positive integers cc for which Dedalus can trap the Minotaur with an expense less than or equal to cc cents of a drachma.

Solutions — 2

Solution 1

Solution:

For every c>0c>0 (not necessarily an integer either), the Lucchese Dedalus, of Genoese origin, can imprison the Minotaur while spending less than cc cents of a drachma.

Lemma. For every pair of positive integers kk and mm, the expense of buying all strings of length mkmk in which kk or more consecutive zeros never occur is less than or equal to
(112k)m \left(1-\frac{1}{2^{k}}\right)^{m}
drachmas.

To prove this, observe first of all that the cost of each of these strings is 2mk2^{-mk} drachmas. This expense must be multiplied by the total number of strings of length mkmk that do not contain kk (or more) consecutive zeros. The exact number of these strings is hard to compute, but in order to obtain an upper bound on the price Dedalus pays, it suffices to find an upper estimate.

To do this, we divide the string of length mkmk into mm blocks of length kk, and observe that each block of length kk can be any binary string except the one made up of only 0 digits. Since the binary strings of length kk number 2k2^{k}, it follows that for each block there are 2k12^{k}-1 possible choices. Having mm blocks, the total number of strings is at most (2k1)m\left(2^{k}-1\right)^{m}, hence the total cost does not exceed
12mk(2k1)m=(112k)m \frac{1}{2^{mk}} \cdot\left(2^{k}-1\right)^{m}=\left(1-\frac{1}{2^{k}}\right)^{m}
as required.

Given now any c>0c>0, we choose kk such that 1/2k<c/2001 / 2^{k}<c / 200, and once kk is fixed we choose mm such that
(112k)m<c200 \left(1-\frac{1}{2^{k}}\right)^{m}<\frac{c}{200}
At this point Dedalus will buy the string consisting of kk digits 0, which is able to block all sequences that contain, somewhere, at least kk consecutive digits 0, plus all strings of length mkmk in which kk or more consecutive zeros never occur, able to block all the remaining possible sequences simply by considering the first block of mkmk characters.

All this while spending less than c/200+c/200=c/100c / 200+c / 200=c / 100 drachmas!

Solution 2

Solution:

Alternative solution

We present an alternative version of the preceding Lemma, which requires the purchase of a smaller number of strings, and still allows us to conclude in an analogous manner. For simplicity, we call a kk-sequence any infinite sequence in which k+1k+1 (or more) consecutive zeros never occur.

Lemma. For every pair of integers k0k \geq 0 and m0m \geq 0, Dedalus can imprison all kk-sequences by buying (k+1)m(k+1)^{m} strings, all ending with the digit 1, spending in total
(112k+1)m \left(1-\frac{1}{2^{k+1}}\right)^{m}
drachmas.

To prove this, suppose kk is a fixed parameter and proceed by induction on mm. The case m=0m=0 is easy, since it suffices to buy the string 1.

Suppose now the statement is true for some mm, and let us prove it for m+1m+1. Let S1S_{1}, \ldots, S(k+1)mS_{(k+1)^{m}} be strings capable of imprisoning any kk-sequence. We now replace each of these strings, say the string SiS_{i}, with the k+1k+1 strings obtained by appending to the right of SiS_{i} the strings
1,01,001,,00k1 1, \quad 01, \quad 001, \quad \ldots, \quad \underbrace{0 \ldots 0}_{k} 1
The key observation is that, if SiS_{i} is able to block a certain kk-sequence, then one of the new strings is also able to block the same kk-sequence, simply by positioning itself at the same point. Indeed, a kk-sequence that contains SiS_{i} can continue to the right of SiS_{i} in only one of the k+1k+1 indicated ways.

Letting LiL_{i} denote the length of SiS_{i}, the cost of the k+1k+1 strings that replace SiS_{i} is
12Li+1++12Li+k+1=12Li(12++12k+1)=12Li(112k+1) \frac{1}{2^{L_{i}+1}}+\cdots+\frac{1}{2^{L_{i}+k+1}}=\frac{1}{2^{L_{i}}}\left(\frac{1}{2}+\cdots+\frac{1}{2^{k+1}}\right)=\frac{1}{2^{L_{i}}} \cdot\left(1-\frac{1}{2^{k+1}}\right)
and hence it equals that of SiS_{i} multiplied by a fixed factor less than one.

If we now apply the same procedure to all (k+1)m(k+1)^{m} strings that were suitable at step mm, we obtain (k+1)m+1(k+1)^{m+1} strings that are still suitable and have a total cost equal to the previous one multiplied by the common fixed factor (11/2k+1)\left(1-1 / 2^{k+1}\right). Moreover, they still all end with the digit 1.

Want a route through all this instead of an archive? The track puts 2,000 problems in a working order, from AMC 10 level to the IMO shortlist.

Source: MathNet, licensed CC-BY-4.0. Statement translated into English from en; metadata (topic, difficulty) added by this project.