Linear Congruences

Linear Congruences
In ordinary algebra, an equation of the form ax = b (where a and b are

given real numbers) is called a linear equation, and its solution x = b/a is
obtained by multiplying both sides of the equation by a−1 = 1/a.
The subject of this lecture is how to solve any linear congruence
ax ≡ b (mod m)
where a, b are given integers and m is a given positive integer.

For a simple example, you can easily check by inspection that the linear
congruence
6x ≡ 4 (mod 10)
has solutions x = 4, 9. Already we see a difference from ordinary algebra:
linear congruences can have more than one solution!
Are these the ONLY solutions? No. In fact, any integer which is congruent
to either 4 or 9 mod 10 is also a solution. You should check this for yourself
now.
So any integer of the form 4 + 10k or of the form 9 + 10k where k ∈ Z is

a solution to the given linear congruence. The above linear congruence has
infinitely many integer solutions.
The is a general principle at work here. Solutions to linear congruences are
always entire congruence classes. If any member of the congruence class
is a solution, then all members are. This is a simple consequence of the
properties of congruences proved in a previous lecture.
This means that although the congruence 6x ≡ 4 (mod 10) had infinitely
many integer solutions, the solutions fall into congruence classes, and there
are only two of those: [4]10 and [9]10 .
Whenever a linear congruence has any solutions, it has infinitely many. The
solutions fall into congruence classes, and there are only a finite number of
congruence classes that solve the congruence.
Here is the key observation which enables us to solve linear congruences.
1
By definition of congruence, ax ≡ b (mod m) iff ax − b is divisible by m.
Hence, ax ≡ b (mod m) iff ax − b = my, for some integer y. Rearranging
the equation to the equivalent form ax − my = b we arrive at the following
result.
Lemma. Solving the congruence ax ≡ b (mod m) is equivalent to solving

the linear diophantine equation ax − my = b.
Since we already know how to solve linear diophantine equations, this means
we can apply that knowledge to solve linear congruences.
Theorem. Let a, b be any integers and let m be a positive integer. Let

d = gcd(a, m). If d - b then the linear congruence ax ≡ b (mod m) has no
solutions. If d | b then the linear congruence ax ≡ b (mod m) has exactly d
solutions, where by “solution” we mean a congruence class mod m.
Comment. Later in this lecture we will see that all the solutions can be
joined together to form a single congruence class mod m/d.
Proof of the theorem. Solving the congruence ax ≡ b (mod m) is equivalent

to solving the linear diophantine equation ax − my = b. If d - b then the
diophantine equation has no solutions, so the congruence has no solutions,
either. If d | b then the solutions of the diophantine equation take the form
x = x0 + (m/d)t, y = y0 + (a/d)t
where (x0 , y0 ) is any particular solution (obtained from the Euclidean algo-
rithm, for instance).
To finish the proof, observe that as t runs through the values 0, 1, . . . , d − 1

(the residues mod d) the congruence classes [x0 + (m/d)t]m run through all
the solutions. (There are no other solutions because the classes just repeat
for higher and lower values of t.)
Example. Returning to the example 6x ≡ 4 (mod 10), we solve it by first

guessing the solution x0 = 4 by trial and error. Then the theorem tells
us that [x0 + (10/2)t]10 for t = 0, 1 gives the complete solution set. Thus,
x = [4]10 and [9]10 is the complete solution.
2
Notice that we could write this as: x ≡ 4, 9 (mod 10). This congruence
describes exactly the same set of integers as the union of the congruence
classes [4]10 , [9]10 .
Even better: we can write the complete solution as: x ≡ 4 (mod 5). This
single congruence describes the set of all integer solutions, as you should
check. In other words, we have
[4]10 ∪ [9]10 = [4]5 .
Example. Let’s solve 230x ≡ 1081 (mod 12167). We start by applying the
Euclidean algorithm to compute d = gcd(230, 12167) = 23. Since d | 1081
there are solutions. The extended Euclidean algorithm gives the particular
solution (s0 , t0 ) = (53, 1) to the diophantine equation 230s − 12167t = 23,
and scaling by 47 = 1081/23 we get the particular solution (x0 , y0 ) =
(2491, 47) to the diophantine equation 230x − 12167y = 1081. So x0 = 2491
solves the original given congruence. In this case, m/d = 529. Thus, with
m = 12167, the set of residue classes
{[2491 + 529t]m : t = 0, 1, 2, . . . , 22}
gives the complete solution set to the congruence.
Thus with m = 12167 we get solutions [a]m for a = 2491, 3020, 3549, 4078,
4607, 5136, 5665, 6194, 6723, 7252, 7781, 8310, 8839, 9368, 9897, 10426,
10955, 11484, 12013, 12542, 13071, 13600, 14129 and no others.
We can also say that (still with m = 12167) we get solutions [a]m for a =
2491, 3020, 3549, 4078, 4607, 5136, 5665, 6194, 6723, 7252, 7781, 8310, 8839,
9368, 9897, 10426, 10955, 11484, 12013, 375, 904, 1433, 1962.
This is because 12542 ≡ 375, 13071 ≡ 904, 13600 ≡ 1433, and 14129 ≡ 1962
(mod m).
When dealing with congruence classes, we can always replace any represen-
tative by another one!
Incidentally, we can also write the complete solution obtained above as a

single congruence class mod 529. The complete solution is given by x ≡ 375
(mod 529). Again, the union of all 23 congruence classes mod m is a single
congruence class mod m/d.
3
The examples suggest a simpler method to solve a linear congruence, which
should always produce a single congruence class mod m/d (assuming d | m).
The remainder of this lectures explores this idea.

As a special case of the theorem, let me point out that if d = gcd(a, m) = 1
then the linear congruance ax ≡ b (mod m) has a unique solution class.
In the special case gcd(a, m) = 1, we can always solve the congruence by

finding the inverse of [a]m and then multiplying both sides of the congruence
by the inverse to obtain the unique solution. This is a satisfying idea because
it is so similar to what we do in ordinary high school algebra to solve linear
equations.
Definition. An inverse of a mod m is any integer c such that a · c ≡ 1

(mod m). We write a−1 mod m = c, or [a]−1 m = [c]m for the modular
inverse just defined, when it exists.
An inverse of a mod m exists iff gcd(a, m) = 1. Proving this is a good

exercise.
Example. Suppose we are given the congruence 11x ≡ 15 (mod 20). Ob-
serve that d = gcd(11, 20) = 1. Thus 11x ≡ 15 (mod 20) has a unique
solution class. Observe that 11 · 11 ≡ 1 (mod 20), so [11]20 · [11]20 = [1]20
and [11]−1
20 = [11]20 . This tells us that we can solve the given congruence
simply by multiplying both sides by 11 and reducing numbers mod 20. Here
we go:
11x ≡ 15 (mod 20)

11 · 11x ≡ 11 · 15 (mod 20)
121x ≡ 165 (mod 20)
x≡5 (mod 20).
This proves that x = [5]20 is the unique solution to the given congruence
11x ≡ 15 (mod 20).
So we can always solve ax ≡ b (mod m) in case gcd(a, m) = 1 simply by

multiplying both sides by the inverse of [a]m (i.e., canceling the a factor).
Of course, to find the inverse of a in general requires the extended Euclidean
algorithm to solve the corresponding diophantine equation as−mt = 1; then
c = s mod m will be an inverse of a mod m. It should be noted that if m is
4
small enough then trial and error works pretty well to find an inverse, since
there are few possibilities to check.
In fact, the technique of multiplying by an inverse can be used to solve any

linear congruence ax ≡ b (mod m), even when d = gcd(a, m) 6= 1.
Let us see why.

Assume that d = gcd(a, m) divides b. Solving the congruence ax ≡ b
(mod m) is equivalent to solving the diophantine equation ax − my = b.
But we can divide both sides of the equation by d to get a reduced diophan-
tine equation
a m b
Ax − M y = B where A = ,M = ,B = .
d d d
The solutions to the reduced diophantine equation are exactly the same as
the solutions to the original one. Thus, solving ax ≡ b (mod m) is equivalent
to solving
Ax ≡ B (mod M ).
This congruence satisfies the condition gcd(A, M ) = 1, and thus can be
solved by finding an inverse of A mod M and multiplying both sides by that
inverse.
Example. Let’s again solve 230x ≡ 1081 (mod 12167) (which we solved
earlier) using this new approach. We start by applying the Euclidean algo-
rithm to compute d = gcd(230, 12167) = 23. Next, we reduce the congruence
to the equivalent congruence 10x ≡ 47 (mod 529) by dividing by d. We now
have gcd(10, 529) = 1, so we can solve by multiplying by 10−1 mod 529. We
can find the inverse by finding any solution to the diophantine equation
10x − 529y = 1 and then throwing away y. For instance, the extended
Euclidean algorithm gives (x, y) = (53, 1) so 10−1 ≡ 53 mod 529.
Multiplying the reduced congruence 10x ≡ 47 (mod 529) by the inverse 53

gives the (unique!) solution x ≡ 53 · 47 ≡ 375 (mod 529).
You should verify for yourself that the union of the congruence classes [a]m
for a = 2491, 3020, 3549, 4078, 4607, 5136, 5665, 6194, 6723, 7252, 7781,
8310, 8839, 9368, 9897, 10426, 10955, 11484, 12013, 375, 904, 1433, 1962
(which we got before) gives exactly the same set of integers as the single
congruence class [375]529 .

Linear Congruences - Article

Uploaded by

Copyright:

Available Formats

Linear Congruences - Article

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Linear Congruences - Article

Uploaded by

Copyright:

Available Formats

In ordinary algebra, an equation of the form ax = b (where a and b are

The subject of this lecture is how to solve any linear congruence

where a, b are given integers and m is a given positive integer.

So any integer of the form 4 + 10k or of the form 9 + 10k where k ∈ Z is

Lemma. Solving the congruence ax ≡ b (mod m) is equivalent to solving

Theorem. Let a, b be any integers and let m be a positive integer. Let

Proof of the theorem. Solving the congruence ax ≡ b (mod m) is equivalent

To finish the proof, observe that as t runs through the values 0, 1, . . . , d − 1

Example. Returning to the example 6x ≡ 4 (mod 10), we solve it by first

[4]10 ∪ [9]10 = [4]5 .

{[2491 + 529t]m : t = 0, 1, 2, . . . , 22}

gives the complete solution set to the congruence.

Incidentally, we can also write the complete solution obtained above as a

The remainder of this lectures explores this idea.

In the special case gcd(a, m) = 1, we can always solve the congruence by

Definition. An inverse of a mod m is any integer c such that a · c ≡ 1

An inverse of a mod m exists iff gcd(a, m) = 1. Proving this is a good

11x ≡ 15 (mod 20)

So we can always solve ax ≡ b (mod m) in case gcd(a, m) = 1 simply by

In fact, the technique of multiplying by an inverse can be used to solve any

Let us see why.

Multiplying the reduced congruence 10x ≡ 47 (mod 529) by the inverse 53

You might also like