Bump function explained

f : \Reals ^{n} \to \Reals on a Euclidean space \Reals ^{n} which is both smooth (in the sense of having continuous derivatives of all orders) and compactly supported. The set of all bump functions with domain \Reals ^{n} forms a vector space, denoted C ^{n)} _{0 (\Reals} or C ^{n).} _{c (\Reals} The dual space of this space endowed with a suitable topology is the space of distributions.

Examples

The function

\Psi : \Reals \to \Reals given byis an example of a bump function in one dimension. It is clear from the construction that this function has compact support, since a function of the real line has compact support if and only if it has bounded closed support. The proof of smoothness follows along the same lines as for the related function discussed in the Non-analytic smooth function article. This function can be interpreted as the Gaussian function \exp\left( - y ^{2\right)} scaled to fit into the unit disc: the substitution y ^{2} = { 1 } / {\left( 1 - x ^{2\right)}} corresponds to sending x = \pm 1 to y = inf ty.

A simple example of a (square) bump function in

n variables is obtained by taking the product of n copies of the above bump function in one variable, so $$\backslash Phi(x\_1,\; x\_2,\; \backslash dots,\; x\_n)\; =\; \backslash Psi(x\_1)\; \backslash Psi(x\_2)\; \backslash cdots\; \backslash Psi(x\_n).$$

A radially symmetric bump function in

n variables can be formed by taking the function \Psi _{n} : \Reals ^{n} \to \Reals defined by \Psi _{n(x)=\Psi(|x|)} . This function is supported on the unit ball centered at the origin.

For another example, take an

h that is positive on (c, d) and zero elsewhere, for example h(x) = \begin{cases} \exp\left( - (1)/((x - c)(d - x)) \right),& c < x < d \\ 0 ,& otherwise \end{cases} .

Smooth transition functions

Consider the function

f(x) = \begin{cases}e ^{- (1)/(x)} & if x> 0 ,\ 0 & if x\le 0 ,\end{cases}

defined for every real number x.

The function

g(x) = (f(x))/(f(x) + f( 1- x)) , x\in R ,

has a strictly positive denominator everywhere on the real line, hence g is also smooth. Furthermore, g(x) = 0 for x ≤ 0 and g(x) = 1 for x ≥ 1, hence it provides a smooth transition from the level 0 to the level 1 in the unit interval [0, 1]. To have the smooth transition in the real interval [a, b] with a < b, consider the function

R \ni x\mapsto g l ( (x - a)/(b - a) r ).

For real numbers, the smooth function

R \ni x\mapsto g l ( (x - a)/(r )g l ( b - a) (d - x)/(d - c) r )

equals 1 on the closed interval [b, c] and vanishes outside the open interval (a, d), hence it can serve as a bump function.

Caution must be taken since, as example, taking

\{a =-1 \} < \{b = c =0 \} < \{d =1 \} , leads to:q(x) = (1)/(1+ e ^{(1-2 |x|)/(x ^{2-|x|})}) which is not an infinitely differentiable function (so, is not "smooth"), so the constraints must be strictly fulfilled.

Some interesting facts about the function:

q(x,a) = (1)/(1+ e ^{(a( 1-2 |x|))/(x ^{2-|x|})}) Are that q\left(x, (\sqrt{ 3)/() }\right) make smooth transition curves with "almost" constant slope edges (a bump function with true straight slopes is portrayed this Another example).

A proper example of a smooth Bump function would be:

u(x) = \begin{cases} 1 , if x =0 , \ 0 , if |x|\geq 1 , \ (1)/(1+ e ^{(1-2 |x|)/(x ^{2-|x|})}) , otherwise , \end{cases}

A proper example of a smooth transition function will be:

w(x) = \begin{cases} (1)/(1+ e ^{(2 x -1)/(x ^{2-x})}) & if 0 <x< 1 ,\ 0 & if x\leq 0 ,\ 1 & if x\geq 1 ,\end{cases}

where could be noticed that it can be represented also through Hyperbolic functions:

(1)/(1+ e ^{(2 x -1)/(x ^{2-x})}) = (1)/(2) \left( 1- \tanh\left( (2 x -1)/(2 (x ^{2-x)}) \right) \right)

Existence of bump functions

It is possible to construct bump functions "to specifications". Stated formally, if

K is an arbitrary compact set in n dimensions and U is an open set containing K, there exists a bump function \phi which is 1 on K and 0 outside of U. Since U can be taken to be a very small neighborhood of K, this amounts to being able to construct a function that is 1 on K and falls off rapidly to 0 outside of K, while still being smooth.

Bump functions defined in terms of convolution

The construction proceeds as follows. One considers a compact neighborhood

V of K contained in U, so K \subseteq V ^{\circ\subseteq} V \subseteq U. The characteristic function \chi _{V} of V will be equal to 1 on V and 0 outside of V, so in particular, it will be 1 on K and 0 outside of U. This function is not smooth however. The key idea is to smooth \chi _{V} a bit, by taking the convolution of \chi _{V} with a mollifier. The latter is just a bump function with a very small support and whose integral is 1 . Such a mollifier can be obtained, for example, by taking the bump function \Phi from the previous section and performing appropriate scalings.

Bump functions defined in terms of a function

c : \Reals \to [ 0 , inf ty) with support ( -inf ty, 0 ]

An alternative construction that does not involve convolution is now detailed. It begins by constructing a smooth function

f : \Reals ^{n} \to \Reals that is positive on a given open subset U \subseteq \Reals ^{n} and vanishes off of U. This function's support is equal to the closure \overline{U} of U in \Reals ^{n,} so if \overline{U} is compact, then f is a bump function.

Start with any smooth function

c : \Reals \to \Reals that vanishes on the negative reals and is positive on the positive reals (that is, c = 0 on ( -inf ty, 0 ) and c > 0 on ( 0 , inf ty), where continuity from the left necessitates c( 0 ) = 0 ); an example of such a function is c(x) : = e ^{-1/x} for x > 0 and c(x) : = 0 otherwise. Fix an open subset U of \Reals ^{n} and denote the usual Euclidean norm by \| ⋅ \| (so \Reals ^{n} is endowed with the usual Euclidean metric). The following construction defines a smooth function f : \Reals ^{n} \to \Reals that is positive on U and vanishes outside of U. So in particular, if U is relatively compact then this function f will be a bump function.

If

U = \Reals ^{n} then let f = 1 while if U = \varnothing then let f = 0 ; so assume U is neither of these. Let \left(U _{k\right)} ^{inf ty} ^{k =1} be an open cover of U by open balls where the open ball U _{k} has radius r _{k} > 0 and center a _{k} \in U. Then the map f _{k} : \Reals ^{n} \to \Reals defined by f _{k(x)} = c\left(r ^{2} _{k} - \left\|x - a ^{2 \right)} _{k\right\|} is a smooth function that is positive on U _{k} and vanishes off of U _{k.} For every k \in N , let$$M\_k\; =\; \backslash sup\; \backslash left\backslash ,$$where this supremum is not equal to +inf ty (so M _{k} is a non-negative real number) because \left(\Reals ^{n} \setminus U _{k\right)} \cup \overline{U _{k}} = \Reals ^{n,} the partial derivatives all vanish (equal 0 ) at any x outside of U _{k,} while on the compact set \overline{U _{k},} the values of each of the (finitely many) partial derivatives are (uniformly) bounded above by some non-negative real number.^[1] The series $$f\; ~:=~\; \backslash sum\_^\; \backslash frac$$converges uniformly on \Reals ^{n} to a smooth function f : \Reals ^{n} \to \Reals that is positive on U and vanishes off of U. Moreover, for any non-negative integers p _{1,} \ldots, p _{n} \in \Z, $$\backslash frac\; f\; ~=~\; \backslash sum\_^\; \backslash frac\; \backslash frac$$where this series also converges uniformly on \Reals ^{n} (because whenever k \geq p _{1} + … + p _{n} then the k ^th term's absolute value is \leq \tfrac{M ^{k} _{k}{ 2} M _{k}} = \tfrac{ 1 }{ 2 ^{k}} ). This completes the construction.

As a corollary, given two disjoint closed subsets

A, B of \Reals ^{n,} the above construction guarantees the existence of smooth non-negative functions f _{A,} f _{B} : \Reals ^{n} \to [ 0 , inf ty) such that for any x \in \Reals ^{n,} f _{A(x)} = 0 if and only if x \in A, and similarly, f _{B(x)} = 0 if and only if x \in B, then the function $$h\; ~:=~\; \backslash frac\; :\; \backslash Reals^n\; \backslash to\; [0,\; 1]$$ is smooth and for any x \in \Reals ^{n,} h(x) = 0 if and only if x \in A, h(x) = 1 if and only if x \in B, and 0 < h(x) < 1 if and only if x \not\in A \cup B. In particular, h(x) ≠ 0 if and only if x \in \Reals ^{n} \smallsetminus A, so if in addition U : = \Reals ^{n} \smallsetminus A is relatively compact in \Reals ^{n} (where A \cap B = \varnothing implies B \subseteq U ) then h will be a smooth bump function with support in \overline{U}.

Properties and uses

While bump functions are smooth, the identity theorem prohibits their being analytic unless they vanish identically. Bump functions are often used as mollifiers, as smooth cutoff functions, and to form smooth partitions of unity. They are the most common class of test functions used in analysis. The space of bump functions is closed under many operations. For instance, the sum, product, or convolution of two bump functions is again a bump function, and any differential operator with smooth coefficients, when applied to a bump function, will produce another bump function.

If the boundaries of the Bump function domain is

\partial x, to fulfill the requirement of "smoothness", it has to preserve the continuity of all its derivatives, which leads to the following requirement at the boundaries of its domain:$$\backslash lim\_\; \backslash frac\; f(x)\; =\; 0,\backslash ,\backslash text\; n\; \backslash geq\; 0,\; \backslash ,n\; \backslash in\; \backslash Z$$

The Fourier transform of a bump function is a (real) analytic function, and it can be extended to the whole complex plane: hence it cannot be compactly supported unless it is zero, since the only entire analytic bump function is the zero function (see Paley–Wiener theorem and Liouville's theorem). Because the bump function is infinitely differentiable, its Fourier transform must decay faster than any finite power of

1 /k for a large angular frequency |k|. ^[2] The Fourier transform of the particular bump function$$\backslash Psi(x)\; =\; e^\; \backslash mathbf\_$$from above can be analyzed by a saddle-point method, and decays asymptotically as$$|k|^\; e^$$for large |k|. ^[3]

Notes and References

1. The partial derivatives (\partial ^{p} f _{k})/(\partial ^{p _{1}} x _{1} … \partial ^{p _{n}} x _{n}) : \Reals ^{n} \to \Reals are continuous functions so the image of the compact subset \overline{U _{k}} is a compact subset of \Reals. The supremum is over all non-negative integers 0 \leq p = p _{1} + … + p _{n} \leq k where because k and n are fixed, this supremum is taken over only finitely many partial derivatives, which is why M _{k} < inf ty.
2. K. O. Mead and L. M. Delves, "On the convergence rate of generalized Fourier expansions," IMA J. Appl. Math., vol. 12, pp. 247–259 (1973) .
3. [Steven G. Johnson]