How this calculation works
Build the finite polynomial around the chosen center and compare its value with the supported function at the target. A center of zero is a Maclaurin expansion. The analytic Lagrange bound measures truncation in exact arithmetic; it does not include floating-point evaluation error. Entered-expression mode restores general bounded formulas, symbolic derivative rows and a printed shifted polynomial through degree ten. Adjust the custom graph span or use the zoom buttons; this mode does not claim the certified bounds available for the four named functions.
Tₙ(x) = Σ from k=0 to n of f⁽ᵏ⁾(a)(x−a)ᵏ/k!A worked example
For exp(x) around zero through degree four at x=1, 1+1+1/2+1/6+1/24 = 65/24 ≈ 2.7083333333. The function value is about 2.7182818285.
Good to know
Four supported real functions only, degree 0–30. Trig inputs are radians. log(1+x) requires center and target above −1, and its infinite series need not converge at a distant target even though the finite polynomial is defined. Near-domain-boundary overflow is rejected; no symbolic arbitrary-function expansion. Entered-expression mode supports the bounded grammar through degree ten with a derivative-size cap. It has no general remainder bound or convergence-radius proof. Nonsmooth points and unresolved removable singularities can reject derivative evaluation.
Reference: OpenStax Calculus: Taylor and Maclaurin series (opens in a new tab) ↗
A center and a degree define the approximation
A Taylor polynomial matches a function’s derivatives at one chosen center. Moving the center changes the polynomial, even when the target x and degree stay the same. Setting the center to zero gives the familiar Maclaurin case.
Degree is the highest included power, rather than the number of nonzero terms. A degree-four sine expansion at zero still has only x−x³/6 because its even-power coefficients vanish. The coefficient table keeps those zero entries visible.
The polynomial tangent calculator isolates the first-order local prediction. A Taylor approximation calculator adds higher-order terms for its stated supported functions.
Compare approximation error with truncation theory
For exp(x) at x=1 around zero, degree four gives 65/24, about 2.7083333333. Subtracting exp(1) gives about −0.0099484951. A negative signed difference means the polynomial is below the function at this particular target.
The analytic truncation bound uses a bound on the next derivative along the real interval from center to target. It is a theoretical exact-arithmetic bound, not a guarantee about a floating-point implementation. Severe cancellation can add numerical error that the displayed bound does not include.
The logarithm needs an extra convergence distinction
A finite log(1+x) Taylor polynomial can be evaluated at a real target greater than −1 even when that target lies beyond the infinite series’ convergence disk around the center. A large degree does not automatically fix an unsuitable expansion center.
The polynomial tangent calculator shows the degree-one local idea through slope and a finite-offset comparison. This page adds higher derivatives for four supported functions; it does not accept arbitrary symbolic expressions or claim that a sampled overlay proves convergence.
Build a finite polynomial for an entered expression
Choose entered-expression mode to supply a bounded formula rather than one of the four named functions. The calculation differentiates that expression symbolically, evaluates each derivative at the selected center and divides by the corresponding factorial. Its polynomial includes shifted powers from zero through n, even when some coefficients are zero.
For sin(x) centered at zero with degree five, the polynomial is x−x³/6+x⁵/120. Its constant, quadratic and quartic coefficients vanish because the corresponding sine derivatives at zero vanish. A degree-five polynomial therefore has powers through five, not necessarily six nonzero terms. The derivative table keeps those zero coefficients visible.
Keep the custom mode’s evidence separate from a general error guarantee
The existing named-function modes retain their analytic truncation bounds under their stated assumptions. The general entered-expression mode does not inherit those bounds merely because it also produces a polynomial. Its target comparison is one numerical check, not a proof that the approximation is accurate everywhere in the plotted window.
A Taylor polynomial is defined from finitely many derivatives at one point. Proving that an infinite Taylor series converges to the original function requires additional analysis. Increasing the degree can improve a local approximation for familiar analytic functions, but does not guarantee improvement at a distant target, establish a radius of convergence or resolve a nonsmooth expansion point.
Inspect the center and the real domain before expanding
The function and every required derivative must be defined at the chosen center. Natural logarithms need positive arguments in their ordinary real domain, and square-root derivatives can be singular at an endpoint even when the function value itself is finite. Absolute value has different behavior on either side of its corner at zero.
Symbolic differentiation here uses explicit rules for the supported grammar rather than a complete algebraic simplifier. A removable singularity can remain in a derivative expression and prevent evaluation until the mathematics is simplified separately. Variable exponents use the ordinary real logarithmic differentiation rule, so a positive base is needed where that rule is used. These limits are reported rather than replaced by invented coefficients.
Change the graph window without changing the expansion
The custom graph span controls a symmetric window around the center. Zoom buttons adjust that span, while the center and derivative coefficients remain determined by the entered function and order. A separate target value gives a concrete comparison point, which need not fall inside the drawing window.
Undefined function evaluations split the sampled function into separate segments instead of drawing through known holes. The polynomial can remain defined at a target where the original expression is undefined; that does not extend the original function’s domain. Export the SVG with the center, order and formula so the shared picture is understood as a finite approximation rather than an identity.
A few questions, answered
Do sine and cosine use degrees?
No. Their arguments, centers and targets are in radians.
Why did degree four add no new sine term?
At center zero, the even-order sine derivatives vanish. Degree four includes that zero coefficient; it does not mean four nonzero terms.
Is the truncation bound the measured error?
No. The signed difference compares two floating-point evaluations; the analytic bound concerns the omitted Taylor remainder in exact arithmetic.