The little theorem and its implication of coprimality.

In summary: Since $a^{n-1}\equiv 1 \pmod n$, we have $a^{n-1}=1+nk$ for some integer $k$. This means that the prime factors of $a^{n-1}$ are the same as those of $1$, which is just $1$. Therefore, $(a,n)=1$.In summary, the Lucas primality test states that if $a^{n-1}\equiv 1 \pmod n$, then $a$ and $n$ are coprime. This can be proven by using different approaches such as Bezout's identity, group theory, or the fact that $a$ and $b$ have the same prime factors when $a\
  • #1
bamuelsanks
3
0
Hi guys.

I'm rather new to number theory, and as part of an assignment I have been learning about various different primality tests.
One of these tests is the Lucas primality test.
As part of the reasoning behind the test, wikipedia states:
"If [$a^{n-1} \equiv 1 \textrm{ (mod }n\textrm{)}$] holds for a, we can deduce that a and n are coprime".
Would anybody be able to help me understand why this is true?

Thanks,
SB
 
Mathematics news on Phys.org
  • #2
A theorem due to Leonhard Euler...

http://mathworld.wolfram.com/EulersTotientTheorem.html

... establishes that for all a coprime to n is...

$\displaystyle a^{\varphi (n)} \equiv 1\ \text{mod}\ n$ (1)

... where $\varphi(n)$ is the 'Euler's Totiens Function'. Now if n is prime is $\displaystyle \varphi(n)=n-1$...

Kind regards

$\chi$ $\sigma$
 
  • #3
chisigma said:
A theorem due to Leonhard Euler...

http://mathworld.wolfram.com/EulersTotientTheorem.html

... establishes that for all a coprime to n is...

$\displaystyle a^{\varphi (n)} \equiv 1\ \text{mod}\ n$ (1)

... where $\varphi(n)$ is the 'Euler's Totiens Function'. Now if n is prime is $\displaystyle \varphi(n)=n-1$...

Kind regards

$\chi$ $\sigma$

That is the converse of what the OP is trying to prove. Secondly, the OP does not say that $n$ is prime.
 
Last edited:
  • #4
What is written by Alexmahone is true. The converse of the Euler's totient theorem established that if...

$\displaystyle a^{\varphi(n)} \equiv 1\ \text{mod}\ n$ (1)

... then a and n are coprime. The important point however is that the identity $\displaystyle \varphi(n)=n-1$ holds only for n prime...

Kind regards

$\chi$ $\sigma$
 
  • #5
chisigma said:
What is written by Alexmahone is true. The converse of the Euler's totient theorem established that if...

$\displaystyle a^{\varphi(n)} \equiv 1\ \text{mod}\ n$ (1)

... then a and n are coprime. The important point however is that the identity $\displaystyle \varphi(n)=n-1$ holds only for n prime...

Kind regards

$\chi$ $\sigma$

Honestly, I don't think you've proved anything. If you think you have, could you please write your proof out?

To be precise, I don't see how you went from $a^{n-1}\equiv 1\pmod{n}$ to $a^{\phi(n)}\equiv 1\pmod{n}$. Note that $n$ need not be prime.
 
Last edited:
  • #6
bamuelsanks said:
Hi guys.

I'm rather new to number theory, and as part of an assignment I have been learning about various different primality tests.
One of these tests is the Lucas primality test.
As part of the reasoning behind the test, wikipedia states:
"If [$a^{n-1} \equiv 1 \textrm{ (mod }n\textrm{)}$] holds for a, we can deduce that a and n are coprime".
Would anybody be able to help me understand why this is true?
If $a^{n-1} \equiv 1 \pmod n$ then $a^{n-1} = 1+kn$ for some integer $k$. Thus $a^{n-1}$ and $n$ cannot have any prime common factor. But the prime factors of $a^{n-1}$ are the same as those of $a$. Hence $a$ and $n$ are coprime.
 
  • #7
Opalg said:
If $a^{n-1} \equiv 1 \pmod n$ then $a^{n-1} = 1+kn$ for some integer $k$. Thus $a^{n-1}$ and $n$ cannot have any prime common factor. But the prime factors of $a^{n-1}$ are the same as those of $a$. Hence $a$ and $n$ are coprime.

You didn't use the fact that the exponent of $a$ is $n-1$, so is it true for any exponent?
 
  • #8
Thanks for the quick replies!
I understand it now, though I couldn't quite follow chisigma's response unfortunately.
Apologies for not making clear that $n$ was not necessarily prime.

Thanks again,
SB
 
  • #9
Uses Bezout's idenity. We have that we can write \(\displaystyle a^{n-1}=kn+1\) Now, we have that 1=-kn+a^{n-1}. As this is the least possible positive number, it follows that it's the least positive positive number writable as a linear combo of n and a (n>1 otherwise, it's trivial). Hence, gcd(n,a)=1.
 
  • #10
a more group-theoretical approach (assuming n ≥ 2):

suppose an-1 = 1 (mod n).

then [a] (the residue of a mod n)

is a unit of Zn (with inverse [an-2]).

this means gcd([a],n) = 1.

since a = [a] + kn, we have:

s([a]) + tn = 1 so:

s(a - kn) + tn = 1, and:

sa + (t-k)n = 1, which shows gcd(a,n) = 1.
 
  • #11
If $a\equiv b\pmod n$, then $(a,n)=(b,n)$. This can be seen by writing $a=b+nk$, for some integer $k$.

Here, $(a^{n-1},n)=(1,n)=1$.
 

Related to The little theorem and its implication of coprimality.

1. What is the little theorem and how does it relate to coprimality?

The little theorem, also known as Fermat's little theorem, states that for any prime number p and any integer a, ap ≡ a (mod p). This means that if a and p are coprime (meaning they have no common factors other than 1), then ap will be congruent to a modulo p. This has important implications for determining coprimality, as it provides a way to quickly check if two numbers are coprime.

2. Can the little theorem be used to prove that two numbers are coprime?

Yes, the little theorem can be used to prove coprimality, but only in one direction. If ap ≡ a (mod p), then we can say that a and p are coprime. However, if a and p are coprime, we cannot necessarily use the little theorem to prove it, as there are other numbers that could produce the same congruence.

3. How does the little theorem relate to Euler's totient function?

The little theorem is a special case of Euler's totient function, which is defined as the number of positive integers less than or equal to n that are coprime to n. In other words, φ(n) is the number of integers a such that 1 ≤ a ≤ n and a is coprime to n. The little theorem can be written as aφ(p) ≡ 1 (mod p), where p is a prime number.

4. What implications does the little theorem have in cryptography?

The little theorem has important implications in public key cryptography. It is used in the RSA algorithm, which relies on the fact that factoring large numbers is difficult. The little theorem allows for efficient encryption and decryption, as the exponentiation involved in the algorithm can be done using modular arithmetic.

5. Are there any generalizations of the little theorem?

Yes, there are several generalizations of the little theorem, including the Carmichael's theorem and the Euler-Fermat theorem. These theorems extend the concept of the little theorem to numbers that are not necessarily prime, but still have certain properties that allow for efficient exponentiation using modular arithmetic.

Similar threads

Replies
2
Views
1K
  • General Math
Replies
1
Views
2K
Replies
6
Views
3K
Replies
1
Views
2K
Replies
1
Views
1K
  • Precalculus Mathematics Homework Help
Replies
6
Views
1K
  • Set Theory, Logic, Probability, Statistics
Replies
1
Views
713
Replies
2
Views
2K
  • Set Theory, Logic, Probability, Statistics
Replies
2
Views
3K
  • Linear and Abstract Algebra
Replies
4
Views
3K
Back
Top