Take the continuous function on . It is Lipschitz continuous with Lipschitz constant one, but is not differentiable at . For , the sample values are , so
This is already a useful warning: the Bernstein polynomial is built from grid values but is not generally their polynomial interpolation.
Figure 1.
Bernstein polynomials approximate the values of an absolute value function, including its corner
. The black target is on ; colored curves are the explicitly computed Bernstein polynomials of degrees , , , and . Each has exact endpoint values and a smooth middle. These are theoretical functions, not measured data. Details.
Every finite approximant is differentiable, and symmetry makes its derivative zero at the center. The target instead has left derivative and right derivative . Smooth curves can therefore approach a corner in supremum norm while their center derivatives have no target value to approach. The Lipschitz error bound for Bernstein approximation quantifies the improvement in values.
At the sampling point , the polynomial inside the Bernstein polynomial has value . The falling factorial identity and the binomial theorem prove the displayed formula. For fixed , both the input and its scalar factor tend uniformly to as grows. The supremum norm contraction of the Bernstein polynomial therefore gives for .
Bernstein monomial recurrence Created 2026-10-05 Updated 2026-10-14
For and , the Bernstein polynomial of the monomial satisfies
It follows by applying and expanding with the binomial theorem. The coefficients sum to one; only the first tends to one as . Induction on proves uniform convergence of these monomial approximants, starting from .
Bernstein polynomial Created 2026-09-24 Updated 2026-10-14
For a function and an integer , its Bernstein polynomial is
The Bernstein basis supplies nonnegative weights whose sum is one. Equivalently, the weights are the probabilities of a binomial distribution, and the result is an expected value. At a given , this averages the sampled values with weights concentrated near .[ref]
The endpoints are exact: and . Interior sampled points need not be interpolated. For example, binomial distribution moments give and . Thus every affine function is reproduced exactly, whereas is generally changed. Bernstein polynomial degree preservation is a statement about degree of a polynomial, not equality of the entire polynomial.
For a continuous , the uniform continuity of and the variance explain uniform convergence. Split the expected value of the error into , where uniform continuity makes the change small, and its complement, whose probability is at most by Chebyshev's inequality. The bounded function bounds the remaining error. This proves convergence without assuming differentiability.
For , the tensor-product Bernstein polynomial is the expected value of for independent random variables . The same uniform continuity and coordinatewise variance estimates give uniform convergence for each fixed finite .
Suppose linear operators on a normed vector space satisfy uniformly and on a dense subspace. The displayed triangle inequality for any element of that subspace first permits to approximate , then lets increase, proving for every . Applied to the Bernstein polynomial, the Weierstrass approximation theorem supplies the dense polynomial subspace and .
The uniform approximation closure of the integer-coefficient polynomials on consists precisely of the continuous functions taking integer values at both endpoints. Necessity follows because evaluations at zero and one are integers and convergent integer sequences are eventually constant. For sufficiency, the rounded Bernstein polynomial has integer monomial coefficients and differs uniformly by a vanishing amount from the Bernstein polynomial, which converges uniformly to the target.
If is Lipschitz continuous on with Lipschitz constant , then Cauchy-Schwarz inequality and the binomial distribution variance give
This is an explicit error guarantee, rather than only an existence theorem. For the Bernstein approximation of a corner, choosing guarantees supremum norm error at most ; guarantees error strictly less than . The guarantee is sufficient, not optimal.
Figure 1. . Left: error profiles for and selected Bernstein polynomials. Right: for even positive degrees, exact maximum error is compared with the sufficient bound . All values are model calculations; no experimental data are represented. Details.
For this specific target, the error reaches its maximum at . Indeed the derivative of is a weighted average of the slopes , all in : see derivatives of Bernstein polynomials. Subtracting the target makes the error nondecreasing to the left of the corner and nonincreasing to its right. For even , the symmetric binomial distribution at gives the exact value
The last equality follows by pairing terms about and telescoping the binomial coefficients. It gives for and for .
The Weierstrass approximation theorem asserts that every real continuous function on a compact interval is a uniform limit of polynomials. An affine change of variable reduces the interval to . There are two complementary routes: construct positive approximation operators, or construct an algebra dense enough to approximate absolute values and hence pointwise maxima and minima.
For the first route, the Korovkin theorem reduces convergence of positive linear operators on continuous functions to the three tests . Here is its proof. Given and , uniform continuity supplies a such that for . For all one then has
Apply positivity to the two real inequalities obtained from this bound. At this gives
The expression in brackets is nonnegative and tends uniformly to zero by the three test assumptions. Also uniformly. Taking the upper limit of the supremum norm and then letting proves convergence for every . This is the quadratic barrier estimate for positive approximation operators.
The Bernstein polynomial supplies the constructive corollary
The binomial theorem makes its nonnegative weights sum to one. Differentiating the binomial theorem, or using the first two binomial distribution moments, gives
The Korovkin theorem proves uniformly, and each is a polynomial, proving the Weierstrass approximation theorem. The same criterion gives the Schoenberg spline operator convergence in Solution 3. For periodic functions its counterpart uses : the nonnegative barrier is bounded away from zero outside any fixed small arc. This periodic Korovkin test set gives convergence of the normalized Fejér sums and density of trigonometric polynomials.
The Lebesgue convergence criterion for approximation operators addresses stability. Suppose bounded linear operators converge to the identity on every polynomial. Then they converge on every continuous function if and only if . Sufficiency is the extension of approximation operators from a dense subspace: for a polynomial with and a common bound ,
First let , then . Necessity follows from the Uniform boundedness principle, since a convergent is bounded for every . Applied to approximation projectors, the Lebesgue inequality for approximation projectors is
Indeed makes ; take the infimum. Thus density of the ranges and a uniform bound on the Lebesgue constants of interpolation imply convergence. Density alone does not control a sequence of interpolation projectors.
There is also a kernel version of this stability criterion, useful for trigonometric polynomial approximation. If integrable periodic kernels have uniformly bounded norms, their integrals tend to one, and for every , then convolution with converges uniformly to every continuous periodic function. Subtract and split the error into a small arc, bounded by , and its complement, bounded by . The remaining mass error tends to zero. This approximate identity argument shows directly why nonnegative unit-mass Fejér kernels work even though ordinary Fourier partial sums need not converge uniformly for every continuous target.
Lebesgue's constructive approach to the Weierstrass approximation theorem starts instead with the absolute value. On define the polynomial recurrence for the absolute value
Induction gives , monotonicity, and
The pointwise limit satisfies and is nonnegative, hence equals . Dini's theorem makes the convergence uniform. More explicitly, the continuous decreasing errors have, for any positive tolerance, open sublevel sets covering the compact interval; a finite subcover supplies one index valid everywhere. Every continuous piecewise linear function has the form
The are its changes in slope. Scaling and translating the absolute-value approximation shows that is uniformly approximable by polynomials. Solution 4, or polygonal interpolation and uniform continuity, makes these piecewise linear functions dense in . Approximating first by and then by a polynomial is another complete proof of the Weierstrass approximation theorem.
The Stone-Weierstrass theorem generalizes from intervals to a compact Hausdorff space : a real subalgebra containing the constants and separating points is uniformly dense. Its proof uses the same absolute-value mechanism. The uniform closure of a function algebra is an algebra. If belongs to this closure, approximation of by polynomials on a bounded interval containing its range proves that also belongs to the closure. Therefore the closure contains
Point separation and constants allow exact interpolation of any two real values: choose with and solve for in . The lattice approximation from two-point approximation proved in Solution 4 now makes the closure all of .
The complex version requires a unital, point-separating subalgebra closed under complex conjugation. Its real-valued members separate points using real or imaginary parts, so the real result approximates both parts of every complex target. Conjugation closure is essential: polynomials in alone on the closed unit disk cannot approximate uniformly, because whereas .
Among the corollaries are density of multivariate polynomials on compact subsets of real Euclidean space, since coordinates separate points, and density of trigonometric polynomials on the circle, since sine and cosine separate points there. On a compact interval, polynomials with rational coefficients are a countable dense family, giving separability of . The real and complex approximation results above concern uniform approximation of continuous targets; neither asserts convergence of arbitrary interpolation schemes or arbitrary Fourier series.
For a continuous function on , its Bernstein polynomial is
The weights are nonnegative and, by the binomial theorem, sum to . Thus at every , including the endpoints. Taking the supremum norm gives
In particular, is a linear operator of norm one, since it preserves the constant function one.
Write the Bernstein polynomial in its partition-normalized Bernstein basis:
For , every weight is nonnegative, and the binomial theorem gives their sum as . Consequently
Taking the supremum norm proves . The construction yields degree at most , possibly lower for special functions.
For , the Bernstein polynomial is
The weights sum to one by the binomial theorem. Equivalently, if has the binomial distribution with parameters , then . This interpretation makes the cancellation of apparent high-degree terms transparent.
For the falling factorial , with , direct cancellation in the binomial weights gives
The polynomials form a monic triangular basis of the degree-at-most- polynomials. Hence , and
For , its leading coefficient is . If has actual degree and leading coefficient , linearity shows that the leading coefficient of is ; lower-degree terms cannot cancel it. Thus the Bernstein polynomial degree preservation property gives
This includes the requested , and also explains why the property cannot simply be extended to .
The same moments give and . Therefore
The first two approximation errors vanish identically, while
All three convergences are consequently uniform convergence on .
For fixed , let have the binomial distribution with parameters . The Bernstein polynomial is . Continuity on the compact interval gives uniform continuity and a finite bound .
Given , choose so whenever . Since and , Chebyshev's inequality gives
The bound is independent of , including both endpoints. Taking the supremum and then letting tend to infinity proves uniformly on . This provides the required polynomial-approximation theorem with its proof.
Write the Bernstein basis as , with for out-of-range indices. The Bernstein polynomial is
Differentiation and the identities and give . Shifting the first sum's index therefore yields
Repeating this calculation takes another forward finite difference of the coefficients and lowers the degree by one. Induction proves the derivatives of Bernstein polynomials formula
For , define a continuous function on by
All sampled arguments lie in , since the largest is . At the degree- sampling point , this gives exactly the coefficient , and hence . The rescaling is essential: the sampling meshes are and , and the function depends on . For , the derivative is the constant , interpreted as a degree-zero Bernstein polynomial.
Now fix and assume . The supplied uniform limit of normalized finite differences, together with and uniform continuity of , shows that
Positivity and imply . For completeness, if then , and . Splitting according to whether and using Chebyshev's inequality gives
First let and then ; this proves uniform convergence for each continuous . Finally,
Thus Bernstein polynomial derivatives converge uniformly to the corresponding derivatives of , for each fixed with . Continuity of alone is sufficient for the definition of , but not for this derivative conclusion.
The Bernstein polynomial is
It has degree at most , since cancellation can lower the degree. For , the Bernstein basis weights are nonnegative and their sum is by the binomial theorem. Consequently
Thus the Bernstein polynomial defines a positive linear operator on continuous functions satisfying . Since , its operator norm in the supremum norm is exactly one.
Write for the falling factorial. At the sampling points, , which vanishes for . For , cancellation of the factorials gives
Insert this into the Bernstein polynomial and set :
The binomial theorem evaluates the remaining sum. For , both sides equal one. This proves the Bernstein falling-factorial identity .
The Bernstein polynomial is the positive weighted average
For , the Bernstein basis weights are nonnegative and the binomial theorem gives . Hence . In particular,
This is the contraction property of the positive linear operator on continuous functions defined by the Bernstein polynomial.
First prove necessity. If polynomials with integer coefficients converge to in the supremum norm, their values at and are convergent sequences of integers. Each such sequence is eventually constant, since a convergent sequence is Cauchy and two distinct integers are at distance at least one. Its limit is an integer. Thus .
For sufficiency we also need uniform convergence of the Bernstein polynomial to . Here is a direct argument. For , write . The binomial distribution has and . For every , uniform continuity of and the Chebyshev inequality give
First make small and then large; the bound is uniform in . Hence .
If the endpoints are integers, part (b) gives . Each coefficient multiplying in is an integer. Expanding by the binomial theorem therefore gives integer coefficients in the ordinary monomial basis. The triangle inequality proves . Thus the integer endpoint condition is necessary and sufficient for integer-coefficient polynomial approximation.
For the constant monomial , the binomial theorem gives . For , the binomial coefficient identity gives
Thus , including by evaluation of the polynomials.
For and , substitute the preceding coefficient identity into the Bernstein polynomial sum and write . This yields the Bernstein monomial recurrence
The coefficients are nonnegative and sum to ; and the others tend to zero. The weights in any Bernstein polynomial sum are nonnegative and sum to , so . Induction on therefore gives : the term has that limit and the other finitely many terms vanish. In fact, the same estimate in the supremum norm proves uniform convergence on .
As an independent quantitative bound, take a binomial distribution variable with parameters . Then . On , factoring the difference of powers gives ; the Cauchy-Schwarz inequality and the variance of give
For the error is exactly zero. Finally the requested sum equals , so
First address the positivity request in part (a), which has no separate Solution heading in the stub. For , each and each coefficient is nonnegative. Their sum is therefore nonnegative, including at the endpoints:
This is the elementary direction of the positive Bernstein coefficients for a strictly positive polynomial criterion; the converse developed later needs strict positivity.
For completeness, the convergence fact listed as (i) also has a short proof. If , the Bernstein polynomial is . Its mean argument is and . Given , uniform continuity of on the compact interval supplies such that whenever . By Chebyshev's inequality, uniformly in ,
Letting and then proves . The two numbered items in the source are assumptions supplied for later parts, not additional exam questions; their stub sections contain these supporting derivations.
For , define its Bernstein polynomial
If , then
Let . By uniform continuity, for any choose such that whenever . Splitting the expectation according to that event and using Chebyshev's inequality gives, uniformly in ,
For sufficiently large this is below . Thus uniformly on . An affine change of variable treats every compact interval, proving the Weierstrass approximation theorem.