Mathematical papers have a very distinctive style, one that became established early in the twentieth century. We intend to give a description of what mathematicians actually produce when they write.
A typical paper is usually a mixture of formal and informal writing. Ideally (but by no means always), the author writes a readable introduction that tells the reader what to expect from the rest of the paper. And if the paper is divided into sections, as most papers are unless they are quite short, then it is also very helpful to the reader if each section can begin with an informal outline of the arguments to follow. But the main substance of the paper has to be more formal and detailed, so that readers who are prepared to make a sufficient effort can convince themselves that it is correct.
The object of a typical paper is to establish mathematical statements. Sometimes this is an end in itself: for example, the justification for the paper may be that it proves a conjecture that has been open for twenty years. Sometimes the mathematical statements are established in the service of a wider aim, such as helping to explain a mathematical phenomenon that is poorly understood. But either way, mathematical statements are the main currency of mathematics.
The most important of these statements are usually called theorems, but one also finds statements called propositions, lemmas, and corollaries. One cannot always draw sharp distinctions between these kinds of statements, but in broad terms this is what the different words mean. A theorem is a statement that you regard as intrinsically interesting, a statement that you might think of isolating from the paper and telling other mathematicians about in a seminar, for instance. The statements that are the main goals of a paper are usually called theorems. A proposition is a bit like a theorem, but it tends to be slightly “boring.” It may seem odd to want to prove boring results, but they can be important and useful. What makes them boring is that they do not surprise us in any way. They are statements that we need, that we expect to be true, and that we do not have much difficulty proving.
Here is a quick example of a statement that one might consider a proposition. The associative law for a binary operation states that
x * (y * z) = (x * y) * z.
One often describes this law informally by saying that “brackets do not matter.” However, while it shows that we can write x * y * z without fear of ambiguity, it does not show quite so obviously that we can also write a * b * c * d * e, for example. How do we know that, just because the positions of brackets do not matter when you have three objects, they do not matter when you have more than three?
Many mathematics students go happily through university without noticing that this is a problem. It just seems obvious that the associative law shows that brackets do not matter. And they are basically right: although it is not completely obvious, it is certainly not a surprise and turns out to be easy to prove. Since we often need this result and could hardly call it a theorem, we might call it a proposition instead. To get a feel for how to prove it, you might wish to show that the associative law implies that
a * ((b * c) * d) * e = a * (b * ((c * d) * e)).
Then you can try to generalize what it is you are doing.
Often, if you are trying to prove a theorem, the proof becomes long and complicated, in which case if you want anybody to read it you need to make the structure of the argument as clear as possible. One of the best ways of doing this is to identify subgoals, which take the form of statements intermediate between your initial assumptions and the conclusion you wish to draw from them. These statements are usually called lemmas. Suppose, for example, that you are trying to give a very detailed presentation of the standard proof that √2 is irrational. One of the facts you will need is that every fraction p/q is a reduced fraction r/s with r and s not both even, and this fact requires a proof. For the sake of clarity, you might well decide to isolate this proof from the main proof and call the fact a lemma. Then you have your task in two separate tasks: proving the lemma, and proving the main theorem using the lemma. One can draw a parallel with computer programming: when writing a complicated program, it is good practice to divide your main task into subtasks and then write small programs for them, which you can then treat as “black boxes,” to be called upon by the main program whenever they are useful.
Some lemmas are difficult to prove and are useful in many different contexts, so the most important lemmas can be more important than the least important theorems. However, a general idea is that a result will be called a lemma if its main reason for proving it is in order to use it as a stepping stone toward the proofs of other results.
A corollary of a mathematical statement is another statement that follows easily from it. Sometimes the main theorem of a paper is followed by several corollaries, which advertise the power of the theorem. Sometimes the main theorem itself is labeled a corollary, because all the work is done in a different, earlier, less punchy statement from which the theorem follows very easily. If this happens, the author may wish to make clear that the corollary is the main result of the paper, and other authors would refer to it as the main theorem.
A mathematical statement is established by means of a proof. It is a remarkable feature of mathematics that proofs are possible: that, for example, an argument invented by Euclid over two thousand years ago can still be accepted today and regarded as a completely convincing demonstration. It took until the late nineteenth and early twentieth centuries for this phenomenon to be properly understood, when the language of mathematics was formalized. Then it became possible to make precise the notion of a proof as well. From a logician’s point of view a proof is a sequence of mathematical statements; each written in formal language, with the following properties: the first few statements are initial assumptions, or premises; each remaining statement in the sequence follows from earlier ones by means of rules that are also simple logical deductions (for instance rules such as “if P and Q then R is true,” where “∧” is the logical symbol for “and”); and the final statement in the sequence is the fact that is to be proved.
The above idea of a proof is a considerable idealization of what actually happens in a normal mathematical paper under the heading “Proof.” That is because a purely formal proof would be very long and almost impossible to read. A shorter, but fact argument-based version of the formal proof provides a very valuable underpinning for the complicated arguments that arise in real mathematical disputes. If a mathematician produces an argument that is hard to understand, then there has to be some way of determining whether it is correct or is too vague to be properly meaningful and might generate confusion. This will usually expose a mistake or make it clearer why the argument works.
Another very important component of mathematical papers is definitions. Some definitions are given simply because they enable one to speak more concisely. For example, if I am proving a result about triangles and I keep needing to consider the distances between the vertices and the opposite sides, then it is a nuisance to have to say “the distances from A, B, and C to the lines BC, AC, and AB, respectively,” so instead I will probably choose a word like “altitude” and write, “Given a vertex of a triangle, define its altitude to be the distance from that vertex to the opposite side.” If I am looking at triangles with obtuse angles, then I will have to be more careful: “Given a vertex A of a triangle ABC, define its altitude to be the distance from A to the unique line that passes through B and C.” From then on, I can use the word “altitude” and the exposition of my proof will be much more crisp.
Definitions like this are mere definitions of convenience. When the need arises, it is pretty obvious what to do and one does it. But the really interesting definitions are ones that are far from obvious and that make you think in new ways once you know them. A very good example is the definition of the derivative of a function. If you do not know this definition, you will have no idea how to find out for which nonnegative x the function
f(x) = 2x³ – 3x² – 6x + 1
takes its smallest value. If you do know it, then the problem becomes a simpler exercise. That is perhaps an exaggeration, since you also need to know that the minimum will occur either at 0 or at a point where the derivative vanishes, and you will need to know how to differentiate f(x), but these are simple facts—propositions rather than theorems—and the real breakthrough is the concept itself.
There are many other examples of definitions like this, but interestingly they are more common in some branches of mathematics than in others. Some mathematicians will tell you that the main aim of their research is to find the right definition, after which the whole area will be illuminated. Yes, they will have to write proofs, but if the definition is the one they are looking for, then these proofs will be fairly straightforward. And yes, there will be problems they can solve with the help of the new definition, but, like the minimization problem above, these will not be central to the theory. Rather, they will demonstrate the power of the definition. For other mathematicians, the main purpose of definitions is to prove theorems, but even very theorem-oriented mathematicians will from time to time.
This brings us to mathematical problems. The main aim of an article in mathematics is usually to prove theorems, but one of the reasons for reading an article is to advance one’s own research. It is therefore very welcome if a theorem is proved by a technique that can be used in other contexts. It is also very welcome if an article contains some good unsolved problems. By way of illustration, let us look at a problem that most mathematicians would not take all that seriously, and try to see what it lacks.
A number is called palindromic if its representation in base 10 is a palindrome: some simple examples are 22, 131, and 548 845. Of these, 131 is interesting because it is also a prime. Let us try to find some more prime palindromic numbers. Single-digit primes are of course palindromic, and two-digit palindromic numbers are multiples of 11, so only 11 itself is also a prime. So let us move quickly on to three-digit numbers. Here there turn out to be several examples: 101, 131, 151, 181, 191, 313, 353, 373, 383, 727, 757, 787, 797, 919, and 929. It is not hard to show that every palindromic number with an even number of digits is a multiple of 11, but the palindromic primes do not stop at 929—for example, 10 301 is the next smallest.
And now anybody with a modicum of mathematical curiosity will ask the question: are there infinitely many palindromic primes? This, it turns out, is an unsolved problem. It is believed (on the combined grounds that the primes should be sufficiently random and that palindromic numbers with an odd number of digits do not seem to have any particular reason to be [factorizable](https://mathworld.wolfram.com/Factorizable.html)) that there are, but nobody knows how to prove it.
This problem has the great virtue of being easy to understand, which makes it appealing in the way that Fermat’s Last Theorem and Goldbach’s Conjecture are appealing. And yet, it is not a central problem in the way that those two are: most mathematicians would put it into a mental box marked “recreational” and forget about it.
What is the reason for this dismissive attitude? Are the primes not central objects of study in mathematics? Well, yes they are, but palindromic numbers are not. And the main reason they are not is that the definition of “palindromic” is extremely artificial. If you know that a number n is palindromic, what you know is less a feature of the number itself than of a feature of the particular way that, for accidental historical reasons, we choose to represent it. In particular, the property depends on our choice of the number 10 as our base. For instance, if we write 131 in base 3, then it becomes 11212, which is no longer the same when written backwards. That is, a prime number is prime however you write it.
Though persuasive, this is not quite a complete explanation, since there could conceivably be interesting facts about the base 10 number 10, or at least about numbers in base 10, in an essential way. For example, the question of whether there are infinitely many primes of the form 2ⁿ − 1 is considered interesting, despite the use of the particular number 2. However, the choice of 2 can be justified here: aⁿ − 1 has a factor a − 1, so for any larger integer the answer would be no. Moreover, numbers of the form 2ⁿ − 1 have special properties that make them more likely to be prime.
But even if we replace 10 by the “more natural” number 2 and look at numbers that are palindromic when written in binary, we still do not obtain a property that would be considered a serious topic for research. Suppose that, given an integer n, we define r(n) to be the reverse of n —that is, the number obtained if you write n in binary and then reverse its digits. Then a palindromic number, in this sense, is a number n such that n = r(n). But the function r(n) is very strange and “unmathematical.” For instance, the reverses of the numbers from 1 to 20 are 1, 1, 3, 1, 5, 3, 7, 1, 9, 5, 13, 3, 11, 7, 15, 1, 17, 9, 25, and 5, which gives us a sequence with no obvious pattern. Indeed, when one calculates this sequence, one realizes that it is even more artificial than it at first seemed. One might imagine that the reverse of the reverse of a number is the number itself, but that is not so. If we take the number 10, for example, it is 1010 in binary, so its reverse is 0101, which is the number 5. But we would normally write as 101, so the reverse of 5 is not 10 but 5. But we cannot solve this problem by deciding to write 5 as 0101, since then we would have the problem that 5 was no longer palindromic, when it clearly ought to be.
This does not mean that nobody would be interested in a proof that there are infinitely many palindromic primes? Not at all. It can be shown quite easily that the number of palindromic numbers less than n is on the order of √n, which is a very small fraction indeed. It is notoriously hard to prove results about primes in sparse sets like this, so a solution to this conjecture would be a big breakthrough. However, the definition of “palindromic” is so artificial that there seems to be no way of using it that is detailed way in a mathematical proof. The only realistic hope of solving this problem would be to prove a much more general result, of which this would be just one of many consequences. Such a result would be wonderful, and undeniably interesting, but you will not discover it by thinking about palindromic numbers. Instead, you would be better off either trying to formulate a more general question, or else looking at a more natural problem of a similar kind. An example of the latter is this: are there infinitely many primes of the form m² + 1 for some positive integer m?
Perhaps the most important feature of a good problem is generality: the solution to a good problem should usually have ramifications beyond the problem itself. A more accurate word for this desirable quality is “generalizability,” since some excellent problems may look rather specific. For example, the statement that √2 is irrational looks as though it is about just one number, but once you know how to prove it, you will have no difficulty in proving that √3 is irrational as well, and in fact the proof can be generalized to a much wider class of numbers. It is quite common for a good problem to look uninteresting until you start to think about it. Then you realize that it has been asked for a reason: it might be the “first difficult case” of a more general problem, or it might be just one well-chosen example of a cluster of problems, all of which appear to run up against the same difficulty.
Sometimes a problem is just a question, but frequently the person who asks a mathematical question has a good idea of what the answer is. A conjecture is a mathematical statement that the author firmly believes but cannot prove. As with problems, some conjectures are better than others, the very best conjectures can have a major effect on the direction of mathematical research.