Derive a field equation from coupled springs and see how a lattice changes short waves.
Before you start: Hooke’s law, acceleration and sine waves. A short section introduces derivatives and Fourier modes.
You will be able to: Derive the chain dispersion and identify when its continuum approximation fails.
Ideal analytic reference; dimensionless reference units. ωlat2=0.25+4sin2(k/2),ωcont2=0.25+k2The curves agree for long wavelengths; the difference grows near the lattice cutoff.
A field describes many local degrees of freedom with one function. A stretched string makes the idea tangible: every small piece can move, and neighboring pieces pull on one another. When we care about waves much longer than the distance between those pieces, their displacements can be described by a smooth function. The field equation keeps the motion that survives at that resolution.
That last condition matters. A continuum description earns its usefulness by making accurate predictions for a chosen range of wavelengths. We can see both its success and its failure in a chain of springs.
Start with a piece you can pull
Put identical masses along the horizontal coordinate x, with spacing a. Let qj(t) be the transverse displacement of mass j, positive upward. Both x and qj have units of length. Write the mass of one piece as μa, where μ is mass per unit length. The spring between adjacent pieces has stiffness κ/a, where κ has units of force. A small spring to the support supplies an additional restoring force with angular frequency Ω.
The two neighboring springs pull a displaced piece toward their average position. Newton’s law gives
μaq¨j=aκ(qj+1−2qj+qj−1)−μaΩ2qj.
The dots mean time derivatives. The first derivative is the rate at which displacement changes; the second is its acceleration. The combination in parentheses measures curvature along the chain. It vanishes for a straight, uniformly tilted row of displacements.
The on-site springs are optional. They will become useful because they give even a very long wave a nonzero oscillation frequency.
A traveling wave reveals the approximation
Try a sinusoidal displacement qj=Acos(kja−ωt). Here A is an amplitude, k=2π/λ is the wave number, λ is the wavelength, and ω is angular frequency. Advancing one site shifts the phase by ka. Substituting the trial wave into Newton’s law and using the addition formula for cosine gives
ω2=Ω2+a24c2sin2(2ka),c=μκ.
This relation between frequency and wave number is the dispersion relation. It tells us which rhythm accompanies each spatial pattern. A Fourier expansion means building a more complicated displacement by adding these sinusoidal patterns. Because the equation is linear, each pattern evolves independently.
For a long wave, the phase changes little from one site to the next. Then sin(ka/2) is close to ka/2, and the relation becomes
ω2≃Ω2+c2k2,∣ka∣≪1.
The figure compares these two predictions in reference units. They agree near the left side because neighboring sites see nearly the same phase. Near the right side, the lattice resolves the wave’s alternating peaks and troughs, and its frequency follows the full sine function.
A numerical example
Take a=0.100m, c=2.00ms−1, and Ω=3.00s−1. A wave with λ=2.00m has k=πm−1. The discrete equation gives about 6.94s−1; the continuum expression gives about 6.96s−1. The difference is roughly a third of a percent.
Keep the material the same and shorten the wavelength to 0.200m. Adjacent sites now have opposite phases. The discrete frequency is about 40.1s−1, and the continuum prediction is about 62.9s−1. The continuum approximation applies to substantially longer waves.
Replace many coordinates with one function
Define a smooth displacement ϕ(x,t) that agrees with qj(t) at the lattice sites. Taylor expansion expresses a nearby value using the value and successive derivatives at the chosen position. The linear terms in the expansions at x+a and x−a have opposite signs and cancel; the quadratic terms add. Keeping the next surviving term gives
a2ϕ(x+a)−2ϕ(x)+ϕ(x−a)=∂x2ϕ+12a2∂x4ϕ+⋯.
The symbol ∂x means differentiating with respect to position while holding time fixed. The first retained term gives the continuum field equation,
∂t2ϕ−c2∂x2ϕ+Ω2ϕ=0.
The omitted fourth derivative explains the discrepancy we just calculated. For a Fourier mode it brings an additional factor of wave number squared relative to the leading gradient term. Reducing the lattice spacing helps a fixed long wave; it does not make every wave on the lattice equally accurate.
We have changed the description from a list of coordinates to a function over space. We still have a classical system. Quantization will come after we understand its independent modes and conserved quantities.
The same equation follows from energy
The Lagrangian is kinetic energy minus potential energy. For the smooth field it is the spatial integral of a density,
L=∫dxL,L=2μ(∂tϕ)2−2κ(∂xϕ)2−2μΩ2ϕ2.
An integral adds the contributions of small intervals. Each term in this density has units of energy per unit length. The gradient term stores energy when neighboring displacements differ; the final term stores energy when the field moves away from the support.
To find the equation of motion, vary the field by a small amount that vanishes at the time endpoints and at fixed spatial boundaries. Integration by parts reverses the product rule inside an integral: it moves a derivative to the other factor and supplies an endpoint term. Here that endpoint term vanishes. Requiring the first-order change of action to vanish for every allowed field variation gives
∂t∂(∂tϕ)∂L+∂x∂(∂xϕ)∂L−∂ϕ∂L=0.
The three terms become μ∂t2ϕ, −κ∂x2ϕ, and +μΩ2ϕ. Dividing by μ recovers Newton’s-law result. The action is useful because the same method extends to several fields and exposes their symmetries.
Compare two mode shapes
Open the field-modes Lab, hold the occupation at one and compare mode numbers one and three. Count the internal nodes, the positions that stay at zero displacement. Predict the frequency and wavelength ratios before changing the mode. In this fixed-end, massless example the third mode has two internal nodes, three times the frequency and one third of the wavelength. The numerical spring chain above is a separate example with an on-site restoring force and finite-spacing corrections.
Exercises
What changes when the masses get heavier?
Double μ while keeping κ, a, and Ω fixed. What happens to the frequency of a nonuniform mode? What happens to the uniform mode?
Answer and reasoning
The wave speed becomes c/2. The spatial contribution to frequency squared is halved, while the on-site contribution remains Ω2. A uniform mode has no spring extension between neighbors, so its frequency stays Ω. This distinguishes two physically different restoring forces.
How would you test a continuum calculation?
A program matches the continuum dispersion for one wavelength. Design a second check that could reveal a lattice error without changing the physical wavelength or length of the chain.
Answer and reasoning
Halve the spacing and double the number of sites. Keep μ, κ, Ω, and the physical chain length fixed. In the long-wave range, the leading frequency-squared error scales as a2, so it should shrink by about four. Increasing only the number of sites at fixed spacing would instead change the volume.
Can a single coordinate describe the field?
Suppose every point is displaced by the same amount at every time. Why can this special motion be described by one oscillator, even though the full field has many degrees of freedom?
Answer and reasoning
All nonuniform Fourier modes have zero amplitude in this preparation. The uniform mode alone remains, and the neighboring springs exert no net force on it. A different initial displacement can excite many independent modes. The full field theory needs a coordinate for each independent mode.