Maths Olympiad Prep

Library / /352 of 397

, 2022

Combinatorics Difficulty 7.0 National Olympiad Prove it Taiwan

A certain country has nn cities, and between any two cities there is a unique road, which is designated to be traveled in only one direction. A path from city XX to city YY is a sequence of one-way roads such that a person can travel from XX to YY using these roads, without visiting the same city twice along the way. A set of paths is called independent if and only if no two paths among them use the same road.

For a city XX, let nXn_X denote the total number of one-way roads leading out from XX. For two distinct cities XX and YY, let NXYN_{XY} denote the maximum possible number of paths from XX to YY among all independent sets of paths. Prove that NXY=NYXN_{XY} = N_{YX} if and only if nX=nYn_X = n_Y.

Solution

Let us use XYX \to Y to denote that XX has a one-way road leading to YY, and use XYX \rightsquigarrow Y to denote a path from XX to YY. Let
FX:={C:XC}FX:={C:XC} \mathcal{F}_X := \{C : X \to C\} \quad \mathcal{F}_X := \{C : X \leftarrow C\}
denote the set of cities from (to) which all one-way roads lead to (from) XX; note that nX=FXn_X = |\mathcal{F}_X|.

We call a path short if and only if the number of roads in it is 2\le 2; otherwise it is called a long path. We first prove the following Lemma:

Lemma: Let P\mathcal{P} be an independent set consisting of paths from XX to YY, with P=p|\mathcal{P}| = p. Then there exists an independent set P\mathcal{P}' consisting of paths from XX to YY, satisfying Pp|\mathcal{P}'| \ge p and containing all short paths from XX to YY.

Proof: Let us perform the following operations on P\mathcal{P}.

1. If XYX \to Y and this path is not in P\mathcal{P}, then add it to P\mathcal{P}.
2. Suppose there exists a city CC satisfying XCYX \to C \to Y and this path is not in P\mathcal{P}. Note that P\mathcal{P} is independent.
- If P\mathcal{P} has only one path containing XCX \to C or CYC \to Y, remove that path and add XCYX \to C \to Y.
- If P\mathcal{P} contains XCYX \to C \rightsquigarrow Y and XCYX \rightsquigarrow C \to Y, remove these two paths, and add XCYX \to C \to Y and XCYX \rightsquigarrow C \rightsquigarrow Y to P\mathcal{P}'.

It is easy to verify that the resulting P\mathcal{P}' is independent, Pp\mathcal{P}' \ge p, and contains all short paths from XX to YY. □

Return to the original problem. Take any two cities XX and YY, and without loss of generality assume XYX \to Y. Take any independent set of paths P\mathcal{P} consisting of paths from XX to YY with size NXYN_{XY}. Our goal is to obtain, through operations, an independent set of paths Q\mathcal{Q} consisting of paths from YY to XX with size NXY(nXnY)N_{XY} - (n_X - n_Y). In this case, by the maximality of NXYN_{XY}, we have
NYXNXY(nXnY)and similarlyNXYNYX(nYnX) N_{YX} \ge N_{XY} - (n_X - n_Y) \quad \text{and similarly} \quad N_{XY} \ge N_{YX} - (n_Y - n_X)

Thus we obtain NXYNYX=nXnYN_{XY} - N_{YX} = n_X - n_Y, which proves the original problem.

To obtain Q\mathcal{Q}, we first use the Lemma to obtain P\mathcal{P}'. Note the following facts:
- P\mathcal{P}' contains all short paths from XX to YY, and the number of these short paths equals FXFY+1|\mathcal{F}_X \cap \mathcal{F}_Y| + 1.
- The paths in P\mathcal{P}' cannot contain any road that would be used by a short path from BB to AA.
- All long paths in P\mathcal{P}' must be of the form XCDYX \to C \sim D \to Y. Moreover, by the previous property:
* CFYC \in \mathcal{F}_Y (otherwise the path would not be independent from XCYX \to C \to Y);
* DFXD \in \mathcal{F}_X (otherwise the path would share the road YCY \to C with YCXY \to C \to X);

We construct Q\mathcal{Q} in the following way:
- Add all short paths from YY to XX into Q\mathcal{Q}.
- For every XCDYX \to C \sim D \to Y in P\mathcal{P}', add YCDXY \to C \sim D \to X to Q\mathcal{Q}.

It is easy to verify that Q\mathcal{Q} is independent. Moreover, since there are FXFY+1|\mathcal{F}_X \cap \mathcal{F}_Y| + 1 short paths from XX to YY, and FXFY|\mathcal{F}_X \cap \mathcal{F}_Y| short paths from YY to XX,
Q=P(FXFY+1)+FXFYNXY{(FXFY+1)FXFY} \begin{aligned} |\mathcal{Q}| &= |\mathcal{P}'| - (|\mathcal{F}_X \cap \mathcal{F}_Y| + 1) + |\mathcal{F}_X \cap \mathcal{F}_Y| \\ &\geq N_{XY} - \{(|\mathcal{F}_X \cap \mathcal{F}_Y| + 1) - |\mathcal{F}_X \cap \mathcal{F}_Y|\} \end{aligned}
Also note that
nXnY=FXFY=FXFY+{XY}FXFY=(FXFY+1)FXFY. \begin{aligned} n_X - n_Y &= |\mathcal{F}_X| - |\mathcal{F}_Y| = |\mathcal{F}_X \cap \mathcal{F}_Y| + |\{X \to Y\}| - |\mathcal{F}_X \cap \mathcal{F}_Y| \\ &= (|\mathcal{F}_X \cap \mathcal{F}_Y| + 1) - |\mathcal{F}_X \cap \mathcal{F}_Y|. \end{aligned}
Combining the two equations above, we have
QNXY(nXnY), |\mathcal{Q}'| \geq N_{XY} - (n_X - n_Y),
which completes the proof of the original problem.

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 zh; metadata (topic, difficulty) added by this project.