Appendix A Some Facts About Generalized Functions
We recall here some basic facts and formulas from the theory of generalized func- tions. There are many mathematical textbooks on this subject. Physicists may find useful concise texts, for example Chaps. 2÷4 in [12], or Chaps. 2÷3 in [11]. Com- prehensive introduction to the subject can be found in [2]. The generalized functions that appear in field theory are of the so called Schwartz class, denoted as S(Rn) or S∗(Rn). The reason is that all such generalized functions have a Fourier transform. The generalized function of the Schwartz class1 (g. f.) is, by definition, a linear and continuous functional on the Schwartz space of functions, denoted as S(Rn). Elements of S(Rn) are called test functions. They are complex ∞ valued functions on Rn of the C class. Moreover, it is assumed that such functions, and all their derivatives, vanish in the limit |x|= (x1)2 + (x2)2 + ...+ (xn)2 → ∞, also when multiplied by any finite order polynomial in the variables x1,...,xn. Here x denotes arbitrary point in Rn and x1,...,xn are its Cartesian coordinates. The space S(Rn) is endowed with a topology, but we shall not describe it here. 2 Examples of test functions from the space S(R1) include e−ax and 1/ cosh(ax), where a > 0 is a real constant. On the other hand, (1 + x2)−1 is not test function from S(R1). The value of a generalized function F ∈ S∗(Rn) on a test function f ∈ S(Rn) is denoted in mathematical literature as F(x), f (x) , but in physics the most popular is the misleading notation dn xF(x) f (x), e.g., dn x δ(x) f (x) in the case of the Dirac delta. One should keep in mind that the integral here is merely a symbol that replaces , from the mathematical notation—it is not the true integral. It may hap- pen however, that a generalized function is represented by an ordinary function F(x) such that the true integral dn xF(x) f (x) exists for all f ∈ S(Rn). Such F is called regular g.f. For example, the step function (x), x ∈ R1, is a regular generalized ∗( 1) ∞ ( ) ( ) = ∞ ( ) function from S R , because the integral −∞dx x f x 0 dx f x exists for every f ∈ S(R1). We show the integration range when we deal with the true integral. The Dirac delta δ(x) is the prominent example of non regular g.f. In the mathematical notation its definition has the form δ(x), f (x)= f (0).
1 Other names are also used: distribution for generalized function, and tempered distribution for generalized functions of the Schwartz class.
H. Arod´z, L. Hadasz, Lectures on Classical and Quantum Theory of Fields, 343 DOI 10.1007/978-3-642-15624-3, C Springer-Verlag Berlin Heidelberg 2010 344 A Some Facts About Generalized Functions
The derivative ∂i F of a g.f. F is defined as follows: n n d x ∂i F(x) f (x) =− d xF(x)∂i f (x) for all test functions f . One should remember that this is the definition, and not the formula of integration by parts. For example, d(x) ∞ df(x) ∞ df(x) dx f (x) =− dx (x) =− dx = f (0), dx −∞ dx 0 dx
hence (x) = δ(x). Derivative of g.f. always exists and is a generalized function. ˜( ) = ( π)n/2 n ( ) ( ) The Fourier transform f k 2 Rn d x exp ikx f x of a test function f also is a test function from the space S(Rn). The g.f. F˜(x) ∈ S∗(Rn) such that for every f ∈ S(Rn) dn x F˜(x) f (x) = dnkF(k) f˜(k), is called the Fourier transform of the g.f. F. It exists for any F ∈ S∗(Rn). The oper- ation of taking the Fourier transform is continuous with respect to F. This property is used in order to facilitate computation of the Fourier transform of (x)—we first compute the Fourier transform of e−εx (x), where ε>0, and take the limit ε → 0+ at the end.2 Thus,