Partial Differential Equations and Boundary-Value Problems

Introduction

Partial differential equations (PDEs) describe systems whose state depends on two or more independent variables, commonly position and time. In electronics and communication engineering, PDEs govern heat dissipation in semiconductor devices, electromagnetic fields in waveguides, vibrations in resonators, signal propagation along transmission lines, and diffusion processes in materials.

Unlike an ordinary differential equation, a PDE generally requires both governing equations and supplementary data. Initial conditions describe the system at a starting time, while boundary conditions describe behavior at spatial boundaries. Together they form an initial-boundary-value problem.

This chapter develops analytical methods for standard linear PDEs, emphasizing classification, separation of variables, eigenfunction expansions, Fourier and Laplace transforms, and engineering applications. Numerical methods are also introduced because realistic geometries and material properties often prevent closed-form solutions.

Learning Objectives

After completing this chapter, students should be able to:

• classify second-order linear PDEs as elliptic, parabolic, or hyperbolic;
• distinguish initial conditions from Dirichlet, Neumann, Robin, and mixed boundary conditions;
• apply separation of variables to heat, wave, and Laplace equations;
• construct Fourier-series and Sturm–Liouville eigenfunction expansions;
• use Fourier and Laplace transforms for unbounded or semi-infinite domains;
• model finite-rod heat flow and vibrating strings;
• derive wave equations from lossless transmission-line equations;
• solve basic electromagnetic boundary and waveguide problems;
• assess analytical assumptions, numerical stability, and engineering limitations.

1. Foundations of Partial Differential Equations

A PDE relates an unknown function to its partial derivatives. For a field u = u(x,y,t), a general second-order PDE may be written as

The order is the order of the highest derivative present. A PDE is linear when the dependent variable and all its derivatives occur only to the first power and are not multiplied together. A general second-order linear PDE in two independent variables is

where the coefficients may depend on the independent variables but not on u or its derivatives. It is homogeneous when G=0.

1.1 Canonical engineering equations

In the heat equation, α is the thermal diffusivity, measured in m²/s. It is often related to thermal conductivity k, density ρ, and specific heat capacity c_p by

Here k has units W/(m·K), ρ has units kg/m³, and c_p has units J/(kg·K). Thus α has units m²/s. The notation α² is not used for diffusivity in this chapter.

2. Classification of Second-Order PDEs

For the principal part

define the discriminant

At a point where the coefficients are defined:

• Δ<0: elliptic;
• Δ=0: parabolic;
• Δ>0: hyperbolic.

This classification assumes a real, nonsingular change of independent variables. If coefficients vary, the type may vary across the domain.

Worked Example 1: Classification

Classify

Comparing with A u_xx+2B u_xy+C u_yy=0 gives A=1, 2B=4, so B=2, and C=3. Hence

The equation is hyperbolic.

3. Initial and Boundary Conditions

A PDE does not normally have a unique solution until appropriate initial and boundary data are prescribed.

3.1 Common boundary conditions

For a scalar field u on a boundary with outward unit normal n̂:

• Dirichlet condition: the field value is specified, such as u=g.
• Neumann condition: the normal derivative is specified, such as ∂u/∂n=h.
• Robin condition: a linear combination is specified:

• Mixed conditions: different boundary types are imposed on different portions of the boundary.

For heat conduction, Fourier’s law gives normal heat flux

where q_n is in W/m². An insulated boundary therefore satisfies ∂T/∂n=0. Convective cooling produces the Robin condition

where h is the convection coefficient in W/(m²·K) and T_∞ is ambient temperature.

3.2 Well-posedness

A problem is well posed in the sense of Hadamard if a solution exists, is unique, and depends continuously on the supplied data. Not every arbitrary combination of boundary conditions is admissible. For example, a pure Neumann problem for Poisson’s equation requires a compatibility condition and determines the solution only up to an additive constant.

4. Separation of Variables and Eigenfunctions

Separation of variables seeks a product solution. For a one-dimensional transient field,

The method is most effective for linear homogeneous PDEs on regular domains with homogeneous boundary conditions. Nonhomogeneous boundary data can often be handled by first subtracting a steady-state or lifting function.

4.1 Sturm–Liouville structure

Many separated spatial equations have the form

with suitable homogeneous boundary conditions. If p(x)>0 and w(x)>0 and the problem is regular and self-adjoint, the eigenvalues are real, and eigenfunctions corresponding to distinct eigenvalues are orthogonal with weight w(x). Thus

Under standard completeness conditions, a sufficiently regular function can be expanded in these eigenfunctions. This mathematical structure explains why Fourier sine and cosine series naturally arise in boundary-value problems.

5. Heat Flow in a Finite Rod

Consider a homogeneous rod of length L with no internal heat generation, constant thermal diffusivity α, and ends maintained at zero reference temperature. The problem is

with

Substitute u=X(x)T(t). Then

After division by αXT,

The separated equations are

The boundary conditions X(0)=X(L)=0 admit nontrivial solutions only for

The temporal factors are

By superposition,

where

Higher spatial frequencies decay more rapidly because the decay rate is proportional to n².

Worked Example 2: Correct finite-rod temperature calculation

Let

• L=1.00\ \text{m},
• α=2.50×10^-4\ \text{m}²/\text{s},
• f(x)=100\sin(πx/L)\ ^\circ\text{C},
• both ends remain at 0 °C.

Because the initial condition is exactly the first eigenfunction,

At the midpoint x=L/2 and t=100\ \text{s},

Therefore,

The exponent is dimensionless because αt/L² has units (\text{m}²/\text{s})(\text{s})/\text{m}²=1.

5.1 Nonzero boundary temperatures

If u(0,t)=T₁ and u(L,t)=T₂ are constant, first write

where the steady solution is

Then v satisfies homogeneous end conditions and can be expanded in sine eigenfunctions.

6. The One-Dimensional Wave Equation

For a taut string with constant tension 𝒯 in newtons and linear mass density μ in kg/m, small transverse displacement y(x,t) satisfies

where c is the wave speed in m/s. The derivation assumes small slopes, constant tension, and negligible damping and bending stiffness.

For a string fixed at x=0 and x=L, separation gives

where

The coefficients are obtained from initial displacement y(x,0)=f(x) and initial velocity y_t(x,0)=g(x).

On an infinite line, sufficiently regular initial data give d’Alembert’s solution:

Unlike the heat equation, the wave equation propagates disturbances at finite speed and does not inherently smooth all initial irregularities.

7. Transform Methods

Transform methods are especially useful on infinite or semi-infinite domains and for nonhomogeneous forcing.

7.1 Fourier transform

Using the convention

the derivative property is

For the heat equation on the real line, u_t=αu_xx, transformation in space gives

Hence

Inverse transformation produces convolution with the heat kernel:

This solution applies to an infinite homogeneous medium with constant α and suitable initial data.

7.2 Laplace transform in time

Define

Because

a time-dependent PDE becomes an ordinary differential equation in space. For a semi-infinite solid initially at zero temperature and suddenly maintained at U₀ at x=0, the result is

where erfc is the complementary error function. This solution assumes x≥0, constant diffusivity, u(x,0)=0, u(0,t)=U₀, and u(x,t)→0 as x→∞.

8. Transmission-Line Equations

Let V(x,t) denote voltage and I(x,t) current on a uniform transmission line. Let R, L′, G, and C′ be resistance, inductance, conductance, and capacitance per unit length, respectively. Their units are Ω/m, H/m, S/m, and F/m.

The telegrapher’s equations are

For a lossless line, R=G=0. Differentiating and eliminating the other field yields

The propagation speed is denoted here by c_tl to avoid confusion with voltage:

The characteristic impedance is

These formulas apply to a uniform lossless line. For a general lossy sinusoidal line,

where γ is the complex propagation constant and ω is angular frequency.

Figure: Incident and reflected waves on a uniform transmission line terminated by load impedance Z_L.

The load-voltage reflection coefficient for a lossless line is

A matched load has Z_L=Z₀ and therefore Γ_L=0.

9. Electromagnetic Boundary-Value Problems

Maxwell’s equations lead to Laplace, Poisson, and Helmholtz equations under different assumptions.

9.1 Electrostatic potential

With electric potential φ defined by E=-∇φ, Gauss’s law gives

where ρ_v is volume charge density in C/m³ and ε is permittivity in F/m. In a charge-free region,

Uniqueness theorems ensure that a solution of Laplace’s equation satisfying prescribed Dirichlet data is unique. A Neumann solution is unique up to an additive constant if its compatibility condition is satisfied.

At an interface between media, Maxwell’s boundary conditions include

assuming no imposed magnetic surface current. Here J_s is electric surface-current density and ρ_s is free surface-charge density. At a perfect electric conductor, the tangential electric field is zero.

9.2 Helmholtz equation and rectangular waveguides

For time-harmonic fields in a homogeneous, source-free, lossless medium, field components satisfy

For a rectangular waveguide of broad dimension a and narrow dimension b, the cutoff frequency of mode TE_mn is

For an air-filled guide, 1/\sqrt{με}≈c₀, the free-space speed of light. If a>b, the dominant mode is normally TE₁₀, with

Worked Example 3: Rectangular waveguide

An air-filled WR-90-type guide has a=22.86\ \text{mm}. Its TE₁₀ cutoff frequency is

At f=10.0\ \text{GHz}, the free-space wavelength is 30.0 mm. The guide wavelength is

This propagating-mode formula requires f>f_c. Below cutoff, the axial propagation constant is imaginary and the mode is evanescent.

10. Applied Case: Reflection on a Digital Interconnect

Illustrative case.

Background

A laboratory transmits a fast voltage step through a 100 m lossless coaxial cable. The line has characteristic impedance Z₀=50\ Ω and propagation speed c_tl=2.00×10⁸\ \text{m/s}. The load is Z_L=100\ Ω. A matched source launches a 1.00 V incident step.

Problem

Determine the one-way delay, load reflection coefficient, reflected-step amplitude, and initial load voltage when the incident wave arrives.

Analysis and solution

The one-way propagation delay is

The load reflection coefficient is

Thus the reflected voltage amplitude is

At the load, the incident and reflected waves add:

Because the source is matched, the returning wave is absorbed rather than reflected again.

Outcome

The load first changes 0.500 μs after launch and reaches 1.333 V. The positive reflection occurs because the load impedance exceeds the characteristic impedance.

Learning points

• Long interconnects must be treated as distributed systems when propagation delay is significant relative to signal rise time.
• A load can experience overvoltage even when the launched step is only 1 V.
• Impedance matching suppresses reflections and improves signal integrity.
• Oscilloscope bandwidth and probe loading must be considered when verifying the prediction experimentally.

11. Numerical Methods, Limitations, and Best Practice

Analytical solutions are limited mainly to simple geometry, linear constitutive laws, and idealized boundary data. Finite-difference, finite-element, and finite-volume methods are therefore widely used.

For the explicit finite-difference heat scheme

stability in one spatial dimension requires

For a standard explicit centered scheme for the one-dimensional wave equation, the Courant condition is

These restrictions are scheme-specific and must not be treated as universal conditions for every numerical method.

Best practices

1. Check dimensions before computation.
2. Verify compatibility among initial and boundary data.
3. Distinguish diffusivity α from conductivity k.
4. Use enough eigenfunctions to resolve the initial profile, especially near discontinuities.
5. Test numerical mesh and time-step convergence.
6. Compare simulations with limiting cases, conservation laws, or analytical benchmarks.
7. Document material properties, boundary assumptions, and solver tolerances.

Challenges and contemporary developments

Real systems may involve anisotropic materials, temperature-dependent properties, dispersive transmission lines, nonlinear media, irregular geometry, or uncertain boundary data. Multiphysics solvers increasingly couple electromagnetic, thermal, structural, and semiconductor equations. Reduced-order models and machine-learning surrogates can accelerate repeated simulations, but they must be validated within their training domain. An AI model does not replace consistency checks, convergence studies, or experimental verification.

Safety and governance matter when PDE models inform high-power RF systems, medical devices, thermal protection, or public communication infrastructure. Engineers should preserve traceable inputs, report uncertainty, respect electromagnetic exposure and spectrum regulations, and avoid presenting unvalidated simulation outputs as measurements.

Chapter Summary

Partial differential equations provide the mathematical foundation for spatially distributed engineering systems. Second-order equations are classified through B²-AC as elliptic, parabolic, or hyperbolic. Boundary and initial conditions are essential to existence, uniqueness, and physical interpretation.

Separation of variables converts suitable linear PDEs into ordinary differential equations and leads naturally to Sturm–Liouville eigenvalue problems. Eigenfunction expansions solve finite-domain heat and wave problems, while Fourier and Laplace transforms are effective on infinite and semi-infinite domains. Heat modes decay exponentially, whereas wave modes oscillate and propagate at finite speed.

The telegrapher’s equations reduce to wave equations for a uniform lossless line, with propagation speed c_tl=1/\sqrt{L′C′}. Maxwell’s equations generate Laplace, Poisson, and Helmholtz boundary-value problems, including waveguide cutoff phenomena. Analytical results remain valuable as physical explanations and verification standards for computational models.

Key Terms

Boundary-value problem: A differential-equation problem supplemented by conditions at spatial boundaries.

Characteristic impedance: The voltage-to-current ratio of a single traveling wave on a transmission line.

Dirichlet condition: A boundary condition specifying the field value.

Eigenfunction: A nonzero function associated with an eigenvalue in an operator equation.

Elliptic PDE: A second-order PDE with B²-AC<0.

Fourier transform: An integral transformation representing a function through spatial or temporal frequencies.

Heat equation: A parabolic PDE describing diffusion of temperature or related quantities.

Helmholtz equation: A spatial PDE governing time-harmonic wave fields.

Hyperbolic PDE: A second-order PDE with B²-AC>0.

Laplace equation: The equation ∇²u=0, describing source-free equilibrium fields.

Neumann condition: A boundary condition specifying a normal derivative or flux.

Parabolic PDE: A second-order PDE with B²-AC=0.

Poisson equation: An elliptic PDE relating a field to a distributed source.

Reflection coefficient: The ratio of reflected-wave amplitude to incident-wave amplitude.

Robin condition: A boundary condition involving a linear combination of field value and normal derivative.

Separation of variables: A method expressing a multivariable solution as a product of single-variable functions.

Sturm–Liouville problem: A self-adjoint eigenvalue problem with weighted orthogonal eigenfunctions.

Thermal diffusivity: The material parameter α=k/(ρc_p), measured in m²/s.

Wave equation: A hyperbolic PDE describing finite-speed propagation.

Well-posed problem: A problem having an existing, unique solution that depends continuously on its data.

Review Questions

Short Answer Questions

1. State the discriminant used to classify a second-order PDE written with principal part A u_xx+2B u_xy+C u_yy.
2. Distinguish Dirichlet, Neumann, and Robin boundary conditions.
3. Why do higher heat-equation modes decay faster than lower modes?
4. State the physical meaning and units of thermal diffusivity.
5. What load condition eliminates reflection on a lossless transmission line?

Long Answer Questions

1. Derive the eigenfunction-series solution for heat conduction in a finite rod with zero end temperatures.
2. Explain the Sturm–Liouville framework and derive the weighted orthogonality relation for distinct eigenfunctions.
3. Derive the normal-mode solution of a string fixed at both ends and explain the roles of initial displacement and velocity.
4. Starting from the lossless telegrapher’s equations, derive the voltage wave equation, propagation speed, and characteristic impedance.
5. Explain how Maxwell’s equations lead to Laplace, Poisson, and Helmholtz equations under different assumptions.

Analytical / Application Questions

1. A 0.50 m rod has α=1.2×10^-5\ \text{m}²/\text{s} and initial temperature 80\sin(πx/L)\ ^\circ\text{C}. With zero-temperature ends, calculate its midpoint temperature after 300 s.
2. A 75 Ω transmission line terminates in 25 Ω. For a 2 V incident wave, calculate the reflection coefficient, reflected voltage, and total load voltage.
3. An air-filled rectangular waveguide has broad dimension 30 mm. Determine the TE₁₀ cutoff frequency and decide whether an 8 GHz signal propagates.

Multiple Choice Questions with Answers

Question 1

For the PDE u_xx+6u_xy+5u_yy=0, what is its classification?

A. Elliptic

B. Parabolic

C. Hyperbolic

D. First-order

Correct Answer: C. Hyperbolic

Question 2

Which quantity has units of m²/s in the heat equation u_t=αu_xx?

A. Thermal conductivity k

B. Thermal diffusivity α

C. Specific heat capacity c_p

D. Heat flux q_n

Correct Answer: B. Thermal diffusivity α

Question 3

For a finite rod with zero end temperatures, which spatial eigenfunctions satisfy the boundary conditions?

A. \sin(nπx/L)

B. \cos(nπx/L) for all positive integers n

C. e^nπx/L

D. 1+x/L

Correct Answer: A. \sin(nπx/L)

Question 4

A 1 m rod has α=2.50×10^-4\ \text{m}²/\text{s} and initial temperature 100\sin(πx/L)\ ^\circ\text{C}. With zero-temperature ends, what is the midpoint temperature after 100 s?

A. 97.6 °C

B. 78.1 °C

C. 24.7 °C

D. 7.81 °C

Correct Answer: B. 78.1 °C

Question 5

What does the Fourier transform convert u_xx into under the convention used in the chapter?

A. ik\hat{u}

B. k²\hat{u}

C. -k²\hat{u}

D. -ik\hat{u}

Correct Answer: C. -k²\hat{u}

Question 6

For a uniform lossless transmission line, the propagation speed is

A. c_tl=\sqrt{L′C′}

B. c_tl=1/\sqrt{L′C′}

C. c_tl=\sqrt{L′/C′}

D. c_tl=1/(L′C′)

Correct Answer: B. c_tl=1/\sqrt{L′C′}

Question 7

A 50 Ω lossless line is terminated by 100 Ω. What is the load reflection coefficient?

A. -1/3

B. 0

C. 1/3

D. 2

Correct Answer: C. 1/3

Question 8

Which boundary condition represents an insulated thermal surface?

A. T=0

B. ∂T/∂n=0

C. T=T_∞

D. ∂T/∂t=0

Correct Answer: B. ∂T/∂n=0

Question 9

For the one-dimensional explicit finite-difference heat scheme described in the chapter, stability requires r=αΔt/(Δx)² to satisfy

A. r>1

B. r=1 only

C. 0≤r≤1/2

D. r<0

Correct Answer: C. 0≤r≤1/2

Question 10

An air-filled rectangular waveguide has broad dimension a. What is the TE₁₀ cutoff frequency?

A. c₀/a

B. c₀/(2a)

C. 2c₀/a

D. a/(2c₀)

Correct Answer: B. c₀/(2a)