ODE: Fourier Sine and Cosine Transforms

When solving partial differential equations (PDEs) on infinite domains, the classic Fourier transform is a powerful tool. But what happens when your physical domain is a semi-infinite interval, like a rod extending from $x=0$ to $x=\infty$? The standard Fourier transform struggles to incorporate boundary conditions at the finite end. This is where the specialized Fourier sine and cosine transforms become indispensable. They are integral transforms specifically designed to solve boundary value problems on half-lines by naturally embedding Dirichlet (fixed value) or Neumann (fixed derivative) conditions at the origin, streamlining the solution process for heat conduction, wave propagation, and other engineering phenomena.

Defining the Sine and Cosine Transforms

The Fourier sine and cosine transforms are derived from the Fourier integral theorem for an odd or even extension of a function defined on $[0,\infty )$. They focus solely on the sine or cosine component of the full Fourier kernel.

The Fourier sine transform of a function $f(x)$, $x>0$, is defined as: $${\mathcal{F}}_s\{f(x)\}=F_s(\omega )=\sqrt{\frac{2}{\pi }}{\int }_0^{\infty }f(x)\sin (\omega x)dx.$$ This transform is ideal for problems where the boundary condition at $x=0$ is of Dirichlet type (e.g., $u(0,t)=0$). Intuitively, since $\sin (0)=0$, any solution reconstructed from a sine transform will automatically satisfy a homogeneous Dirichlet condition.

Conversely, the Fourier cosine transform is defined as: $${\mathcal{F}}_c\{f(x)\}=F_c(\omega )=\sqrt{\frac{2}{\pi }}{\int }_0^{\infty }f(x)\cos (\omega x)dx.$$ This transform naturally handles Neumann boundary conditions at the origin (e.g., $u_x(0,t)=0$). The derivative of $\cos (\omega x)$ with respect to $x$ is $-\omega \sin (\omega x)$, which is zero at $x=0$.

The factor $\sqrt{2/\pi }$ is a normalization constant that ensures the transforms, together with their inverses, form a symmetric pair, much like the standard Fourier transform.

Inverse Transforms and Operational Properties

To recover the original function from its transformed state, we apply the corresponding inverse transform. The inverses are remarkably similar to the forward definitions, which is a key property that simplifies calculations.

The inverse Fourier sine transform is given by: $$f(x)=\sqrt{\frac{2}{\pi }}{\int }_0^{\infty }{F}_s(\omega )\sin (\omega x)d\omega .$$

The inverse Fourier cosine transform is: $$f(x)=\sqrt{\frac{2}{\pi }}{\int }_0^{\infty }{F}_c(\omega )\cos (\omega x)d\omega .$$

Their true power for solving differential equations lies in their operational properties—how they handle derivatives. These properties convert derivatives with respect to $x$ into algebraic multiplication in the $\omega$-domain, subject to the boundary conditions at $x=0$.

For the Fourier sine transform:

• First derivative: ${\mathcal{F}}_s\{f^{\prime }(x)\}=-\omega F_c(\omega )$.
• Second derivative: ${\mathcal{F}}_s\{f^{\prime \prime }(x)\}=-{\omega }^2{F}_s(\omega )+\sqrt{\frac{2}{\pi }}\omega f(0)$.

For the Fourier cosine transform:

• First derivative: ${\mathcal{F}}_c\{f^{\prime }(x)\}=\omega F_s(\omega )-\sqrt{\frac{2}{\pi }}f(0)$.
• Second derivative: ${\mathcal{F}}_c\{f^{\prime \prime }(x)\}=-{\omega }^2{F}_c(\omega )-\sqrt{\frac{2}{\pi }}{f}^{\prime }(0)$.

Notice how the boundary values $f(0)$ and $f^{\prime }(0)$ appear explicitly. This is the mechanism by which the boundary condition is incorporated directly into the transformed equation.

Solving PDEs on the Half-Line

Let's see these transforms in action by solving a classic heat conduction problem on a semi-infinite rod. Consider the PDE: $$\frac{\partial u}{\partial t}=k\frac{{\partial }^{2}u}{\partial x^2},x>0,t>0,$$ with boundary condition $u(0,t)=0$ and initial condition $u(x,0)=f(x)$.

Step 1: Choosing the Transform. The Dirichlet condition $u(0,t)=0$ makes the Fourier sine transform the natural choice, as it will force this condition automatically. We apply ${\mathcal{F}}_s$ to both sides of the PDE with respect to $x$, denoting $U_s(\omega ,t)={\mathcal{F}}_s\{u(x,t)\}$.

Step 2: Transforming the PDE. Using the operational property for the second derivative: $${\mathcal{F}}_s\left\{\frac{{\partial }^{2}u}{\partial x^2}\right\}=-{\omega }^2{U}_s(\omega ,t)+\sqrt{\frac{2}{\pi }}\omega u(0,t).$$ Because our boundary condition is $u(0,t)=0$, the second term vanishes. The time derivative transforms simply: ${\mathcal{F}}_s\{u_t\}=\frac{d}{dt}{U}_s(\omega ,t)$. Thus, the PDE becomes an ordinary differential equation in the transform domain: $$\frac{d{U}_s}{dt}=k(-{\omega }^2{U}_s).$$

Step 3: Solving in the Transform Domain. This is a simple first-order ODE: $\frac{d{U}_s}{dt}=-k{\omega }^2{U}_s$. Its solution is: $$U_s(\omega ,t)=C(\omega )e^{-k{\omega }^{2}t}.$$ To find $C(\omega )$, we use the transformed initial condition: $U_s(\omega ,0)={\mathcal{F}}_s\{f(x)\}\equiv F_s(\omega )$. Therefore, $C(\omega )=F_s(\omega )$, so: $$U_s(\omega ,t)=F_s(\omega )e^{-k{\omega }^{2}t}.$$

Step 4: Applying the Inverse Transform. Finally, we recover $u(x,t)$ by applying the inverse Fourier sine transform to $U_s(\omega ,t)$: $$u(x,t)=\sqrt{\frac{2}{\pi }}{\int }_0^{\infty }{F}_s(\omega )e^{-k{\omega }^{2}t}\sin (\omega x)d\omega .$$ Often, $F_s(\omega )$ is itself an integral from the forward transform, leading to a solution expressed as a double integral that can sometimes be simplified.

Choosing Between Sine and Cosine Transforms

The decision to use the sine or cosine transform is not arbitrary; it is dictated by the boundary condition at $x=0$. This is a critical strategic step.

• Use the Fourier Sine Transform when the problem has a homogeneous Dirichlet condition at $x=0$: $u(0,t)=0$. This condition makes the extra boundary term in the transform of the second derivative vanish, simplifying the algebra. It is also used if the condition is non-homogeneous (e.g., $u(0,t)=T_0$), but this requires a more involved procedure, often involving a shift in the dependent variable.

• Use the Fourier Cosine Transform when the problem has a homogeneous Neumann condition at $x=0$: $u_x(0,t)=0$ (insulated end). This condition eliminates the boundary term in the transform of the second derivative for the cosine transform. It is the default choice for problems involving the gradient or flux being zero at the boundary.

The goal is always to select the transform whose operational property causes the boundary term to disappear, given your known condition. This directly incorporates the physical constraint into the mathematical machinery, leading to a cleaner, more straightforward solution.

Common Pitfalls

1. Misapplying the Operational Property: A frequent error is forgetting the boundary term in the transform of a second derivative. Remember:

• For ${\mathcal{F}}_s\{f^{\prime \prime }(x)\}$, the term is $+\sqrt{2/\pi }\omega f(0)$.
• For ${\mathcal{F}}_c\{f^{\prime \prime }(x)\}$, the term is $-\sqrt{2/\pi }{f}^{\prime }(0)$.

Always write this term down first. It will only be zero if the corresponding boundary condition ($f(0)$ for sine, $f^{\prime }(0)$ for cosine) is zero.

1. Incorrect Transform Choice for Mixed Conditions: If the boundary condition at $x=0$ is neither pure Dirichlet nor pure Neumann (e.g., a Robin condition like $u_x(0,t)=hu(0,t)$), neither transform simplifies the boundary term directly. In such cases, the standard Fourier transform on a full line might be more appropriate, or a different solution method like Laplace transforms may be required.

1. Neglecting the Solution's Domain: The Fourier sine/cosine transform solution is formally valid for $x>0$. While the solution may be extended to $x<0$ as an odd or even function, one must be cautious when interpreting results or calculating derivatives at $x=0$ from the integral formula, as smoothness conditions matter.

1. Confusing Normalization Constants: Different textbooks and engineering fields use different normalization constants (e.g., $\sqrt{2/\pi }$, $2/\pi$, or even $1$). The most critical rule is consistency: the constant in the forward transform and the inverse transform must multiply to $2/\pi$. Always check your source's definition and stick with it throughout a single problem.

Summary

• The Fourier sine and cosine transforms are specialized tools for solving PDEs on semi-infinite domains $[0,\infty )$, defined by integrating $f(x)$ against $\sin (\omega x)$ or $\cos (\omega x)$.
• Their inverse transforms have nearly identical forms, allowing for symmetric analysis and solution recovery.
• The key to their utility lies in their operational properties, which transform spatial derivatives into algebraic operations while explicitly incorporating boundary values $f(0)$ or $f^{\prime }(0)$.
• They are applied by transforming a PDE into a simpler ODE in the transform domain, solving that ODE, and then applying the inverse transform to find the solution in the original physical variables.
• The choice between them is strategic: use the sine transform for Dirichlet conditions ($u=0$ at $x=0$) and the cosine transform for Neumann conditions ($u_x=0$ at $x=0$) to make the boundary terms vanish and simplify the solution process.