Bessel function

For $w^{\prime \prime }+[1-A/z^2]w=0$, with $A=n^2-1/4$, substitution gives $s=\pm i$, exponent zero and $c_{k+1}=[k(k+1)-A]c_k/[2s(k+1)]$. Thus the first correction is $c_1=sA{c}_0/2$. For integer $n$ no coefficient truncation occurs and coefficient ratios grow as $k/(2s)$; this is an asymptotic expansion, not a convergent series at a finite $z$. The leading Liouville–Green approximation $q^{-1/4}{e}^{\pm i\int \sqrt{q}}$ has the same $1/z$ correction.

The two Hankel functions $H_{\nu }^{(1)}=J_{\nu }+i{Y}_{\nu }$ and $H_{\nu }^{(2)}=J_{\nu }-i{Y}_{\nu }$ solve the Bessel differential equation. Their large-argument forms distinguish outgoing and incoming oscillations and have sector-dependent asymptotic expansions.

For fixed order, the standard Hankel function expansions in $e^{\pm iz}{z}^{-1/2}$ are valid in different sectors on a chosen argument sheet. Analytic continuation can mix exponentials; the sectors must be specified rather than treating a formal inverse-power series as global.

For noninteger order, $Y_{\nu }=(\cos (\pi \nu )J_{\nu }-J_{-\nu })/\sin (\pi \nu )$; the integer-order function is defined by a limit. It gives a second independent solution of the Bessel differential equation. At order zero, $Y_0(z)\sim (2/\pi )[\log (z/2)+\gamma ]$ near positive zero and $Y_0(z)\sim \sqrt{2/(\pi z)}\sin (z-\pi /4)$ at large positive argument, where $\gamma$ is the Euler--Mascheroni constant.

The spherical second-kind function is $n_{\ell }(z)=\sqrt{\pi /(2z)}{Y}_{\ell +1/2}(z)$. Together with the first-kind Spherical Bessel function, it spans the radial free-wave solutions; $n_{\ell }(z)\sim -\cos (z-\ell \pi /2)/z$ at large positive $z$.

The spherical Bessel functions solve the radial wave equation in three dimensions. In particular,The apparent singularity at the origin is removable, as follows from the power series of the Bessel function of the first kind.

A scalar field solving the Helmholtz equation inside a sphere of radius $R$ and Laplace's equation outside has a regular interior mode $b{j}_{\ell }(\sqrt{s}r)P_{\ell }(\cos \theta )$ and a decaying exterior mode $a{r}^{-\ell -1}{P}_{\ell }(\cos \theta )$. Continuity of the field and its radial derivative gives, for $x=\sqrt{s}R>0$,The Spherical Bessel function recurrence turns this into $j_{\ell -1}(x)=0$, not $j_{\ell }(x)=0$. This elimination does not require division by a possibly vanishing boundary value.

For $\ell =0$, $j_0(x)=\sin x/x$ and $j_{-1}(x)=\cos x/x$, so the first positive root is $\pi /2$. For $\ell =1$, the matching condition is $j_0(x)=0$, whose first positive root is $\pi$. The distinction matters when a scalar model permits the monopole sector but a physical multipole model excludes it.

To compare angular sectors, put $u=r{j}_{\ell }(\sqrt{s}r)$. Its equation and boundary conditions are $-u^{\prime \prime }+\ell (\ell +1)u/r^2=su$, $u(0)=0$ and $u^{\prime }(R)+\ell u(R)/R=0$. Its Rayleigh quotient isBoth added terms are nonnegative and increase with $\ell$. Thus the lowest scalar eigenvalue is ${\pi }^2/(4R^2)$ in the $\ell =0$ sector; if only $\ell \ge 1$ is admitted, the lowest is ${\pi }^2/R^2$ in the $\ell =1$ sector. Zeros at $x=0$ caused by the vanishing normalization of $j_{\ell }$ are not additional eigenvalues.

For $x^2{y}^{\prime \prime }+2x{y}^{\prime }+[x^2-\ell (\ell +1)]y=0$ with integer $\ell \ge 0$, the Frobenius method gives $\sigma =\ell$ or $-\ell -1$. The general coefficient equation is $[(j+\sigma )(j+\sigma +1)-\ell (\ell +1)]a_j=-a_{j-2}$, taking negative-index coefficients as zero. The regular branch has $a_1=0$ and $a_j=-a_{j-2}/[j(j+2\ell +1)]$. The singular branch has $a_j=-a_{j-2}/[j(j-2\ell -1)]$ except at $j=2\ell +1$. That resonant equation is compatible because its earlier odd coefficient vanishes; the newly free coefficient adds a multiple of the regular solution. Setting it to zero gives a pure even series times $x^{-\ell -1}$, independent of the regular branch, with no logarithmic term. For $\ell =0$ the resonance occurs already at $j=1$.

The normalization $a_0=2^{\ell }\ell !/(2\ell +1)!$ on the regular branch gives the Spherical Bessel functionIn particular $j_0(x)=\sin x/x$, with removable value one at zero.

For integer $\ell \ge 1$, put $I_{\ell }={\int }_0^{\infty }{j}_{\ell }(x{)}^{2}dx/x$ and $J_{\ell }={\int }_0^{\infty }{j}_{\ell }(x)j_{\ell -1}(x)dx$. Integrating $j_{\ell }^{\prime }=j_{\ell -1}-(\ell +1)j_{\ell }/x$ gives $J_{\ell }=(\ell +1)I_{\ell }$. For $\ell >1$, integrating $j_{\ell -1}^{\prime }=(\ell -1)j_{\ell -1}/x-j_{\ell }$ also gives $J_{\ell }=(\ell -1)I_{\ell -1}$. For $\ell =1$, the boundary values $j_0(0)=1$ and $j_0(\infty )=0$ give $J_1=1/2$, hence $I_1=1/4$. The recurrence proves the displayed identity. The integral without the square is different; for example ${\int }_0^{\infty }{j}_1(x)dx/x=\pi /4$.

These combinations of Spherical Bessel functions solve the radial Helmholtz equation. For time dependence $e^{-i\omega t}$, $h_n^{(1)}(kr)$ is outgoing and $h_n^{(2)}(kr)$ incoming. At fixed order and large argument, $h_n^{(1)}(t)\sim (-i{)}^{n+1}{e}^{it}/t$. Their recurrence is $h_{n+1}=(2n+1)h_n/t-h_{n-1}$.

For fixed $t>0$ and large integer order, $y_n(t)\sim -(2n-1)!!/t^{n+1}$ while $j_n(t)$ is negligible. Stirling's formula gives $(2n-1)!!=(2n)!/(2^{n}n!)\sim \sqrt{2}(2n/e{)}^n$, proving the displayed asymptotic. In particular its magnitude is bounded above and below by constant multiples of $(2n/(et){)}^n$. This two-sided comparison is stronger than a one-sided big-O estimate and is what converts a finite spherical trace norm into a weighted condition on far-field coefficients.

For integer $\ell \ge 0$ and $r,R>0$, $r_{<}=\min (r,R)$ and $r_{>}=\max (r,R)$. A proof compares the multipole expansion of $1/|\mathbf{r}-\mathbf{R}|$ with its Fourier transform of inverse distance in three dimensions. Applying the Rayleigh plane-wave expansion to the Fourier representation gives coefficient $8{\int }_0^{\infty }{j}_{\ell }(kr)j_{\ell }(kR)dk$ multiplying $Y_{\ell m}(\hat{r})Y_{\ell m}^{\ast }(\hat{R})$; the Coulomb multipole coefficient is $4\pi r_{<}^{\ell }/[(2\ell +1)r_{>}^{\ell +1}]$. Equating the coefficients proves the identity. The powers also match continuously at $r=R$.

The Bessel function of the first kind $J_{\nu }$ is the solution of the Bessel differential equation selected by its regular power-series behavior at the origin. For $\mathrm{Re}\nu >-1/2$, it has the Poisson integral representation

For integer $m\ge 0$, $J_{-m}(z)=(-1{)}^m{J}_m(z)$. Thus negative integer orders remain regular at zero. The familiar $J_m(z)\sim z^m/(2^{m}m!)$ formula is used for nonnegative order; a negative factorial is not a valid continuation.

For $x>0$, the Bessel differential equation of order $1/2$ has independent solutions $x^{-1/2}\sin x$ and $x^{-1/2}\cos x$. Substituting $y=x^{-1/2}u$ reduces it to $u^{\prime \prime }+u=0$. The normalization of the Bessel function of the first kind gives the displayed formulas for orders $1/2$ and $-1/2$. The Frobenius method obtains the same pair from exponents $1/2$ and $-1/2$: their even-coefficient series reduce to the sine and cosine power series. In the negative-exponent recurrence, the free coefficient at the next exponent can add the positive-exponent solution, an example of an undetermined coefficient at Frobenius resonance.

With the Fourier transform convention $\hat{f}(\omega )={\int }_{\mathbb{R}}f(t)e^{-i\omega t}dt$, the regular order-zero Bessel function of the first kind has the tempered distribution transformThe density is integrable despite its endpoint singularities. Fourier inversion gives the equivalent integral representationThe second formula follows by $\omega =\cos \theta$ and cancellation of the odd imaginary part. It is the standard real integral representation of $J_0$.^[ref]

Indeed, transforming the order-zero Bessel differential equation $t{y}^{\prime \prime }+y^{\prime }+ty=0$ gives $(1-{\omega }^2)Y^{\prime }-\omega Y=0$. Inside $(-1,1)$, its solutions are $C(1-{\omega }^2{)}^{-1/2}$. The compactly supported choice also solves the equation as a distribution: rewrite it as $[(1-{\omega }^2)Y{]}^{\prime }+\omega Y=0$, and observe that $(1-{\omega }^2)Y=C\sqrt{1-{\omega }^2}{1}_{(-1,1)}$ is continuous and zero at the two endpoints. Its distributional derivative has no endpoint Dirac delta terms. The inverse is regular at $t=0$, and $y(0)=1$ fixes $C=2$.

The transformed equation alone also allows separate constants on $\omega <-1$ and $\omega >1$. Selecting the compactly supported density amounts to selecting the solution regular at the origin, rather than every possible singular solution of the Bessel differential equation.

A Fourier series expansion of a sinusoidally modulated complex wave phase in Bessel functions of the first kind.

For real $z$, the probabilities $J_m(z{)}^2$ over integer angular-momentum modes satisfy ${\sum }_m{m}^2{J}_m(z{)}^2=z^2/2$. Differentiate the unit-modulus function $e^{iz\cos \theta }$ and use Parseval to obtain the identity; its zero first moment follows from the symmetry of opposite orders.

The regular power series has $J_0(\rho )=1-{\rho }^2/4+{\rho }^4/64-\cdots$. Its quartic truncation has a double zero at $\rho =\sqrt{8}$, giving a rough disc-well energy estimate $4{\hslash }^2/(\mu a^2)$. A truncation zero is not an exact Bessel zero or a rigorous eigenvalue bound. The double-root behavior particularly warns against treating a low-order local expansion as accurate far from its center.

For fixed $0<\alpha <\pi /2$, the Bessel function of the first kind has an interior nondegenerate stationary point in its cosine integral. The stationary phase method gives the displayed oscillatory leading term and an additive $O(n^{-3/2})$ remainder. The approximation loses uniformity near $\alpha =0$, where the stationary point reaches the endpoint and the curvature vanishes. For negative integer-order arguments, parity supplies a factor $(-1{)}^n$.

For fixed $\nu$, putThenas $|z|\to \infty$, uniformly for $|\arg z|\le \pi -\delta$. In particular,

The Bessel differential equation of order $\nu$ isFor noninteger $\nu$, $J_{\nu }$ and $J_{-\nu }$ form a linearly independent pair of solutions.

Choose the square-root branch asymptotic to $\zeta$ outside the cut $[-i,i]$. The Laurent series of its reciprocal gives the displayed integral by termwise residue calculation. Adding any entire function to the weight leaves the integral unchanged, since its product with $e^{z\zeta }$ is entire. Thus a nonzero solution represented by a contour integral does not by itself determine the weight uniquely; an inference from the differential equation to an equation for the weight needs an injective representation or an explicit construction assumption.

For the Bessel differential equation $x^2{y}^{\prime \prime }+x{y}^{\prime }+(x^2-{\nu }^2)y=0$, substitute $y=x^s{\sum }_{n\ge 0}{a}_n{x}^n$. The indicial equation is $s^2-{\nu }^2=0$, and coefficients obey $[(n+s{)}^2-{\nu }^2]a_n=-a_{n-2}$. For $s=\nu \ge 0$, $a_1=0$ and all odd coefficients vanish. Iteration gives $a_{2m}=(-1{)}^m{a}_0\Gamma (\nu +1)/[4^{m}m!\Gamma (m+\nu +1)]$. Choosing $a_0=1/[2^{\nu }\Gamma (\nu +1)]$ gives the Bessel function of the first kind $J_{\nu }$.

For noninteger $\nu$, choosing the even-power series with $s=-\nu$ replaces $m+\nu$ by $m-\nu$ and yields $J_{-\nu }$ with its standard normalization. These solutions are independent. For integer $\nu \ge 0$, the first solution has coefficients $(-1{)}^m/[2^{2m+\nu }m!(m+\nu )!]$ and defines an entire function: the term ratio tends to zero at every finite complex argument. For noninteger order, the even analytic series converges everywhere, while the prefactor $x^{\pm \nu }$ requires a branch off the positive real axis.

For fixed $\beta$, move the order term of the Bessel equation to the right: $x^2{y}^{\prime \prime }+x{y}^{\prime }+{\beta }^2{x}^{2}y=n^{2}y$. The integrating factor for its leading differential expression is $1/x$, giving the displayed Sturm-Liouville theory form with coefficient $p=x$, potential $q={\beta }^{2}x$, weight $w=1/x$ and eigenvalue $n^2$. The origin is a singular endpoint. The integrating factor must be normalized from a positive reference point; its defining logarithmic integral cannot literally start at zero. Boundary conditions are needed before this differential identity becomes a specified self-adjoint spectral problem.

The modified Bessel equation of order zero isIts solution regular at zero and normalized by $I_0(0)=1$ is the modified Bessel function $I_0$.

For fixed real $\nu$ as $z\to +\infty$, the Modified Bessel function of the first kind and Modified Bessel function of the second kind have asymptotic expansionsThe expansion and its error bounds establish actual growing and decaying solutions, beyond formal matching of a differential equation. These formulas are documented in NIST DLMF, equations 10.40.1–2.

The solution of the Modified Bessel differential equation with power series $I_{\nu }(z)={\sum }_{j\ge 0}(z/2{)}^{2j+\nu }/[j!\Gamma (j+\nu +1)]$, interpreted through the reciprocal gamma function at its zeros. For fixed real order it grows exponentially along the positive real axis.

The differential equation defining modified Bessel functions of order $\nu$. For real fixed $\nu$ and positive $z$, its independent standard solutions are $I_{\nu }(z)$ and $K_{\nu }(z)$.

The modified Bessel function of the second kind is the solution $K_{\nu }(x)$ of the modified Bessel differential equation that decays exponentially as $x\to +\infty$.

The integral representation of the modified Bessel function of the second kind gives $K_0(x)={\int }_0^{\infty }{e}^{-x\cosh t}dt$ for $x>0$. Taking the Fourier transform of the even continuation and setting $u=\sinh t$ gives $2{\int }_0^{\infty }(1+k^2+u^2{)}^{-1}du=\pi /\sqrt{1+k^2}$. The large-wavenumber terms $\pi /|k|-\pi /(2|k{|}^3)$ are also obtained from the local logarithmic singularity $-\log |x|-(x^2/4)\log |x|$ using the Fourier transform of the absolute logarithm.

For real $x>0$ and $\nu$, one useful representation isIt turns large-order asymptotics into a Laplace integral with a moving maximum.

For fixed $x>0$,as $\nu \to +\infty$. Applying Laplace's method to the integral representation of the modified Bessel function of the second kind proves the formula.

Laplace's integral method seeks $y(z)={\int }_C{e}^{zt}f(t)dt$. Substitution into the differential equation and integration by parts transfer multiplication by $z$ into differentiation with respect to $t$, producing a first-order equation for $f$ and endpoint conditions for $C$.

For $x>0$, substituting $y(x)={\int }_{\gamma }{e}^{xt}f(t)dt$ into $x{y}^{(3)}+2y=0$ and integrating by parts givesThe amplitude equation $(t^{3}f{)}^{\prime }=2f$ has solution $f(t)=C{t}^{-3}{e}^{-1/t^2}$. On $\gamma =(-\infty ,0)$ both endpoint terms vanish, and hence

For the Laguerre equationthe ansatz $y={\int }_{\gamma }{e}^{zt}f(t)dt$ givesThe contour must make the endpoint contributionvanish.

A finite contour based at an integrable branch point, or a closed contour around an integer-order pole, gives a solution analytic at $z=0$. It is therefore a constant multiple of the unique Frobenius power-series solution there.

For $\lambda =N\ge 0$, the amplitude $(t-1{)}^N/t^{N+1}$ has a pole at zero whose contour residue is a degree-$N$ Laguerre polynomial in $z$. For $\lambda =-m<0$, the amplitude $t^{m-1}/(t-1{)}^m$ has a pole at one whose residue is $e^z$ times a polynomial of degree $m-1$.

The normalized order-zero function has the representation