Algebraic Equation under Permutation


Disclaimer: This text is neither a formal proof of why algebraic equations of degree over 5 are not solvable by radicals, nor a rigorous text with its content carefully cross validated. It is just a text the author write for fun. You may safely skip this text without worrying about missing some important concepts, and you must judge by your own if you decide to read.


Some of us might have already heard of the famous theorem that algebraic equations of degree over are not solvable by radicals, and by the text of permutation group we've known that is simple when it is on over letters, should these two fact have connection? The answer is, yes.

Given an algebraic equation of degree :

When it comes to you, what are you thinking about? Are you meditating on some magical term to transform and simplify it into form like , just as what we were taught in middle school?

According to the fundamental theorem of algebra, an algebraic equation of degree has roots. Noted that:

By associating terms of corresponding degree, we have the Vieta's formula:

While is itself an linear equation, if we can find some method to generate more linear equations in the form of , to form a system of linear equations of rank , the equation is solved equivalently. One way to do so is to assign values to that is linearly independent from all s and previously assigned value. When an equation is solvable by radicals, each is representable by formula of combined by finite times of addition, subtraction, multiplication, division and root extraction. Substituting these formula into each will yield us each equation in the system of linear equations.

Conversely, for any algebraic equation, if you can prove it is impossible to find such a system of linear equations, the algebraic equation is not solvable by radicals.

Symmetric polynomial

Inspired by how we solve the quadratic equation, we would like to construct such a linearly independent equation by expanding the and see how we can use terms in Vieta's formula to express it.

In the beginning, things go quiet fluently for the quadratic equation, since there's:

However, when it comes to the cubic equation, we encounters the first obstable that:

I won't expand any further since it's enough for demonstrating the problem. Unlike what we've done so easily with the quadratic case, there're terms related with that are not possible to get rid of.

While you might still disbelieve and argue that might still be expressed if more exhaustive reduction is taken, or another choice of coefficient for is tried, let's see how the case of and are fundamentally different from each other.

For a polynomial of roots 1, define the permutation group action on polynomial as , and obviously there's . Polynomial is a symmetric polynomial if it is unconditionally invariable, that is, .

All terms from the Vieta's formula are symmetric polynomials, which can be proved by swapping roots in and see how coefficients are invariable and so do terms in Vieta's formula. We call the symmetric polynomials from Vieta's formula elementary symmetric polynomials, since they are directly connected with coefficients of the equation we want so solve.

The is also symmetric since . While is not symmetric: randomly choose some cubic equation with root , there're and .

What's more, for two symmetric polynomials , their combination , where defines and depends only on the result of , is still symmetric since . The combination could be addition, subtraction, multiplication, division, or functions depending only on value of . So basically you cannot express a asymmetric polynomial in combination of symmetric ones.

If we want to construct an equation linearly independent from , there can't be , and among them we're always able to select some . Consider the polynomial with which we construct a linearly independent equation, we would like to show when the polynomial cannot be symmetric.

Let's rewrite the polynomial as , where contains all the terms besides and . After swapping and , the polynomial turns into . The equation has solution , where . So if we want the polynomial to preserve its origin value after swapping and , the following condition must hold:

To simplify, we denote as , which means the portion of equals unconditionally, and denote as its opposition.

When , the polynomial is in the form of , by it is not possible to be symmetric except for : just choose an equation with solution , swap and you will see the result.

When , since merely swapping and does not touch coefficients of terms in , it is not possible for . When , the condition turns into , in order for the condition to hold regardless of the choice of , there must be , while is requried as a precondition. So the polynomial cannot be symmetric when .

When , since it's always possible to choose another in . Take out from and put back to and we get . Repeat the discussion of above and we will get the same result.

By now, we conclude the proof of polynomial in the form of will never be symmetric when .

Lagrange resolvent

The statement above seem to be telling us equations of degree over are not solvable by radicals, while in fact the equations of degree and are actually solvable, so there must be something wrong with the reasoning. What is it?

While symmetric polynomial can be constructed from composition of symmetric polynomials, there's nothing prohibiting composition of asymmetric polynomials to be symmetric.

In a cubic equation, consider the polynomial , under the permutations of there're:

And under the permutations of there're:

Expand the polynomial and we get:

Noted that , so let's combine and we get:

Similarily, let's combine and we get:

Phew, after verbose mechanical deductions of and , we find their values fit into roots of quadratic equation . Although we are not able to represent and using terms from Vieta's formula directly, we are still able to construct symmetric polynomials from them, solve their values from a quadratic equation, and finally form linear equations and , combined with we are able to solve the values of .

Consider another asymmetric polynomial , under the permutations of there're:

Although we can still construct symmetric polynomials , , and , we will need to solve another cubic equation in order to extract their values. Transforming from a cubic equation to another cubic equation is not reducing the problem and is useless.

So the asymmetric polynomials and are deliberately constructed, they have special symmetries inside, reducing the problem of solving a cubic equation to solving a quartic equation.

Lagrange, who carefully studied the solution for cubic equation published by Cardano, and the solution for quartic equation by Ferrari, was the first to reveal the affinity between solution by radicals and symmetries of roots, and came up with the method of finding a solution by radicals as above. Such method is also called the Lagrange resolvent. In Lagrange's theory, he also came up with the Lagrange resolvents for cubic and quartic equations.

Criteria of symmetries

So what kind of symmetries should the asymmetric polynomials in a Lagrange resolvent has? For an algebraic equation of degree , first of all, we need a polynomial of linear combination of roots, whose th root contributes an equation to the system of linear equations we used to extract the roots. The polynomial must be deliberatedly selected so that it is helpful for reducing the problem.

For the symmetry group acting on , it either preserve its original form, or transform into another form. The permutations preserving 's original form forms a group 2:

  1. (Closed) .
  2. (Identity) .
  3. (Invertible) .
  4. (Associative) By definition of and 's associativity.

For each left coset , it's obvious that . and . For convienience, denote the polynomials result from different left cosets as , and each one is distinct.

Assume 's acting on results in different polynomials . Then over these polynomials we build a symmetric polynomial depending only on asymmetric polynomials' values and masking out their asymmetries. That is:

Which means 's acting on symmetric polynomial is the same as permutating asymmetric polynomials in , and due to 's permutation invariant nature the result of keeps unchanged. But none of these s may be dropped or duplicated, otherwise by combining the duplicate 's asymmetry, and 's nature of allowing only permutation of inputs, we can easily construct contradiction to 's being symmetric.

Have you noticed this is an invitation for constructing a group homomorphism? It's the joint force of these polynomials to define a map such that permutation of the roots are mapped to permutation of these polynomials.

Let's focus on what happens to paramters of in . For any , there's some , when acts on , it permutates parameters of that . Then for any , there's some , when acts on , it permutates the parameters again that . Since and , there's , and is a group homomorphism.

So the symmetrizing requires each polynomial corresponds to an element in quotient set . Noted that since and is amomg these , there's . Finally by we have . So in order to create appropriate polynomials for , must be held.

This is what we have exactly done with cubic equation. Consider the subnormal series of :

The polynomial is fixed under , and under there're and corresponding to each one of . With this setup we derive the solution by radicals for cubic equation.

While for the quartic equation, things go complex. Consider the subnormal series of :

If we are able to (though in fact not able to) directly construct polynomials in the form of which is symmetric under and linearly independent from , there're only linearly independent equations generated and is obviously insufficient.

So instead of creating polynomial symmetric under directly, we first create polynomial which is symmetric under and generate asymetric polynomials under :

Under normal circumstance, we will need to solve a sextic equation to extract their values. But we are lucky this time since we have in between. Let's bundle them three by three so that:

Now we are able to build which is symmetric under . The value of can be extracted with a quadratic equation. Then to extract from and from , we set to Vieta's formula of cubic equation, and extract the value of s with cubic equations.

Since the work of solution to quartic equation is too complex however well studied, we will not talk too much about it in this text. You can get more detailed information from here 3.

Finally, for algebraic equation of degree over , consider its subnormal series:

Similarily, if we are able to create linearly independent polynomials which is symmetric under , we only get more linearly independent equations. This is far from the requirement of collecting more linearly independet equations. We also can't follow the route of solving quartic equations by symmetrizing into intermediate polynomials since they also require normal subgroups in the middle to work but is simple. So finally algebraic equation of degree over is not solvable by radicals.

Conclusion

In this text, we characterize the step of solving an algebraic equation as follow:

  1. Come up with some linearly independent polynomial of roots .
  2. Find what other polynomials can transform into under permutations of , so we have a set of polynomials .
  3. Symmetrize these polynomials and see how the coefficients of the polynomial can be matched to them. Since the values of each are distinct, in order to extract their values, we will need to symmetrize them into Vieta's formula of equation of order . Sometimes intermediate polynomials are needed and we would have solved multiple algebraic equations before we can extract the value for .
  4. With the extracted values of these , we generate more linearly independent equations.
  5. Solve the system of linear equations to extract the values of .

For an algebraic equation of degree over , to extract the value of any linearly independent polynomial we create for it, we will need to solve an algebraic equation of order or . The former is not reducing the problem, the latter is not providing sufficient equations to the system of linear equations. So there's no way we can solve an algebraic equation of degree over by radicals.

Although this text is far from a formal proof, actually when compared with the formal Galois theory you will find there're many points missing, but I hope you can see in what way group theory objects can be connected to other mathematical objects from this text.


  1. We abbreviate the parameter list of function as when what the function takes in inferrable. We also use where is eligible operator concatenating . [return]
  2. We are not introducing concepts like stabilizer subgroup or orbit, they will be addessed in the text of group action, and not knowing them in detail does not affect the comprehension of this text. [return]
  3. Svante Janson, Roots of polynomials of degrees 3 and 4. https://arxiv.org/pdf/1009.2373.pdf [return]
March 30, 2023