ODE: Bessel Functions and Applications

Bessel functions are not mere mathematical curiosities; they are indispensable tools for engineers and physicists. When you solve partial differential equations in systems with cylindrical or circular symmetry—think of heat spreading through a metal rod, vibrations on a drumhead, or electromagnetic waves in a waveguide—these special functions emerge as the natural language of the solutions. Mastering them unlocks your ability to model and analyze a vast array of real-world phenomena.

The Bessel Equation and Its Fundamental Solutions

The journey begins with a specific second-order linear ordinary differential equation. The Bessel equation of order n is given by: $$x^2{y}^{\prime \prime }+x{y}^{\prime }+(x^2-n^2)y=0$$ where $n$ is a constant (usually a non-negative real number or integer). This equation arises directly when applying the separation of variables technique to the heat equation or wave equation in cylindrical coordinates $(r,\theta ,z)$. The variable $x$ in the Bessel equation often corresponds to a scaled radial coordinate.

This equation has two linearly independent solutions. The most common and well-behaved is the Bessel function of the first kind, denoted $J_n(x)$. For integer order $n$, it can be defined by a power series: $$J_n(x)=\sum _{m=0}^{\infty }\frac{(-1{)}^m}{m!\Gamma (m+n+1)}{\left(\frac{x}{2}\right)}^{2m+n}$$ where $\Gamma$ is the Gamma function. For $n=0$ and $n=1$, these functions are particularly important. $J_0(x)$ starts at 1 and oscillates with decreasing amplitude, resembling a damped cosine wave. $J_1(x)$ starts at 0, rises to a peak, and then exhibits similar damped oscillations. Graphically, all $J_n(x)$ functions are oscillatory and bounded, crossing the x-axis at an infinite number of points—their zeros.

However, $J_n(x)$ and $J_{-n}(x)$ are not linearly independent for integer $n$. To form a complete basis for solutions, we need a second, linearly independent function. This is the Bessel function of the second kind (or Neumann function), denoted $Y_n(x)$. It is defined as a linear combination of $J_n(x)$ and $J_{-n}(x)$ in a specific way that ensures independence. Crucially, $Y_n(x)$ is singular (goes to $-\infty$) as $x\to 0$. This singularity is physically meaningful: it often corresponds to an infinite or unrealistic condition at the center (r=0) of a cylinder, helping us reject this solution in problems where the center is included.

Key Properties and Recurrence Relations

Understanding the behavior of Bessel functions is aided by their properties and graphs. As mentioned, $J_n(x)$ are bounded and oscillatory, with zeros that are not periodically spaced. The amplitude of oscillation decays roughly as $1/\sqrt{x}$. The functions $Y_n(x)$ share the oscillatory nature for large $x$ but blow up at the origin. These graphs inform physical interpretations: the oscillatory nature corresponds to standing waves or modal shapes in a cylinder, while the boundedness or singularity dictates which solution is physically admissible.

Powerful tools for manipulation and integration are the recurrence relations. These are formulas that relate Bessel functions of different orders. Two of the most useful are: $$\frac{d}{dx}[x^n{J}_n(x)]=x^n{J}_{n-1}(x)$$ $$\frac{d}{dx}[x^{-n}{J}_n(x)]=-x^{-n}{J}_{n+1}(x)$$ From these, you can derive other essential relations, such as: $$J_{n-1}(x)+J_{n+1}(x)=\frac{2n}{x}{J}_n(x)$$ $$J_{n-1}(x)-J_{n+1}(x)=2J_n^{\prime }(x)$$ Identical relations hold for $Y_n(x)$. These relations allow you to compute higher-order Bessel functions from known lower-order ones, evaluate derivatives, and solve integrals that appear in engineering applications.

Orthogonality and Series Expansions

A pivotal property for solving boundary value problems is orthogonality. For a fixed order $n$, the set of functions $\{J_n({\alpha }_{n,m}r/a)\}$ is orthogonal on the interval $[0,a]$ with respect to the weight function $r$. Here, ${\alpha }_{n,m}$ represents the $m$-th positive zero of $J_n(x)$. The orthogonality condition is: $${\int }_0^{a}r{J}_n\left(\frac{{\alpha }_{n,i}r}{a}\right)J_n\left(\frac{{\alpha }_{n,j}r}{a}\right)dr=0\text{for }i\ne j.$$ When $i=j$, the integral yields a specific positive value, which serves as a normalization constant.

This property is the cylindrical analogue of the orthogonality of sine functions used in Fourier series. It allows you to expand any reasonably well-behaved function $f(r)$ defined on $0\le r\le a$ as a Fourier-Bessel series: $$f(r)=\sum _{m=1}^{\infty }{c}_m{J}_n\left(\frac{{\alpha }_{n,m}r}{a}\right),$$ where the coefficients $c_m$ are calculated using the weighted inner product provided by the orthogonality relation. This series expansion is the final key you need to construct complete solutions to physical problems.

Application to Heat Conduction in a Cylinder

Consider the classic problem of heat conduction in cylinders. Imagine a long solid cylinder of radius $a$ with a prescribed initial temperature distribution $f(r)$ and whose surface is held at zero temperature for time $t>0$. Assuming no $\theta$ or $z$ dependence, the temperature $u(r,t)$ satisfies the heat equation in cylindrical coordinates. Separation of variables leads to a radial equation that is precisely Bessel's equation of order zero. The boundary condition $u(a,t)=0$ forces the solution to be $J_0(\lambda r)$, with the requirement $J_0(\lambda a)=0$. Hence, the eigenvalues are ${\lambda }_m={\alpha }_{0,m}/a$, where ${\alpha }_{0,m}$ are the zeros of $J_0$.

The final solution is a series expansion: $$u(r,t)=\sum _{m=1}^{\infty }{c}_m{J}_0\left(\frac{{\alpha }_{0,m}r}{a}\right)e^{-k({\alpha }_{0,m}/a{)}^{2}t},$$ where the coefficients $c_m$ are found by expanding the initial condition $f(r)$ in a Fourier-Bessel series using the orthogonality of $J_0$. The Bessel function describes the radial temperature profile of each mode, and the exponential term describes its decay over time.

Application to Vibrating Circular Membranes

The problem of a vibrating circular membrane (like a drumhead) clamped at its edge $r=a$ follows a similar pattern but uses the wave equation. For a mode vibrating with a fixed frequency, separation of variables again produces Bessel's equation for the radial part, with the order $n$ arising from the angular dependence (e.g., $\cos (n\theta )$ or $\sin (n\theta )$). The condition that the displacement is finite at $r=0$ rejects the $Y_n$ solution. The fixed-edge boundary condition requires $J_n(\lambda a)=0$.

Thus, the eigenvalues are ${\lambda }_{n,m}={\alpha }_{n,m}/a$, and the natural frequencies of vibration are proportional to these ${\alpha }_{n,m}$. The spatial mode shape (eigenfunction) is: $${\varphi }_{n,m}(r,\theta )=J_n\left(\frac{{\alpha }_{n,m}r}{a}\right)\{\begin{array}{l}\cos (n\theta )\\ \sin (n\theta )\end{array}.$$ The indices $n$ and $m$ describe the number of nodal diameters and nodal circles, respectively. The complete motion of the membrane is a superposition of all such modes. This direct application shows how the zeros of Bessel functions determine the fundamental and harmonic tones of a drum.

Common Pitfalls

1. Misapplying the Second Kind Solution: A frequent error is including $Y_n(x)$ in a physical solution where the domain includes the point $x=0$ (the center of the cylinder). Since $Y_n(x)$ is singular at the origin, its presence would imply an infinite displacement, temperature, or potential at the center, which is usually non-physical. Always check your domain: use $Y_n$ only for problems involving annular regions (like a pipe) where $r=0$ is excluded.

1. Incorrect Boundary Condition Application: When applying a boundary condition like $u(a,t)=0$, it is not enough to set $J_n(\lambda a)=0$. You must remember that this equation has an infinite number of solutions: the zeros ${\alpha }_{n,m}$. The correct step is to define your eigenvalues as ${\lambda }_m={\alpha }_{n,m}/a$ for $m=1,2,3,...$. Forgetting to include all these eigenvalues means you miss most of the solution modes in your series expansion.

1. Confusing Recurrence Relations: The recurrence relations for derivatives, such as $d/dx[x^n{J}_n(x)]$, are often misremembered. A reliable strategy is to derive the needed relation from the two fundamental ones listed earlier, or to consistently use a known reference. Misapplying these can lead to significant errors in simplifying integrals or solving ODEs that contain Bessel functions.

1. Overlooking the Weight in Orthogonality: When computing Fourier-Bessel series coefficients, you must use the correct inner product with the weight function $r$. The orthogonality integral is ${\int }_0^{a}r{J}_n({\lambda }_{i}r)J_n({\lambda }_{j}r)dr$. Omitting the factor of $r$ (the weight inherent to cylindrical coordinates) will yield incorrect coefficients and an invalid series expansion.

Summary

• Bessel functions, $J_n(x)$ and $Y_n(x)$, are the standard solutions to the Bessel equation of order $n$, which arises from separation of variables in cylindrical geometries.
• $J_n(x)$ is finite at $x=0$ and oscillatory, while $Y_n(x)$ is singular at the origin; the choice between them depends on whether the physical domain includes the centerline.
• Recurrence relations provide crucial tools for differentiating, integrating, and relating Bessel functions of different orders.
• Bessel functions of a fixed order satisfy an orthogonality condition with weight $r$, enabling the Fourier-Bessel series expansion of functions, analogous to Fourier series.
• These functions are directly applied in solving engineering problems such as transient heat conduction in solid cylinders and the modal vibration analysis of circular membranes, where boundary conditions determine eigenvalues based on the zeros of Bessel functions.