A WAY TO WRITE PROOFS

MITYA BOYARCHENKO

In scientific work it is right to

hold every author guilty until
he proves himself innocent.
Oscar Zariski
Mathematics is the art of
saying the right words in the
correct order.
Vladimir Drinfeld

1. Introduction
This text is intended for those who are currently taking a proof-based course (such as
Mathematics 185 or 295 at the University of Michigan) and who have never seen proofs
before and/or feel uncomfortable with proofs.
To illustrate the goal of this text, consider the following two passages. The first one is
from Eleonora by E.A. Poe:
I am come of a race noted for vigor of fancy and ardor of passion. Men
have called me mad; but the question is not yet settled, whether madness
is or is not the loftiest intelligence whether much that is glorious
whether all that is profound does not spring from disease of thought
from moods of mind exalted at the expense of the general intellect. They
who dream by day are cognizant of many things which escape those who
dream only by night. In their gray visions they obtain glimpses of eternity,
and thrill, in awakening, to find that they have been upon the verge of the
great secret. In snatches, they learn something of the wisdom which is of
good, and more of the mere knowledge which is of evil.
If you dont know what the second one is paraphrasing, congratulations: you slept
through the last two weeks of March 2011:
I woke up at 7am in the morning. Then I got fresh. Then I went downstairs.
Then I had my bowl of cereal. [...] Then I got down to the bus stop to
catch my bus. Then I saw my friends. [...]
1

MITYA BOYARCHENKO

Both of these passages are coherent, understandable and grammatically correct. Only
one of them constitutes an example of good literature.
My goal is to teach you to write proofs in the style of the second passage rather than
the first. It is far easier to learn how to write correct proofs than it is to learn how
to write elegant ones. In fact, the one model I am going to present can be adapted to
any situation that will come up in a first-year mathematics course, and using this model
in every proof that you write will make your graders life much easier and your scores
significantly higher.
2. The model
Exercise. If a is a real number and a 0, prove that a has a real square root.
Solution.
Step 1. Consider the following set:



S = x R  x2 a .

(1)

a 0.

(2)

0 S.

(3)

S is nonempty.

(4)

By assumption,
By (1) and (2), we have
(3) implies that
Step 2. (Optional sentence: Let us show that S has an upper bound.) We have
a1

or

a < 1.

(5)

These two cases are considered in Steps 2.1 and 2.2 below.
Step 2.1. Suppose that
a 1.

(6)

a is an upper bound for S.

(7)

We claim that if (6) holds, then

To prove (7) we argue by contradiction. Thus we assume that
a is not an upper bound for S.

(8)

By the definition of an upper bound, (8) implies that

there exists x S such that x > a.

(9)

So let us choose x S such that

x > a.

(10)

x 1.

(11)

By (6) and (10), we have

A WAY TO WRITE PROOFS

Therefore
x2 = x x x 1 = x.

(12)

By (10) and (12), we have

x2 > a.
(13)
But (13) contradicts (1) and the fact that x S. Since assuming (6) and (8) led us to a
contradiction, we conclude that if (6) holds, then so does (7).
Step 2.2. Suppose that
a < 1.

(14)

1 is an upper bound for S.

(15)

We claim that if (14) holds, then

To prove (15) we argue by contradiction. Thus we assume that
1 is not an upper bound for S.

(16)

By the definition of an upper bound, (16) implies that

there exists x S such that x > 1.

(17)

So let us choose x S such that

x > 1.
(18)
In Step 2.1 above we showed that if x 1, then x2 x. In particular, (18) implies that
x2 x > 1.

(19)

By (14) and (19), we have

x2 > 1 > a.
(20)
But (20) contradicts (1) and the fact that x S. Since assuming (14) and (16) led us to
a contradiction, we conclude that if (14) holds, then so does (15).
Step 2.3. Because of (5), we showed in Steps 2.1 and 2.2 that in all possible cases,
S has an upper bound.
Step 3. According to the least upper bound axiom (or property, if you like),
if X R is a nonempty subset that has an upper bound,
then X has a least upper bound.

(21)

(22)

Using (4), (21) and (22), we see that

S has a least upper bound.

(23)

Let us define
b = sup(S).
Note that since b is in particular an upper bound for S, (3) implies that
b 0.

(24)
(25)

MITYA BOYARCHENKO

To complete the solution it suffices to prove that

b2 = a.

(26)

Step 4. To prove (26) we argue by contradiction. There are three possibilities:

b2 < a

or

b2 > a

or

b2 = a.

(27)

Step 4.1. Assume that

b2 < a.

(28)

We will show that (28) leads to a contradiction. By (25), we have

2b + 1 > 0.
In particular, 2b + 1 6= 0, which allows us to define1


a b2
 = min 1,
.
2b + 1

(29)

(30)

By (28), we have
a b2 > 0.

(31)

a b2
> 0.
2b + 1

(32)

0 <  1.

(33)

(b + )2 = b2 + 2b + 2 = b2 +  (2b + ).

(34)

By (29) and (31), we have

By (32) and (30), we have

Now observe that
By (33) and (34), we have
(b + )2 b2 +  (2b + 1).

(35)

 (2b + 1) a b2 .

(36)

(b + )2 b2 + a b2 = a.

(37)

b +  S.

(38)

b +  > b.

(39)

By (29) and (30), we have

By (35) and (36), we have
By (1) and (37), we have
By (33), we have
But (38) and (39) together contradict (24). Thus (28) cannot hold.
1Recall

that min{y, z} denotes the smaller of the two numbers y and z.

A WAY TO WRITE PROOFS

Step 4.2. Assume that

b2 > a.
We will show that (40) leads to a contradiction. By (40) and (2),

(40)

b2 > 0.

(41)

b 6= 0.

(42)


 2
b a
.
 = min b,
2b

(43)

b > 0.

(44)

b2 a > 0.

(45)

b2 a
> 0.
2b

(46)

 > 0.

(47)

b  < b.

(48)

(b )2 = b2 2b + 2 b2 2b.

(49)

In particular,
In view of (42), we are allowed to define

By (25) and (42), we have

By (40), we have
By (44) and (45), we have

By (43), (44) and (46), we have

Therefore
2

Now observe that since  0, we have

The last inequality can be rewritten as follows:


(b )2 a + (b2 a) 2b .

(50)

2b b2 a.

(51)

(b )2 a.

(52)

b  is not an upper bound for S.

(53)

By (43) and (44), we have

By (50) and (51), we have
By (24) and (48), we have
By (53), we can choose an element x S such that
x > b .

(54)

b  0.

(55)

x2 > (b )2 .

(56)

By (43), we have  b, so that

By (54) and (55), we have

MITYA BOYARCHENKO

By (52) and (56), we have

x2 > a.

(57)

But (57) contradicts (1) and the assumption that x S. Thus (40) cannot hold.
Step 4.3. In Steps 4.1 and 4.2 we showed that assuming that either b2 < a or b2 > a leads
to a contradiction. In view of (27), the only remaining possibility is that b2 = a. This
completes the solution.
3. Analysis of the model
3.1. General structure. The proof given in the previous section consists of a sequence
of mathematical statements joined to each other by connectives. Each statement is
written out in a separate line and is labeled for future reference. The connectives can be
used for several different purposes, including the following ones:
To introduce a new definition; for example, look at the sentences preceding (1),
(30) and (43).
To introduce a new assumption, such as an assumption used in an argument by
contradiction; for example, look at the sentences preceding (16), (28) and (40).
To explain why the next statement follows from what was said earlier.
To explain what is going to happen in the next step of the proof; for example,
look at the sentence preceding (26).
The proof is split up into steps (each of which can then be split up into sub-steps, etc.)
This makes it easier to understand the structure of the proof. For instance, each step
can have its own set of assumptions, which is especially important when you use proofs
by contradiction. To see a specific example, look at Step 4.1 above. In it we made the
assumption (28), which was shown to lead to a contradiction. One should prevent the
reader from thinking that assumption (28) will be used in any other parts of the proof.
Since that particular argument by contradiction is contained in its own step, this helps
the reader see that (28) will not be used in any other steps.
3.2. Things to look out for. Here is a checklist for making sure that you dont make
any mistakes while using the standard model of a proof presented above.
Do not use any letters (or other mathematical symbols) unless they have already
been introduced previously. For example, in the proof above, it is acceptable to
use the letter S in (1) because (1) is the definition of S. It is acceptable to use
x in (1) because there x is used as a dummy variable. It is acceptable to use
R because it is the universal notation for the set of real numbers and was already
defined in class. Finally, it is acceptable to use a in (1) because the meaning of a
was already explained in the statement of the problem.

A WAY TO WRITE PROOFS

Whenever you make a new statement in your proof, always make it clear what the
purpose of that statement is. In other words, indicate whether you are making
a definition, or an assumption, or a statement that follows from the previous
statements (and if so, why), etc.
Always be clear about your assumptions (i.e., never implicitly assume a statement
that you did not write down explicitly). Moreover, as a rule, you should not
simply assume something that hasnt already been shown (and does not appear as
a hypothesis in the result you are proving). However, there are several standard
exceptions:
You can make an assumption if you are giving a proof by contradiction. In
this case you would assume the opposite of the statement that you are trying
to prove, and then use this assumption (along with the other statements you
proved earlier) to derive a contradiction.
You can make assumptions in a case-by-case argument. For example, if you
are proving something about a real number x, you are allowed to first assume
that x > 0 and prove your result; then assume that x = 0 and prove your
result; and finally assume that x < 0 and prove your result.
You can make an assumption as a part of a proof by induction. Indeed, the
induction step typically begins with a certain assumption about some natural
number, say n, and your goal is to then prove that the same property holds
for n + 1 in place of n.
Always make sure that when you claim that a certain statement follows from some
other collection of statements (and/or definitions), your claim is really correct
beyond any shadow of doubt. As a simple illustration, suppose that in the proof
above, before (56), I said: By (54), we have x2 > (b )2 . That would be a gap
in the proof, because (54) alone does not suffice to guarantee (56). However, (54)
and (55) together do.
Make sure that the structure of your proof is clear to the reader. This is especially
relevant when you use an argument by contradiction or a case-by-case analysis as
a part of your proof. Always be clear about when one case ends and the next one
begins, or when a proof by contradiction is finished, and so on.
To sum up all the points above leave absolutely no room for complaints about
your proof. After you finish your solution, re-read every single sentence and make sure
that its meaning is 100% unambiguous. Also make sure that the entire logical structure
of your sequence of sentences is sound and there are no gaps in your arguments.

MITYA BOYARCHENKO

4. Criticisms of the model

4.1. Length. If you follow the model presented above, your proofs will be longer than the
proofs one normally sees in lectures and mathematical textbooks. This is OK. Remember
that the goal is to understand what you are doing and receive full scores for your solutions.
From this viewpoint the benefits of the model definitely outweigh its disadvantages.
4.2. Difficulty for you. One may object that it is difficult to write all your proofs in
the way presented above. However, I dont think that this is actually the case. Indeed,
if you have a proof written in some other form, the only true test of correctness of that
proof is whether you can easily translate it into the form of the model above or not. In
particular, if you produced a proof and encounter an obstacle while trying to rewrite it
in the form above, chances are that you have a gap in your argument or that you made
a subtle mistake somewhere.
4.3. Difficulty for the grader. One may also complain that a proof such as the one
presented above is difficult to read. This is true in the sense that you wouldnt want a
math textbook (or a math course) where all proofs are presented in this way. However,
following the model presented above makes it much easier to check whether a proof is
correct or not; it also makes it very easy to identify the mistakes if any are present.
5. Further advice
5.1. I do not recommend that you try to look for a proof by immediately putting it into
the form presented above. The standard model should be the last stage of the process
of solving a problem. Once you figured out what the correct argument is and convinced
yourself that you can get all the details right, then you should write up your proof in the
form given above. This is by far the best way to avoid mistakes.
5.2. How long should you keep writing your solutions in the form of the model above?
If you begin using this model, after solving a number of problems you will stop making
mistakes, and eventually, with every proof that we do in class, it will be instantly clear
to you how to rewrite it in the form of the standard model. Once you get to this point,
you can stop using this model and start working on developing your own style of writing
proofs. (If at a later point you encounter a particularly difficult problem or confusing
proof, its OK to switch back to the standard model to make sure that all the steps of the
argument are right and presented in the correct logical order.)