Unit guide and mathematical standard
These notes are intended for undergraduate study. Work through the derivations before memorising results. Unless stated otherwise, use SI units, state assumptions, retain units through calculations and round final values to a precision justified by the data. Check limiting cases and dimensions as part of every solution.
Method for numerical problems: identify known and unknown quantities; draw a labelled diagram where useful; select a physical model; state assumptions; derive or rearrange the governing relation; substitute values with units; calculate; then interpret and check the answer.
1. Physical quantities, dimensions and units
A physical equation must be dimensionally homogeneous: every term added or equated must have the same dimensions. The SI base dimensions are mass [M], length [L], time [T], electric current [I], temperature [Θ], amount [N] and luminous intensity [J]. Velocity has dimensions [LT⁻¹], acceleration [LT⁻²], force [MLT⁻²], energy [ML²T⁻²] and pressure [ML⁻¹T⁻²]. Dimensional analysis detects impossible equations and helps construct scaling laws, but cannot determine dimensionless constants such as 1/2 or 2π.
2. Vectors and components
A vector A in Cartesian coordinates is A = Aₓi + Aᵧj + A_zk. Its magnitude is |A| = √(Aₓ²+Aᵧ²+A_z²). For a vector of magnitude A making angles θ and φ in a plane, components are Aₓ=A cosθ and Aᵧ=A sinθ. The dot product A·B=AB cosθ gives a scalar and is useful for work; the cross product A×B has magnitude AB sinθ and gives an oriented vector, as in torque.
3. Differential calculus in physics
The derivative dx/dt is the instantaneous rate of change of x. If position is x(t), velocity v=dx/dt and acceleration a=d²x/dt². For a scalar field f(x,y,z), the gradient ∇f points in the direction of greatest increase. Differentiation rules include the product rule d(uv)/dx=u'v+uv' and chain rule df/dx=(df/du)(du/dx).
4. Integral calculus and physical accumulation
A definite integral ∫ₐᵇ f(x)dx accumulates a quantity over an interval. If force varies with position, work is W=∫ F(x)dx. If power varies with time, energy transferred is E=∫P(t)dt. Integration constants are fixed by initial or boundary conditions.
5. First-order differential equations
A separable equation dy/dx=g(x)h(y) can be rearranged as dy/h(y)=g(x)dx and integrated. Exponential relaxation obeys dN/dt=−kN, giving N=N₀e^(−kt). The constant k has units of inverse time, and the time constant is τ=1/k.
6. Complex numbers and oscillatory physics
A complex number is z=x+iy, with modulus r=√(x²+y²) and argument θ=tan⁻¹(y/x), with quadrant chosen correctly. Euler's identity e^(iθ)=cosθ+i sinθ allows sinusoidal signals to be represented compactly. The real physical signal is obtained by taking the real part of a complex representation.
7. Worked calculation: vector resultant
Let A=(3i+4j) m and B=(−i+2j) m. Then R=A+B=(2i+6j) m. Its magnitude is |R|=√(2²+6²)=√40=6.32 m. Its direction relative to +x is θ=tan⁻¹(6/2)=71.6°. The components are added algebraically before calculating magnitude.
8. Worked calculation: derivative and acceleration
A particle moves as x(t)=4t³−3t²+2 m, where x is in metres and t in seconds. Differentiate: v=dx/dt=12t²−6t m s⁻¹; a=dv/dt=24t−6 m s⁻². At t=2 s, v=48−12=36 m s⁻¹ and a=48−6=42 m s⁻².
9. Worked calculation: exponential decay
A sensor response follows N(t)=N₀e^(−0.20t), with t in seconds. At t=5.0 s, N/N₀=e^(−1)=0.3679. Thus 36.8% of the initial response remains. The time constant is τ=1/0.20=5.0 s; after one time constant, the response is 1/e of its initial value.
10. Laboratory and computational application
For a pendulum experiment, measure length L and period T for several trials. The model T=2π√(L/g) gives T²=(4π²/g)L. Plot T² against L; the gradient m=4π²/g, hence g=4π²/m. Use repeated timings, uncertainty in length and period, a best-fit line and residual inspection. The small-angle approximation sinθ≈θ (θ in radians) is essential.
Inside diagrams: labelled physical models
Use these diagrams together with the equations. Arrows show vector directions or energy/heat transfer; labels identify the quantities used in the mathematical models. Diagrams are schematic and not to scale.
Expanded concept-by-concept explanations and worked examples
Every concept below is followed immediately by a worked numerical example. Calculations use SI units and state the model assumptions.
1. Dimensions and dimensional homogeneity
For T=2π√(L/g), [L/g]=L/(LT⁻²)=T²; the square root has dimension T. The period expression is dimensionally consistent. Dimensional consistency is necessary, but does not prove an equation correct.
2. Cartesian vectors and components
A 50 N force acts 30° above +x. Fₓ=50cos30°=43.3 N; Fᵧ=50sin30°=25.0 N. Recombining gives √(43.3²+25.0²)=50.0 N.
3. Dot product and work
A 12 N force moves a body 5.0 m at 60° to displacement. W=Fs cosθ=12×5×0.5=30 J. The perpendicular force component does no work.
4. Cross product and torque
A 20 N force acts perpendicular to a 0.30 m spanner. τ=rF=6.0 N m. At 30° to the spanner, τ=0.30×20×sin30°=3.0 N m.
5. Differentiation: velocity and acceleration
For x=2t³−5t²+4t m, v=6t²−10t+4 and a=12t−10. At t=2 s, v=8 m s⁻¹ and a=14 m s⁻².
6. Integration: displacement and work
If v=3t² m s⁻¹, displacement from 0 to 2 s is ∫₀²3t²dt=[t³]₀²=8.0 m.
7. Partial derivatives and gradients
For T=4x²+3y, ∇T=8xi+3j. At x=2, ∇T=16i+3j; this vector points toward the steepest temperature increase.
8. First-order differential equations
dN/dt=−0.15N gives N=N₀e^(−0.15t). At 4 s, N/N₀=e^(−0.6)=0.549. Half-life is ln2/0.15=4.62 s.
9. Complex numbers and Euler representation
(3+4i)(2−i)=10+5i. Its modulus is √125=11.18 and argument tan⁻¹(0.5)=26.6°.
10. Uncertainty propagation
For A=lw, l=2.00±0.01 m and w=1.00±0.02 m, A=2.00 m². Relative uncertainty=√[(0.01/2)²+(0.02/1)²]=0.0206, so A=2.00±0.04 m².
Structured revision questions with model answers
- State the central physical principle behind the unit and give one condition under which a commonly used model fails.
Model answer: State the relevant conservation law or governing equation, define its variables and identify assumptions such as negligible friction, small oscillations, ideal-gas behaviour or a uniform field. Explain how relaxing the assumption changes the model.
- Starting from a fundamental law, derive one equation used in this unit. Define every symbol and state the boundary or initial conditions.
Model answer: Begin with the stated law, substitute the constitutive relation, separate variables or integrate as appropriate, apply the initial/boundary conditions, and check dimensions.
- A measured quantity is reported with uncertainty. Explain how the uncertainty should be propagated through a calculated result.
Model answer: For independent uncertainties, use first-order propagation: σ_f²≈Σ(∂f/∂xᵢ)²σᵢ². For products and powers, fractional uncertainties combine in quadrature after multiplying each fractional uncertainty by the absolute power.
- Design a laboratory experiment to test one relationship from this unit. Identify apparatus, independent and dependent variables, controls, safety issues and graph expected.
Model answer: Describe a repeatable method, collect repeated measurements over a useful range, control confounding variables, plot the linearised relationship, obtain the physical parameter from the gradient and report uncertainty and limitations.
- Explain how the relevant theory connects to at least one other undergraduate Physics unit.
Model answer: Identify the shared mathematical structure or conservation principle and explain how the variables or physical interpretation differ in the second context.
Textbook expansion: mathematical foundations for physical reasoning
How to use this chapter. Mathematical methods are not separate from Physics: they are the language used to express measurable relationships, predict motion, quantify fields and compare theory with experiment. For every model, identify the physical system, define coordinates and sign conventions, state assumptions, choose a mathematical representation, solve it, interpret the result and test whether the result is physically plausible. Keep units attached to numerical quantities throughout the calculation.
11. Coordinate systems, position vectors and changes of reference frame
A coordinate system assigns numbers to the location of an event or object. In a Cartesian frame, a point in three-dimensional space is represented by r = xi + yj + zk, where the unit vectors identify mutually perpendicular axes. The coordinate values depend on the chosen origin and orientation, but the physical separation between two points does not. If A and B have position vectors rA and rB, the displacement from A to B is Δr = rB − rA. Its magnitude is the straight-line distance, while its direction describes the orientation of the change.
Changing the origin translates all position vectors by the same constant vector. Rotating the axes changes their components but not scalar quantities such as distance, speed or the dot product of two geometrical vectors. This distinction is essential: components are coordinate descriptions; vectors represent geometrical or physical quantities. In a rotating or accelerating frame, additional inertial terms may be required in the equations of motion, so a frame should be stated explicitly.
12. Vector algebra, projections and physical components
Vector addition follows the parallelogram rule and is independent of the order of addition. Scalar multiplication changes magnitude and may reverse direction if the scalar is negative. The component of vector A along a unit direction n is A·n. The corresponding projection vector is (A·n)n. This operation is useful when resolving a force along an inclined plane, extracting the component of velocity along a track, or determining the part of a field measured by a directional sensor.
For A = Axi + Ayj + Azk and B defined similarly, A·B = AxBx + AyBy + AzBz. Geometrically it is AB cosθ. A positive dot product indicates an angle below 90°, a zero dot product indicates perpendicular vectors, and a negative dot product indicates an obtuse angle. Do not confuse the dot product with multiplication of magnitudes: direction matters.
13. Cross products, orientation and torque
The cross product A×B is a vector perpendicular to the plane containing A and B. Its magnitude is AB sinθ and its direction follows the right-hand rule. In Cartesian components, A×B = (AyBz−AzBy)i + (AzBx−AxBz)j + (AxBy−AyBx)k. The operation is anti-commutative: A×B = −B×A. It appears in torque τ = r×F, angular momentum L = r×p and magnetic force F = qv×B.
The lever arm is the perpendicular distance between the pivot and the line of action of the force. Therefore τ = rF sinθ = Fd⊥. A force directed through the pivot produces no torque, even if its magnitude is large. A force perpendicular to the radius vector produces maximum torque for a fixed r and F.
14. Differentiation from first principles and local linear behaviour
The derivative is the limiting rate of change of a function. For y=f(x), f′(x)=limh→0[f(x+h)−f(x)]/h, provided the limit exists. The derivative is the slope of the tangent to the graph and the best local linear approximation: f(x+Δx)≈f(x)+f′(x)Δx for sufficiently small Δx. In mechanics, differentiating position with respect to time gives velocity; differentiating velocity gives acceleration. In thermodynamics and electromagnetism, derivatives describe response coefficients and spatial or temporal variation.
For a power law y=Cxⁿ, differentiation gives dy/dx=nCxⁿ⁻¹. The product rule, quotient rule and chain rule are not merely algebraic shortcuts; they encode how combined dependencies change. If y=f(g(x)), then dy/dx=(df/dg)(dg/dx). Always identify the independent variable. A derivative with respect to time has different dimensions from a derivative with respect to distance.
15. Second derivatives, curvature and extrema
The second derivative measures the rate of change of the first derivative. For a position coordinate x(t), d²x/dt² is acceleration. For a graph y(x), the sign of y″ indicates local concavity: positive curvature corresponds to a locally convex graph and negative curvature to a locally concave graph. A stationary point satisfies y′=0. If y″ is positive there, the point is a local minimum; if y″ is negative, it is a local maximum. If y″=0, the test is inconclusive and higher derivatives or a sign analysis are needed.
In physical optimisation, extrema must be interpreted within constraints. A stationary value may be unstable, inaccessible or outside the domain of the model. For example, a calculated minimum energy is meaningful only if the system can occupy the corresponding configuration and the assumed potential remains valid there.
16. Definite integrals, signed area and accumulation
A definite integral ∫abf(x)dx is the limit of a sum of contributions f(xi)Δx. It represents signed area for a plotted function and accumulated physical quantity when f is a density or rate. If v(t) is velocity, integrating it over time gives displacement, not necessarily total distance. If power P(t) is integrated over time, the result is energy transferred. If force F(x) is integrated along a path, the result is work for a one-dimensional conservative force component.
The fundamental theorem of calculus links differentiation and integration: if F′(x)=f(x), then ∫abf(x)dx=F(b)−F(a). The limits and variable carry units. A common error is to integrate a rate but omit the differential, or to confuse signed area with the sum of absolute areas.
17. Multivariable functions and partial derivatives
Many physical quantities depend on more than one independent variable. Temperature may be T(x,y,z,t), pressure may depend on volume and temperature, and electric potential may vary throughout space. A partial derivative measures the rate of change with respect to one variable while holding the others fixed. For example, (∂U/∂V)T describes how internal energy changes with volume at fixed temperature. The subscript is important because different experimental constraints can produce different derivatives.
The total differential of f(x,y) is df=(∂f/∂x)dx+(∂f/∂y)dy for small changes. This is the basis of first-order uncertainty propagation and local sensitivity analysis. The gradient ∇f is a vector of partial derivatives; it points in the direction of greatest increase of a scalar field, and its magnitude gives the maximum directional rate of change.
18. Ordinary differential equations: modelling a physical process
A differential equation relates an unknown function to its derivatives. It is a compact statement of a physical law. The equation must be accompanied by initial or boundary conditions to select a particular solution. For a first-order linear equation dy/dt+ay=b, where a and b are constants, the integrating-factor method uses μ(t)=eat. Multiplication gives d(μy)/dt=bμ, which can be integrated directly. If b=0, the solution is exponential decay or growth depending on the sign of a.
For a cooling body in an environment at fixed temperature Te, Newton's law of cooling gives dT/dt=−k(T−Te). Separating variables and integrating yields T−Te=(T₀−Te)e−kt. The model assumes a constant environment, a uniform body temperature and a heat-transfer rate proportional to the temperature difference. Real experiments may depart from these assumptions.
19. Complex numbers, phasors and harmonic signals
A complex number z=a+ib has real part a and imaginary part b, where i²=−1. Its modulus is |z|=√(a²+b²), and its argument is the angle it makes with the positive real axis. Euler's identity eiθ=cosθ+i sinθ gives a compact representation of sinusoidal behaviour. A harmonic signal can be represented as the real part of a complex exponential, for example x(t)=Re[X₀eiωt]. This notation simplifies differentiation because d(eiωt)/dt=iωeiωt.
Phasors encode amplitude and phase. They are particularly useful for sinusoidal steady-state circuits and wave superposition. The physical measured signal is real; the complex representation is a mathematical tool. When taking a final observable, select the real part or magnitude as required by the model. Do not interpret the imaginary component as a second physical displacement.
20. Fourier ideas and why physical signals are decomposed
Fourier analysis represents a suitable periodic signal as a sum of sinusoidal components. A complicated waveform can therefore be studied in terms of its frequencies, amplitudes and phases. The fundamental frequency sets the repetition rate; harmonics occur at integer multiples of that frequency. The method underpins sound analysis, vibration diagnostics, optical spectroscopy, signal processing and the solution of linear differential equations with periodic forcing.
For a periodic function with period T, the angular fundamental frequency is ω₀=2π/T. The Fourier series contains a constant term and sine and cosine terms at nω₀. The coefficients quantify the contribution of each harmonic. A sharp discontinuity requires many harmonics and can produce Gibbs overshoot near the discontinuity. Experimental sampling must also be fast enough to avoid aliasing; the sampling frequency should exceed twice the highest frequency of interest in the ideal Nyquist condition, with practical systems allowing a safety margin and anti-alias filtering.
21. Dimensional analysis, scaling and model testing
Dimensional analysis tests whether an equation is physically homogeneous and can reveal how variables combine. Suppose a pendulum period T depends only on length L and gravitational acceleration g. Assume T∝Lᵃgᵇ. Since [T]=T, [L]=L and [g]=LT⁻², dimensional balance gives T=La+bT−2b. Thus −2b=1 and a+b=0, giving b=−1/2 and a=1/2. Therefore T∝√(L/g). Dimensional analysis cannot determine the dimensionless constant 2π; that requires solving the equation of motion under the small-angle approximation.
Scaling arguments help identify dominant effects. A dimensionless ratio such as Reynolds number compares inertial and viscous effects; a small or large value indicates which contribution is likely to dominate. Such reasoning is useful before attempting a detailed calculation, but it does not replace the assumptions and boundary conditions of a complete model.
22. Numerical methods: roots, integration and differential equations
Many physical equations cannot be solved exactly. Numerical methods approximate a solution and must be accompanied by an error estimate or convergence check. The bisection method brackets a root between values of opposite sign and repeatedly halves the interval. Newton–Raphson iteration uses xn+1=xn−f(xn)/f′(xn), which can converge rapidly near a suitable root but may fail if the derivative is small or the initial guess is poor.
Numerical integration approximates an area using rectangles, trapezia or higher-order quadrature. For a measured data set, the trapezoidal rule is often appropriate: ∫f(x)dx≈Σ[(fᵢ+fᵢ₊₁)/2]Δx. The Euler method advances a differential equation y′=F(t,y) by yn+1=yn+hF(tn,yn). Reducing step size should improve accuracy for a well-behaved problem, but rounding error and model error remain.
23. Measurement uncertainty and significant figures
A measured value is incomplete without an uncertainty and unit. Random uncertainty describes scatter among repeated measurements; systematic uncertainty shifts measurements in a consistent direction. Repetition reduces the standard error of the mean but does not automatically remove calibration error, zero offset, parallax or environmental bias. Record instrument resolution, calibration status, repeat count and the method used to estimate uncertainty.
For independent variables x and y and a derived quantity f(x,y), first-order propagation gives σf²≈(∂f/∂x)²σx²+(∂f/∂y)²σy². Correlated variables require covariance terms. For multiplication and powers, fractional uncertainties are often convenient: if q=xᵃyᵇ, then (σq/q)²≈a²(σx/x)²+b²(σy/y)² for independent inputs.
24. Graphs, linearisation and extraction of physical constants
A graph is an analytical tool, not decoration. Plot the dependent variable vertically against the independent variable horizontally unless there is a reason to choose otherwise. Include variable names, units, uncertainty bars where available and a scale that uses the plotting area well. A fitted gradient and intercept should be interpreted through the theoretical model. Linearisation can reveal a parameter: if y=Ae−kt, plotting ln y against t gives gradient −k, provided the measured quantity is positive and the model assumptions hold.
Do not force a straight line through data that show curvature. Inspect residuals, compare fit quality, and distinguish a statistically good fit from a physically valid model. The intercept may represent a background signal, initial condition or instrument offset. A non-zero intercept should not be discarded without justification.
25. Common mathematical modelling errors and how to prevent them
- Undefined symbols: define each symbol, its unit and sign convention before substitution.
- Inconsistent units: convert to SI units before combining quantities, unless a clearly stated alternative system is used.
- Missing initial conditions: a differential equation's general solution is not a unique physical prediction until conditions are applied.
- Sign errors: draw axes and indicate positive directions before writing vector equations.
- Over-precision: calculator digits do not imply measurement accuracy.
- Unstated assumptions: identify idealisations such as negligible friction, constant acceleration, small angle or uniform field.
- Uninterpreted answers: state what the sign, magnitude and units mean for the physical system.
26. Extended laboratory: determining gravitational acceleration with a pendulum
Objective: estimate local gravitational acceleration by measuring the period of a simple pendulum at several lengths. Apparatus: light inextensible string, dense bob, clamp stand, metre rule, stopwatch or photogate, balance and protractor. Model: for small angular displacement θ, sinθ≈θ and the restoring torque leads to d²θ/dt²+(g/L)θ=0. The angular frequency is ω=√(g/L), so T=2π√(L/g) and T²=(4π²/g)L.
Method: measure the distance from pivot to the bob's centre; use a small initial angle, preferably below about 10°; release without pushing; time 20 complete oscillations; repeat at least three times for each length; and repeat for several lengths. Calculate the mean period T from total time divided by the number of oscillations. Plot T² against L. The gradient m should be 4π²/g, so g=4π²/m.
Data analysis: include uncertainty in length and timing, use the spread of repeated trials to estimate random uncertainty, and determine the uncertainty in the fitted gradient. Propagate the gradient uncertainty using σg/g≈σm/m. Discuss reaction time, pivot friction, finite bob size, string stretch, air resistance and departures from the small-angle approximation. A photogate reduces timing uncertainty but does not correct a length measurement from the wrong reference point.
Safety: secure the clamp stand, keep the swing path clear, use a compact bob and prevent the bob from striking people or equipment. Do not stand beneath a suspended mass while adjusting the clamp.
27. Extended laboratory: testing Hooke's law and uncertainty in a gradient
Objective: determine the spring constant k from the relation F=kx within the elastic limit. Suspend known masses, measure extension from the unloaded reference length, add loads gradually and repeat readings while avoiding parallax. Plot force F against extension x. The gradient is k in N m⁻¹. If the graph curves, the spring may have left its linear elastic range, the zero may have shifted or the measurement may be faulty.
Use a suitable range of loads and do not exceed the spring's safe working limit. Allow oscillations to settle before reading. Repeat each measurement and consider unloading data to check for hysteresis. The intercept should be consistent with the chosen zero within uncertainty. A spring with a non-zero intercept may indicate pre-load or a systematic length offset.
28. Further worked synthesis: choosing the right mathematical tool
Consider a particle whose position is r(t)=(2t²)i+(3t)j+(t³)k m. Differentiate each component to obtain v(t)=4t i+3j+3t²k m s⁻¹ and a(t)=4i+6tk m s⁻². At t=1 s, v=(4,3,3) m s⁻¹, with speed √34=5.83 m s⁻¹. The acceleration is (4,0,6) m s⁻². The scalar speed is the magnitude of velocity; it is not generally equal to the rate of change of a single coordinate.
At t=1 s, the instantaneous rate of change of speed is (v·a)/|v|=(4×4+3×0+3×6)/√34=34/√34=5.83 m s⁻². The component of acceleration perpendicular to velocity changes the direction of motion, while the parallel component changes the speed. This decomposition is a useful bridge to curvilinear motion and classical mechanics.
29. Concept checks for self-assessment
- Why is displacement a vector while distance is a scalar?
Displacement describes the directed change in position; distance is the non-negative path length.
- Can a particle have zero velocity but non-zero acceleration?
Yes. At a turning point in one-dimensional motion, instantaneous velocity can be zero while acceleration remains non-zero.
- What does a negative gradient on a position-time graph mean?
The velocity component along the chosen positive axis is negative.
- Why does dimensional agreement not prove an equation is correct?
Many incorrect equations are dimensionally homogeneous; dimensional analysis cannot usually determine dimensionless constants or the full functional form.
- When is a complex representation useful?
For linear sinusoidal systems, because differentiation, phase shifts and superposition become algebraically simpler.
- Why must a numerical solution be checked for convergence?
A computed value can depend on step size or iteration tolerance; stability under refinement provides evidence that discretisation error is controlled.
30. Connections to the rest of the undergraduate Physics curriculum
Vector calculus is used in classical mechanics to express force, momentum and angular momentum; in electromagnetism it defines divergence, curl and field flux; in fluid mechanics it describes velocity and pressure fields. Differential equations describe oscillators, heat flow, electrical circuits, radioactive decay and wave propagation. Complex numbers simplify AC circuit analysis and quantum amplitudes. Integration connects local quantities to global observables such as work, charge, mass, probability and energy. Numerical methods support systems where exact analytic solutions are unavailable. The methods in this unit should therefore be revisited as each later unit introduces its own physical variables and boundary conditions.
Advanced textbook treatment: mathematical methods in physical science
This extended section develops the reasoning behind the mathematical tools introduced earlier. The objective is not to memorise a collection of formulae, but to understand why a method is appropriate, what assumptions make it valid, how its result should be interpreted, and how to recognise when a calculation has been misapplied. Throughout, a mathematical model is treated as a bridge between a physical question and a testable prediction.
31. Scalars, vectors, tensors and the hierarchy of physical quantities
A scalar is specified by magnitude and unit alone. Mass, time, temperature, energy and electric charge are common examples. A vector requires magnitude, direction and a rule for combining components. Displacement, velocity, acceleration, force, momentum and electric field are vectors. A tensor generalises these ideas by describing a relation between directions in a way that cannot be represented by a single scalar or vector. Stress, strain, moment of inertia and the electromagnetic field in relativistic formulations are examples of tensorial quantities.
The distinction is not simply about how many numbers are written down. A vector remains the same geometrical object when coordinates are rotated, although its components change. A second-rank tensor transforms with two coordinate factors. This transformation property is why tensor notation is useful in continuum mechanics and general relativity: it separates the physical law from the arbitrary coordinate description. In introductory work, Cartesian vector components are often sufficient, but recognising tensor structure prevents the mistaken assumption that every multi-component quantity is a vector.
For example, the moment of inertia of a rigid body depends on the axis of rotation. A single scalar moment of inertia is sufficient for rotation about a specified fixed axis, but a general three-dimensional body is described by an inertia tensor. The tensor links angular velocity to angular momentum and can contain off-diagonal components when the chosen axes are not principal axes.
32. Polar and cylindrical coordinates
Cartesian coordinates are convenient for rectangular geometry, but radial symmetry is often easier to express with polar or cylindrical coordinates. In a plane, a point is described by radius r and angle θ, with x=r cosθ and y=r sinθ. The corresponding unit vectors are er and eθ. Unlike Cartesian unit vectors, these directions vary with position. Their time derivatives are der/dt=θ̇eθ and deθ/dt=−θ̇er.
The position vector is r= r er. Differentiating carefully gives velocity v=ṙer+rθ̇eθ, and acceleration a=(r̈−rθ̇²)er+(rθ̈+2ṙθ̇)eθ. The radial term includes centripetal acceleration; the transverse term includes angular acceleration and a contribution when the radius changes. These expressions are essential for circular motion, orbital mechanics and motion in central-force fields.
33. Coordinate transformations and invariants
A coordinate transformation changes the mathematical description of a point or field. Under a rotation of Cartesian axes through angle θ, the two-dimensional components transform as x′=x cosθ+y sinθ and y′=−x sinθ+y cosθ for one common passive-rotation convention. The signs depend on whether the axes or the vector are rotated, so the convention must be stated. The squared distance x²+y² is unchanged. Such invariant quantities are valuable checks on algebra and numerical code.
In physics, a change of inertial reference frame may also alter measured positions and velocities. In Newtonian mechanics, two frames moving at constant relative velocity are related by Galilean transformations: x′=x−ut and t′=t. Acceleration is unchanged under this transformation, which is consistent with Newton's laws having the same form in all inertial frames. At speeds comparable to the speed of light, Galilean transformations are inadequate and Lorentz transformations are required; this is one reason mathematical methods must be matched to the physical regime.
34. Taylor series and controlled approximations
A sufficiently smooth function can be approximated near x=a by a Taylor series: f(x)=f(a)+f′(a)(x−a)+f″(a)(x−a)²/2!+… . The first term gives the value at the expansion point; the next term gives the local linear change; higher terms correct the approximation as the interval grows. A truncated series is useful only when the neglected terms are small compared with the required accuracy.
Physics frequently uses approximations such as sinθ≈θ, cosθ≈1−θ²/2 and eˣ≈1+x for small θ or x, with angles in radians. The approximation sinθ≈θ is central to the simple pendulum model. It is not an identity: the fractional error grows as the amplitude grows. A good derivation states the small parameter and estimates the size of the neglected term.
35. Linear algebra: matrices, systems and eigenvalues
Many physical problems involve several coupled unknowns. A system of linear equations can be written A x=b, where A is a matrix of coefficients, x is a vector of unknowns and b is a vector of known terms. If A is invertible, x=A⁻¹b, although numerical algorithms such as Gaussian elimination or LU decomposition are usually preferable to explicitly calculating an inverse. A singular matrix indicates that the equations are not independent or that a unique solution does not exist.
Eigenvalues and eigenvectors solve A v=λv. They identify directions that a linear transformation scales without changing direction. In mechanics, eigenvectors of the inertia tensor identify principal axes. In coupled oscillations, eigenvectors describe normal modes and eigenfrequencies. In quantum mechanics, eigenvalues of an observable's operator correspond to possible measurement outcomes in the idealised theory. The same algebra therefore appears in several apparently different areas of Physics.
36. Eigenvalue example: normal modes of two coupled coordinates
For a system of coupled linear equations, assume a trial solution proportional to eiωt. The differential equations become an algebraic eigenvalue problem. The allowed angular frequencies are obtained by setting the determinant of the coefficient matrix to zero. Each eigenvector gives the relative amplitude and phase of the coordinates in a normal mode. This approach is more systematic than guessing separate motions for each mass or oscillator.
For two identical masses coupled by identical springs in a symmetric arrangement, one mode may have the masses moving together while another has them moving oppositely. Symmetry makes the mode shapes easy to recognise. In more complicated systems, matrix methods are used to find the frequencies and mode shapes computationally. The eigenvectors can be normalised so that their squared components have a convenient total, but the physical meaning of the normalisation must be stated.
37. Probability distributions and statistical descriptions
Not every physical prediction is deterministic. In statistical physics and measurement science, outcomes may be described by probability distributions. A probability density p(x) is non-negative and integrates to one over its full domain. The probability that a continuous variable lies between a and b is ∫abp(x)dx. The expectation value is E[x]=∫xp(x)dx, and the variance is E[(x−E[x])²]. The standard deviation is the square root of the variance and measures spread, not the uncertainty of every individual measurement.
For repeated independent measurements drawn from a distribution with finite variance σ², the standard error of the sample mean decreases approximately as σ/√N. This explains why averaging repeated measurements improves precision. It does not eliminate a systematic offset shared by every measurement. When data are strongly non-Gaussian, contain outliers or have small sample size, a mean and standard deviation may not adequately describe the distribution; medians, robust spread measures or an explicit probability model may be preferable.
38. Least-squares fitting and residual analysis
Experimental data rarely lie exactly on a theoretical line. Least-squares fitting selects parameters that minimise the sum of squared residuals, where each residual is the difference between an observed value and the model prediction. For a straight line y=mx+c, the fitted gradient m and intercept c summarise the trend. If measurement uncertainties differ substantially, weighted least squares gives greater influence to more precise observations. The assumptions behind the fitting procedure should be checked rather than accepted automatically.
Residual plots are an important diagnostic. Random scatter around zero is consistent with a simple model, while curvature may indicate a missing nonlinear term, a changing physical regime or an incorrect transformation. A funnel-shaped residual pattern may indicate that the variance changes with the independent variable. A small residual sum does not prove that the physical model is correct; compare the model with the underlying mechanism and, where possible, independent measurements.
39. Fourier transforms and the frequency-domain view
A Fourier series is suited to periodic functions. For a non-periodic signal, a Fourier transform describes how its content is distributed over continuous frequency. A time-domain signal x(t) can be represented through frequency components X(ω), subject to the chosen transform convention. The exact factors of 2π differ between conventions, so the convention must be stated when comparing formulae. The frequency-domain view is especially useful for filters, pulses, transient signals, spectroscopy and imaging.
A narrow pulse in time generally contains a broad range of frequencies, while a long nearly sinusoidal signal has a more concentrated spectrum. This time-frequency trade-off appears in signal processing and wave physics. The transform does not create new physical information; it reorganises the same signal in a form that makes frequency-dependent behaviour easier to analyse.
40. Boundary conditions, initial conditions and uniqueness
Initial conditions specify the state of a system at a starting time, such as position and velocity at t=0. Boundary conditions specify behaviour at spatial boundaries, such as a fixed end of a string or a specified temperature at the surfaces of a conducting slab. A differential equation together with suitable conditions may determine a unique solution. Without these conditions, a family of mathematically valid solutions can remain.
For a second-order equation in time, two independent initial conditions are usually required. For a second-order equation in space, two boundary conditions are commonly required, although the exact number and type depend on the equation and physical domain. Boundary conditions are not decorative additions: they encode how the system interacts with its surroundings. A string fixed at both ends has a different set of allowed modes from a string fixed at one end and free at the other.
41. Green's functions and the response of linear systems
For a linear differential equation, a Green's function describes the response to a localised impulse or point source. Once the Green's function is known for a particular operator and set of boundary conditions, the response to a more general forcing can often be obtained by integrating the Green's function against that forcing. This is the continuous analogue of constructing a solution from the responses to many small inputs.
The idea appears in electrostatics, wave propagation, diffusion, quantum theory and engineering. It also clarifies the distinction between a homogeneous solution, which satisfies the equation with zero forcing, and a particular solution, which responds to the applied source. The method can be mathematically advanced, but its physical interpretation is intuitive: determine how the system responds to a simple impulse, then combine impulses to represent a complicated input.
When using a Green's function, check the operator, domain, boundary conditions and normalisation convention. A Green's function for an infinite line is not automatically valid for a finite interval or a bounded three-dimensional region. Boundary conditions are part of the definition of the response function.
42. Dimensional consistency of calculus and mathematical operators
Dimensional reasoning applies to derivatives and integrals. If x has dimension L and t has dimension T, then dx/dt has dimension LT⁻¹ and d²x/dt² has dimension LT⁻². If a force F has dimension MLT⁻², then ∫F dx has dimension ML²T⁻², which is energy. The gradient of a scalar field has the field's dimension divided by length. Divergence and curl each introduce an inverse-length dimension, though their physical interpretation differs.
When adding terms in an equation, every term must have the same dimensions. Arguments of exponential, logarithmic and trigonometric functions must be dimensionless. Thus e−t/τ is meaningful because t/τ is dimensionless, whereas e−t is incomplete if t carries units. A logarithm of a dimensional quantity must be formed relative to a reference quantity, for example ln(x/x₀).
43. A systematic method for solving unfamiliar Physics problems
Begin by translating the verbal statement into a clearly defined physical system. Draw a diagram and identify the object, field, boundaries, interactions and relevant time interval. Define symbols, coordinates and positive directions. List known values with units and identify the unknown quantity. Then select the governing physical law and the mathematical method appropriate to its structure. Do not choose an equation merely because it contains a familiar symbol.
Next, state assumptions and check whether the model's conditions are met. Resolve vectors into components when necessary. Keep algebra symbolic until the final substitution where practical; this makes cancellations and dimensions easier to inspect. Solve the equations, apply initial or boundary conditions, and calculate the requested quantity. Finally, check dimensions, limiting cases, sign, order of magnitude and physical interpretation. If a result is implausible, return to the diagram and assumptions before changing arithmetic at random.
For experimental problems, distinguish measured data from derived quantities. State the measurement uncertainty, explain how it propagates, and use an appropriate graph or fit. Report the result with units and justified significant figures. A strong Physics solution communicates the reasoning so another reader can reproduce it, not merely the final calculator output.
44. Integrated problem: motion with a time-dependent force
Consider a particle of mass m moving along one axis under a force F(t)=F₀e−t/τ, with initial velocity v₀. Newton's second law gives m dv/dt=F₀e−t/τ. Integrating from 0 to t gives v(t)−v₀=(F₀/m)∫₀ᵗe−s/τds=(F₀τ/m)(1−e−t/τ). Thus v(t)=v₀+(F₀τ/m)(1−e−t/τ). Integrating once more gives x(t)=x₀+v₀t+(F₀τ/m)[t−τ(1−e−t/τ)].
The force decays to zero, but the particle does not necessarily return to its original velocity. The total impulse is finite: J=∫₀∞F(t)dt=F₀τ. The final velocity is v(∞)=v₀+F₀τ/m, consistent with impulse equalling change in momentum. This example combines differential equations, integration, exponential functions, initial conditions and dimensional analysis.
45. Integrated problem: uncertainty in a derived physical constant
Suppose a physical constant is inferred from a fitted gradient. If g=K/m, where K is an exact constant and m is a measured gradient, then differentiating gives dg/dm=−K/m². The magnitude of the propagated uncertainty is σg≈(K/m²)σm, or σg/g≈σm/m. This relative form is often the clearest way to understand how uncertainty in a fitted parameter affects the final result.
When several measured quantities contribute, use partial derivatives and include covariance if the measurements are correlated. If the fitted gradient and intercept are correlated, treating their uncertainties as independent may understate or overstate the final uncertainty. Statistical software can provide a covariance matrix, but the analyst remains responsible for understanding the model and reporting the assumptions.
46. Review of assumptions and validity domains
Every model has a domain of validity. A constant-acceleration model assumes acceleration does not vary appreciably over the interval. A small-angle pendulum model assumes sinθ≈θ. A linear spring model assumes extension remains within the proportional regime. A first-order uncertainty formula assumes uncertainties are sufficiently small for local linearisation. A Fourier reconstruction assumes the signal is sampled and represented under a suitable mathematical framework. A numerical method assumes the step size and algorithm resolve the relevant scales.
To evaluate a model, compare its assumptions with the physical system, estimate the size of neglected effects and test predictions against data. A more complicated model is not automatically better: extra parameters can be poorly constrained and can obscure the dominant physics. Prefer the simplest model that explains the observations within their uncertainties, and state what the model cannot explain.
47. Extended examination practice: multi-step university problems
- A particle has r(t)=(t²−2t)i+(3t−t²)j m. Derive velocity and acceleration, find the speed at t=2 s, and determine whether the acceleration is parallel to the velocity at that instant.
Model answer: v=(2t−2)i+(3−2t)j and a=2i−2j. At t=2, v=2i−j, so speed=√5 m s⁻¹. Since a·v=2(2)+(−2)(−1)=6 and a is not a scalar multiple of (2,−1), acceleration is not parallel to velocity.
- A measured response follows y=Ae−kt. Explain how a linear graph can determine k, identify two sources of systematic error and state the dimensions of k.
Model answer: Plot ln y against t; the gradient is −k and the intercept is ln A. Timing calibration and sensor zero/background offset are possible systematic errors. Since kt is dimensionless, k has dimension T⁻¹.
- Use dimensional analysis to show that the speed of a wave on a stretched string can depend on tension F and linear mass density μ through v∝√(F/μ). State what dimensional analysis cannot determine.
Model answer: Assume v∝Fᵃμᵇ. With [F]=MLT⁻² and [μ]=ML⁻¹, equate dimensions to [v]=LT⁻¹. The mass exponent gives a+b=0; time gives −2a=−1 so a=1/2 and b=−1/2. It cannot determine the dimensionless proportionality constant or verify all physical assumptions.
- A data set is fitted by y=mx+c. The residuals show a smooth U-shaped pattern. Explain what this suggests and propose a next step.
Model answer: The linear model may be missing curvature or a relevant variable. Plot the residuals, inspect the physical mechanism and compare a justified nonlinear model without overfitting.
- Explain why the eigenvectors of a symmetric matrix can be chosen orthogonal and why this is useful in coupled oscillations.
Model answer: Real symmetric matrices have real eigenvalues and orthogonal eigenvectors for distinct eigenvalues. Orthogonal mode vectors allow independent normal coordinates, simplifying the coupled equations into separate modal equations.
- For x(t)=A cos(ωt+φ), derive velocity and acceleration and show that acceleration is proportional to negative displacement.
Model answer: v=−Aω sin(ωt+φ); a=−Aω² cos(ωt+φ)=−ω²x. The restoring acceleration is proportional to displacement and directed oppositely.
- A result is calculated from q=x²/y. Derive the approximate relative uncertainty for independent x and y.
Model answer: ln q=2 ln x−ln y. Therefore (σq/q)²≈4(σx/x)²+(σy/y)² for independent small uncertainties.
- Explain the difference between a boundary condition and an initial condition, giving one physical example of each.
Model answer: An initial condition specifies the state at a starting time, such as initial displacement and velocity. A boundary condition specifies behaviour at a spatial boundary, such as zero displacement at a fixed end of a string.
- Why can a high-order numerical method still produce a poor physical prediction?
Model answer: Numerical accuracy does not correct an incorrect physical model, wrong parameters, poor initial data, inappropriate boundary conditions or unresolved experimental uncertainty.
- Describe how to test whether a measured exponential decay is consistent with a single time constant.
Model answer: Plot the logarithm of the background-corrected signal against time. A straight line with random residuals is consistent with a single exponential over that range. Curvature or structured residuals may indicate multiple processes or a changing environment.
48. Chapter synthesis and mastery checklist
Before progressing to the next unit, a student should be able to move fluently between physical descriptions, diagrams, equations, graphs and numerical results. The key skills are: represent scalar and vector quantities correctly; resolve vectors in appropriate coordinates; differentiate and integrate with correct variables and units; solve basic differential equations with conditions; use complex and Fourier representations for oscillatory signals; apply dimensional analysis and scaling; solve small matrix systems and interpret eigenvalues; fit experimental data and inspect residuals; propagate uncertainty; and explain the assumptions and limitations of each model.
Mastery is demonstrated by applying these methods to unfamiliar physical situations, not simply repeating a worked example. When revising, cover the worked solution, attempt the problem independently, compare each step and explain why each mathematical operation is physically justified. Keep a record of recurring errors, especially sign conventions, missing units, inappropriate approximations and confusion between vector magnitude and components.
49. Introduction to vector calculus: gradient, divergence and curl
Vector calculus extends ordinary differentiation and integration to fields that vary through space. A scalar field assigns a scalar to every point, such as temperature T(x,y,z) or electric potential V(x,y,z). A vector field assigns a vector, such as fluid velocity v(x,y,z) or electric field E(x,y,z). The del operator is ∇=i∂/∂x+j∂/∂y+k∂/∂z in Cartesian coordinates. Applied to a scalar field, it produces the gradient ∇f. The gradient points in the direction of greatest increase and is perpendicular to a surface of constant f.
The divergence of a vector field is ∇·A=∂Ax/∂x+∂Ay/∂y+∂Az/∂z. It measures local net outward flux per unit volume in the limiting sense. Positive divergence indicates a local source-like behaviour; negative divergence indicates sink-like behaviour. The curl is ∇×A and measures local circulation or rotational tendency. These operations underpin Gauss's law, Faraday's law, fluid continuity and many field theories. Their precise physical meaning depends on the field and its units.
50. Line, surface and volume integrals
A line integral accumulates a scalar or vector contribution along a path. Work by a force along a curve C is W=∫CF·dr. A surface integral accumulates flux through a surface, such as Φ=∫SA·dS. A volume integral accumulates a density through a three-dimensional region, such as total mass M=∫Vρ dV. The orientation of a path or surface matters for vector integrals; reversing the orientation can reverse the sign of the result.
For a conservative force, the work between two points is independent of the path and can be written as the negative change in potential energy. For a non-conservative force, the work may depend on the route taken. This distinction is important when deciding whether a potential-energy function can be defined globally. In practical calculations, parametrise the path or surface, write the differential element explicitly and check the resulting dimensions.
51. Euler and Runge–Kutta methods for initial-value problems
The Euler method approximates the next value of a function using the slope at the beginning of a step. For y′=F(t,y), yn+1=yn+hF(tn,yn). It is simple and useful for understanding numerical integration, but its global error is generally proportional to the step size h for smooth problems. A smaller step can improve accuracy, but it increases computation and may expose round-off or stability problems.
Higher-order Runge–Kutta methods evaluate the slope at intermediate points and combine these slopes to obtain a more accurate update. The classical fourth-order method uses four slope estimates per step. Its global truncation error is typically proportional to h⁴ for sufficiently smooth problems. Adaptive solvers estimate local error and adjust the step size, which is useful when the solution changes rapidly in some intervals and slowly in others. Numerical stability must still be considered, particularly for stiff differential equations.
52. Stability, conditioning and numerical reliability
Stability describes whether small perturbations, rounding errors or numerical errors remain controlled as a computation proceeds. Conditioning describes how sensitive the exact mathematical solution is to small changes in the input data. A well-conditioned problem can still be solved poorly by an unstable algorithm; a stable algorithm cannot remove the intrinsic sensitivity of an ill-conditioned problem. These are different sources of numerical difficulty.
For a linear system A x=b, a matrix with nearly dependent rows or columns can be ill-conditioned. Small errors in b or in the coefficients may cause large changes in x. The condition number provides a measure of this sensitivity. In experimental Physics, uncertainty in the input measurements should therefore be considered alongside numerical precision. Reporting many decimal places from an ill-conditioned calculation can be misleading.
A robust computational workflow includes unit tests for simple limiting cases, comparison with an analytic solution when available, refinement of grid or step size, conservation checks where applicable, and independent verification of a representative result. Document software version, algorithm, tolerances and parameter choices so another researcher can reproduce the calculation.
53. Further problem-solving exercise: field gradients and physical direction
Suppose a temperature field is T(x,y,z)=T₀+ax+by+cz. Its gradient is ∇T=ai+bj+ck. For a small displacement dr, the first-order temperature change is dT=∇T·dr. The directional derivative along a unit vector n is ∇T·n. This gives the rate of temperature change per unit distance in the chosen direction. The direction of steepest increase is n=∇T/|∇T| when the gradient is non-zero.
54. Final independent-study assignment
Choose a measurable physical process that can be represented by a mathematical model: cooling of a heated object, oscillation of a spring, motion under a resistive force, charging of a capacitor or attenuation of a signal. Write a concise research question and identify the dependent and independent variables. Derive a prediction from first principles, define all symbols and units, state the assumptions and identify at least one limitation that could affect the experiment.
Design a repeatable measurement procedure, select an appropriate range and interval, identify calibration requirements and plan how uncertainty will be estimated. Decide whether the raw data should be plotted directly or transformed to test a linear relationship. Explain how a fitted parameter will be extracted and how its uncertainty will propagate to the final physical quantity. Include a labelled apparatus diagram and a sample calculation using clearly identified illustrative values; never present invented values as actual experimental observations.
In the discussion, compare the trend with the model, inspect residuals, identify systematic and random effects, and state whether the evidence supports the prediction within uncertainty. Suggest one improvement that would most reduce the dominant uncertainty. Conclude with a short explanation of how the mathematical method used in the project appears in another area of undergraduate Physics.
Further study and academic integrity
Compare these notes with your university's module descriptor and recommended textbook. In assessed work, cite sources, show intermediate steps, distinguish measured values from theoretical predictions and never report fabricated experimental data.
CBE Nexus · Independent undergraduate learning resource · Mathematical Methods for Physics I