The Central Idea
A polynomial can be described in several ways:
- by its coefficients;
- by a formula;
- by its factors;
- by its values at selected inputs;
- by points on its graph.
These descriptions are connected because polynomial structure is highly constrained.
Two central facts guide this lesson:
\[ \boxed{\text{equal polynomials have equal corresponding coefficients}} \]and
\[ \boxed{\text{a degree-}n\text{ polynomial is determined by }n+1 \text{ distinct input-output pairs}}. \]Polynomial Equality
Suppose
\[ P(x)=a_nx^n+\cdots+a_1x+a_0 \]and
\[ Q(x)=b_nx^n+\cdots+b_1x+b_0. \]If
\[ P(x)=Q(x) \]for every value of \(x\), then
\[ \boxed{ a_n=b_n,\ldots,a_1=b_1,\ a_0=b_0 }. \]Identity Versus Equation
An equation may be true only for certain values of the variable.
For example,
\[ x^2=9 \]is true only when
\[ x=\pm3. \]A polynomial identity is true for every value of the variable in the intended domain.
For example,
\[ \boxed{(x+1)^2=x^2+2x+1} \]is true for every real or complex \(x\).
The Identity Symbol
Some writers use
\[ \equiv \]to emphasize that two expressions are identically equal.
Thus one may write
\[ (x+1)^2\equiv x^2+2x+1. \]Ordinary equality is also widely used when the context already makes the identity clear.
Why Coefficients Must Match
Suppose
\[ P(x)=Q(x) \]for every \(x\).
Then
\[ P(x)-Q(x)=0 \]for every \(x\).
A nonzero polynomial of degree \(n\) can have at most \(n\) distinct zeros.
Since \(P-Q\) is zero at infinitely many values, it cannot be a nonzero polynomial.
Therefore,
\[ P-Q \]must be the zero polynomial, so all corresponding coefficients agree.
Example: Compare Coefficients
Suppose
\[ 3x^2+ax+7 = 3x^2-5x+b \]for every \(x\).
Compare corresponding coefficients:
\[ a=-5 \]and
\[ b=7. \]Therefore,
\[ \boxed{a=-5,\qquad b=7}. \]Example: Expand Before Comparing
Suppose
\[ (x+p)(x+q) = x^2+7x+10 \]for every \(x\).
Expand the left side:
\[ x^2+(p+q)x+pq. \]Compare coefficients:
\[ p+q=7 \]and
\[ pq=10. \]Therefore,
\[ \boxed{\{p,q\}=\{2,5\}}. \]Identities Can Determine Parameters
Suppose
\[ ax^2+bx+4 = 2(x-1)^2+3x+2 \]for every \(x\).
Expand:
\[ 2(x^2-2x+1)+3x+2. \] \[ =2x^2-4x+2+3x+2. \] \[ =2x^2-x+4. \]Therefore,
\[ \boxed{a=2,\qquad b=-1}. \]Checking an Identity
To verify a claimed polynomial identity, simplify both sides to standard form.
For example, verify
\[ (x+3)(x-3)+9=x^2. \]Expand:
\[ x^2-9+9=x^2. \]Therefore, the identity is valid for every \(x\).
One Numerical Check Is Not a Proof of Identity
Suppose two expressions happen to agree at \(x=2\).
That alone does not prove they agree for all \(x\).
For example,
\[ x^2 \]and
\[ 2x \]both equal \(4\) when \(x=2\), but
\[ x^2\not\equiv2x. \]An identity is a statement about every input, not one successful test.
How Many Agreement Points Are Enough?
If two polynomials both have degree at most \(n\) and agree at \(n+1\) distinct input values, then they must be identical.
In symbols, if
\[ P(x_k)=Q(x_k) \]for \(n+1\) distinct values
\[ x_0,x_1,\ldots,x_n, \]then
\[ \boxed{P(x)\equiv Q(x)} \]provided both degrees are at most \(n\).
Why \(n+1\) Points Are Enough
Consider
\[ R(x)=P(x)-Q(x). \]If both \(P\) and \(Q\) have degree at most \(n\), then \(R\) also has degree at most \(n\).
If \(P\) and \(Q\) agree at \(n+1\) distinct inputs, then \(R\) has \(n+1\) distinct zeros.
A nonzero polynomial of degree at most \(n\) cannot have more than \(n\) distinct zeros.
Therefore,
\[ R(x)\equiv0, \]so
\[ P(x)\equiv Q(x). \]Examples of the Uniqueness Rule
- A constant polynomial is determined by \(1\) point.
- A linear polynomial is determined by \(2\) points with distinct \(x\)-values.
- A quadratic polynomial is determined by \(3\) points with distinct \(x\)-values.
- A cubic polynomial is determined by \(4\) points with distinct \(x\)-values.
More generally,
\[ \boxed{\deg P\le n\Rightarrow n+1\text{ distinct data points determine }P}. \]What Is Polynomial Interpolation?
Polynomial interpolation is the process of constructing a polynomial that passes exactly through specified data points.
If the points are
\[ (x_0,y_0),\ (x_1,y_1),\ldots,(x_n,y_n) \]with distinct \(x\)-values, then there is exactly one polynomial of degree at most \(n\) satisfying
\[ P(x_k)=y_k \]for every listed point.
Linear Interpolation from Two Points
Suppose a line must pass through
\[ (1,3) \]and
\[ (4,9). \]Write
\[ P(x)=mx+b. \]The slope is
\[ m=\frac{9-3}{4-1}=2. \]Therefore,
\[ P(x)=2x+b. \]Finish the Linear Interpolation
Use the point \((1,3)\):
\[ 3=2(1)+b. \]Hence
\[ b=1. \]Therefore,
\[ \boxed{P(x)=2x+1}. \]Checking:
\[ P(4)=8+1=9. \]Quadratic Interpolation from Three Points
Suppose a quadratic passes through
\[ (0,1),\qquad(1,3),\qquad(2,9). \]Write
\[ P(x)=ax^2+bx+c. \]Each point produces one equation.
Build the System
From \((0,1)\):
\[ c=1. \]From \((1,3)\):
\[ a+b+c=3. \]Since \(c=1\),
\[ a+b=2. \]From \((2,9)\):
\[ 4a+2b+c=9. \]Therefore,
\[ 4a+2b=8. \]Solve the Interpolation System
We have
\[ a+b=2 \]and
\[ 4a+2b=8. \]Double the first equation:
\[ 2a+2b=4. \]Subtract:
\[ 2a=4. \]Thus
\[ a=2. \]Then
\[ b=0. \]Therefore,
\[ \boxed{P(x)=2x^2+1}. \]Interpolation as a Linear System
For a quadratic
\[ P(x)=ax^2+bx+c, \]three data points produce three linear equations in the three unknown coefficients
\[ a,\ b,\ c. \]For a cubic, four data points produce four equations in four unknown coefficients.
Polynomial interpolation therefore connects directly to systems of linear equations and matrix methods.
Matrix View
For points
\[ (x_0,y_0),\ (x_1,y_1),\ (x_2,y_2), \]a quadratic interpolation problem can be written as
\[ \begin{bmatrix} x_0^2 & x_0 & 1\\ x_1^2 & x_1 & 1\\ x_2^2 & x_2 & 1 \end{bmatrix} \begin{bmatrix} a\\ b\\ c \end{bmatrix} = \begin{bmatrix} y_0\\ y_1\\ y_2 \end{bmatrix}. \]This structured coefficient matrix is a small example of a Vandermonde-type system.
Why Distinct \(x\)-Values Matter
A function cannot assign two different outputs to the same input.
Therefore, data such as
\[ (2,5) \]and
\[ (2,8) \]cannot both lie on the graph of a function \(P(x)\).
Polynomial interpolation assumes the input values are distinct.
Lagrange Interpolation
There is a direct formula that constructs the interpolating polynomial without first solving for ordinary coefficients.
For points
\[ (x_0,y_0),\ldots,(x_n,y_n), \]define basis polynomials
\[ L_k(x) = \prod_{\substack{j=0\\j\ne k}}^{n} \frac{x-x_j}{x_k-x_j}. \]Then
\[ \boxed{ P(x)=\sum_{k=0}^{n}y_kL_k(x) }. \]This is the Lagrange interpolation formula.
How a Lagrange Basis Polynomial Works
Each \(L_k(x)\) is constructed so that
\[ L_k(x_k)=1 \]while
\[ L_k(x_j)=0 \]for every
\[ j\ne k. \]Therefore, at \(x=x_k\), every term in the interpolation sum disappears except
\[ y_kL_k(x_k)=y_k. \]The polynomial is built to hit each required data value exactly.
Two-Point Lagrange Example
Interpolate the points
\[ (1,3) \]and
\[ (4,9). \]The basis polynomials are
\[ L_0(x)=\frac{x-4}{1-4} \]and
\[ L_1(x)=\frac{x-1}{4-1}. \]Therefore,
\[ P(x) = 3\frac{x-4}{-3} + 9\frac{x-1}{3}. \]Simplify the Lagrange Example
Simplify:
\[ P(x)=-(x-4)+3(x-1). \] \[ P(x)=-x+4+3x-3. \] \[ \boxed{P(x)=2x+1}. \]This is the same polynomial found earlier using the slope-intercept method.
Three-Point Lagrange Structure
For three points
\[ (x_0,y_0),\quad(x_1,y_1),\quad(x_2,y_2), \]the interpolating quadratic is
\[ P(x) = y_0 \frac{(x-x_1)(x-x_2)} {(x_0-x_1)(x_0-x_2)} \] \[ + y_1 \frac{(x-x_0)(x-x_2)} {(x_1-x_0)(x_1-x_2)} \] \[ + y_2 \frac{(x-x_0)(x-x_1)} {(x_2-x_0)(x_2-x_1)}. \]Each term is activated at one data point and vanishes at the others.
Interpolation Does Not Require Expanded Form
A Lagrange interpolating polynomial may initially look complicated.
It is still a valid polynomial even before expansion.
Expanded form is useful when coefficients are needed, but factored or basis form can reveal the construction more clearly.
Interpolation from Known Zeros
Root information can also determine much of a polynomial.
Suppose a quadratic has zeros
\[ 1 \]and
\[ 4. \]Then
\[ P(x)=a(x-1)(x-4) \]for some nonzero constant \(a\).
One additional point determines \(a\).
Example: Roots Plus One Point
Suppose the zeros are \(1\) and \(4\), and
\[ P(0)=8. \]Write
\[ P(x)=a(x-1)(x-4). \]Substitute \(x=0\):
\[ 8=a(-1)(-4)=4a. \]Therefore,
\[ a=2. \]Hence
\[ \boxed{P(x)=2(x-1)(x-4)}. \]Multiplicity Information Reduces the Unknown Structure
Suppose a cubic has a root \(2\) of multiplicity \(2\) and root \(-1\) of multiplicity \(1\).
Then
\[ P(x)=a(x-2)^2(x+1). \]Only the leading scale factor \(a\) remains unknown.
One nonzero function value can determine it.
Finite Differences and Polynomial Degree
When input values are equally spaced, finite differences can reveal polynomial structure.
For a polynomial of degree \(n\), the \(n\)th finite differences are constant.
For example:
- linear data have constant first differences;
- quadratic data have constant second differences;
- cubic data have constant third differences.
Example: Constant Second Differences
| \(x\) | \(P(x)\) | First Difference | Second Difference |
|---|---|---|---|
| 0 | 1 | 2 | 4 |
| 1 | 3 | 6 | 4 |
| 2 | 9 | 10 | |
| 3 | 19 |
The constant second difference suggests quadratic behavior.
Indeed,
\[ P(x)=2x^2+1. \]Interpolation Versus Extrapolation
Interpolation estimates or constructs values within the range of known data.
Extrapolation uses the model beyond the observed range.
Even when a polynomial exactly fits known data, predictions far outside the data range may be unreliable if the polynomial is only a model rather than a known exact law.
Interpolation Versus Regression
Interpolation requires the polynomial to pass exactly through the supplied data points.
Regression instead seeks an approximate relationship, often when data contain measurement error or natural variation.
Therefore:
\[ \boxed{\text{interpolation}=\text{exact fit}} \]while
\[ \boxed{\text{regression}=\text{approximate fit}}. \]More Points Do Not Always Mean a Better Model
Given \(n+1\) distinct points, a degree-at-most-\(n\) interpolating polynomial exists and is unique.
But a high-degree interpolating polynomial may bend strongly between data points.
Exact algebraic fit and good practical modeling are different goals.
In applied work, the simplest useful model is often preferable to a high-degree polynomial that merely reproduces every observed value.
Identity from Enough Data
Suppose \(P\) and \(Q\) are both quadratics.
If
\[ P(0)=Q(0), \] \[ P(1)=Q(1), \]and
\[ P(2)=Q(2), \]then the two quadratics agree at three distinct inputs.
Since degree is at most \(2\),
\[ \boxed{P(x)\equiv Q(x)}. \]Why Two Points Are Not Enough for a Quadratic
Many different quadratics can pass through the same two points.
For example, both
\[ P(x)=x \]and
\[ Q(x)=x+(x-1)(x-2) \]agree at
\[ x=1 \]and
\[ x=2. \]But they are not identical.
Three distinct points are needed to determine a quadratic uniquely.
Polynomial Identities as Coefficient Constraints
Suppose
\[ A(x)(x-1)+B(x)(x+1)=2x^2+3 \]and the forms of \(A\) and \(B\) contain unknown coefficients.
Expanding and comparing coefficients converts the identity into a system of algebraic conditions.
This method appears throughout algebra, partial fractions, approximation, and symbolic computation.
A Simple Parameter Identity
Find \(a\), \(b\), and \(c\) if
\[ a(x+1)^2+b(x+1)+c = 2x^2+7x+8 \]for every \(x\).
Expand:
\[ a(x^2+2x+1)+bx+b+c. \] \[ =ax^2+(2a+b)x+(a+b+c). \]Compare coefficients:
\[ a=2, \] \[ 2a+b=7, \] \[ a+b+c=8. \]Finish the Parameter Identity
Since
\[ a=2, \]we have
\[ 4+b=7, \]so
\[ b=3. \]Then
\[ 2+3+c=8, \]giving
\[ c=3. \]Therefore,
\[ \boxed{a=2,\quad b=3,\quad c=3}. \]Common Error: Comparing Unlike Forms Too Early
Coefficients should be compared only after both sides are expressed in compatible standard form.
For example,
\[ (x+2)^2 \]should first become
\[ x^2+4x+4. \]Only then should its coefficients be compared with another quadratic.
Common Error: Treating Agreement at One Point as Identity
Two different polynomials may share one, two, or many values without being identical.
Degree information determines how many agreement points force identity.
Without a degree bound, finitely many matching values do not establish an identity.
Common Error: Using Repeated \(x\)-Values as Separate Interpolation Conditions
Standard point interpolation requires distinct input values.
The points
\[ (2,3) \]and
\[ (2,5) \]are incompatible with any function.
Repeated inputs with additional derivative information lead to more advanced interpolation methods, but that is a different problem.
Common Error: Confusing Exact Interpolation with Prediction
An interpolating polynomial is guaranteed to match the supplied points.
It is not automatically guaranteed to describe the real process that produced them.
Mathematical exactness at known data points and scientific reliability of a model are separate questions.
A Reliable Procedure for Coefficient Comparison
- Expand both sides if necessary.
- Combine like terms.
- Write both polynomials in standard form.
- Insert zero coefficients for missing powers when useful.
- Match coefficients of equal powers.
- Solve the resulting equations for unknown parameters.
- Substitute back or re-expand to verify.
A Reliable Procedure for Direct Interpolation
- Choose the lowest degree allowed by the problem.
- Write a general polynomial with unknown coefficients.
- Substitute each data point.
- Build the resulting linear system.
- Solve for the coefficients.
- Write the polynomial.
- Check every original data point.
A Reliable Procedure for Lagrange Interpolation
- Confirm that the \(x\)-values are distinct.
- Construct one basis polynomial \(L_k(x)\) for each data point.
- Make each \(L_k\) equal \(1\) at its own input and \(0\) at all others.
- Multiply \(L_k(x)\) by the corresponding output \(y_k\).
- Add the terms.
- Expand only if standard form is needed.
- Verify the interpolation values.
Practice 1: Compare Coefficients
Find \(a\) and \(b\) if
\[ 4x^2+ax+b = 4x^2-7x+9 \]for every \(x\).
Practice 2: Expand and Compare
Find \(p\) and \(q\) if
\[ (x+p)(x+q) = x^2+x-12 \]identically.
Practice 3: Test a Claimed Identity
Determine whether
\[ (x-4)^2+8x = x^2+16 \]is an identity.
Practice 4: Linear Interpolation
Find the linear polynomial passing through
\[ (2,5) \]and
\[ (6,13). \]Practice 5: Quadratic Interpolation
Find the quadratic
\[ P(x)=ax^2+bx+c \]passing through
\[ (0,2),\qquad(1,5),\qquad(2,12). \]Practice 6: Roots Plus One Point
A quadratic has roots \(2\) and \(-3\), and
\[ P(0)=-12. \]Find \(P(x)\).
Practice 7: Uniqueness
Two cubic polynomials agree at four distinct input values.
What can you conclude?
Practice 8: Enough Points?
Are two distinct points sufficient to determine a quadratic polynomial uniquely?
Explain.
Practice 9: Finite Differences
A table with equally spaced inputs has constant third differences.
What polynomial degree does this suggest?
Practice 10: Interpolation or Regression?
A scientist has noisy measurements and wants a simple curve that describes the overall trend without passing through every point.
Is interpolation or regression the more appropriate idea?
Practice Answers
Practice 1
Corresponding coefficients must match:
\[ \boxed{a=-7,\qquad b=9}. \]Practice 2
Expand:
\[ (x+p)(x+q)=x^2+(p+q)x+pq. \]Therefore,
\[ p+q=1 \]and
\[ pq=-12. \]The pair is \(4\) and \(-3\).
Thus
\[ \boxed{\{p,q\}=\{4,-3\}}. \]Practice 3
Expand:
\[ (x-4)^2+8x = x^2-8x+16+8x. \] \[ =x^2+16. \]Therefore,
\[ \boxed{\text{yes, it is an identity}}. \]Practice 4
The slope is
\[ m=\frac{13-5}{6-2}=2. \]Write
\[ P(x)=2x+b. \]Use \((2,5)\):
\[ 5=4+b. \]Therefore,
\[ \boxed{P(x)=2x+1}. \]Practice 5
From \((0,2)\),
\[ c=2. \]From \((1,5)\),
\[ a+b=3. \]From \((2,12)\),
\[ 4a+2b=10. \]Double the first equation:
\[ 2a+2b=6. \]Subtract:
\[ 2a=4, \]so
\[ a=2,\qquad b=1. \]Therefore,
\[ \boxed{P(x)=2x^2+x+2}. \]Practice 6
Write
\[ P(x)=a(x-2)(x+3). \]Use \(P(0)=-12\):
\[ -12=a(-2)(3)=-6a. \]Thus
\[ a=2. \]Therefore,
\[ \boxed{P(x)=2(x-2)(x+3)}. \]Practice 7
Two degree-at-most-\(3\) polynomials agreeing at \(4\) distinct inputs must be identical.
\[ \boxed{P(x)\equiv Q(x)}. \]Practice 8
No.
A quadratic has three independent coefficients, and infinitely many quadratics can pass through two distinct points.
Three distinct input-output points are required for unique quadratic interpolation.
Practice 9
Constant third differences suggest a polynomial of degree
\[ \boxed{3}. \]Practice 10
Because the data are noisy and an approximate trend is desired,
\[ \boxed{\text{regression}} \]is more appropriate than exact interpolation.
What to Remember
Polynomial identities hold for every input.
If two polynomials are identical, their corresponding coefficients are equal.
A nonzero polynomial of degree \(n\) has at most \(n\) distinct zeros. Therefore, two polynomials of degree at most \(n\) that agree at \(n+1\) distinct inputs must be identical.
This leads to the interpolation principle:
\[ \boxed{ n+1\text{ points with distinct }x\text{-values determine a unique polynomial of degree at most }n }. \]Interpolating polynomials can be constructed by solving for coefficients, by using known roots and factors, or by formulas such as Lagrange interpolation.
Interpolation is an exact algebraic fit. It should not be confused with regression, which seeks an approximate model.
Section Perspective
Polynomial and power relationships now connect several views of the same mathematical objects:
- power terms and polynomial structure;
- addition, multiplication, and division;
- factoring and special forms;
- quadratic equations and complex roots;
- zeros, factors, and multiplicity;
- end behavior and turning points;
- identities, coefficients, and interpolation from data.
Together these ideas make polynomials not merely expressions to manipulate, but structured functions that can be built, analyzed, solved, compared, and reconstructed from partial information.