Partial Differential Equations

Introduction and engineering context

A partial differential equation (PDE) relates an unknown function of two or more independent variables to its partial derivatives. In mechatronics engineering, PDEs model temperature in machine components, vibration of beams and membranes, diffusion of species, electrostatic potential, fluid motion, and electromagnetic fields. Unlike an ordinary differential equation, whose solution varies with one independent variable, a PDE describes a field distributed over space and time.

A typical field is temperature T(x,t), where x is position and t is time. A complete mathematical model requires three elements:

1. the governing PDE;
2. boundary conditions describing behavior at spatial boundaries; and
3. initial conditions describing the field at the initial time, when the PDE is time-dependent.

This chapter develops the classical one-dimensional heat and wave equations and the two-dimensional Laplace equation. The principal analytical method is separation of variables, supported by Fourier series.

Learning objectives

By the end of this chapter, you should be able to:

• classify a second-order PDE as elliptic, parabolic, or hyperbolic;
• derive and interpret the heat, wave, and Laplace equations;
• formulate suitable initial and boundary conditions;
• solve standard rectangular-domain problems by separation of variables;
• construct Fourier sine and cosine coefficients;
• interpret eigenvalues, eigenfunctions, convergence, and physical meaning;
• select appropriate models for thermal, structural, and field applications;
• recognize assumptions and limitations of analytical PDE solutions.

1. PDE terminology and classification

For an unknown field u(x,y,t), the order of a PDE is the order of its highest derivative. A PDE is linear if the dependent variable and its derivatives occur only to the first power and are not multiplied together. For example,

is linear, whereas u_t = u u_x is nonlinear.

For a general second-order PDE in two spatial variables,

classification is based on B² - 4AC:

This classification is local for equations with variable coefficients, but it provides the fundamental modeling distinction for elementary engineering problems.

Boundary and initial conditions

For a spatial interval 0 < x < L, common boundary conditions are:

• Dirichlet: prescribe the field, such as u(0,t)=U₀.
• Neumann: prescribe the derivative or flux, such as u_x(L,t)=0.
• Robin or mixed: relate field and derivative, such as k u_x + h u = r(t).

For a time-dependent problem, an initial condition has the form

A second-order time equation requires two initial conditions, normally u(x,0) and u_t(x,0). Boundary conditions must be compatible with the governing equation and the physical device.

2. The one-dimensional heat equation

Consider a slender homogeneous rod of length L, constant cross-sectional area A, density ρ, specific heat capacity c, and thermal conductivity k. Let T(x,t) denote temperature. An energy balance on a small element gives

Dividing by A Δx and taking the limit produces

Here α² is thermal diffusivity, measured in m²/s. The notation α is conventional; the diffusivity itself is α², not α. The model assumes one-dimensional conduction, constant properties, no internal heat generation, and negligible radiation and convection within the rod.

For a rod held at zero temperature at both ends,

the solution is

where

The sine functions are eigenfunctions satisfying the homogeneous endpoint conditions. The exponential factor shows that high spatial frequencies decay more rapidly because their decay rate is proportional to n². Thus heat conduction smooths temperature variations.

Separation-of-variables derivation

Assume a product solution T(x,t)=X(x)G(t). Substitution gives

After division by α² XG,

The two sides depend on different variables and must therefore equal a constant. For zero-temperature endpoints,

Nontrivial solutions exist only for

The time equation is G_n' + α² λ_n G_n=0, yielding the exponential term in the series solution. Superposition is valid because the heat equation and boundary conditions are linear.

Nonhomogeneous boundary temperatures

Separation of variables is simplest when boundary conditions are homogeneous. Suppose the ends are maintained at constant temperatures T(0,t)=T_a and T(L,t)=T_b. First write

where the steady state satisfies w''=0, w(0)=T_a, and w(L)=T_b. Therefore,

The transient field v then satisfies homogeneous endpoint conditions and can be expanded in a sine series. This steady-plus-transient decomposition is widely used in thermal design.

Worked example 1: cooling of a rod

An insulated lateral surface rod has L=0.50 m, thermal diffusivity α²=1.0×10^-5 m²/s, and initial temperature

Both ends are suddenly maintained at 0 °C. Because the initial condition is already the first eigenfunction, only n=1 is present:

At the midpoint, x=L/2, and at t=1000 s,

Since (π/0.50)²≈39.478, the exponent is approximately -0.3948. Hence

The result has correct units: the exponent is dimensionless because α² has units m²/s, (π/L)² has units 1/m², and t has units s.

3. The wave equation

The transverse displacement of a taut, uniform string can be modeled by

where u(x,t) is displacement, c=√(T₀/μ) is wave speed, T₀ is tension in newtons, and μ is mass per unit length in kilograms per metre. The assumptions include small transverse displacement, constant tension, negligible damping, and linear motion.

For a string fixed at both ends,

with initial displacement u(x,0)=f(x) and initial velocity u_t(x,0)=g(x), the solution is

where

and

The coefficients follow by matching displacement and velocity at t=0. Unlike the heat equation, the wave equation contains oscillatory terms rather than decaying exponentials. In the ideal model, energy is conserved and disturbances propagate at finite speed c.

Worked example 2: a single-mode string vibration

Let a string of length L=1 m have wave speed c=20 m/s. Suppose

Only A₁=0.01 m is nonzero. Therefore,

At the midpoint and at t=0.025 s,

The midpoint is instantaneously at equilibrium, although the string is moving there. Its velocity is not zero, illustrating why both initial displacement and initial velocity are required for a second-order time equation.

Damping and forced vibration

Real mechatronic structures dissipate energy. A simplified damped wave model is

where γ≥0 is a damping coefficient with units s^-1. External actuators may add a forcing term F(x,t). Separation of variables remains useful for some linear systems, but modal analysis, Laplace transforms, numerical methods, or finite-element models are often required for complex geometry and forcing.

4. Laplace’s equation and steady fields

Laplace’s equation in two dimensions is

It is elliptic and describes steady temperature in a region without internal heat generation, electrostatic potential in a charge-free region, and certain incompressible potential flows. The solution is determined by boundary data throughout the boundary; there is no initial condition because time does not appear.

For a rectangular region 0<x<L, 0<y<H, consider

Assume u=X(x)Y(y). Substitution gives

and hence

The side conditions yield

The corresponding Y equation is Y''-λ_nY=0. Applying Y(0)=0 gives a hyperbolic sine function. Thus

where

At y=H, the factor involving hyperbolic sines becomes one and reproduces the specified boundary function in the Fourier sense.

Interpretation of the Laplace solution

Laplace solutions possess the mean-value property: under suitable regularity assumptions, the value at an interior point equals the average of values over any sufficiently small surrounding circle. Consequently, a nonconstant solution has no isolated interior maximum or minimum. This property explains why boundary conditions control the entire steady field and why abrupt boundary changes are smoothed in the interior.

5. Fourier series and eigenfunction methods

A Fourier series represents a function by orthogonal trigonometric modes. On 0,L, sine functions satisfy

and

These relations produce the coefficient formula. Cosine series are appropriate for Neumann-type conditions, while sine series are natural for homogeneous Dirichlet conditions. More general boundary conditions lead to Sturm–Liouville eigenvalue problems.

If f is piecewise smooth, its Fourier series converges to f(x) at continuity points and to the midpoint of the one-sided limits at a jump. Near discontinuities, partial sums may exhibit Gibbs oscillations. This is important when representing sharply switched thermal or mechanical inputs.

6. Boundary-condition selection and physical modeling

The mathematical type of boundary condition must reflect the device:

For heat flow, Fourier’s law is q_x=-kA T_x, where q_x is heat-transfer rate in watts. The negative sign means heat flows down the temperature gradient. In a Robin condition, h is a convection coefficient in W/(m² K), and T_∞ is ambient temperature.

A model is well posed when a solution exists, is unique, and depends continuously on the data. Incorrect or incomplete boundary conditions can produce nonunique or physically meaningless solutions. For example, specifying temperature and heat flux independently at every point may overconstrain a problem unless the data are compatible.

7. Illustrative case: thermal management of a motor housing

Background. Consider an illustrative motor-housing wall idealized as a one-dimensional slab of thickness L. Heat is generated internally by electrical losses, while the outer surfaces exchange heat with coolant.

Problem. The design team wants to estimate whether the housing exceeds an allowable temperature during a transient load. A first model assumes constant material properties, one-dimensional conduction, and no radiation.

Analysis. With volumetric heat generation Q̇ measured in W/m³, the heat equation becomes

Convective boundaries may be represented by

and

The signs depend on the chosen positive x direction and must be derived from outward heat flux at each surface. A steady solution can first be calculated, followed by a transient expansion in eigenfunctions satisfying the homogeneous form of the Robin conditions.

Approach and solution. Engineers may use the analytical solution to identify dominant time scales and verify a finite-element model. If the first eigenvalue is λ₁, the slowest transient generally decays approximately as e^-α2λ₁t. A characteristic diffusion time is

This estimate is dimensionally consistent and helps determine whether temperature changes occur over milliseconds, seconds, or hours.

Outcome. If the predicted peak temperature exceeds the material limit, design changes may include increasing conductivity, reducing wall thickness, improving coolant flow, or lowering electrical loss. The one-dimensional result is not a final certification because corners, contact resistances, anisotropy, and temperature-dependent properties may be important.

Learning points. PDE analysis supports design decisions by explaining the dominant physics, supplying benchmark solutions, and revealing which assumptions need experimental validation. It should complement, not automatically replace, computational simulation and measurement.

8. Numerical and computational considerations

Analytical series solutions are valuable but limited to simple geometries, coefficients, and boundary conditions. Real mechatronic systems often require finite-difference, finite-element, finite-volume, or spectral methods. Numerical models must still preserve the PDE’s physical structure.

For a simple explicit finite-difference heat scheme with spatial step Δx and time step Δt,

For the standard one-dimensional explicit scheme, stability requires approximately r≤1/2. This is a numerical restriction, not a physical law. Violating it can produce oscillations or divergence even when the actual heat process is stable.

For computational practice:

• verify units and sign conventions before coding;
• test against an analytical solution;
• refine mesh and time step to check convergence;
• monitor conservation of energy where appropriate;
• document material properties, boundary data, and uncertainty;
• avoid interpreting numerical precision as physical accuracy.

AI-assisted coding or surrogate modeling can accelerate parameter studies, but engineering responsibility remains with the analyst. Training data, extrapolation, uncertainty, and traceability must be assessed before using an AI model for safety-critical thermal or structural decisions.

9. Limitations and extensions

The classical equations assume idealizations. Thermal conductivity may vary with temperature; radiation is nonlinear; wave speed may depend on strain; damping may be frequency-dependent; and geometry may require two- or three-dimensional models. Nonlinear PDEs generally do not permit direct superposition, and separation of variables may fail.

Important extensions include the diffusion equation with reaction, the damped and forced wave equation, Poisson’s equation ∇²u=f, cylindrical and spherical-coordinate models, and coupled thermoelastic systems. Boundary-element, finite-element, and model-reduction methods extend PDE analysis to industrial geometries.

Chapter Summary

Partial differential equations describe fields varying over space and, often, time. Elliptic equations such as Laplace’s equation model steady spatial fields; parabolic equations such as the heat equation model diffusion; and hyperbolic equations such as the wave equation model finite-speed propagation. Separation of variables converts a PDE into ordinary differential equations and produces eigenfunctions selected by boundary conditions. Fourier series then combine these modes to satisfy arbitrary compatible initial or boundary data. Heat modes decay exponentially, whereas ideal wave modes oscillate. Laplace solutions are controlled entirely by boundary values and exhibit strong smoothing properties. Correct physical assumptions, units, boundary conditions, convergence interpretation, and numerical verification are essential for reliable engineering use.

Key Terms

Partial differential equation: An equation involving partial derivatives of a function of several independent variables.

Elliptic equation: A PDE type associated commonly with steady spatial fields.

Parabolic equation: A PDE type associated commonly with diffusion and heat conduction.

Hyperbolic equation: A PDE type associated commonly with wave propagation.

Boundary condition: A prescribed field value or derivative on the boundary of a domain.

Initial condition: A prescribed state at the initial time.

Thermal diffusivity: The quantity α²=k/(ρc) governing the rate of thermal diffusion.

Eigenvalue: A parameter for which a boundary-value problem has a nontrivial eigenfunction.

Eigenfunction: A mode satisfying a differential equation and its homogeneous boundary conditions.

Separation of variables: A method assuming a product of functions of individual variables.

Fourier series: An expansion of a function in orthogonal trigonometric modes.

Dirichlet condition: A condition prescribing the dependent variable.

Neumann condition: A condition prescribing a derivative or flux.

Robin condition: A linear combination of a field and its derivative.

Laplace equation: The equation ∇²u=0.

Superposition: Addition of solutions of a linear homogeneous PDE.

Wave speed: The propagation speed c in the ideal wave equation.

Gibbs phenomenon: Oscillatory overshoot near a Fourier representation’s jump discontinuity.

Well-posed problem: A problem having existence, uniqueness, and continuous dependence on data.

Review Questions

Short Answer Questions

1. What distinguishes a PDE from an ordinary differential equation?
2. State the classification condition involving B²-4AC for a second-order PDE.
3. What physical quantity does thermal diffusivity describe?
4. Why are two initial conditions required for the classical wave equation?
5. Distinguish Dirichlet and Neumann boundary conditions.

Long Answer Questions

1. Derive the one-dimensional heat equation from an energy balance on a differential rod element.
2. Derive the separated solution for a rod with zero temperature at both ends.
3. Explain how Fourier sine coefficients are obtained using orthogonality.
4. Derive the rectangular-domain solution of Laplace’s equation for three zero sides and one prescribed side.
5. Compare the physical interpretations of heat-equation and wave-equation modes.

Analytical / Application Questions

1. A rod has L=0.2 m and α²=4×10^-5 m²/s. Estimate its diffusion time using t_d≈L²/α², and explain what the estimate means.
2. A string has T₀=100 N and mass per unit length μ=0.01 kg/m. Calculate its wave speed and identify the assumptions behind the result.
3. Design a boundary-condition model for a metal plate cooled by air on one side and attached to an ideal temperature reservoir on the other. Explain whether Dirichlet, Neumann, or Robin conditions are most appropriate and why.

Multiple Choice Questions with Answers

Question 1

Which classification corresponds to the one-dimensional heat equation?

A. Elliptic

B. Parabolic

C. Hyperbolic

D. Algebraic

Correct Answer: B. Parabolic

Question 2

For a rod with zero temperature at both ends, which eigenfunctions satisfy the boundary conditions?

A. cos(nπx/L)

B. exp(nπx/L)

C. sin(nπx/L)

D. x^n

Correct Answer: C. sin(nπx/L)

Question 3

The thermal diffusivity in the heat equation is defined by:

A. α²=ρc/k

B. α²=k/(ρc)

C. α²=kρc

D. α²=ρ/(kc)

Correct Answer: B. α²=k/(ρc)

Question 4

For the ideal string equation, the wave speed is:

A. c=T₀μ

B. c=√(μ/T₀)

C. c=√(T₀/μ)

D. c=T₀/μ²

Correct Answer: C. c=√(T₀/μ)

Question 5

Which condition prescribes the normal derivative or flux of a field?

A. Dirichlet

B. Neumann

C. Initial

D. Periodic

Correct Answer: B. Neumann

Question 6

In a heat-equation Fourier solution, higher spatial modes generally:

A. decay more slowly because their index is larger

B. remain constant in time

C. decay more rapidly because the rate is proportional to n²

D. oscillate without decay

Correct Answer: C. decay more rapidly because the rate is proportional to n²

Question 7

Laplace’s equation without a time variable is normally classified as:

A. Elliptic

B. Parabolic

C. Hyperbolic

D. Non-differential

Correct Answer: A. Elliptic

Question 8

For the explicit finite-difference heat scheme stated in the chapter, the standard stability restriction is approximately:

A. r≥2

B. r=1

C. r≤1/2

D. r≤−1/2

Correct Answer: C. r≤1/2

Question 9

A wave problem requires two initial conditions because it is:

A. first order in time

B. second order in time

C. independent of time

D. nonlinear in space

Correct Answer: B. second order in time

Question 10

For a string with tension 100 N and mass per unit length 0.01 kg/m, the wave speed is:

A. 1 m/s

B. 10 m/s

C. 100 m/s

D. 10,000 m/s

Correct Answer: C. 100 m/s