Modular arithmetic

From Mersennewiki

Modular arithmetic is the set of operations that can be done when working modulo N, where N is an integer greater than 1.

Since almost all factorization methods and primality tests use modular arithmetic, reading this article is a prerequisite to understanding those methods.

We can visualize this arithmetic using a clock. Suppose that the number 12 in the clock is replaced by zero. Then when we have to add an hour, we get, for example: 3+1 = 4, 10+1 = 11, 11+1 = 0. If we have to add three hours we get: 5+3 = 8, 11+3 = 2 (as in 11 AM + 3 hours = 2 PM). We can also subtract: 2-3 = 11 and even multiply: 5×4 = 8 (because 5+5+5+5 = 8). This is arithmetic modulo 12 and the set of numbers representing the hours 0, 1, 2, 3,..., 11 is known as Z/12Z.

When making these modular operations, the equal symbol is not used. We use the congruence symbol (\eq ) instead. Note that two numbers A and B are said to be congruent modulo n if A-B is a multiple of n.

The set Z/nZ of numbers modulo n contains the numbers 0, 1, 2, 3, ..., n-2 and n-1. The following operations are defined:


Modular addition

Let A and B be numbers in Z/nZ. Then the addition is defined as:

 A + B \eq C (mod n)


 C = A + B if A + B < n
 C = A + B - n if A + B \geq n 

Modular subtraction

The subtraction is defined as:

 A - B \eq C (mod n)


 C = A - B if A \geq B
 C = A - B + n if A < B 

Modular multiplication

The multiplication is defined as:

 A * B \eq C (mod n)

where C is the remainder of the division between A × B and n.

Modular exponentiation

The exponentiation is defined as:

 A^B \eq C (mod n)

where C\eq A*A*A*A*...*A (B times). An efficient method to perform modular exponentiation is the binary method.

Since reducing modulo n takes a lot of time, Montgomery multiplication is used in this context where only one modular reduction is needed.

Modular inversion

The inverse is defined as:

 A^{-1} \eq B (mod n)

where B is the number such that A * B \eq 1 (mod n).

This operation is only defined when the numbers A and n are coprime, i.e., when gcd(A, n) = 1.

The following algorithm to compute the multiplicative inverse n-1 mod q for n with 0 < n < q, where 0 < n-1 < q is based on the Extended Euclidean algorithm. All variables are integers (some programming languages call them Big Integers).

  1. Set i = q, h = n, v = 0, and d = 1.
  2. Set t = i DIV h, where DIV is defined as integer division.
  3. Set x = h.
  4. Set h = i - tx.
  5. Set i = x.
  6. Set x = d.
  7. Set d = v - tx.
  8. Set v = x.
  9. If h > 0, go to step 2.
  10. Let n-1 = v mod q.

When the modulus is a power of 2, there is an easier method. Let the modulus be 2^m, let the number to be inverted be N and let k = \log_2 m rounded to the next integer. Then the method is:

  1. Set x = 1.
  2. Perform k times: Set x = x(2-Nx) mod 2^m

Modular division

The division is defined as:

 A / B \eq C (mod n)


 C \eq A * B^{-1}
Personal tools