Knowing a polynomial at several points gives equations for its coefficients. How much information do those values determine? The degree bound is decisive: without one, a nonzero polynomial can vanish at all the selected points and be added without changing any of the data. Interpolation constructs the unique low-degree representative and connects evaluation with polynomial quotients.
Build a polynomial from its values
Definition 9.1.1
(Lagrange basis polynomials)
Let be a field and let be distinct elements of (the nodes). For each define
Each denominator is nonzero because the nodes are distinct; it is invertible because is a field. At every factor equals 1, while at any other node one numerator vanishes. Thus .
Theorem 9.1.2
(Existence and uniqueness of the interpolating polynomial)
Let be a field, let be distinct, and let . There is exactly one polynomial of degree at most (the zero polynomial included) with for every , namely
Proof
(Lagrange basis for existence, factor theorem for uniqueness)
Existence: the sum has degree at most , and because . Uniqueness: if is another such polynomial, then vanishes at all distinct nodes. Repeated application of the factor theorem gives . A nonzero polynomial divisible by that product has degree at least , whereas has degree at most . Therefore . This argument includes the zero polynomial without assigning it a degree.
Proof X-ray
What the uniqueness argument uses
The non-routine step is the passage from separate roots to the single divisibility . Write . At this gives , and the factor is nonzero, so only because a field has no zero divisors; this cancellation is repeated for each further node, which uses that the nodes are distinct. The degree count then closes the argument: a nonzero multiple of a degree- polynomial cannot have degree below . Without the degree bound, the conclusion is only with , and uniqueness fails.
Worked example 9.1.3
(Interpolation over )
Over , prescribe , , . Set , and . The first two basis polynomials are
where in . Hence . Direct substitution gives modulo 5. The third basis polynomial is not needed in this sum because its coefficient is zero. Uniqueness does not depend on the ; it follows from the factor-theorem argument above.
Worked example 9.1.4
(Constant data (an edge case))
Take any field with at least three elements, three distinct nodes, and the data . The theorem promises a polynomial of degree at most , but the constant polynomial already has these values and degree . By uniqueness, , so . The bound is an upper bound, not the actual degree.
Counterexample mode
Interpolation needs a field
Claim: interpolation at distinct nodes is unique over any commutative ring. Work in with nodes and , which are distinct. The polynomial is nonzero and has degree , yet and . So the data is matched both by and by . The proof fails exactly where a field was used: the denominator is not invertible in , so do not exist, and is a zero divisor relation that the cancellation in the X-ray cannot rule out.
Quick check
(Over , find the polynomial of degree at most with and .)
Use .
Solution · Answer
and the second term has coefficient , so in . Check: and .
Quick check
(A polynomial of degree at most over a field vanishes at four distinct points. What is it?)
Count the roots against the degree.
Solution · Answer
It is the zero polynomial: four distinct roots give a factor of degree , which a nonzero polynomial of degree at most cannot have.
What all the other solutions look like
Proposition 9.1.5
(All interpolating polynomials)
Put and let be the interpolating polynomial of degree at most . A polynomial satisfies for all if and only if
Proof
(Every solution differs from by a multiple of )
Every polynomial has the same values as at the selected nodes. Conversely, if has those values, has every as a root, so by the same factor-theorem argument.
Worked example 9.1.6
(A solution of degree 3)
Over with the data of the first example, and . Taking gives . Then , and modulo 5, so has the same values as although its degree is 3, in agreement with the proposition.
In quotient language, evaluation induces a ring homomorphism
where operations in are coordinatewise. It is well defined because a multiple of vanishes at every node. It is injective by the factor-theorem argument and surjective by interpolation, hence an isomorphism. In particular, the quotient has zero divisors when : for example, although neither class is zero. A quotient ring need not be a field to be useful.
Finite fields and a necessary distinction
Over , interpolation at all field elements represents every function by a unique polynomial of degree below . But the formal polynomials and are distinct and define the same function on : for nonzero , Lagrange's theorem in the multiplicative group gives , and . Thus equality of values on a finite field is not, by itself, equality of formal polynomials. The degree bound is what makes the representative unique.
Common mistake
(Equal as functions does not mean equal as polynomials)
Over , the polynomials and take the same values at (namely ), yet in . Concluding from agreement at every field element that two polynomials coincide is valid only when both have degree below the number of elements of the field, here below .
Quick check
(How many polynomials of degree at most over are there, and how does the theorem explain this count?)
Count coefficient choices, then count value triples at three nodes.
Solution · Answer
There are of them (three coefficients, five choices each). Interpolation at three fixed nodes gives a bijection between these polynomials and the triples of values , which is the same count.