Section 5.2 A Strategy For the First Solution
ΒΆThe previous proposition always works. However, it can be very tedious to find that first solution if the modulus is not small. This section is devoted to strategiesβ1βThe reader should note that we roughly follow [C.2.1, pp. 50-51] in this, but that an alternate (or supplemental?) approach using the Bezout identity is followed in texts like [C.2.4] or [C.2.13]. for simplifying a congruence so that finding such a solution is easier.Fact 5.2.1. Strategies that work for simplifying congruences.
We can do two main types of simplification. First, there are two types of cancellation.
If a, b, and n all are divisible by a common divisor, we can cancel that divisor out (keeping in mind that we still will need our final solution to be modulo n).
If a and b share a common divisor which is coprime to the modulus, we can cancel that divisor from a,b (only).
See Propositions 5.2.6 and 5.2.7 for precise statements and proofs.
Secondly, there are two counterintuitive operations that may lead to a simpler congruence (using least nonnegative residues).
We could multiply a and b by something coprime to n. If, after reducing modulo n, that makes a or b smaller, then that was a good idea!
We can add some multiple of n to b. Again, if that happens to make a and (the new) b share a factor, then it was a good idea!
These four steps may be applied in any order, though typically the first two are done as often as possible. See Example 5.2.5 for why coprime is necessary in two of the steps.
Example 5.2.2. A big example.
Let's do a big problem exemplifying all the strategies; we will break it up into possible steps you might do.
First, note that all three of the coefficients and modulus are divisible by 3. So right away we should simplify by dividing by 3. But keep in mind that our final solution will need to be modulo 33, not modulo eleven! We should still end up with gcd(30,33)=3 total solutions, and if we don't, we have messed up somewhere.
-
Now we have 10xβ‘6 (mod 11). (Again, although this will have one solution modulo 11, we will need to get the other two solutions modulo 33.) Since 10 and 6 are both divisible by 2, and since gcd(2,11)=1, we can divide the coefficients (not modulus) by 2 without any other muss.
5xβ‘3 (mod 11) -
So take 5xβ‘3 (mod 11), and let's try to replace 3 by another number congruent to 3 modulo 11 which would allow me to use the above steps again.
-
I could try 3+11=14, but that gives
5xβ‘14 (mod 11)and 14 doesn't share a divisor with 5 (from the 5x).
-
If I try 3+22=25, giving
5xβ‘25 (mod 11)then 25 does share a divisor with 5.
-
-
Now I can go back and reduce 5xβ‘25 (mod 11) to
xβ‘5 (mod 11)And that's the answer!
-
Or is it? Remember in the first step that we started modulo 33, and that all the answers will be equivalent modulo 11. So we see that
x=5+11k for kβZwill be the answer, which is the three equivalence classes {[5],[16],[27]}.
Does it check out?
xxxxxxxxxx
[mod(30*x,33)==18 for x in [5,16,27]]
One final observation is that we avoided trial and error as long as possible. At various points we could have done so, but x=1 and x=2 wouldn't have worked right away, and I am lazyβ¦
Example 5.2.3.
Let's finish the previous example again, but using the other possible counterintuitive strategy. That was the trick to multiply a and b by something which would reduce; ideally it would reduce [a]β‘[1].
We were at 5xβ‘3 (mod 11).
-
Multiplying a=5 and b=3 by 9, which is coprime to 11, gives us
45xβ‘27 (mod 11). This reduces to xβ‘5, and gives the same answer as before (provided we remember to get all possible answers modulo 33).
Example 5.2.4.
Try completely solving one of the following two congruences (Exercise 5.6.3) on your own now, before moving on. The rest of the Exercises provide other interesting practice.
7xβ‘8 (mod 15)
6xβ‘8 (mod 14)
Example 5.2.5.
Finally, let's see examples of using the strategies poorly.
First, suppose 6xβ‘12 (mod 4). Then we could divide all terms by 2, yielding 3xβ‘6 (mod 2), and then reducing everything modulo two we obtain xβ‘0, or that the solution is all even x. If we had instead canceled the 2 from only the 6xβ‘12 portion, we would have gotten 3xβ‘6 (mod 4), which is βxβ‘2 or xβ‘2 modulo four, which is only half of the true solutions.
As a similar example, suppose we want to solve 7xβ‘7 (mod 12). If we used cancellation the solution would obviously be xβ‘1. Set this aside and instead multiply 7x and 7 by 2 in order to obtain 14xβ‘14 which simplifies to 2xβ‘2 (mod 12), which now looks like an easy target for cancelling 2 from all three numbers to obtain xβ‘1 (mod 6), which is twice the true solutions.
The moral of the story is that while some structure is preserved when we don't stick to numbers coprime to the modulus, it's very easy to remove or add spurious solutions, so it must be avoided.
Proposition 5.2.6. Canceling, Part I.
If dβ 0, then adβ‘bd (mod nd) precisely for the same a,b,n as when aβ‘b (mod n).
Proof.
Like many such proofs, you basically follow your nose.
First write \(ad\equiv bd\) (mod \(nd\)) as \(nd \mid ad-bd\text{,}\) or \(ad-bd=k(nd)\) for some \(k\in\mathbb{Z}\text{.}\) We rewrite this as \(d(a-b)=d(kn)\text{.}\)
Since \(d\neq 0\text{,}\) asserting \(d(a-b)=d(kn)\) is equivalent to saying \(a-b=kn\text{,}\) which is of course by definition saying that \(a\equiv b\) (mod \(n\)).
Since all steps were equivalences, both statements are equivalent.
Proposition 5.2.7. Canceling, Part II.
If dβ 0 and gcd(d,n)=1, then adβ‘bd (mod n) precisely for the same a,b,n as when aβ‘b (mod n).
Proof.
We'll only sketch the proof; see Exercise 5.6.2.
Use the definitions as above, starting with the \(ad\) situation.
You should have that \(n\) divides some stuff, which is itself a product of \(d\) and other stuff.
We had a proposition somewhere about coprimeness and division; what remains should yield us \(a\equiv b\) (mod \(n\))