A Taylor polynomial matches a function's value and successive derivatives at one center. Near that center, the polynomial can approximate a complicated function using only powers. An error bound tells you how much trust to place in the approximation.
Matching derivatives determines the coefficients
The degree-n Taylor polynomial centered at a is
The factorial compensates for repeated differentiation of a power. At all higher powers vanish, allowing each derivative to match the corresponding derivative of . A Maclaurin polynomial is simply a Taylor polynomial centered at zero.
The degree is the highest power retained, not necessarily the number of nonzero terms. For sine, the degree Maclaurin polynomial has only three nonzero terms:
Example: approximate an exponential
Every derivative of equals so at zero every derivative value is 1. The degree polynomial is
At this gives The approximation is close to because the input is near the center and the omitted terms are small.
A tangent line is the degree Taylor polynomial. Higher degrees extend the same local idea by matching curvature and further derivative information.
exponential
- (dashed)
cubic model
The exponential curve and its cubic Maclaurin polynomial are shown, meeting at x = 0 with matching first three derivatives.
The cubic Taylor polynomial closely follows eˣ near the center . Its error grows farther from that center.
Describe this figure
The exponential curve and its cubic Maclaurin polynomial are shown, meeting at with matching first three derivatives.
Curves: exponential; cubic model.
Marked points: center: filled marker at (0, 1).
The Lagrange bound controls the remainder
If the derivative of order exists and its magnitude is at most M between a and , then
M bounds the next derivative over the entire interval between the center and evaluation point, not just at the center. The distance factor explains why approximations generally worsen farther from the center.
For the exponential example, on The error is therefore less than The bound certifies accuracy without needing the exact value of the function.
Example: choose a degree for a guarantee
Approximate using A general Lagrange bound uses giving error at most
A tighter argument uses that the degree Taylor polynomial is the same polynomial, because the fourth derivative at zero is zero. The next derivative is cosine, bounded by 1, so the degree remainder bound is
The polynomial's expression has not changed, but a higher-order remainder theorem exploits the zero coefficient. Alternatively, the alternating sine series supplies the same next-nonzero-term bound when its conditions hold.
Centers other than zero
For centered at 1, the first values are and Thus
At the estimate is 0.095. On so the error is at most The expansion uses powers of not powers of , because 1 is the matching center.
Taylor polynomials are local models, not global identities
Matching several derivatives at a does not make a finite polynomial equal to f everywhere. Its quality depends on both distance from the center and the size of higher derivatives. Choosing a center near the requested input, where derivative values are manageable, can be more effective than increasing the degree around a distant center.
Also distinguish a finite Taylor approximation from claiming that an infinite Taylor series equals a function. That stronger claim requires the remainder to tend to zero as the degree grows. For familiar exponential and trigonometric functions this can be established, but it is not a consequence of simply writing formal derivative coefficients.
Try it yourself
Write the degree Maclaurin polynomial for and bound its error at using
Answer: giving 0.995. The bound is A sharper bound is possible, but this stated derivative bound already gives a valid guarantee.
What to remember
Match derivatives at the chosen center, keep the factorials and shifted powers, and bound the next derivative throughout the relevant interval. An approximation and its justified error bound belong together.