35 Mathematical Proofs
35 Mathematical Proofs
35 Mathematical Proofs
Mathematical Proofs
John C. Mitchell & Maggie Johnson Department of Computer Science Stanford University
To try or to ascertain by an experiment, or by a test or standard.... To evince, establish, or ascertain, as truth, reality, or fact, by argument ...
These two definitions reflect a distinction developed by philosophers, namely, the difference between analytic and synthetic statements. Putting it briefly, some statements are best confirmed by experiment and other statements are best confirmed by argument. If I were to tell you, while sitting in a room with no windows, that it is raining outside right now, then there is no amount of argument that would be as convincing as stepping outside to see for yourself. The statement it is raining'' is not an analytical statement about the relationship between concepts, but a synthetic proposition about the world that might or might not be true at any given time. In contrast, mathematical statements, such as x y(x+y = y+x) are analytical statements that are better proved by argument than by experiment.
2 Generally speaking, analytic statements can be established by reasoning about the definitions of words and the relationships between the concepts involved. For example, a proof of the Pythagorean Theorem relies on the definitions of hypotenuse, right triangle, square of a number, and so on.
3 The interesting thing here is that it was gold itself that was valued. When you paid with a gold coin, you weren't providing a promise to deliver value; you were paying with the valuable thing.
Gold standard: The true, undoubted proof of a statement is its formal proof. The proofs we did in Fitch were true formal proofs. As we saw in Fitch, the definition of a formal proof system is clear and unambiguous. We can mechanically check that a formal proof is written correctly, without using any creativity or ingenuity. Since the formal proof system we use has been proved sound, a correctly written formal proof demonstrates without a doubt that the conclusion follows from the assumptions.
Silver certificates: A semi-formal proof should be a simplified way of conveying the same information as a formal proof. A skilled mathematician, logician or computer scientist should accept a semi-formal proof if that person can see how the argument could be converted to a formal proof, given enough time and energy. We think of semi-formal proofs as statement/reason chart proofs or detailed prose proofs. Federal reserve notes: In practice, many mathematical proofs do not contain all of the details that would be needed to construct a formal proof. Experienced mathematicians and computer scientists who do analytical work write informal proofs that are convincing to their audience. They have learned what steps can be left out so that the soundness of the argument and the flow of the proof are not affected.
In this class, please give formal proofs when the problem asks you to do so, and silver certificate proofs otherwise. In other words, even if you write your proof out in English, think about the problem carefully enough so that you are convinced that you could produce a formal proof if you needed to. As you gain more experience, this will become easier to do. Here are some things you can omit from proofs (unless you are asked to give a formal proof), assuming that an intelligent teaching assistant or grader is able to fill in the missing details (thats your audience): 1. Simple algebraic reasoning, of the sort you would do in a high-school algebra class. For example, you could simplify 3(x + 2x)) + 4(x 1) to 13x 4 in a single step without going through the calculation. 2. Simple tautologies and logical identities can be omitted, as long as the problem is not to prove a propositional tautology. 3. Anything previously proven on a problem set, in a handout or in the text, can simply be cited as a step in a proof.
4. Examples of Form
Formal Proof (Gold Standard)
Given: (P^R) Q P Prove: R Q 1. (P^R) Q 2. P 3. R 4. P ^ R 5. Q 6. R Q Therefore, R Q ^ Intro 2, 3 Elim 1, 4 Intro, 3-5
n3 - m3 is even
"Helper proof": The product of an even integer and an integer is even. m=2*k m is even
6 x=m*n x = (2k) * n x = 2(kn) kn is an integer x is even given substitution associative k and n are integers and so is their product definition of even
Therefore we have proven: For all integers m and n, if n - m is even, then n3 - m3 is even. Example: Prose Form Prove: The set of prime numbers is infinite. Proof by Contradiction: Suppose not; i.e. suppose the set of prime numbers is finite. Then all prime numbers can be listed, say, in ascending order: p1 = 2, p2 = 3, p3 = 5, p4 = 7, p5 = 11, , pn. Consider the integer N = (p1 * p2 * p3 * * pn) + 1. Then, N > 1. We will show elsewhere that any integer n>1 is divisible by a prime, so N is divisible by some prime number p. But none of the primes p1...pn divides N, because dividing N by each of them produces a remainder of 1. So p must be some prime not in the list p1, ..., pn, which contradicts the assumption that the list includes every prime. So the set of prime numbers is infinite.
Informal Proof (Federal Reserve Note) (from Rosen, Discrete Mathematics and its
Application) Prove: For every positive integer n, there is an integer divisible by more than n primes. Proof: Let p1, p2, , pn+1 be the first n+1 prime numbers. Then p1*p2**p n+1 is divisible by more than n primes. Whats missing from this proof?
Remember: Give formal proofs when the problem asks you to do so, and semi-formal proofs otherwise. In other words, even if you write your proof out in English, think about the problem carefully enough so that you are convinced that you could produce a formal proof if you needed to. Do not give informal proofs unless we specifically ask you to do so.