Preface


This section is devoted to expansions of real-valued functions into series over Laguerre polynomials
\[ f(x) = \sum_{k\ge 0} f_k L_k (x) , \qquad f_k = \int_0^{\infty} L_k (x) \,e^{-x} f(x) \, {\text d} x , \tag{7} \]
and Sonin polynomials
\[ f(x) = \sum_{k\ge 0} f_k^{(\alpha )} L_k^{(\alpha )} (x) , \qquad f_k^{(\alpha )} = \frac{k!}{\Gamma (k + \alpha + 1)}\int_0^{\infty} L_k^{(\alpha )} (x) \,e^{-x} x^{\alpha} f(x) \, {\text d} x , \qquad k = 0,1,2,\ldots . \tag{8} \]
It is known that these series converge in 𝔏² sense for functions f ∈ 𝔏²(ℝ+, xαe−x). We do not discuss a delicate topic of pointwise and uniform convergence of these series. Instead, we present some examples of these series for demonstration.

 

Laguerre Polynomials


Laguerre's polynomials,
\begin{equation} \label{EqLaguerre.1} L_n (x) = \sum_{k=0}^n (-1)^k \binom{n}{k} \frac{x^k}{k!} , \qquad n=0,1,2,\ldots , \end{equation}
are eigenfunctions of the singular Sturm--Liouville problem on the half-line (0, ∞):
\[ x\,y'' + \left( 1 -x \right) y' + \lambda\,y = 0, \qquad \lambda = n. \]
Here \( \displaystyle \binom{n}{k} = \frac{n^{\underline{k}}}{k!} = \frac{n \left( n-1 \right) \left( n-2 \right) \cdots \left( n-k+1 \right)}{1\cdot 2 \cdot 3 \cdots k} \) is the binomial coefficient. The polynomials \eqref{EqLaguerre.1} were invented by the Russian mathematician Pafnuty Chebyshev (1821--1894) in 1859. Therefore, these polynomials were known in nineteen century as Chebyshev--Laguerre polynomials.

The Laguerre equation has one regular singular point at the origin and irregular singular point at infinity. So the Laguerre polynomial is a bounded at the origin solution to the Chebyshev--Laguerre equation

\begin{equation} \label{EqLaguerre.2} x\,y'' + \left( 1-x \right) y' +n\,y =0, \qquad\mbox{or in self-adjoint form} \qquad \frac{{\text d}}{{\text d} x} \left( x\,e^{-x} \,y' \right) + n\,e^{-x}\,y =0 , \qquad x\in (0,\infty ). \end{equation}

 

Sonin Polynomials


In 1880, the Russian mathematician Nikolay Yakovlevich Sonin (1849--1915) introduced a generalization of the Laguerre equation:
\begin{equation} \label{EqLaguerre.3} x\,y'' + \left( 1-x + \alpha \right) y' +n\,y =0, \qquad\mbox{or in self-adjoint form} \qquad \frac{{\text d}}{{\text d} x} \left( x^{1 + \alpha} e^{-x} \,y' \right) + n\, x^{\alpha} e^{-x}\,y =0 , \qquad x\in (0,\infty ), \end{equation}
where α > −1 is a real parameter. It has a polynomial solution
\begin{equation} \label{EqLaguerre.4} L_n^{(\alpha )} (x) = \sum_{k=0}^n \frac{\Gamma (n+ \alpha + 1)}{\Gamma (k + \alpha + 1)} \cdot \frac{(-x)^k}{k! \left( n-k \right) !} = \frac{1}{n!} \sum_{k=0}^n \frac{\Gamma (n+ \alpha + 1)}{\Gamma (k + \alpha + 1)} \binom{n}{k} (-x)^k , \end{equation}
known as the Sonin polynomial of degree n. This function, also denoted as Ln(α, x), is usually referred to as the generalized or associated Laguerre polynomial.

 

Orthogonality of Laguerre Polynomials


A definition of orthogonality requires a special bilinear form, called an inner product, denoted with angle brackets such as in ⟨ 𝑎, b ⟩. A vector space of functions equipped with an inner product is called a Hilbert space subject that it is complete. An importatnt example of Hilbert space presents the space of square Lebesque integrabler real- or complex-valued functions on some interval, denoted by 𝔏². Then the inner product (with weight ρ ≥ 0)
\[ \left\langle f(x), g(x) \right\rangle = \int_a^b \overline{f(x)}\,g(x)\,\rho(x)\,{\text d}x , \]
generats a norm \( \| f \| = \left\langle f(x), f(x) \right\rangle^{1/2} . \) Here \( \overline{f(x)} \) denotes the complex conjugate of f(x). In particular, we are intereted in a semi-infinite interval (when 𝑎 = 0 and b = +∞) and an inner product involving a weight function ρ(x ≥ 0:
\[ \left\langle f(x), g(x) \right\rangle = \int_0^{\infty} \overline{f(x)}\,g(x)\,x^{\alpha} e^{-x} {\text d}x \qquad \Longrightarrow \qquad \| f \|^2 = \left\langle f(x), f(x) \right\rangle , \]
where α > −1. The corresponded complete space is denoted as 𝔏²(ℝ+; xαe−x). A Hilbert space is always complete meaning that every Cauchy sequence converges.

The Sonin polynomials or associated Laguerre polynomials are orthogonal over [0, ∞) with respect to the weighting function \( \rho (x) = x^{\alpha} e^{-x} : \)

\begin{equation} \label{EqLaguerre.5} \left\langle L_n^{(\alpha )} (x) , L_m^{(\alpha )} (x) \right\rangle = \int_0^{\infty} x^{\alpha} e^{-x} L_n^{(\alpha )} (x) \, L_m^{(\alpha )} (x) \,{\text d}x = \frac{\Gamma (n+\alpha + 1)}{n!}\, \delta_{n,m} , \end{equation}
where \( \Gamma (\nu ) = \int_0^{\infty} t^{\nu -1} e^{-t} {\text d}t \) is the gamma function of Euler and
\[ \delta_{n,m} = \begin{cases} 0, & \ \mbox{when} \quad n \ne m , \\ 1, & \ \mbox{for} \quad n = m, \end{cases} \]
is the Kronecker delta symbol. In particular,

\begin{equation} \label{EqLaguerre.6} \left\langle L_n (x) , L_m (x) \right\rangle = \int_0^{\infty} e^{-x} L_n (x) \, L_m (x) \,{\text d}x = \delta_{n,m} . \end{equation}
Note that Hilbert spaces 𝔏²(ℝ+; xαe−x) or 𝔏²(ℝ+; e−x) contain unbounded (not integrable) functions on semi-infinite line ℝ+ = [0, ∞), including polynomials.

 

Fourier--Laguerre Series


Arbitrary function f ∈ 𝔏²((0,∞); e−x), for which the integral
\[ \| f \|^2 = \int_0^{\infty} e^{-x} |f(x)|^2 {\text d} x < \infty \]
is finite, can be expanded into Fourier--Laguerre series:
\begin{equation} \label{EqLaguerre.7} f(x) = \sum_{i\ge 0} f_i L_i (x) , \qquad f_i = \int_0^{\infty} L_i (x) \,e^{-x} f(x) \, {\text d} x , \end{equation}
which converges in 𝔏p(ℝ+, e−x) for \( p \in \left( \frac{4}{3} , 4 \right) . \) Such expansion is based on the orthogonal property of Laguerre polynomials, Eq.\eqref{EqLaguerre.6}.

More generally, introducing the inner product and norm in 𝔏²((0,∞, xαe−x)

\[ \left\langle f(x), g(x) \right\rangle = \int_0^{\infty} x^{\alpha} e^{-x} f(x)\, g(x)\,{\text d} x \qquad \Longrightarrow \qquad \| f \|^2 = \int_0^{\infty} x^{\alpha} e^{-x} f^2(x)\,{\text d} x , \]
we expand a real-valued function f(x)∈𝔏²(ℝ+, xαe−x) into Sonin series
\begin{equation} \label{EqLaguerre.8} f(x) = \sum_{k\ge 0} f_k^{(\alpha )} L_k^{(\alpha )} (x) , \qquad f_k^{(\alpha )} = \frac{k!}{\Gamma (k + \alpha + 1)}\int_0^{\infty} L_k^{(\alpha )} (x) \,e^{-x} x^{\alpha} f(x) \, {\text d} x , \qquad k = 0,1,2,\ldots . \end{equation}
Theorem (Parseval): Let f: [0, ∞) → ℝ (or ℂ) belongs to the Hilbert space 𝔏²(ℝ+, xαe−x). Then Parseval's identity holds
\[ \int_0^{\infty} \left\vert f(x) \right\vert^2 x^{\alpha} e^{-x} {\text d}x = \sum_{n\ge 0} \left( f_k^{(\alpha )} \right)^2 \frac{\Gamma (n+\alpha +1)}{n!} = \sum_{n\ge 0} \frac{n!}{\Gamma (n+\alpha +1)} \,\left\langle f(x), L_n^{(\alpha )} (x) \right\rangle^2 . \]
If function f∈𝔏²((0,∞); xαe−x) is sufficiently smooth, then coefficients in Eq.\eqref{EqLaguerre.8} can be expressed as
\begin{equation} \label{EqLaguerre.9} f_n = \frac{1}{\| L_n (\alpha ,x \|^2}\,\langle f, L_n^{(\alpha )} \rangle = \frac{n!}{\Gamma (n + \alpha + 1)}\,\int_0^{\infty} f(x) \, L_n^{(\alpha )} (x)\, x^{\alpha} e^{-x} {\text d} x = \frac{1}{\| L_n (\alpha ,x \|^2} \cdot \frac{(-1)^n}{n!}\,\int_0^{\infty} f^{(n)} (x) \, x^{n+\alpha} e^{-x} {\text d} x . \end{equation}

When f(x) is a polynomial of degree k < n, then

\[ \left\langle f(x), L_n^{(\alpha )} (x) \right\rangle = \int_0^{\infty} f(x) \, L_n^{(\alpha )} (x)\, x^{\alpha} e^{-x} {\text d} x = 0. \]
The first example follows from the generating function:
\[ x^{-\alpha} e^x \Gamma (\alpha , x) = \sum_{n\ge 0} \frac{L_n^{(\alpha )} (x)}{n+1} , \]
where
\[ \Gamma (\nu , A) = \int_A^{\infty} t^{\nu -1} e^{-t} {\text d} t \]
is incomplete gamma function. This formula can be used for numerical evaluation of the incomplete gamma function as well as the complete gamma function.

Example 1: First, we expand the upper incomplete gamma function, known as Exponential integral:

\[ \Gamma (0 , x) = \int_x^{\infty} t^{-1} e^{-t} {\text d}t = -\mbox{Ei}(-x) = e^{-x} \sum_{n\ge 0} \frac{L_n (x)}{n+1} . \]
Of course, Mathematica has a dedicated command, ExpIntegralEi, but we apply the Laguerre series for its approximation. So we build partial sums:
      We plot the exponential integral (in blue) and its Laguerre approximations with 10 and 20 terms:
S10[x_] = Exp[-x]*Sum[LaguerreL[n, x]/(n + 1), {n, 0, 10}];
S20[x_] = Exp[-x]*Sum[LaguerreL[n, x]/(n + 1), {n, 0, 20}];
Plot[{-ExpIntegralEi[-x], S10[x], S20[x]}, {x, 0.5, 5}, PlotStyle -> {{Thickness[0.01], Blue}, {Thick, Orange}, {Thick, Purple}}]
       Laguerre approximations with 10 and 20 terms.            Mathematica code.

We also check the accuracy by evaluating approximate values at x = 5.0:
S10[5.0]
0.0031435
S20[5.0]
0.00191203
-ExpIntegralEi[-5.0]
0.0011483
NIntegrate[Exp[-x]/x, {x, 5.0, Infinity}]
0.0011483
So 20-term Laguerre approximation gives 3 correct decimal places.

We expand the incomplete gamma function into Laguerre series:

\[ \Gamma (\alpha , x) = \int_{x}^{\infty} t^{\alpha -1} e^{-t} {\text d} t = x^{\alpha} e^{-x} \sum_{n\ge 0} \frac{L_n^{(\alpha )} (x)}{n+1} \]
We take α = 3/2, and use the Sonin expansion:
\[ \Gamma \left(\frac{3}{2} , x\right) = \int_{x}^{\infty} t^{1/2} e^{-t} {\text d} t = x^{3/2} e^{-x} \sum_{n\ge 0} \frac{L_n^{(3/2 )} (x)}{n+1} . \]
      We plot the incomplete gamma function (in blue) and its Sonin approximation (in orange) with 50 terms along with the corresponding Cesàro regularization (in purple):
\[ C50(x) = x^{3/2} e^{-x} \sum_{n= 0}^{50} \frac{L_n^{(3/2 )} (x)}{n+1} \left( 1 - \frac{n}{51} \right) . \]
S50[x_] = x^(3/2)*Exp[-x]*Sum[LaguerreL[n, 3/2, x]/(n + 1), {n, 0, 50}];
C50[x_] = x^(3/2)*Exp[-x]* Sum[LaguerreL[n, 3/2, x]*(1 - n/51)/(n + 1), {n, 0, 50}];
Plot[{Gamma[3/2, x, Infinity], S50[x], C50[x]}, {x, 0.5, 5}, PlotStyle -> {{Thickness[0.01], Blue}, {Thick, Orange}, {Thick, Purple}}]
       Sonin and corresponding Cesàro approximations with 50 terms.            Mathematica code.

We also check the accuracy by evaluating approximate values at x = 5.0:
S50[5.0]
-0.0121446
C50[5.0]
0.0187587
Gamma[3/2, 5.0, Infinity]
0.0164538
NIntegrate[t^(1/2)*Exp[-t], {t, 5.0, Infinity}]
0.0164538
So we see that 50-term Sonin approximation has a poor accuracy. However, its Cesàro regularization gives much better approximation.    ■

Example 2: Our next example is about expansions that follow from the corresponding generating function. Upon changing the variable of expansion \( w = a/(a+1) , \) we find

\[ e^{-ax} = \frac{1}{(1+a)^{1 + \alpha}} \,\sum_{k\ge 0} \left( \frac{a}{1+a} \right)^k L_k^{(\alpha )} (x) \qquad \Re\alpha > ½. \]
In particular,
\[ e^{-x} = \sum_{k\ge 0} \frac{1}{2^{k + \alpha + 1}} \, L_k^{(\alpha )} (x) , \qquad 0 < x < \infty . \]
Multiplying the former series by \( (a+1)^{\alpha -1} \) and integrate, we get
\[ x^{-\alpha} e^x \Gamma (\alpha , x) = \sum_{n\ge 0} \frac{1}{n+1} \, L_n^{(\alpha )} (x) , \qquad \alpha > -1, \]
where \( \displaystyle \Gamma (\nu , A) = \int_A^{\infty} t^{\nu -1} e^{-t} {\text d} t \) is the incomplete gamma function.

With exponential function, we verify Parseval's identiy:
\[ \int_0^{\infty} \left( e^{-x} \right)^2 e^{-x} {\text d} x = \int_0^{\infty} e^{-3x} {\text d} x = \frac{1}{3} = \sum_{k\ge 0} \left( \frac{1}{2^{k + 1}} \right)^2 . \]
Integrate[Exp[-3*x], {x, 0, Infinity}]
1/3
Sum[1/4^(k + 1), {k, 0, Infinity}]
1/3

We also have expansion for the natural logarithm function:

\[ \ln x = \frac{\Gamma' (\alpha +1)}{\Gamma (\alpha +1)} - \Gamma (\alpha +1) \sum_{n\ge 1} \frac{(n-1)!}{\Gamma (\alpha +n +1)}\, L_n^{(\alpha )} (x) . \]
Here the logarithmic derivative of the gamma function
\[ \psi (x) = \frac{\text d}{{\text d}x}\,\ln\Gamma (x) = \frac{\Gamma' (x)}{\Gamma (x)} \]
is called the digamma function. Mathematica has a dedicated command: PolyGamma[x]. For α = 0, we get the Laguerre series
\[ \ln x = \psi (1) - \sum_{n\ge 1} \frac{1}{n}\, L_n (x) . \]
      We plot two Laguerre approximations with 10 and 50 terms.
ln[n_] = PolyGamma[1] - Sum[(1/k)*LaguerreL[k, x], {k, 1, n}];
Plot[{Log[x], ln[10], ln[50]}, {x, 0.5, 10}, PlotStyle -> {{Thickness[0.01], Blue}, {Thick, Orange}, {Thick, Purple}}]
       Laguerre approximations of the logarithm function with 10 and 50 terms.            Mathematica code.

We verify Parseval's identity for logarithmic function expansion:
\[ \int_0^{\infty} \left( \ln x \right)^2 e^{-x} {\text d}x = \left( \psi (1) \right)^2 + \sum_{n\ge 1} \frac{1}{n^2} . \]
Mathematica confirms
NIntegrate[(Log[x])^2 *Exp[-x], {x, 0, Infinity}]
1.97811
N[PolyGamma[1]^2 + Sum[1/n^2, {n, 1, Infinity}]]
1.97811
Now we check Parseval's identity for Sonin expansion:
\[ \| \ln x \|^2 = \int_0^{\infty} \left( \ln x \right)^2 x^{\alpha} e^{-x} {\text d}x = \left( \Gamma (\alpha +1)\,\psi (\alpha +1) \right)^2 + \Gamma (\alpha +1)^2 \sum_{n\ge 1} \frac{(n-1)!}{\Gamma (n+\alpha +1)\,n} \approx 0.829627 . \]
Upon taking α = ½, we use Mathematica
NIntegrate[(Log[x])^2 *Exp[-x]*Sqrt[x], {x, 0, Infinity}]
0.829627
N[(Gamma[3/2]*PolyGamma[3/2])^2 + (Gamma[3/2])^2 * Sum[Factorial[n - 1]/Gamma[3/2 + n]/n, {n, 1, Infinity}]]
0.829493
   ■

Example 3: Consider a power function \( f(x) = 4\,x^3 -1 . \) This function has a finite square norm with weight \( e^{-x} : \)

\[ \| 4\,x^3 -1 \|_2^2 = \int_0^{\infty} \left( 4\,x^3 -1 \right)^2 \, e^{-x} \, {\text d} x = 11473. \]
Therefore, this function can be expanded into convergent Laguerre series (which is actually a finite sum):
\[ 4\,x^3 -1 = \sum_{k\ge 0} c_k L_k (x) , \]
where
\[ c_0 = \int_0^{\infty} \left( 4\,x^3 -1 \right) \,e^{-x} \, {\text d} x = 23 , \quad c_1 = -72, \quad c_2 =72, \quad c_3 = -24 . \]
All other coefficients are zeroes, and we get the identity:
\[ 4\,x^3 -1 = 23\,L_0 (x) -72\,L_1 (x) + 72\,L_2 (x) -24\,L_3 (x) , \]
   ■

Example 4: Find the Fourier--Laguerre series expansion of the power function \( f(x) = x^p , \) where parameter p satisfies the condition:

\[ \begin{split} p & > - \frac{1}{2} \left( \alpha + \frac{3}{2} \right) \quad \mbox{if} \quad \alpha > 0, \\ p & > - \left( \frac{\alpha}{2} + \frac{1}{4} \right) \quad \mbox{if} \quad -1 < \alpha \le 0. \end{split} \]

To answer this question, we need to find coefficients ck in the Laguerre expansion:

\[ c_k = \int_0^{\infty} x^p e^{-x} L_k (x)\,{\text d} x = \frac{1}{k!} \, \int_0^{\infty} x^p \,\frac{{\text d}^k}{{\text d} x^k} \left( x^k e^{-x} \right) , \quad k=0,1,2,\ldots . \]
Starting with k = 0, we have
\[ c_0 = \int_0^{\infty} x^p e^{-x} \,{\text d} x = \Gamma (p+1) , \]
where Γ(ν) is the gamma function of Euler. For k > 0, we integrate by parts in the integral
\begin{align*} c_k &= \frac{1}{k!} \, \int_0^{\infty} x^p \,\frac{{\text d}^k}{{\text d} x^k} \left( x^k e^{-x} \right) \,{\text d} x = \left. \frac{1}{k!} \, x^p \,\frac{{\text d}^{k-1}}{{\text d} x^{k-1}} \left( x^k e^{-x} \right) \right\vert_{x=0}^{\infty} - \frac{p}{k!} \, \int_0^{\infty} x^{p-1} \,\frac{{\text d}^{k-1}}{{\text d} x^{k-1}} \left( x^k e^{-x} \right) \,{\text d} x \\ &= -\left. \frac{p}{k!} \, x^{p-1} \,\frac{{\text d}^{k-2}}{{\text d} x^{k-2}} \left( x^k e^{-x} \right) \right\vert_{x=0}^{\infty} + \frac{p(p-1)}{k!} \, \int_0^{\infty} x^{p-2} \,\frac{{\text d}^{k-2}}{{\text d} x^{k-2}} \left( x^k e^{-x} \right) \,{\text d} x \\ &= (-1)^k \,\frac{1}{k!} \, p^{\underline{k}} \, \Gamma (p+1) = \Gamma (p+1) (-1)^k \binom{p}{k} , \end{align*}
where \( p^{\underline{k}} = p(p-1) \cdots (p-k+1) \) is kth falling factorial. Hence, the Fourier-Laguerre series expansion of the power function is given by
\[ x^p = \Gamma (p+1) + \Gamma (p+1) \,\sum_{k\ge 1} \frac{(-1)^k}{k!} \, p^{\underline{k}} \, L_k (x) = \Gamma (p+1) + \Gamma (p+1) \,\sum_{k\ge 1} (-1)^k \binom{p}{k} L_k (x) . \]
We check some first coefficients with Mathematica:
Assuming[p > 0, Integrate[x^p*Exp[-x]*LaguerreL[1, x], {x, 0, Infinity}]]
Gamma[1 + p] - Gamma[2 + p]
Assuming[p > 0, Integrate[x^p*Exp[-x]*LaguerreL[2, x], {x, 0, Infinity}]]
1/2 (-1 + p) p Gamma[1 + p]
Assuming[p > 0, Integrate[x^p*Exp[-x]*LaguerreL[3, x], {x, 0, Infinity}]]
-(1/6) (-2 + p) (-1 + p) p Gamma[1 + p]
      We plot two Laguerre approximations of the square root function with 10 and 50 terms for p = ½.
root[n_] = Gamma[3/2] + Gamma[3/2]* Sum[((-1)^k *Binomial[1/2 , k]*LaguerreL[k, x], {k, 1, n}];
Plot[{Sqrt[x], root[10], root[50]}, {x, 0.5, 5}, PlotStyle -> {{Thickness[0.01], Blue}, {Thick, Orange}, {Thick, Purple}}]
       Laguerre approximations of the square root function with 10 and 50 terms.            Mathematica code.

We check Parseval's identity for the square root:

\[ \| \sqrt{x} \|^2 = \int_0^{\infty} x\,e^{-x} {\text d}x = 1 = \Gamma^2 (3/2) \left[ 1 + \sum_{k\ge 1} \binom{1/2}{k}^2 \right] . \]
Integrate[x*Exp[-x], {x, 0, Infinity}]
1
N[Gamma[3/2]^2*(1 + Sum[Binomial[1/2, k]^2, {k, 1, Infinity}])]
1.

If p is a positive integer, the above series becomes a polynomial of degree p because falling factorial \( p^{\underline{k}} =0 \) for k > p. Also \( \Gamma (p+1) = p! \) for positive integer p. In particular,

\[ x^n = n! \,\sum_{k=0}^n (-1)^k \binom{n+\alpha}{n-k} \, L_k^{(\alpha )} (x) . \]

The binomial coefficients have the parametrization

\[ \binom{n+x}{n} = \sum_{k=0}^n \frac{\alpha^k}{k!} \, L_{n-k}^{(x+k )} (\alpha ) . \]

Example 5: Consider a rational function \( \displaystyle f(x) = \frac{4\,x^3 -1}{x^2 +1} . \) This function has a finite square norm with weight \( e^{-x} : \)

\[ \left\| \frac{4\,x^3 -1}{x^2 +1} \right\|_2^2 = \int_0^{\infty} \left( \frac{4\,x^3 -1}{x^2 +1} \right)^2 \, e^{-x} \, {\text d} x \approx 21.3606. \]
Therefore, this function can be expanded into convergent Laguerre series
\[ \frac{4\,x^3 -1}{x^2 +1} = \sum_{k\ge 0} c_k L_k (x) , \]
where
\begin{align*} c_0 &= \int_0^{\infty} \left( \frac{4\,x^3 -1}{x^2 +1} \right) \, e^{-x} \, {\text d} x \approx 2.00504, \\ c_1 &= \int_0^{\infty} \left( \frac{4\,x^3 -1}{x^2 +1} \right) \, e^{-x} \,L_1 (x) \, {\text d} x \approx -4.13738, \\ c_2 &= \int_0^{\infty} \left( \frac{4\,x^3 -1}{x^2 +1} \right) \, e^{-x} \,L_2 (x) \, {\text d} x \approx 0.217678, \\ c_3 &= \int_0^{\infty} \left( \frac{4\,x^3 -1}{x^2 +1} \right) \, e^{-x} \,L_3 (x) \, {\text d} x \approx 0.260623, \end{align*}
and so on, getting \( c_4 \approx 0.223731, \ c_5 \approx 0.17164. \) Now we build Laguerre approximation with six terms:
c3 = NIntegrate[ LaguerreL[3, x]*(4*x^3 - 1)*Exp[-x]/(x*x + 1), {x, 0, Infinity}]
laguerre = c0 + c1*LaguerreL[1, x] + c2*LaguerreL[2, x] + c3*LaguerreL[3, x] + c4*LaguerreL[4, x] + c5*LaguerreL[5, x]
Plot[{(4*x^3 - 1)/(x*x + 1), laguerre}, {x, 0, 5}, PlotStyle -> {Blue, Orange}]

Example 5B: Let us consider another rational function \( \displaystyle g(x) = \frac{x^2 +1}{4\,x^3 +1} \) that decays at infinity. First, we check whether the given function belongs to the Hilbert space 𝔏²(ℝ+, e−x)

NIntegrate[(x^2 + 1)/(4*x^3 + 1)^2*Exp[-x], {x, 0, Infinity}]
0.424348
Then we calculate first few coefficients in Fourier--Laguerre series
c0 = NIntegrate[(x^2 + 1)/(4*x^3 + 1)*Exp[-x], {x, 0, Infinity}]
c = Table[ NIntegrate[(x^2 + 1)/(4*x^3 + 1)*Exp[-x]*LaguerreL[i, x], {x, 0, Infinity}], {i, 1, 10}];
Then we build a 10-term approximation
rat[x_] = c0+N[Sum[c[[n]]*LaguerreL[n, x], {n, 1, 10}]];
      We plot a Laguerre approximations of the rational function with 10 terms.
Plot[{(x^2 + 1)/(4*x^3 + 1), rat[x]}, {x, 0.4, 5}, PlotStyle -> {{Thickness[0.01], Blue}, {Thick, Orange}}]
       Laguerre approximations of the rational function with 10 terms.            Mathematica code.

Finally, we check validity of Paeseval's identity:
\[ 0.424348 = \| g(x) \|^2 = \int_0^{\infty} \left( \frac{x^2 +1}{4\,x^3 +1} \right)^2 e^{-x} {\text d}x = \sum_{n\ge 0} c_n^2 . \]
N[Sum[c[[n]]^2, {n, 1, 10}]] + c0^2
0.485588
   ■

Example 6: Let us consider the function

\[ F(x) = \left( xt \right)^{-\alpha /2} J_{\alpha} \left( 2 \sqrt{xt} \right) , \qquad a > 0, \quad \alpha > -1, \quad x > 0. \]
Expanding this function into Sonin polynomials, we obtain
\[ \left( xt \right)^{-\alpha /2} J_{\alpha} \left( 2 \sqrt{xt} \right) = e^{-t} \sum_{n\ge 0} \frac{t^n}{\Gamma \left( \alpha + n +1 \right)}\, L_n^{(\alpha )} (x) . \]
So
\[ e^t \left( tx \right)^{-\alpha /2} J_{\alpha} \left( 2\sqrt{xt} \right) = \sum_{n\ge 0} \frac{L_n^{(\alpha )} (x)}{\Gamma (n+ \alpha + 1)} \, t^n . \tag{6.1} \]
In particular,
\[ \left( x \right)^{-\alpha /2} J_{\alpha} \left( 2\sqrt{x} \right) e = \sum_{n\ge 0} \frac{L_n^{(\alpha )} (x)}{\Gamma (n+ \alpha + 1)} . \tag{6.2} \]
For α = 0, we have
\[ J_{0} \left( 2\sqrt{x} \right) e = \sum_{n\ge 0} \frac{L_n (x)}{n!} . \tag{6.3} \]
      We plot a Laguerre approximations of the Bessel function with 10 terms.
bessel10[x_] = Sum[LaguerreL[n, x]/Factorial[n], {n, 0, 10}];
Plot[{BesselJ[0, 2*Sqrt[x]]*Exp[1], bessel10[x]}, {x, 0, 6}, PlotStyle -> {{Thickness[0.01], Blue}, {Thick, Orange}}]
       Laguerre approximations of the Bessel function with 10 terms.            Mathematica code.

Finally, we check validity of Paeseval's identity:
\[ 2.27959 = \| F(x) \|^2 = \int_0^{\infty} \left( J_{0} \left( 2\sqrt{x} \right) \right)^2 e^{2-x} {\text d}x = \sum_{n\ge 0} \frac{1}{\left( n! \right)} . \]
NIntegrate[(BesselJ[0, 2*Sqrt[x]]*Exp[1])^2*Exp[-x], {x, 0, 100}]
2.27959
N[Sum[1/(Factorial[n])^2 , {n, 0, 50}]]
2.27959
   ■

Example 7: First, we expand the cosine function into Laguerre series

\[ \cos x = \sum_{k\ge 0} a_{2k} L_{2k} (x) + \sum_{k\ge 0} a_{2k+1} L_{2k+1} (x) , \]
where coefficients are
\[ a_{n} = \int_0^{\infty} \cos x \, e^{-x} L_n (x)\,{\text d} x , \qquad n=0,1,2,\ldots . \]
First, we perform a computational experiment.
Integrate[Cos[x]*Exp[-x]*LaguerreL[0, x], {x, 0, Infinity}]
1/2
Integrate[Cos[x]*Exp[-x]*LaguerreL[1, x], {x, 0, Infinity}]
1/2
Integrate[Cos[x]*Exp[-x]*LaguerreL[2, x], {x, 0, Infinity}]
1/4
Integrate[Cos[x]*Exp[-x]*LaguerreL[3, x], {x, 0, Infinity}]
0
Integrate[Cos[x]*Exp[-x]*LaguerreL[4, x], {x, 0, Infinity}]
-(1/8)
Integrate[Cos[x]*Exp[-x]*LaguerreL[5, x], {x, 0, Infinity}]
-(1/8)
Integrate[Cos[x]*Exp[-x]*LaguerreL[6, x], {x, 0, Infinity}]
-(1/16)
Integrate[Cos[x]*Exp[-x]*LaguerreL[7, x], {x, 0, Infinity}]
0
Integrate[Cos[x]*Exp[-x]*LaguerreL[8, x], {x, 0, Infinity}]
1/32
Integrate[Cos[x]*Exp[-x]*LaguerreL[9, x], {x, 0, Infinity}]
1/32
Integrate[Cos[x]*Exp[-x]*LaguerreL[10, x], {x, 0, Infinity}]
1/64
Therefore, we conclude that
\[ \cos x = \sum_{k\ge 0} \frac{(-1)^{ \lfloor k/2 \rfloor}}{2^{k+1}}\, L_{2k} (x) + \sum_{k\ge 0} \frac{(-1)^k}{2^{1+2k}}\, L_{4k+1} (x) . \]

 

Similarly, for sine function, we get Laguerre expansion
\[ \sin x = \sum_{k\ge 0} \frac{(-1)^{ \lfloor (k+1)/2 \rfloor}}{2^{k+1}}\, L_{2k} (x) + \sum_{k\ge 0} \frac{(-1)^{k+1}}{2^{2k+2}}\, L_{4k+3} (x) . \]
We plot these approximations
sinL10[x_] = Sum[(-1)^(Floor[(k+1)/2])*LaguerreL[2*k,x]/2^(k+1), {k,0,10}] + Sum[(-1)^(k+1)*LaguerreL[4*k+3,x]/4^(k+1), {k, 0, 10}];
Plot[{Sin[x], sinL10[x]}, {x, 0, 10}, PlotStyle -> Thickness[0.01]]
sinL20[x_] = Sum[(-1)^(Floor[(k + 1)/2])*LaguerreL[2*k, x]/2^(k + 1), {k, 0, 20}] + Sum[(-1)^(k + 1)*LaguerreL[4*k + 3, x]/4^(k + 1), {k, 0, 20}];
Plot[{Sin[x], sinL20[x]}, {x, 0, 15}, PlotStyle -> Thickness[0.01]]
     
       Sine approximation with 10 terms.            ;Sine approximation with 20 terms.

   ■

Example 8: We considered previously in section ii of Tutorial I the Heaviside and Dirac delta functions. The Laguerre expansion of the Dirac delta function is

\[ \delta (x-a) = e^{-(x+a)/2} \,\sum_{k\ge 0} \, L_k (x) \, L_k (a) . \]

Upon choosing a positive number 𝑎, we consider the shifted Heaviside function:

\[ H(t-a) = \begin{cases} 1, & \ \mbox{for} \quad t > a, \\ 1/2, & \ \mbox{for} \quad t = a, \\ 0, & \ \mbox{for} \quad t < a. \end{cases} \]
Let us find a partial sum with N + 1 terms of the corresponding Laguerre expansion:
\[ S_N (t) = \sum_{k=0}^N c_k L_k (t) . \]
First, we calculate coefficients
\[ c_k = \int_a^{\infty} e^{-t} L_k (t) \,{\text d}t , \qquad k=0,1,2,\ldots . \]
With Mathematica, we find a few first terms:
Assuming[a > 0, Integrate[Exp[-x]*LaguerreL[0, x], {x, a, Infinity}]]
E^-a
Assuming[a > 0, Integrate[Exp[-x]*LaguerreL[1, x], {x, a, Infinity}]]
-a E^-a
Assuming[a > 0, Integrate[Exp[-x]*LaguerreL[2, x], {x, a, Infinity}]]
1/2 (-2 + a) a E^-a
Assuming[a > 0, Integrate[Exp[-x]*LaguerreL[3, x], {x, a, Infinity}]]
-(1/6) a (6 - 6 a + a^2) E^-a
Assuming[a > 0, Integrate[Exp[-x]*LaguerreL[4, x], {x, a, Infinity}]]
1/24 a (-24 + (-6 + a)^2 a) E^-a
Assuming[a > 0, Integrate[Exp[-x]*LaguerreL[5, x], {x, a, Infinity}]]
-(1/120) a (120 + a (-240 + a (120 + (-20 + a) a))) E^-a
Let us set 𝑎 = 1, we expand the shifted Heaviside function H(t - 1) into Laguerre series. First, we calculate the coefficients:

Do[ c[k] = Integrate[Exp[-x]*LaguerreL[k, x], {x, 1, Infinity}], {k, 0, 10}]
S10[x_] = Sum[c[k]*LaguerreL[k, x], {k, 0, 10}];
Plot[S10[x], {x, 0, 10}, PlotTheme -> "Web", PlotLabel -> "10 terms approximation"]
Then we repeat calculations with 20 terms and 30 terms:
Do[ c[k] = Integrate[Exp[-x]*LaguerreL[k, x], {x, 1, Infinity}], {k, 11, 20}]
S20[x_] = Sum[c[k]*LaguerreL[k, x], {k, 0, 20}];
Plot[S20[x], {x, 0, 10}, PlotTheme -> "Web", PlotLabel -> "20 terms approximation"]
and
Do[ c[k] = Integrate[Exp[-x]*LaguerreL[k, x], {x, 1, Infinity}], {k, 21, 30}]
S30[x_] = Sum[c[k]*LaguerreL[k, x], {k, 0, 30}];
Plot[S30[x], {x, 0, 10}, PlotTheme -> "Web", PlotLabel -> "30 terms approximation"]
       
       
Laguerre approximation with 20 terms         Laguerre approximation with 35 terms         Laguerre approximation with 30 terms

Since finite sums exhibit Gibbs phenomenon at point x = 1, we apply Cesàro summation.

C10[x_] = Sum[c[k]*LaguerreL[k, x]*(1 -k/11), {k, 0, 10}];
C20[x_] = Sum[c[k]*LaguerreL[k, x]*(1 -k/21), {k, 0, 20}];
C30[x_] = Sum[c[k]*LaguerreL[k, x]*(1 -k/31), {k, 0, 30}];
Plot[C10[x], {x, 0, 10}, PlotTheme -> "Web", PlotLabel -> "10 terms Cesaro approximation"]
Plot[C20[x], {x, 0, 10}, PlotTheme -> "Web", PlotLabel -> "20 terms Cesaro approximation"]
Plot[C30[x], {x, 0, 10}, PlotTheme -> "Web", PlotLabel -> "30 terms Cesaro approximation"]
       
       
Cesàro--Laguerre approximation with 10 terms         Cesàro--Laguerre approximation with 20 terms         Cesàro--Laguerre approximation with 30 terms

   ■

Example 9: Consider piecewise step function

\[ \mbox{sign}(x-a) = \begin{cases} \phantom{-}1 , & \ a < x , \\ -1 , & \ 0 < x < a , \end{cases} \]
where 𝑎 is a positive number. The Fourier coefficients are evaluated according to Eq.\eqref{EqLaguerre.8}
\[ f_k = - \frac{k!}{\Gamma (k + \alpha + 1 )} \int_0^a L_k^{(\alpha )} (x)\,x^{\alpha} e^{-x} {\text d} x + \frac{k!}{\Gamma (k + \alpha + 1 )} \int_a^{\infty} L_k^{(\alpha )} (x)\,x^{\alpha} e^{-x} {\text d} x . \]
The signum function has the expansion:
\[ \mbox{sign}(x-a) = \sum_{k\ge 0} f_k L_k^{(\alpha )} (x) . \]
In case of α = 0, we have
\[ \mbox{sign}(x-a) = \sum_{k\ge 0} f_k L_k (x) , \qquad f_k = -\int_0^a L_k (x) \, e^{-x} {\text d} x + \int_a^{\infty} L_k (x)\, e^{-x} {\text d} x . \]
.

Here is Mathematica code for 𝑎 = 1 and α = ½:

Do[ f[k] = (Integrate[ Exp[-x]*LaguerreL[k, 1/2, x]*x^(1/2), {x, 1, Infinity}] - Integrate[Exp[-x]*LaguerreL[k, 1/2, x]*x^(1/2), {x, 0, 1}])* k! /Gamma[k + 3/2], {k, 0, 10}] S10[x_] = Sum[f[k]*LaguerreL[k,1/2. x], {k, 0, 10}]; Plot[S10[x], {x, 0, 10}, PlotTheme -> "Web", PlotLabel -> "Sonin 10 terms approximation"]
Then we repeat the calculation with 20 terms
Do[ f[k] = (Integrate[ Exp[-x]*LaguerreL[k, 1/2, x]*x^(1/2), {x, 1, Infinity}] - Integrate[Exp[-x]*LaguerreL[k, 1/2, x]*x^(1/2), {x, 0, 1}])* k! /Gamma[k + 3/2], {k, 11, 20}] S20[x_] = Sum[f[k]*LaguerreL[k,1/2. x], {k, 0, 20}]; Plot[S20[x], {x, 0, 10}, PlotTheme -> "Web", PlotLabel -> "Sonin 20 terms approximation"]
and then with 30 terms
Do[ f[k] = (Integrate[ Exp[-x]*LaguerreL[k, 1/2, x]*x^(1/2), {x, 1, Infinity}] - Integrate[Exp[-x]*LaguerreL[k, 1/2, x]*x^(1/2), {x, 0, 1}])* k! /Gamma[k + 3/2], {k, 21, 30}] S30[x_] = Sum[f[k]*LaguerreL[k,1/2. x], {k, 0, 10}]; Plot[S30[x], {x, 0, 10}, PlotTheme -> "Web", PlotLabel -> "Sonin 30 terms approximation"]
       
       
Sonin approximation with 10 terms         Sonin approximation with 20 terms         Sonin approximation with 30 terms

Since finite sums exhibit Gibbs phenomenon at point x = 1, we apply Cesàro summation:

\[ C_N (x) = \sum_{k=0}^N f_k L_n^{(\alpha )} (x) \left( 1 - \frac{k}{N+1} \right) . \]
C10[x_] = Sum[f[k]*LaguerreL[k, 1/2, x]*(1 -k/11), {k, 0, 10}];
C20[x_] = Sum[c[k]*LaguerreL[k, 1/2, x]*(1 -k/21), {k, 0, 20}];
C30[x_] = Sum[c[k]*LaguerreL[k, 1/2, x]*(1 -k/31), {k, 0, 30}];
Plot[C10[x], {x, 0, 10}, PlotTheme -> "Web", PlotLabel -> "Sonin 10 terms Cesaro approximation"]
Plot[C20[x], {x, 0, 10}, PlotTheme -> "Web", PlotLabel -> "Sonin 20 terms Cesaro approximation"]
Plot[C30[x], {x, 0, 10}, PlotTheme -> "Web", PlotLabel -> "Sonin 30 terms Cesaro approximation"]
       
       
Cesàro--Sonin approximation with 10 terms         Cesàro--Sonin approximation with 20 terms         Cesàro--Sonin approximation with 30 terms

Module[{a}, coef = {};
Do[coef = Append[coef, ((0.73575888234) ((LaguerreL[x - 1, 1] -
LaguerreL[x, 1]) - ((0.13533528323) (LaguerreL[x - 1, 2] - LaguerreL[x, 2]))))], {x, 1, 200}]]
coef[[2]] lagsum[m_, x_] :=
Module[{a}, 0.36787944117 + Sum[N[coef[[a]]]*LaguerreL[a, x], {a, 1, m}]]
lg[x_] = Piecewise[{{1, 0 <= x < 1}, {-1, 1 < x <= 2}, {0, x > 2}}];
lagraph[m_] := Plot[{lg[x], lagsum[m, x]}, {x, 0, 2}, PlotRange -> {-1.3, 1.3},
PlotStyle -> {{RGBColor[0, 0, 1], Thickness[0.005]}, {RGBColor[1, 0, 0], Thickness[0.005]}}]
lagraph[50]
lagraph[200]
Module[{a}, coef = {};
Do[coef =
Append[coef, ((0.73575888234) ((LaguerreL[x - 1, 1] - LaguerreL[x, 1]) - ((0.13533528323) (LaguerreL[x - 1, 2] - LaguerreL[x, 2]))))], {x, 1, 200}]]
coef[[2]] lagsum[m_, x_] :=
Module[{a},
0.36787944117 + Sum[N[coef[[a]]]*LaguerreL[a, x], {a, 1, m}]] lg[x_] = Piecewise[{{1, 0 <= x < 1}, {-1, 1 < x <= 2}, {0, x > 2}}];
lagraph[m_] := Plot[{lg[x], lagsum[m, x]}, {x, 0, 2}, PlotRange -> {-1.3, 1.3}, PlotStyle -> {{RGBColor[0, 0, 1], Thickness[0.005]}, {RGBColor[1, 0, 0], Thickness[0.005]}}]
"Plot of piecewise and laguerre with n=50 terms"
lagraph[50]
"Plot of piecewise and laguerre with n=200 terms"
lagraph[200]
"Plot of piecewise-laguerre with n=50 terms"
f[x_] = lg[x] - lagsum[50, x];
Plot[f[x], {x, 0, 2}, PlotRange -> {-1.3, 1.3}, PlotStyle -> {Thick, Red}]
"Plot of piecewise-laguerre with n=200 terms"
f2[x] = lg[x] - lagsum[200, x];
Plot[f2[x], {x, 0, 2}, PlotRange -> {-1.3, 1.3}, PlotStyle -> {Thick, Red}]
"Max of difference with n=50 terms"
Maximize[{f[x], 0 <= x <= 2}, x] "Min of difference with n=50 terms" Minimize[{f[x], 0 <= x <= 2}, x]
"Max of difference with n=200 terms"
Maximize[{f2[x], 0 <= x <= 2}, x]
"Min of difference with n=200 terms"
Minimize[{f2[x], 0 <= x <= 2}, x]
"Integral of (piecewise-laguerre)^2 over interval [0,2] for n=50 terms"
Integrate[(f[x]^2), {x, 0, 2}]
"Integral of (piecewise-laguerre)^2 over interval [0,2] for n=200 terms"
Integrate[(f2[x]^2), {x, 0, 2}]

================================================= to be checked
Consider piecewise step function
\[ f(x) = \begin{cases} \phantom{-}1 , & \ 0 < t < 1 , \\ -1 , & \ 1 < t . \end{cases} \]
Module[{a}, coef = {};
Do[coef = Append[coef, ((0.73575888234) ((LaguerreL[x - 1, 1] -
LaguerreL[x, 1]) - ((0.13533528323) (LaguerreL[x - 1, 2] - LaguerreL[x, 2]))))], {x, 1, 200}]]
coef[[2]] lagsum[m_, x_] :=
Module[{a}, 0.36787944117 + Sum[N[coef[[a]]]*LaguerreL[a, x], {a, 1, m}]]
lg[x_] = Piecewise[{{1, 0 <= x < 1}, {-1, 1 < x <= 2}, {0, x > 2}}];
lagraph[m_] := Plot[{lg[x], lagsum[m, x]}, {x, 0, 2}, PlotRange -> {-1.3, 1.3},
PlotStyle -> {{RGBColor[0, 0, 1], Thickness[0.005]}, {RGBColor[1, 0, 0], Thickness[0.005]}}]
lagraph[50]
lagraph[200]
Module[{a}, coef = {};
Do[coef =
Append[coef, ((0.73575888234) ((LaguerreL[x - 1, 1] - LaguerreL[x, 1]) - ((0.13533528323) (LaguerreL[x - 1, 2] - LaguerreL[x, 2]))))], {x, 1, 200}]]
coef[[2]] lagsum[m_, x_] :=
Module[{a},
0.36787944117 + Sum[N[coef[[a]]]*LaguerreL[a, x], {a, 1, m}]] lg[x_] = Piecewise[{{1, 0 <= x < 1}, {-1, 1 < x <= 2}, {0, x > 2}}];
lagraph[m_] := Plot[{lg[x], lagsum[m, x]}, {x, 0, 2}, PlotRange -> {-1.3, 1.3}, PlotStyle -> {{RGBColor[0, 0, 1], Thickness[0.005]}, {RGBColor[1, 0, 0], Thickness[0.005]}}]
"Plot of piecewise and laguerre with n=50 terms"
lagraph[50]
"Plot of piecewise and laguerre with n=200 terms"
lagraph[200]
"Plot of piecewise-laguerre with n=50 terms"
f[x_] = lg[x] - lagsum[50, x];
Plot[f[x], {x, 0, 2}, PlotRange -> {-1.3, 1.3}, PlotStyle -> {Thick, Red}]
"Plot of piecewise-laguerre with n=200 terms"
f2[x] = lg[x] - lagsum[200, x];
Plot[f2[x], {x, 0, 2}, PlotRange -> {-1.3, 1.3}, PlotStyle -> {Thick, Red}]
"Max of difference with n=50 terms"
Maximize[{f[x], 0 <= x <= 2}, x] "Min of difference with n=50 terms" Minimize[{f[x], 0 <= x <= 2}, x]
"Max of difference with n=200 terms"
Maximize[{f2[x], 0 <= x <= 2}, x]
"Min of difference with n=200 terms"
Minimize[{f2[x], 0 <= x <= 2}, x]
"Integral of (piecewise-laguerre)^2 over interval [0,2] for n=50 terms"
Integrate[(f[x]^2), {x, 0, 2}]
"Integral of (piecewise-laguerre)^2 over interval [0,2] for n=200 terms"
Integrate[(f2[x]^2), {x, 0, 2}]
   ■

Example 10: Let us consider a characteristic function of the interval [𝑎, b]

\[ \chi_{[a,b]} (x) = \begin{cases} 1, & \ \mbox{ when} \quad a \le x \le b , \\ 0, & \ \mbox{ otherwise}, \end{cases} \]
where 0 ≤ 𝑎 < b. Expanding this function into Fourier--Laguerre series, we get
\[ \chi_{[a,b]} (x) = \sum_{n\ge 0} c_n L_n (x) , \qquad c_n = \int_a^b L_n (x)\,e^{-x} {\text d}x . \]
   ■

 

Connection to Hermite expansion


Suppose we know a Hermite expansion for some function
\[ \phi (x) = \sum_{n\ge 0} c_{2n} H_{2n} (x) . \]
Using the formula
\begin{equation} \label{EqLaguerre.10} L_n^{(\alpha )} (x) = \frac{(-1)^n \Gamma \left( n+ \alpha + 1 \right)}{(2n)! \,\sqrt{\pi}\,\Gamma \left( n + \frac{1}{2} \right)} \int_{-1}^1 \left( 1 - t^2 \right)^{\alpha - 1/2} H_{2n} \left( t\sqrt{x} \right) {\text d}t , \end{equation}
we get another function that we expand into Sonin series
\[ f(x) = \int_{-1}^1 \left( 1 - t^2 \right)^{\alpha - 1/2} \phi \left( t\sqrt{x} \right) {\text d}t = \sum_{n\ge 0} a_n L_n^{(\alpha )} (x) . \]
This expansion is valid for α > −½ and its coefficients are
\[ a_n = (-1)^n \frac{\sqrt{\pi} \Gamma \left( n + \frac{1}{2} \right) (2n)!}{\Gamma \left( n+ \alpha + 1 \right)}\, c_{2n} , \qquad n=0,1,2,\ldots . \]