Unit overview
Mathematical Methods for Physics II develops the tools used to solve field equations, wave problems, diffusion, quantum systems and boundary-value problems. Emphasis is placed on the meaning of mathematical operations, the assumptions behind solution methods and the physical interpretation of boundary and initial conditions.
Level: Undergraduate university physics. Prerequisites: introductory mechanics, algebra, trigonometry, calculus and basic laboratory or mathematical reasoning appropriate to the topic. Study approach: work through each concept, reproduce the derivations, attempt the numerical examples independently, then complete the examination questions.
Learning outcomes
- Define the central physical quantities and use SI units consistently.
- Derive and apply governing equations while stating assumptions and validity limits.
- Interpret diagrams, graphs, mathematical models and measurable consequences.
- Solve multi-step numerical and conceptual problems with clear working.
- Evaluate uncertainty, approximations, boundary conditions and experimental limitations where relevant.
- Communicate conclusions using correct notation, units and physical reasoning.
Core textbook content
Mathematical Methods for Physics II develops the tools used to solve field equations, wave problems, diffusion, quantum systems and boundary-value problems. Emphasis is placed on the meaning of mathematical operations, the assumptions behind solution methods and the physical interpretation of boundary and initial conditions. The sections below develop the main concepts from definitions to applications. For each topic, identify the system, list known quantities, choose a model, state approximations, and test the result against units and physical limits.
1. Vector fields and differential operators
A scalar field assigns one value to each point, such as temperature T(x,y,z); a vector field assigns a vector, such as fluid velocity or electric field. The gradient ∇φ points in the direction of steepest increase and has magnitude equal to the maximum directional derivative. Divergence ∇·A measures local source strength per unit volume, while curl ∇×A measures local circulation density. In Cartesian coordinates these operators have simple component forms; curvilinear coordinates require metric factors. Coordinate systems should be selected to match geometry, not merely to simplify notation.
2. Line, surface and volume integrals
A line integral ∫A·dl accumulates the tangential component of a vector field along a path. A surface integral ∫A·dS measures flux through an oriented surface, while a volume integral accumulates a scalar density over a region. Orientation matters: reversing a path or surface normal changes the sign. Physical examples include work, circulation, electric flux, mass and charge. The integration domain and coordinate measure must be stated explicitly.
3. Divergence theorem and Gauss's theorem
The divergence theorem relates outward flux through a closed surface to the volume integral of divergence: ∮A·dS=∫(∇·A)dV. It converts a local differential description into a global flux statement and is central to Gauss's laws. The theorem requires a sufficiently smooth field and a properly oriented closed boundary; singularities inside the volume require care. It is especially powerful when symmetry makes the surface integral simple.
4. Stokes' theorem and circulation
Stokes' theorem states ∮C A·dl=∫S(∇×A)·dS, where C is the positively oriented boundary of surface S. It connects circulation around a closed curve to curl over the spanning surface. The right-hand rule fixes orientation. This relation underpins Faraday's law and Ampère–Maxwell law in integral and differential forms. If two surfaces share the same boundary and the field is regular, the fluxes of curl agree.
5. Fourier series and orthogonality
A periodic function can be represented by a Fourier series when suitable convergence conditions hold: f(x)=a₀/2+Σ[a_n cos(nπx/L)+b_n sin(nπx/L)] on an interval of length 2L. Orthogonality isolates coefficients through integrals. Even functions contain cosine terms; odd functions contain sine terms. Discontinuities lead to Gibbs overshoot near jumps, which does not vanish in height as more terms are added although it becomes confined to a narrower region. Fourier series are used for vibrating strings, heat conduction and periodic signals.
6. Fourier transforms and spectra
The Fourier transform decomposes a non-periodic or suitably integrable signal into spatial or temporal frequencies. One convention is F(k)=∫f(x)e^(−ikx)dx, with inverse f(x)=(1/2π)∫F(k)e^(ikx)dk. Other conventions distribute factors of 2π differently; always state the convention. A narrow feature in one domain spreads in the conjugate domain. This uncertainty-like trade-off links pulse duration to bandwidth and aperture width to diffraction spread. Parseval's theorem relates total energy in the original and transformed domains.
7. Complex analysis and analytic functions
A complex function f(z)=u(x,y)+iv(x,y) is analytic when it is complex differentiable in a neighbourhood. The Cauchy–Riemann equations u_x=v_y and u_y=−v_x are necessary, and with suitable continuity conditions sufficient, for analyticity. Analytic functions have powerful integral properties. Complex methods simplify oscillatory integrals, potential theory and wave analysis, but branch points and branch cuts must be handled explicitly for multivalued functions such as log z and z^α.
8. Residues and contour integration
The residue theorem evaluates contour integrals of meromorphic functions: ∮f(z)dz=2πiΣRes(f,z_k) for positively oriented contours enclosing isolated singularities. A simple pole at z₀ has residue lim(z→z₀)(z−z₀)f(z). Higher-order poles use derivatives or Laurent expansions. The method is useful for real integrals, inverse transforms and response functions. The contour must be chosen so that contributions from auxiliary arcs vanish or are otherwise evaluated; pole locations and contour orientation determine signs.
9. Ordinary differential equations and Sturm–Liouville systems
A Sturm–Liouville problem has form −d/dx[p(x)y′]+q(x)y=λw(x)y with boundary conditions. Under standard regularity and self-adjoint conditions, eigenvalues are real and eigenfunctions with distinct eigenvalues are orthogonal under weight w(x). This structure appears in normal modes, quantum mechanics and separation of variables. Boundary conditions determine the allowed spectrum; changing them changes eigenvalues and physical modes.
10. Partial differential equations and classification
Second-order linear PDEs in two variables can be classified by the discriminant B²−AC for A u_xx+B u_xy+C u_yy+…: elliptic, parabolic or hyperbolic. Laplace's equation models steady potential fields; the diffusion equation models dissipative spreading; the wave equation models finite-speed propagation. Classification predicts qualitative behaviour and guides solution methods. Initial data are central for evolution equations, while boundary data are central for spatial boundary-value problems.
11. Separation of variables and eigenfunction expansions
Assume a product solution such as u(x,t)=X(x)T(t), substitute into the PDE and separate variables by equating expressions that depend on independent variables to a constant. Boundary conditions produce an eigenvalue problem for X; initial conditions determine expansion coefficients. The method is not a universal guess: it relies on separable geometry, coefficients and boundary conditions. Superposition allows linear combinations of eigenmodes for linear PDEs.
12. Laplace and inverse Laplace transforms
The unilateral Laplace transform F(s)=∫₀∞f(t)e^(−st)dt converts derivatives into algebraic expressions while encoding initial conditions. It is useful for linear ODEs, circuits, damped oscillators and control systems. Poles determine natural modes and stability properties. Inverse transforms may use partial fractions, tables or contour methods. The region of convergence matters, especially for causal signals and stability; the same algebraic expression can correspond to different time-domain functions under different convergence regions.
13. Green functions and boundary-value response
A Green function is the response of a linear operator to a point source, subject to specified boundary conditions. If L G(x,ξ)=δ(x−ξ), then a solution to L y=f can often be written y(x)=∫G(x,ξ)f(ξ)dξ plus a homogeneous contribution for boundary data. The Green function depends on the operator and boundary conditions; it is not a universal kernel. Its symmetry follows under appropriate self-adjoint conditions.
14. Special functions in physical models
Legendre polynomials, Bessel functions, Hermite polynomials and Laguerre polynomials arise from separation of variables in common coordinate systems and operators. Legendre functions appear in spherical angular problems; Bessel functions arise in cylindrical symmetry; Hermite functions occur for the quantum harmonic oscillator; Laguerre polynomials appear in radial hydrogenic solutions. Their orthogonality, recurrence relations and asymptotic behaviour are computational tools as well as mathematical properties. Normalisation and domain must be stated.
15. Numerical solution and convergence checks
Analytical solutions are not always available. Finite differences approximate derivatives on grids; finite elements use local basis functions and weak formulations; spectral methods expand solutions in global basis functions. Numerical answers require grid refinement, convergence studies, stability analysis and comparison with limiting cases or known solutions. A small residual in a discretised equation does not alone prove a small physical error. Boundary implementation, conditioning and round-off must be considered.
16. Dimensional and physical interpretation
Mathematical transformations must preserve physical dimensions. A Fourier wavenumber has inverse-length units; angular frequency has inverse-time units; eigenvalues inherit units from the differential operator. Boundary conditions encode physical constraints such as fixed temperature, zero displacement or specified flux. Check solutions for symmetry, conservation laws, regularity, limiting behaviour and dimensional consistency. Mathematical elegance does not compensate for a solution that violates the physical problem.
Summary equations and principles
Model → assumptions → governing equation → boundary / initial data → solution → dimensional and physical checks.
Use the equations introduced in each section as a connected framework rather than as isolated formulae. Define every symbol, retain units through calculations and identify whether quantities are vectors, scalars, fields, rates or integrated totals. Check special cases and conservation statements before accepting a numerical answer.
University examination and problem set
Attempt all questions. Show derivations, diagrams, units and intermediate steps. Answers should distinguish exact results from approximations and should state assumptions explicitly.
- 1. Define the central quantities and state the assumptions used in mathematical methods for physics ii.
Answer guidance: Define symbols and assumptions; show the governing equation; substitute values with units; check dimensions and limiting behaviour; state the physical interpretation. For explanatory items, support the argument with a labelled sketch or a relevant equation.
- 2. Derive the principal governing relation for one of the systems discussed in this unit.
Answer guidance: Define symbols and assumptions; show the governing equation; substitute values with units; check dimensions and limiting behaviour; state the physical interpretation. For explanatory items, support the argument with a labelled sketch or a relevant equation.
- 3. Explain how the choice of model or coordinate system affects the solution.
Answer guidance: Define symbols and assumptions; show the governing equation; substitute values with units; check dimensions and limiting behaviour; state the physical interpretation. For explanatory items, support the argument with a labelled sketch or a relevant equation.
- 4. A measured or calculated quantity changes by 8%. Determine the effect on a dependent quantity using the relevant sensitivity relation.
Answer guidance: Define symbols and assumptions; show the governing equation; substitute values with units; check dimensions and limiting behaviour; state the physical interpretation. For explanatory items, support the argument with a labelled sketch or a relevant equation.
- 5. Compare two limiting cases and explain the physical meaning of the difference.
Answer guidance: Define symbols and assumptions; show the governing equation; substitute values with units; check dimensions and limiting behaviour; state the physical interpretation. For explanatory items, support the argument with a labelled sketch or a relevant equation.
- 6. Draw and label a diagram showing the geometry, directions and relevant variables for a representative problem.
Answer guidance: Define symbols and assumptions; show the governing equation; substitute values with units; check dimensions and limiting behaviour; state the physical interpretation. For explanatory items, support the argument with a labelled sketch or a relevant equation.
- 7. Identify three sources of systematic limitation and explain how each could be reduced.
Answer guidance: Define symbols and assumptions; show the governing equation; substitute values with units; check dimensions and limiting behaviour; state the physical interpretation. For explanatory items, support the argument with a labelled sketch or a relevant equation.
- 8. Construct a graph that would test a predicted linear relationship. State the expected gradient and intercept.
Answer guidance: Define symbols and assumptions; show the governing equation; substitute values with units; check dimensions and limiting behaviour; state the physical interpretation. For explanatory items, support the argument with a labelled sketch or a relevant equation.
- 9. Solve a multi-step numerical problem using consistent SI units and show intermediate working.
Answer guidance: Define symbols and assumptions; show the governing equation; substitute values with units; check dimensions and limiting behaviour; state the physical interpretation. For explanatory items, support the argument with a labelled sketch or a relevant equation.
- 10. Explain where the approximation used in the unit would fail and propose a more complete model.
Answer guidance: Define symbols and assumptions; show the governing equation; substitute values with units; check dimensions and limiting behaviour; state the physical interpretation. For explanatory items, support the argument with a labelled sketch or a relevant equation.
- 11. Relate a mathematical result to an observable experimental consequence.
Answer guidance: Define symbols and assumptions; show the governing equation; substitute values with units; check dimensions and limiting behaviour; state the physical interpretation. For explanatory items, support the argument with a labelled sketch or a relevant equation.
- 12. Evaluate a proposed experimental method, including controls, uncertainty and safety where relevant.
Answer guidance: Define symbols and assumptions; show the governing equation; substitute values with units; check dimensions and limiting behaviour; state the physical interpretation. For explanatory items, support the argument with a labelled sketch or a relevant equation.
- 13. Show how a conservation law or boundary condition constrains the solution.
Answer guidance: Define symbols and assumptions; show the governing equation; substitute values with units; check dimensions and limiting behaviour; state the physical interpretation. For explanatory items, support the argument with a labelled sketch or a relevant equation.
- 14. Interpret a non-ideal result and distinguish measurement effects from failure of the theoretical model.
Answer guidance: Define symbols and assumptions; show the governing equation; substitute values with units; check dimensions and limiting behaviour; state the physical interpretation. For explanatory items, support the argument with a labelled sketch or a relevant equation.
- 15. Write a concise scientific conclusion that includes the result, uncertainty or limitations, and physical interpretation.
Answer guidance: Define symbols and assumptions; show the governing equation; substitute values with units; check dimensions and limiting behaviour; state the physical interpretation. For explanatory items, support the argument with a labelled sketch or a relevant equation.
Independent study and further reading
Use a recognised university-level physics textbook and laboratory manual alongside these notes. Compare notation and sign conventions across sources, but do not combine formulae until their conventions and assumptions have been reconciled. For deeper study, locate peer-reviewed papers or university lecture material on the specific physical model, then reproduce at least one derivation or numerical result independently.
For revision, prepare a one-page concept map linking definitions, governing equations, experimental observables, limiting cases and common failure modes. Rework the examples with altered numerical values and explain why the qualitative outcome does or does not change.
CBE Nexus · Undergraduate Physics · Unit 8 of 36 · Educational resource. Check the assumptions and conventions stated in each derivation before applying equations to a different physical system.
17. Vector calculus: derivations, identities and physical meaning
17.1 Deriving the gradient from a directional change
Let a scalar field be φ(x,y,z), and let a particle move through a small displacement d r=(dx,dy,dz). The first-order change in the field is obtained from the multivariable Taylor expansion: dφ=(∂φ/∂x)dx+(∂φ/∂y)dy+(∂φ/∂z)dz. Introducing the Cartesian unit vectors gives d r=i dx+j dy+k dz and ∇φ=i∂φ/∂x+j∂φ/∂y+k∂φ/∂z. Therefore dφ=∇φ·d r. If the displacement has unit direction n and length ds, the directional derivative is dφ/ds=∇φ·n. By the Cauchy–Schwarz inequality, this is largest when n is parallel to ∇φ and smallest when n is antiparallel. Thus the gradient is not merely a collection of partial derivatives: it is a geometrical vector that identifies the steepest spatial change. Its magnitude has units of the scalar quantity divided by length. For temperature, it has units K m⁻¹; for electric potential, V m⁻¹.
A surface of constant φ is a level surface. Along a tangent displacement on that surface dφ=0, so ∇φ·d r=0. The gradient is consequently normal to every regular level surface. This fact explains why electric field lines are perpendicular to equipotential surfaces in electrostatics and why heat flows down a temperature gradient in an isotropic conductor. A zero gradient identifies a stationary point, but not necessarily a maximum or minimum; it may be a saddle point. Classification requires second derivatives, commonly assembled into the Hessian matrix. In physical applications, a stationary point can represent equilibrium, an extremum of potential energy, or a steady spatial profile.
17.2 Divergence from flux density
Consider a small rectangular box of sides dx, dy and dz in a vector field A. The net outward flux through the two faces normal to x is approximately [A_x(x+dx/2)−A_x(x−dx/2)]dy dz. Taylor expansion reduces this to (∂A_x/∂x)dx dy dz. Adding the corresponding y and z contributions gives net outward flux ≈(∂A_x/∂x+∂A_y/∂y+∂A_z/∂z)dV. Dividing by dV and taking the limit defines ∇·A. The divergence is therefore the local outward flux per unit volume. Positive divergence indicates a local source-like tendency; negative divergence indicates sink-like behaviour. Zero divergence means no net local production of flux, not that the vector itself is zero.
For an incompressible fluid of constant density, conservation of mass gives ∇·v=0. For a compressible fluid, the correct continuity equation is ∂ρ/∂t+∇·(ρv)=0. In electrostatics, Gauss's law is ∇·E=ρ/ε₀. These equations share a mathematical structure but describe different physical quantities. The divergence theorem follows by summing the flux through small volume elements: internal shared faces cancel, leaving only the outer boundary. This cancellation argument is the geometric reason that a local divergence integrates to total boundary flux.
17.3 Curl and infinitesimal circulation
The curl can be derived by considering circulation around a tiny oriented rectangle. For a rectangle in the xy-plane, the circulation per unit area tends to ∂A_y/∂x−∂A_x/∂y, which is the z-component of curl A. Repeating for the other coordinate planes yields ∇×A. The direction follows the right-hand rule and the magnitude measures the limiting circulation per unit area. In fluid mechanics, curl velocity is vorticity. In electromagnetism, curl E is related to the time rate of change of magnetic flux density. A field can have nonzero magnitude and zero curl; a conservative electrostatic field is an important example.
Two identities are particularly useful. First, ∇×(∇φ)=0 for a sufficiently smooth scalar field: mixed partial derivatives cancel. Second, ∇·(∇×A)=0 for a sufficiently smooth vector field: every mixed derivative appears twice with opposite signs. These identities are not arbitrary algebraic tricks. They express compatibility between scalar potentials, circulation and flux. Singularities, excluded regions or multivalued potentials can complicate global conclusions, so the domain of the field must always be considered.
17.4 Coordinate systems and scale factors
Cartesian coordinates are ideal for rectangular boundaries, but cylindrical coordinates (r,φ,z) suit pipes, wires and circular membranes, while spherical coordinates (r,θ,φ) suit central forces, isolated charges and spherical waves. The differential volume elements are respectively dx dy dz, r dr dφ dz and r² sinθ dr dθ dφ. These factors arise because equal coordinate increments do not correspond to equal physical lengths everywhere. In cylindrical coordinates, a small azimuthal displacement has length r dφ; in spherical coordinates, the azimuthal length is r sinθ dφ. Forgetting these scale factors causes incorrect integrals and differential operators.
For a scalar field in cylindrical coordinates, the Laplacian is ∇²φ=(1/r)∂/∂r(r∂φ/∂r)+(1/r²)∂²φ/∂φ²+∂²φ/∂z². In spherical coordinates it is ∇²φ=(1/r²)∂/∂r(r²∂φ/∂r)+(1/[r² sinθ])∂/∂θ(sinθ∂φ/∂θ)+(1/[r² sin²θ])∂²φ/∂φ². The apparent singular factors at r=0 or sinθ=0 may be coordinate singularities rather than physical divergences. A physically acceptable solution must be regular in actual space, not merely finite in a particular coordinate expression.
18. Fourier series: coefficient derivation, convergence and physical use
18.1 Deriving the coefficients by orthogonality
On −L<x<L, represent a sufficiently well-behaved periodic function as f(x)=a₀/2+Σ[a_n cos(nπx/L)+b_n sin(nπx/L)]. Multiply both sides by cos(mπx/L) and integrate from −L to L. Orthogonality gives zero for cross terms and L for the square of a nonzero cosine mode. The constant term also integrates to zero against cosine. Therefore a_m=(1/L)∫₋Lᴸ f(x)cos(mπx/L)dx. Multiplying by sine instead gives b_m=(1/L)∫₋Lᴸ f(x)sin(mπx/L)dx. Integrating the original expansion gives a₀=(1/L)∫₋Lᴸ f(x)dx. These coefficients are projections of the function onto mutually orthogonal basis functions.
Orthogonality is the mathematical reason each coefficient can be calculated independently. It also explains why Fourier modes are natural eigenfunctions for many boundary-value problems. For a string fixed at x=0 and x=L, sine functions vanish at both ends and form the appropriate spatial basis. For a system with periodic boundary conditions, complex exponentials e^(ikx) are often more convenient. The basis is selected to satisfy the geometry and boundary conditions; using the wrong basis can make a simple problem unnecessarily difficult.
18.2 Piecewise functions and endpoint behaviour
For a piecewise smooth periodic function, the Fourier series converges to the function at points of continuity and to the average of the left and right limits at a jump. At a jump from f₋ to f₊, the series approaches (f₋+f₊)/2. The Gibbs phenomenon is the persistent overshoot near the discontinuity. Increasing the number of harmonics narrows the region over which the oscillation occurs, but the maximum overshoot approaches a fixed fraction of the jump. This behaviour matters when representing square waves, switching signals and idealised discontinuous boundary data.
Even and odd symmetry can greatly reduce the calculation. If f(−x)=f(x), the product with every sine mode is odd, so b_n=0. If f(−x)=−f(x), the cosine products are odd and a_n=0, including a₀. For a function defined only on 0<x<L, an even extension produces a cosine series and an odd extension produces a sine series. These extensions encode different endpoint behaviour and should be chosen to match the physical boundary conditions.
18.3 Parseval's theorem and mean-square content
Parseval's theorem equates the mean-square value of a periodic function to the sum of the squared Fourier coefficients. For the convention above, (1/L)∫₋Lᴸ f²(x)dx=a₀²/2+Σ(a_n²+b_n²). This is useful for calculating signal power, average energy and the contribution of individual harmonics without reconstructing the waveform. A truncated Fourier series can be assessed by the omitted coefficient energy. If high-frequency coefficients decay slowly, many terms are required for accurate representation, especially near sharp transitions.
18.4 Worked Fourier derivations
18.5 Fourier methods in heat conduction and vibration
For a rod with fixed endpoint temperatures, subtract a steady-state profile to obtain homogeneous boundary conditions. Expand the remaining initial temperature distribution in sine eigenfunctions. Each mode decays as exp[−κ(nπ/L)²t]. High spatial frequencies decay faster because the decay rate is proportional to n². The solution therefore becomes smoother with time. For a vibrating string, each mode oscillates at ω_n=nπc/L and does not decay in the ideal undamped model. A measured string motion can be decomposed into harmonics, and the modal amplitudes reveal how the string was plucked or struck.
In experimental signal processing, finite-duration sampling and finite instrument bandwidth alter the observed spectrum. A window function reduces leakage caused by truncating a signal, but changes amplitude and spectral width. Sampling below twice the highest frequency of interest risks aliasing; an anti-alias filter should remove frequencies above the usable Nyquist range before digitisation. A plotted spectrum must state sampling frequency, window, amplitude normalisation and whether the frequency axis is in hertz or radians per second.
19. Fourier and Laplace transforms: properties and derivations
19.1 Translation, differentiation and convolution
With F(k)=∫f(x)e^(−ikx)dx, the transform of f′ is ikF(k), provided boundary terms vanish or are handled distributionally. This follows from integration by parts: ∫f′e^(−ikx)dx=[fe^(−ikx)]₋∞∞+ik∫fe^(−ikx)dx. A spatial translation f(x−a) transforms to e^(−ika)F(k). Multiplication by x in the original domain corresponds to i dF/dk in transform space. These properties turn differential equations into algebraic or ordinary differential equations in the transformed variable.
The convolution of f and g is (f*g)(x)=∫f(ξ)g(x−ξ)dξ. Its Fourier transform is F(k)G(k) under the stated convention. Conversely, multiplication in x corresponds to a scaled convolution in k. This is the basis of linear filtering: an impulse response h(t) convolved with an input produces the output, and in frequency space the output spectrum is the product of input spectrum and transfer function. The convention and any 2π factors must be tracked carefully.
19.2 Laplace transform of derivatives
For the unilateral transform F(s)=∫₀∞f(t)e^(−st)dt, integration by parts yields L{f′}=sF(s)−f(0⁺). Applying the same operation again gives L{f″}=s²F(s)−sf(0⁺)−f′(0⁺). Initial values therefore enter directly into the algebraic equation. For a damped oscillator m x″+c x′+kx=F(t), transformation gives [ms²+cs+k]X(s)−m[sx(0)+v(0)]−cx(0)=F(s). Solving for X(s) includes both the driven response and the natural response determined by initial displacement and velocity.
The characteristic polynomial ms²+cs+k determines poles. For positive m,c,k, both poles have negative real parts, giving decaying natural motion. Underdamping occurs when c²<4mk; critical damping when c²=4mk; overdamping when c²>4mk. The pole structure reveals these regimes without first returning to the time domain. However, pole interpretation must include the region of convergence and the causal assumptions of the model.
19.3 Transfer functions and frequency response
For a linear time-invariant system with zero initial conditions, the transfer function is H(s)=Y(s)/X(s), where X and Y denote input and output transforms. Evaluating H(iω) gives the steady sinusoidal frequency response when the system is stable and the convention is appropriate. Its magnitude gives amplitude gain and its argument gives phase shift. Resonance appears as a peak in magnitude, although damping prevents an infinite response in a physical oscillator. Bode plots display magnitude and phase against logarithmic frequency and are useful for circuits, sensors and control systems.
20. Complex variables: Cauchy theory, Laurent series and residues
20.1 Cauchy–Riemann equations in polar coordinates
Write z=re^(iθ) and f(z)=u(r,θ)+iv(r,θ). The Cauchy–Riemann equations become u_r=(1/r)v_θ and v_r=−(1/r)u_θ for r>0. These relations express the compatibility required for a single complex derivative independent of direction. They are coordinate representations of the same analytic condition, not additional physical laws. Analyticity implies harmonic real and imaginary parts: ∇²u=0 and ∇²v=0. Thus two-dimensional electrostatic potential problems can be linked to complex analytic functions in source-free regions.
If f=u+iv is analytic, curves u=constant and v=constant intersect orthogonally wherever the derivative is nonzero. This provides a useful graphical construction for two-dimensional potential flow and electrostatics. The complex potential combines a velocity potential and stream function, or an electric potential and its harmonic conjugate. Boundary conditions can sometimes be transformed into simple geometric curves in the complex plane, where conformal maps preserve local angles.
20.2 Cauchy's integral formula
If f is analytic on and inside a simple positively oriented contour C, then f(z₀)=(1/[2πi])∮C f(z)/(z−z₀) dz for z₀ inside C. Differentiating with respect to z₀ gives formulae for derivatives. This result means that values throughout an analytic region are constrained by boundary values. It also explains why analytic functions are infinitely differentiable within their domains and why their Taylor coefficients are controlled by contour integrals.
20.3 Laurent expansion and classification of singularities
Near an isolated singularity z₀, a function can be expanded as Σ from n=−∞ to ∞ a_n(z−z₀)^n. If all negative-power coefficients vanish, the singularity is removable. If there are finitely many negative powers, it is a pole; the largest negative power determines its order. If infinitely many negative powers occur, the singularity is essential. The coefficient a_−1 is the residue. A simple pole is often evaluated by a limit; a pole of order m can be evaluated using a derivative of order m−1.
The residue theorem is powerful because the integral around a contour depends only on enclosed residues, not on the detailed path, provided the function is meromorphic in the relevant region and the contour avoids singularities. For real integrals, one chooses a semicircular or keyhole contour according to the integrand's decay, branch structure and pole locations. The contribution from the large arc must be justified by an estimate; it must not simply be discarded because the arc looks large.
20.4 Branch cuts and logarithms
The complex logarithm is multivalued because z=re^(iθ) can use θ+2πn. A branch is selected by restricting the argument to an interval, commonly (−π,π), with a cut along the negative real axis. Fractional powers inherit this ambiguity. Contour integration around a branch point requires a keyhole contour or another contour that accounts for both sides of the cut. State the chosen branch before evaluating a complex power or logarithm.
21. Sturm–Liouville theory, orthogonality and normalisation
Consider the regular Sturm–Liouville equation −(p y′)′+q y=λw y on a finite interval, with real coefficient functions, p>0 and w>0, and separated self-adjoint boundary conditions. Let y_m and y_n correspond to distinct eigenvalues λ_m and λ_n. Multiply the equation for y_m by y_n and the equation for y_n by y_m, subtract and integrate. The q terms cancel. The remaining left-hand side is a total derivative: [p(y_m y_n′−y_n y_m′)] at the endpoints. Self-adjoint boundary conditions make this boundary term zero. Thus (λ_m−λ_n)∫w y_m y_n dx=0. If λ_m≠λ_n, the weighted integral must vanish. This is weighted orthogonality.
Orthogonality allows a function to be expanded as y(x)=Σc_n y_n(x), with c_n determined by projection: c_n=∫w f y_n dx/∫w y_n² dx. The denominator is the norm squared. A normalised eigenfunction has unit weighted norm. In quantum mechanics, normalisation is tied to probability; in classical field problems it provides a convenient amplitude convention. Degenerate eigenvalues require an orthogonal basis to be chosen within the degenerate subspace, for example by Gram–Schmidt orthogonalisation.
Boundary conditions determine the spectrum. A fixed string has zero displacement at each endpoint and sine modes. A free-end string has derivative conditions and a different family of modes. A membrane has a two-dimensional eigenvalue problem whose spectrum depends on shape. For a circular membrane, Bessel functions appear because the radial equation in polar coordinates is Bessel's equation. The zeros of the appropriate Bessel function determine allowed radial wavenumbers. The physical mode frequencies then follow from the wave speed and geometry.
21.1 Completeness and limitations
Orthogonality alone does not guarantee that every admissible function can be represented by the eigenfunctions. Completeness is a separate property. For regular self-adjoint Sturm–Liouville problems, the eigenfunctions form a complete basis in the relevant weighted function space. Singular endpoints and non-self-adjoint operators require more careful analysis. In applied calculations, a truncated expansion is an approximation; convergence depends on regularity, boundary compatibility and the norm used to assess error.
22. Partial differential equations: derivation from conservation laws
22.1 Diffusion equation from conservation of energy
Take a uniform rod of cross-sectional area A and density ρ, specific heat c and thermal conductivity k. Let T(x,t) be temperature. Fourier's law gives heat flux q=−k∂T/∂x. For a slice between x and x+dx, the net rate of heat entering is A[q(x)−q(x+dx)]≈−A(∂q/∂x)dx. The thermal energy change per unit time is ρcA dx ∂T/∂t. Equating the two gives ρc∂T/∂t=−∂q/∂x=k∂²T/∂x² for constant k. Hence ∂T/∂t=κ∂²T/∂x², where κ=k/(ρc) is thermal diffusivity. Its units are m² s⁻¹. This derivation assumes one-dimensional conduction, constant material properties, no internal heat source and negligible transverse gradients.
If there is volumetric heating Q̇, the equation becomes ρc∂T/∂t=k∂²T/∂x²+Q̇. If k varies with temperature or position, the conduction term is ∂/∂x(k∂T/∂x), not simply kT_xx. This distinction is important in composite materials and strongly temperature-dependent systems. Boundary conditions may prescribe temperature (Dirichlet), heat flux (Neumann), or a convective exchange law (Robin). Initial temperature T(x,0) is also required because the equation describes time evolution.
22.2 Wave equation from Newton's second law
For a taut string with tension τ and linear mass density μ, consider a small element dx. At small transverse slopes, the vertical force is τ[∂y/∂x at x+dx−∂y/∂x at x]≈τ y_xx dx. The mass is μdx and transverse acceleration is y_tt. Newton's second law gives μdx y_tt=τ y_xx dx, so y_tt=c²y_xx with c=√(τ/μ). The small-slope approximation permits sinθ≈tanθ≈∂y/∂x. Large-amplitude motion requires a more complete nonlinear treatment.
The wave equation differs fundamentally from diffusion. Its solutions carry disturbances at finite speed c and preserve wave-like features in the ideal nondispersive model. The diffusion equation smooths gradients and has no finite propagation front in its classical mathematical formulation. The difference is reflected in PDE classification: hyperbolic versus parabolic. Physical diffusion at extremely short time scales may require non-Fourier models, because classical Fourier conduction predicts instantaneous influence at arbitrarily large distances.
22.3 Laplace equation and uniqueness
In a source-free electrostatic region, ∇·E=0 and E=−∇V, so ∇²V=0. This is Laplace's equation. With charge density ρ, Poisson's equation is ∇²V=−ρ/ε₀. A solution to Laplace's equation cannot have a strict interior maximum or minimum unless it is constant, under the usual regularity assumptions. This maximum principle supports uniqueness: if two solutions have the same Dirichlet boundary values, their difference is harmonic and zero on the boundary, so it must be zero throughout the region. Thus the boundary-value problem has at most one solution.
22.4 Separation of variables for the heat equation
Let u_t=κu_xx on 0<x<L with u(0,t)=u(L,t)=0. Set u=X(x)T(t). Dividing by κXT gives T′/(κT)=X″/X=−λ. The spatial equation is X″+λX=0; homogeneous endpoint conditions give λ_n=(nπ/L)² and X_n=sin(nπx/L). The temporal equation is T′+κλ_nT=0, so T_n=e^(−κλ_nt). Superposition gives u(x,t)=Σb_n sin(nπx/L)e^(−κn²π²t/L²). Coefficients are determined from the initial profile u(x,0)=f(x) by b_n=(2/L)∫₀ᴸf(x)sin(nπx/L)dx.
23. Green functions: construction, jump conditions and examples
For a one-dimensional operator L=−d²/dx² on 0<x<L with Dirichlet conditions, define G(x,ξ) by −∂²G/∂x²=δ(x−ξ), G(0,ξ)=G(L,ξ)=0. Away from x=ξ, the delta function is zero, so G is linear on each side: G=A x for x<ξ and G=B(L−x) for x>ξ. Continuity at ξ requires Aξ=B(L−ξ). Integrate the defining equation from ξ−ε to ξ+ε: −[G_x]₋₊=1, hence G_x(ξ⁺)−G_x(ξ⁻)=−1. Solving gives G(x,ξ)=x(L−ξ)/L for x≤ξ and ξ(L−x)/L for x≥ξ.
For a source f(x), the solution is y(x)=∫₀ᴸG(x,ξ)f(ξ)dξ. The kernel weights the influence of a source at ξ on the response at x. Its symmetry G(x,ξ)=G(ξ,x) follows from self-adjointness and the chosen boundary conditions. If the boundary values are nonzero, first construct a function satisfying those boundary values and subtract it, leaving a homogeneous-boundary problem for the Green-function integral. This separation avoids confusing source response with boundary forcing.
In electrostatics, a Green function represents the potential due to a point charge in a specified geometry. In wave theory, the retarded Green function encodes causal propagation. In quantum mechanics, Green operators are related to resolvents and propagators. The kernel always depends on the operator, geometry, boundary conditions and, for time-dependent problems, causal prescription. Reusing a Green function from a different domain without checking these features is a serious modelling error.
23.1 Green functions and impulse response
For a time-invariant linear system, the impulse response h(t) is the output produced by a unit impulse. Any sufficiently well-behaved input x(t) can be represented as a superposition of shifted impulses, giving y(t)=∫h(t−τ)x(τ)dτ. This is convolution. Causality requires h(t)=0 for t<0. The Fourier or Laplace transform of h is the transfer function. The same mathematical idea appears in spatial boundary-value problems: a distributed source is assembled from point-source responses.
24. Special functions: Bessel, Legendre, Hermite and Laguerre equations
24.1 Bessel functions in cylindrical geometry
Separation of the two-dimensional Helmholtz equation in polar coordinates, ∇²ψ+k²ψ=0, with ψ(r,θ)=R(r)Θ(θ), gives Θ″+m²Θ=0 and r²R″+rR′+(k²r²−m²)R=0. The radial equation is Bessel's equation of order m. Its independent solutions are J_m(kr) and Y_m(kr). The Y_m solution is singular at r=0 for integer m≥0, so a solution regular at the centre usually retains J_m. Circular drums, cylindrical waveguides and diffusion in cylinders all produce Bessel functions, although their eigenvalue conditions differ according to the physical boundary.
For a circular membrane of radius a with fixed rim, R(a)=0, so J_m(ka)=0. The allowed radial wavenumbers are k_mn=j_mn/a, where j_mn is the nth positive zero of J_m. The mode frequencies follow from ω=ck. A table of Bessel zeros is often used numerically, but the equation and boundary condition explain why those values occur. Orthogonality relations include a radial weight r, reflecting the polar area element r dr dθ.
24.2 Legendre polynomials in spherical symmetry
Separation of Laplace's equation in spherical coordinates leads to the angular equation (1/sinθ)d/dθ(sinθ dΘ/dθ)+l(l+1)Θ=0 for azimuthally symmetric cases. With μ=cosθ, this becomes d/dμ[(1−μ²)dP/dμ]+l(l+1)P=0. Solutions regular at both poles are Legendre polynomials P_l(μ) for non-negative integer l. The first few are P₀=1, P₁=μ, P₂=(3μ²−1)/2 and P₃=(5μ³−3μ)/2. They describe multipole angular structure: monopole, dipole, quadrupole and higher orders.
Orthogonality on −1≤μ≤1 is ∫₋₁¹P_l(μ)P_m(μ)dμ=2δ_lm/(2l+1). This allows an angular distribution to be expanded in Legendre polynomials. For a point outside a bounded source distribution, the multipole expansion often provides a controlled approximation: the monopole dominates far away, while higher multipoles encode finer structure and decay more rapidly with distance.
24.3 Hermite functions and the oscillator
The one-dimensional quantum harmonic oscillator equation can be written in dimensionless coordinate ξ=x√(mω/ℏ). After factoring out a Gaussian exp(−ξ²/2), the remaining function satisfies Hermite's differential equation. Normalisable solutions require a polynomial of degree n, giving energies E_n=ℏω(n+1/2). The Hermite polynomials begin H₀=1, H₁=2ξ, H₂=4ξ²−2 and H₃=8ξ³−12ξ. Their orthogonality weight is exp(−ξ²). The Gaussian factor ensures normalisability at large |ξ|.
24.4 Associated Laguerre functions and radial problems
After separating the hydrogen atom Schrödinger equation into radial and angular parts, the radial equation can be transformed using a dimensionless radius and an exponential factor. The remaining polynomial solutions are associated Laguerre functions. The quantisation condition arises from requiring a normalisable solution at large radius. This is an example of how mathematical regularity and boundary behaviour generate discrete physical spectra. The associated Laguerre polynomials have a weight involving x^αe^(−x), and their orthogonality relation depends on the index α.
25. Numerical methods for mathematical physics: finite differences, stability and error
25.1 Finite-difference approximations
Expand f(x+h) and f(x−h) in Taylor series about x. Subtracting gives f′(x)≈[f(x+h)−f(x−h)]/(2h), with truncation error O(h²). Adding and subtracting twice f(x) gives f″(x)≈[f(x+h)−2f(x)+f(x−h)]/h², also O(h²). A forward difference f′≈[f(x+h)−f(x)]/h has O(h) error. Central differences are often more accurate for smooth interior points, but one-sided stencils may be needed at boundaries.
Reducing h decreases truncation error but increases sensitivity to floating-point round-off and cancellation. The total error is therefore not guaranteed to decrease indefinitely. A convergence study evaluates the numerical result at successively refined grids and estimates the observed order. If a method is second order, halving h should reduce the leading discretisation error by approximately a factor of four, once the asymptotic regime has been reached. Failure to observe the expected rate can indicate boundary errors, insufficient smoothness or implementation mistakes.
25.2 Explicit heat-equation scheme and stability
For u_t=κu_xx, forward Euler in time and central difference in space gives u_i^{n+1}=u_i^n+r(u_{i+1}^n−2u_i^n+u_{i−1}^n), where r=κΔt/(Δx)². A Fourier-mode stability analysis substitutes an error mode proportional to e^(iqiΔx) and finds amplification factor g=1−4r sin²(qΔx/2). Requiring |g|≤1 for all wavenumbers gives 0≤r≤1/2. This is the familiar stability condition for the one-dimensional explicit scheme. It is not a physical law; it is a numerical restriction of this discretisation.
In two spatial dimensions with equal grid spacing, the corresponding explicit restriction is more stringent, commonly r≤1/4. Implicit schemes such as backward Euler can be unconditionally stable for the linear diffusion equation, but they require solving a system of equations at each time step and still have accuracy constraints. Stability means errors do not grow uncontrollably; accuracy means the computed solution approximates the intended continuum solution. A stable answer can be inaccurate.
25.3 Finite elements and weak forms
Finite-element methods divide a domain into elements and approximate the unknown field using local basis functions. Rather than requiring the differential equation to hold pointwise, a weak formulation multiplies by a test function and integrates. Integration by parts lowers the derivative order and naturally introduces boundary flux terms. This is valuable for complicated geometry, mixed materials and boundary conditions. Mesh refinement should be concentrated where gradients are large, but poor element shape can degrade conditioning and accuracy.
25.4 Conditioning, residuals and verification
A problem is ill-conditioned if small perturbations in input data can produce large changes in the solution. The condition number quantifies sensitivity for a specified norm. Residuals measure how well a computed vector satisfies the discretised equations, but a small residual does not guarantee a small forward error when the matrix is ill-conditioned. Verification asks whether the equations were implemented correctly; validation asks whether the equations adequately represent the real system. These are distinct tasks.
26. Integrated applications across physics
26.1 Electrostatic boundary-value problem
A long cylindrical conductor or coaxial geometry suggests cylindrical coordinates. In a charge-free annular region with cylindrical symmetry, Laplace's equation reduces to (1/r)d/dr(r dV/dr)=0. Integrating once gives rV′=C₁, and again V=C₁ ln r+C₂. Two boundary potentials determine the constants. The radial electric field is E_r=−dV/dr=−C₁/r. This logarithmic potential is specific to cylindrical symmetry and an idealised infinite or sufficiently long geometry. Near the ends of a finite coaxial structure, fringing fields make the one-dimensional radial approximation less accurate.
26.2 Vibrating membrane and modal spectra
A circular membrane is described by a two-dimensional wave equation. Separation into radial and angular parts yields Bessel functions and angular sine/cosine factors. The boundary condition at the rim selects Bessel zeros. Degeneracy can occur because sine and cosine angular modes share the same radial zero. A physical tap excites a combination of modes; the sound spectrum contains peaks at modal frequencies. Damping and nonuniform tension shift and broaden those peaks. A microphone spectrum can be compared with predicted mode ratios, but the instrument response and room acoustics must be considered.
26.3 Diffusion and relaxation spectra
Any initial temperature profile can be decomposed into eigenmodes. The slowest mode dominates at late time because all faster modes decay more rapidly. A semilogarithmic plot of late-time amplitude against time becomes approximately linear, with slope −κπ²/L² for the fixed-end one-dimensional rod. This provides a way to estimate thermal diffusivity. In practice, contact resistance, heat loss from the sides, finite sensor response and uncertain boundary temperatures can bias the estimate. A good experiment varies rod length or uses independent measurements of material properties.
26.4 Quantum wavefunctions and boundary constraints
In quantum mechanics, acceptable wavefunctions are typically single-valued, finite where required, square-integrable for bound states and compatible with boundary conditions. Eigenfunction orthogonality allows states to be expanded in a basis. Fourier transforms connect position and momentum representations. The uncertainty relation is not a statement about poor instruments; it follows from the non-commutation of position and momentum operators. The mathematical methods in this unit therefore provide the language for both classical fields and quantum states, while physical interpretation differs between the theories.
26.5 Signal processing and experimental inference
Experimental time series often contain deterministic signals, random noise, drift and transient events. Fourier analysis can identify periodic components; Laplace methods model transient response; Green functions describe instrument impulse response. Sampling, quantisation and filtering must be reported. A frequency peak alone does not prove a physical resonance: it may arise from electrical interference, aliasing or an instrument artefact. Repeat measurements, alter the system's parameters and check whether the peak shifts as the physical model predicts.
27. Laboratory and computational investigations
Practical A: Fourier analysis of a periodic signal
Aim: determine the harmonic content of a periodic waveform and compare measured amplitudes with a Fourier-series prediction. Apparatus: function generator, oscilloscope or data-acquisition interface, computer with a documented analysis package, connecting leads and suitable termination. Method: generate a square or triangular wave at a safe low voltage; record several complete periods; record sampling frequency and number of samples; remove any DC offset only if the analysis explicitly requires it; apply a stated window; calculate the discrete Fourier transform; convert bin number to frequency; and compare harmonic amplitudes with theory. Avoid clipping and ensure the sample rate comfortably exceeds twice the highest harmonic retained.
Analysis: plot amplitude against frequency and label the fundamental and harmonics. For a square wave, odd harmonics dominate and ideal amplitudes scale as 1/n. For a triangular wave, odd harmonics decay more rapidly, approximately as 1/n². Explain why a finite sample and non-integer number of periods cause spectral leakage. Repeat with a different record length and window. Evaluation: identify noise floor, generator distortion, oscilloscope bandwidth, sample-clock uncertainty and termination effects. Do not report an amplitude spectrum without stating its normalisation convention.
Practical B: thermal diffusion and eigenmode decay
Aim: estimate thermal diffusivity from transient temperature measurements. Apparatus: uniform metal rod or approved teaching apparatus, controlled heater, temperature sensors at documented positions, data logger, thermal insulation, thermometer reference and appropriate heat-resistant gloves. Method: establish safe operating conditions and avoid exposed hot surfaces; record initial temperatures; apply a reproducible heating pulse or initial profile; monitor temperature as the system relaxes; and record sensor positions and response times. If boundary temperatures are held fixed, compare the late-time data with the first eigenmode decay model.
Analysis: plot the logarithm of the fitted first-mode amplitude against time. The gradient is −κπ²/L², so κ=−slope×L²/π². Use the actual effective length between boundary conditions, not simply the visible rod length. Propagate uncertainty from slope and length: for independent uncertainties, (u_κ/κ)²=(u_s/s)²+(2u_L/L)². Evaluation: discuss heat loss, imperfect insulation, contact resistance, nonuniform material and sensor lag. A systematic drift in the late-time slope may indicate that a single-mode approximation has not yet become valid.
Practical C: numerical solution of the one-dimensional wave equation
Aim: investigate how boundary conditions and initial shape determine normal modes. Implement a finite-difference scheme for a string with fixed endpoints. Use a grid in space and time, define initial displacement and velocity, and keep endpoints fixed at zero. Choose the time step to satisfy the stability condition of the selected scheme. Plot displacement against position at several times and compare the numerical oscillation period with f_n=nc/(2L).
Quality checks: verify that endpoints remain fixed, that the initial condition is represented correctly and that the numerical energy behaves consistently with the scheme. Repeat with a finer grid and smaller time step. A numerical movie can look plausible even when phase error accumulates, so compare frequencies quantitatively. Save the code, parameters, version of software and raw data. Safety: keep any physical string apparatus tension within the rated limit and protect eyes from snapping strings or moving components.
Practical D: solving a boundary-value problem with finite differences
Discretise y″=−f(x) on a uniform grid. At interior node i, use (y_(i+1)−2y_i+y_(i−1))/h²=−f_i. This produces a tridiagonal linear system after boundary values are moved to the right-hand side. Solve the system using a stable linear algebra routine and compare against an analytic solution for a test source. Calculate residuals, refine the mesh and estimate convergence order. Include a graph of numerical and exact solutions and a table of maximum absolute error versus grid spacing.
Scientific reporting checklist
- State the physical question, model, assumptions and boundary/initial conditions.
- Identify instrument ranges, calibration status, sampling and software versions.
- Retain raw observations and document every transformation applied to data.
- Show equations before numerical substitution and preserve units.
- Separate random scatter, systematic bias, numerical error and model inadequacy.
- Use labelled axes, meaningful significant figures and uncertainty bars where justified.
- Compare with a limiting case, analytic solution or independent measurement.
- State whether the evidence supports the model within the uncertainty of the investigation.
28. Common mathematical errors and how to avoid them
Confusing a vector with its components: a gradient is a vector, while its Cartesian partial derivatives are components. In a curvilinear coordinate system, the unit vectors vary with position and the scale factors matter. Ignoring orientation: reversing a surface normal changes flux sign; reversing a contour direction changes circulation sign. Draw the normal and use the right-hand rule before integrating.
Dropping boundary terms: integration by parts in Fourier or Laplace transforms produces endpoint terms. They vanish only under justified decay or boundary assumptions. Mixing transform conventions: definitions differ in signs and 2π factors. Write the forward and inverse transform pair at the beginning of a calculation. Misusing Fourier coefficients: the constant coefficient is a₀/2 in the convention used here. A different convention may define it differently.
Assuming every PDE separates: separation requires compatible geometry, coefficients and boundary conditions. Failure to find a product solution does not imply that no solution exists. Other approaches include Green functions, integral transforms, numerical methods and variational formulations. Ignoring initial data: an evolution equation generally needs an initial condition in addition to spatial boundary conditions. Confusing a boundary-value problem with an initial-value problem: the former specifies behaviour at the boundary of a domain; the latter specifies the state at an initial time.
Accepting a formal eigenfunction without checking admissibility: solutions may diverge at the origin, fail periodicity, violate endpoint conditions or be non-normalisable. Assuming an eigenvalue is dimensionless: its units follow from the operator and coordinate. Using a numerical result without convergence testing: a stable computation can still be inaccurate. Interpreting a small residual as proof: conditioning and model validity also matter.
Sign mistakes in physical laws: Fourier heat flux points down the temperature gradient; electrostatic field points down potential; a positive restoring force is opposite displacement. State the coordinate direction and physical convention. Overlooking approximations: small-slope string theory, constant thermal conductivity, ideal boundary conditions and infinite-domain contour limits each have a domain of validity. Quantify whether neglected terms are small relative to retained terms.
29. Extended university examination questions with worked model answers
- Derive the gradient from the total differential of a scalar field.
Model answer: Expand dφ=φ_x dx+φ_y dy+φ_z dz and write d r=(dx,dy,dz). Then dφ=∇φ·d r, with ∇φ=(φ_x,φ_y,φ_z). For a unit direction n, the directional derivative is ∇φ·n. Cauchy–Schwarz shows the maximum is |∇φ| when n is parallel to ∇φ. Therefore the gradient gives both the direction and rate of steepest increase and is normal to a regular level surface.
- Use the divergence theorem to find the flux of A=(x,y,z) through a sphere of radius R.
Model answer: ∇·A=3. The volume is 4πR³/3, so the flux is ∫3dV=4πR³. On the sphere A=Rn and the direct flux is R(4πR²)=4πR³. The agreement verifies orientation and the theorem.
- Explain the physical meaning of curl and state two vector identities.
Model answer: Curl measures local circulation per unit area with direction set by the right-hand rule. For smooth fields, curl(grad φ)=0 and div(curl A)=0. These follow from cancellation of mixed partial derivatives and express compatibility of scalar potentials and vector circulation.
- Derive Fourier coefficients for a function on −L to L.
Model answer: Multiply the Fourier expansion by the relevant sine or cosine and integrate over the interval. Orthogonality removes cross terms. This gives a₀=(1/L)∫f dx, a_n=(1/L)∫f cos(nπx/L)dx and b_n=(1/L)∫f sin(nπx/L)dx.
- Obtain the Fourier series of a periodic odd square wave taking values ±1.
Model answer: Odd symmetry eliminates all cosine coefficients. Integration gives b_n=2[1−(−1)^n]/(nπ), so only odd harmonics remain: f=(4/π)(sin x+sin3x/3+sin5x/5+…). At each jump the series converges to the midpoint zero.
- What is the Gibbs phenomenon and why does it persist?
Model answer: A truncated Fourier series overshoots near a jump. As more terms are included, the oscillatory region narrows but the limiting maximum overshoot remains a fixed fraction of the jump. It is a consequence of representing a discontinuity with a finite sum of smooth global basis functions.
- Show that the Fourier transform of a derivative is ikF(k).
Model answer: Integrate ∫f′(x)e^(−ikx)dx by parts. The boundary term vanishes for a sufficiently decaying f, leaving ik∫f(x)e^(−ikx)dx=ikF(k). State the decay or distributional assumptions.
- Solve y′+ay=0, y(0)=y₀ using a Laplace transform.
Model answer: sY−y₀+aY=0, so Y=y₀/(s+a). Inversion gives y=y₀e^(−at). Substitution confirms the differential equation and initial condition.
- Explain how poles of a transfer function relate to stability.
Model answer: For a causal continuous-time linear system, poles in the left half-plane produce decaying natural modes and indicate asymptotic stability under standard assumptions. A pole in the right half-plane gives a growing mode. Imaginary-axis poles require a more careful marginal-stability analysis, including multiplicity and input-output definitions.
- State the Cauchy–Riemann equations and explain their significance.
Model answer: For f=u+iv, u_x=v_y and u_y=−v_x. With appropriate continuity of first partial derivatives, these conditions ensure complex differentiability. They also imply that u and v are harmonic in a regular analytic domain.
- Evaluate the contour integral of e^z/(z−a)² around a positive contour enclosing a.
Model answer: The residue at the second-order pole is d(e^z)/dz at a, equal to e^a. The integral is 2πi e^a by the residue theorem.
- Derive orthogonality of eigenfunctions of a self-adjoint Sturm–Liouville operator.
Model answer: Multiply the two eigenvalue equations by the opposite eigenfunction, subtract and integrate. The differential part becomes a boundary term that vanishes under self-adjoint conditions. Thus (λ_m−λ_n)∫w y_m y_n dx=0; distinct eigenvalues imply weighted orthogonality.
- Derive the one-dimensional heat equation from Fourier's law.
Model answer: Heat flux q=−kT_x. Conservation gives ρcT_t=−q_x=kT_xx for constant k and no internal source. Hence T_t=κT_xx with κ=k/(ρc), dimensions m² s⁻¹.
- Derive the wave speed on a stretched string.
Model answer: For a small string element, the net transverse tension is τy_xx dx. Its mass is μdx and acceleration is y_tt. Newton's law gives μy_tt=τy_xx, or y_tt=c²y_xx with c=√(τ/μ).
- Classify the wave, diffusion and Laplace equations and interpret the classes.
Model answer: The wave equation is hyperbolic and describes finite-speed propagation; diffusion is parabolic and describes smoothing/relaxation; Laplace's equation is elliptic and describes static equilibrium fields. Classification guides data requirements and solution methods.
- Explain the role of boundary and initial conditions in separation of variables.
Model answer: Boundary conditions select admissible spatial eigenfunctions and eigenvalues. Initial conditions determine the expansion coefficients for the state at the initial time. Both are needed for a well-posed evolution problem.
- Construct the Green function for −y″=δ(x−ξ) with zero endpoint values.
Model answer: Use linear solutions on either side, continuity at ξ and derivative jump G′(ξ⁺)−G′(ξ⁻)=−1. The result is x(L−ξ)/L for x≤ξ and ξ(L−x)/L for x≥ξ.
- Why do Bessel functions occur in circular membranes?
Model answer: The radial part of the Helmholtz equation in polar coordinates is Bessel's differential equation. Regularity at r=0 selects J_m, while the fixed rim condition selects its zeros, which determine allowed radial wavenumbers and mode frequencies.
- Derive the explicit finite-difference stability parameter for the heat equation.
Model answer: Substitution of a Fourier error mode into the explicit update yields g=1−4r sin²(qΔx/2), where r=κΔt/Δx². Requiring |g|≤1 for all modes gives r≤1/2 in one dimension.
- Distinguish numerical verification from physical validation.
Model answer: Verification checks that the mathematical model is implemented and solved correctly, using tests, residuals and convergence. Validation assesses whether the model represents the physical system adequately, by comparison with observations and uncertainty.
- A heat-decay experiment gives a fitted slope of −0.0030 s⁻¹ for a rod of length 0.40 m. Estimate κ.
Model answer: For the first fixed-end mode, slope=−κπ²/L². Thus κ=0.0030×0.40²/π²=4.86×10⁻⁵ m² s⁻¹. Report uncertainty from the fitted slope and effective length, and verify that the first-mode regime applies.
- Explain why the coordinate singularity at r=0 does not always imply a physical singularity.
Model answer: Cylindrical and spherical coordinates are degenerate at the origin or axis. Terms such as 1/r can appear because the coordinate scale factors vanish. A physical field may remain smooth when expressed in Cartesian components. Test scalar invariants and the limit from different directions.
- Describe how a Fourier spectrum can be corrupted by aliasing.
Model answer: Sampling maps frequencies separated by integer multiples of the sampling frequency onto indistinguishable discrete frequencies. Components above the Nyquist frequency can appear as false lower-frequency components. Use an analogue anti-alias filter and sample at a rate comfortably above twice the maximum retained frequency.
- Compare finite-difference and finite-element approaches.
Model answer: Finite differences approximate derivatives on a grid and are straightforward on regular domains. Finite elements use local basis functions and a weak form, making complex geometry and material interfaces more manageable. Both require mesh refinement, stability/conditioning checks and validation.
- Explain why a small residual does not guarantee a small solution error.
Model answer: In an ill-conditioned system, a small perturbation or residual can correspond to a large forward error. Assess conditioning, discretisation error, round-off, boundary implementation and comparison with a trusted solution.
30. Formula sheet, notation and final revision map
Vector calculus: ∇φ=(φ_x,φ_y,φ_z); ∇·A=A_x,x+A_y,y+A_z,z; ∇×A=(A_z,y−A_y,z, A_x,z−A_z,x, A_y,x−A_x,y); ∇²φ=φ_xx+φ_yy+φ_zz. Divergence theorem: ∮A·dS=∫∇·A dV. Stokes theorem: ∮A·dl=∫(∇×A)·dS.
Fourier series: f=a₀/2+Σ[a_n cos(nπx/L)+b_n sin(nπx/L)], a_n=(1/L)∫₋Lᴸf cos(nπx/L)dx, b_n=(1/L)∫₋Lᴸf sin(nπx/L)dx. Fourier transform convention: F(k)=∫f(x)e^(−ikx)dx, f(x)=(1/2π)∫F(k)e^(ikx)dk.
Laplace transform: F(s)=∫₀∞f(t)e^(−st)dt; L{f′}=sF−f(0⁺); L{f″}=s²F−sf(0⁺)−f′(0⁺). Convolution in time corresponds to multiplication of transforms.
Complex analysis: Cauchy–Riemann u_x=v_y and u_y=−v_x; residue theorem ∮f dz=2πiΣRes. For a pole of order m, Res(f,a)=(1/(m−1)!)lim(z→a)d^(m−1)/dz^(m−1)[(z−a)^m f(z)].
Sturm–Liouville: −(py′)′+qy=λwy; distinct eigenfunctions are orthogonal with weight w under self-adjoint boundary conditions. Diffusion: u_t=κu_xx. Wave: u_tt=c²u_xx. Laplace: ∇²V=0; Poisson: ∇²V=−ρ/ε₀. Green function: LG(x,ξ)=δ(x−ξ).
Numerics: central first derivative error O(h²); central second derivative error O(h²); explicit 1D diffusion stability parameter r=κΔt/h²≤1/2. Always distinguish discretisation error, round-off, conditioning and model error.
Revision strategy
First reproduce the derivations of gradient, divergence, Fourier coefficients, Laplace derivative transforms, Sturm–Liouville orthogonality, heat diffusion and string wave speed without looking at the notes. Next solve one boundary-value problem in Cartesian coordinates and one in cylindrical or spherical coordinates. Then calculate a Fourier series for an odd and an even function. Finally, solve a numerical problem at three grid resolutions and discuss convergence. For every answer, define notation, state assumptions, show the algebra, include units where physical quantities occur and interpret the result in words.
Mathematical methods are transferable tools. The same operator may describe temperature, electric potential or a quantum amplitude, but the physical meaning, units and boundary conditions differ. The strongest solutions make this distinction explicit. An equation is not fully interpreted until the variables, domain, initial data, boundary data, approximation and observable consequences have been identified.