Hensel’s lemma gives necessary conditions to lift a solution mod of a polynomial equation to a solution mod . Collecting all these solutions mod , we get a solution in the p-adic integers.
Basic Version:
If an integer polynomial satisfies
then there is unique solution to . All these solutions can be put together to get a solution in such that
Examples:
Consider the equation
There is a solution given by .
, but we can add a multiple of to get a solution, and continue similarly.
Therefore
is a solution to lifting .
It’s not clear that we can continue the above process indefinitely, but Hensel lemma guarantees that there is a way to do it, and in fact a unique way to lift at each stage.
The other solution is obtained by lifting .
Example 2: in
The solution lifting is given by
This is the unique solution in because is the unique solution. Note that
Example 3:
There are two solutions . But
lifts to a unique solution
But doesn’t lift because, in fact for , is the only solution.
Example 4:
The only root is . But So Hensel’s lemma doesn’t guarantee a lift, but is a solution and hence is a solution in .
Example 5: , where .
If , then is a solution modulo . And we have
So if , there is unique solution to in with .
In fact, if we have , we are trying to find roots of unity. If is co-prime to , we have a unique solution lifting a solution . We can see that has a unique solution lifting the solution , where is any non-zero residue. So we have roots of unity for . What about other ?
When is coprime to , then has solutions that lift any residue class, but these solutions because of the uniqueness of the lifting should be exactly the solutions to . It can also seen that for divisible by (odd), there are no solutions to other than the trivial solution . (For , we also get )
Example 6:
The only solution turns out to be because But , we cannot decide using the lemma if there is a lifting.
We will come back to this example later.
Proof of Lemma: The proof is easy. If we have lifted the solution to , then to get a solution mod , we need
, and we want to determine up to such that
So we can find and hence find a lift to the solution. Note that we crucially use that the derivative doesn’t vanish in the above equation.
Newton’s method of approximations: We can think of the above method as starting from an approximate solution and applying Newton’s iteration to get better approximations. The difference from classical Newton’s method is that our notion of approximation or closeness is p-adic. (Number divisible by large powers of are small) Newton infact used this method to solve equation in power series, the p-adic integers are somewhat like the power series in
What happens when .? We can see examples of cases where there are no lifts, cases where there are unique lifts and also cases with multiple lifts from the same solution .
We now give a stronger version of the lemma.
Stronger version: If a polynomial
for some , then we have a unique such that and is close to , that is with
The case of reduces to the basis version mentioned before.
As mentioned before we want to prove the convergence of the Newton iteration
The conditions of the theorem are exactly designed to get the convergence. We skip the calculations in the proof. The strong theorem can be deduced from the weak version by applying the theorem to a modification of the polynomial involved, so the abstract content of both the results is the same although the strong weaker seems to have the basic version as a special case.
IMPORTANT: In the weaker version, we lifted each solution uniquely to higher powers and constructed the solution. But for the strong version, the uniqueness needn’t be at every finite level , the solution is unique only in the limit in .That is there could be multiple solutions lifting solutions to , for a large . but only one of these solutions extends to all the higher powers indefinitely. Some solutions lift to higher powers and then stop extending further. To get uniqueness we have to look at solution not just mod , but a high power . There is a high power after which there is a unique lift provided the conditions of the theorem are satisfied.
has only one solution . But , and hence we cannot say that the lift are unique or even that they exist. But
are the solutions. So we see that there are multiple lifts of the solution .
,
so the conditions of the second version hold and we have convergence to a unique solution with
So has to be .
If we start with , then
,
so the solution converges to such that
Hence the solution in this case has to be
Example 6: We will look at the previous example.
doesn’t satisfy the conditions of the basic version and even this version, but satisfies the strong version:
Hence there is a unique solution such that which is given by
Finally, the conditions of the stronger version are always true if we choose to be close to a root of in provided the root is a simple root. That is the conditions are necessary and not just sufficient to find a simple root.
Multivariable Hensel’s Lemma:
The strong version can be generalized as follows.
If a polynomial is such that
for some , then we have a unique such that and is close to , that is with
Where are the norms are p-adic and refers to taking maximum of the absolute values of coordinates. That is
Here is the gradient.
More generally if are polynomials in such that
for some , then we have a unique such that and is close to , that is with
Here is the Jacobian, the determinant of the Derivative matrix.