Surface Plasmon

What a surface plasmon actually is

Inside a metal, the outer electrons are not bound to individual atoms — they form a dense gas (the conduction electrons, roughly 10²²–10²³ per cm³) free to slosh against a stationary background of positive ions. Displace this electron sea and the exposed charge sets up a restoring electric field that pulls it back. Like any restoring force, that produces an oscillation: the plasmon, the quantized collective oscillation of the electron gas, with a natural bulk frequency called the plasma frequency ω_p.

A surface plasmon is the version of this oscillation that lives at a boundary — a metal-dielectric interface — rather than in the bulk. There, the electron oscillation couples to an electromagnetic wave, and the two lock together into a single hybrid excitation. Because it is part charge oscillation and part photon, the propagating mode is properly called a surface plasmon polariton (SPP). Its electric field is largest right at the surface and falls off exponentially as you move away on either side. That is the defining trait: the energy clings to the interface instead of radiating away.

The payoff is that an SPP can carry light in a sheet only nanometers thick, beating the usual rule that you cannot focus light tighter than about half its wavelength (the diffraction limit). Squeezing the same energy into a smaller volume amplifies the local field, which is exactly what makes surface plasmons useful: a strong, tightly confined field that any nearby molecule, defect, or quantum emitter feels intensely.

Where the surface mode comes from

To see why a metal supports a bound surface wave at all, you write down Maxwell's equations at a flat interface (z = 0) between a metal of permittivity ε_m and a dielectric of permittivity ε_d, and look for a wave that travels along the surface (in x) but decays away from it (in ±z). A transverse-magnetic (TM, or p-polarized) wave does this only if the dielectric functions on the two sides have opposite signs. Metals below their plasma frequency provide exactly that: ε_m is negative while ε_d > 0.

Matching the fields across the boundary gives the SPP dispersion relation — the link between frequency ω and in-plane wavevector k_x:

k_x = (ω/c) · √( ε_m·ε_d / (ε_m + ε_d) )

The same matching produces the decay constants into each medium, κ_i, with κ_i² = k_x² − ε_i (ω/c)². Because k_x exceeds the photon wavevector in the dielectric, κ is real and the field is evanescent — exponentially bound to the surface. A bound, non-radiating solution requires ε_m + ε_d < 0, i.e. |ε_m| > ε_d with ε_m negative.

Two limits are worth remembering. At low frequency the SPP wavevector hugs the light line k_x ≈ (ω/c)√ε_d, so it behaves almost like an ordinary grazing photon and is weakly confined. As ε_m → −ε_d, the denominator vanishes, k_x blows up, the wavelength shrinks, and the field becomes extremely confined — but the wave also slows and damps heavily. That special frequency is the surface plasmon resonance.

The resonance condition and the √2

The resonance condition is compact:

ε_metal(ω_sp) = −ε_dielectric

Model the metal with the Drude dielectric function, the simplest description of a free-electron gas:

ε_m(ω) = 1 − ω_p² / (ω² + iγω)

where ω_p is the plasma frequency and γ the electron damping rate. Ignoring damping and setting ε_m = −ε_d gives the asymptotic surface plasmon frequency:

ω_sp = ω_p / √(1 + ε_d)

For a metal against vacuum or air (ε_d = 1) this is the famous result ω_sp = ω_p/√2 — about 71% of the bulk plasma frequency. The surface oscillation costs less energy than the bulk one because, at a surface, the electrons only have to push the field into half-space, so the restoring force is weaker.

Real metals, real numbers

The free-electron picture is a good first cut, but real metals have interband transitions and damping that move and broaden the resonance. The table collects representative values.

Quantity	Silver (Ag)	Gold (Au)
Bulk plasma energy ħω_p	≈ 9.6 eV	≈ 8.9 eV
Observed SPP resonance (air)	≈ 3.5 eV (~350 nm)	≈ 2.4 eV (~520 nm, capped by interband)
Nanosphere LSPR (in water)	≈ 400 nm	≈ 520 nm
SPP propagation length (visible)	tens of µm	a few µm
SPP propagation length (1550 nm)	hundreds of µm	tens–hundreds of µm
Practical strength	lowest loss in visible	chemically inert, biocompatible

Watch out for one trap when you plug numbers in: the bare free-electron formula ω_p/√2 with ħω_p ≈ 9.6 eV would predict a silver surface plasmon near 6.8 eV, yet the observed value is ~3.5 eV. The difference is screening by silver's bound d-electrons, which add a large background permittivity ε_∞ (≈ 4–5). Folding that in (ε_m = ε_∞ − ω_p²/ω²) pulls the effective plasma energy down and brings ω_sp into agreement with experiment — a reminder that real metals are not pure free-electron gases.

Silver gives the sharpest, lowest-loss plasmons across the visible, which is why it is favoured for surface-enhanced Raman and color-pure nanostructures; but it tarnishes. Gold is duller plasmonically because gold's d-band interband absorption begins around 2.4 eV (520 nm) and damps the resonance, yet its chemical inertness and easy bio-functionalization make it the workhorse of commercial sensors.

The momentum mismatch — and how to bridge it

Here is the catch that makes plasmonics an engineering discipline. At any given frequency the SPP carries more momentum than a freely propagating photon of the same frequency in the dielectric: its dispersion curve lies to the right of the "light line" ω = ck/√ε_d. The two curves never intersect, so you cannot simply shine light at a flat metal and launch a surface plasmon — energy conserves but momentum does not.

You must supply the missing in-plane momentum, Δk = k_SPP − k_light. The standard tricks:

• Prism coupling (Kretschmann geometry). Light totally internally reflects inside a high-index prism; the evanescent tail leaks through a thin (~50 nm) gold film and matches the SPP momentum at a precise angle. This is the heart of commercial SPR sensors.
• Otto geometry. A small air gap separates prism and metal; the prism's evanescent field tunnels across the gap to the metal surface.
• Grating coupling. A periodic corrugation of period Λ adds reciprocal-lattice momentum in steps of 2π/Λ: k_SPP = k_light·sinθ ± m·2π/Λ. This is why ruled gratings show sharp absorption dips (Wood's anomalies).
• Near-field / defect coupling. A sharp tip, slit, or nanoparticle scatters light into a broad spread of momenta, some of which match the SPP.

Localized surface plasmons in nanoparticles sidestep this entirely. A particle much smaller than the wavelength has no well-defined momentum to conserve, so light couples directly to its dipolar resonance — which is why gold colloids paint stained glass without any prism or grating.

Concrete examples

• The Lycurgus Cup (4th century Rome). Embedded gold-silver nanoparticles make the glass appear green in reflected light and ruby red in transmitted light — a localized surface plasmon resonance, engineered 1,500 years before the physics was understood.
• SPR biosensors. A 50 nm gold film in the Kretschmann geometry; the resonance angle shifts by a fraction of a degree when a protein monolayer binds, detecting roughly 1 pg/mm² of bound mass in real time, label-free.
• Surface-enhanced Raman spectroscopy (SERS). Roughened silver or gold nanostructures concentrate the field into "hot spots" between particles; the Raman signal of an adsorbed molecule is amplified by 10⁶–10¹¹, enough to detect single molecules.
• Plasmonic waveguides. Metal stripes, slots, and wedges guide light in cross-sections of tens of nanometers, far below what dielectric waveguides allow — promising for on-chip optical interconnects, at the cost of metal loss.
• Solar and photothermal. Plasmonic nanoparticles trap light in thin-film solar cells and convert absorbed light to heat for photothermal cancer therapy and steam generation.

Propagating vs. localized surface plasmons

Feature	Surface plasmon polariton (SPP)	Localized surface plasmon (LSPR)
Geometry	Extended flat metal-dielectric interface	Sub-wavelength metal nanoparticle
Motion	Propagating wave (microns)	Standing, non-propagating oscillation
Excitation	Needs prism / grating (momentum match)	Direct illumination, no matching needed
Resonance set by	ε_m = −ε_d, plus k-matching geometry	Particle size, shape, and surroundings
Typical use	Sensing films, waveguides	Colorimetric sensing, SERS, color, therapy
Field concentration	Thin sheet at surface	Hot spots at tips and gaps

Computing the surface plasmon dispersion

// Drude dielectric function of a free-electron metal (lossless for simplicity)
// energies in eV; omega_p is the bulk plasma energy
function epsilonDrude(omega, omega_p) {
  return 1 - (omega_p * omega_p) / (omega * omega);
}

// Asymptotic surface plasmon frequency: eps_m = -eps_d
function surfacePlasmonEnergy(omega_p, eps_d = 1) {
  return omega_p / Math.sqrt(1 + eps_d);  // ω_sp = ω_p / √(1 + ε_d)
}

// Silver: ħω_p ≈ 9.6 eV (bare free-electron value)
console.log(`Ag bare ω_sp (air): ${surfacePlasmonEnergy(9.6).toFixed(2)} eV`); // 6.79 eV; d-band screening pulls the observed value down to ~3.5 eV
// Vacuum, free-electron: ω_p/√2
console.log(`ω_p/√2 fraction: ${(1/Math.SQRT2).toFixed(3)}`);               // 0.707

// SPP in-plane wavevector (real part) at photon energy E (eV) for metal vs dielectric
// returns k_x in units of (ω/c); multiply by ω/c (1/m) for SI
function sppWavevectorScaled(E, omega_p, eps_d = 1) {
  const eps_m = epsilonDrude(E, omega_p);
  const num = eps_m * eps_d;
  const den = eps_m + eps_d;
  return Math.sqrt(num / den);   // = k_x / (ω/c)
}

// Near resonance the wavevector diverges (field gets very confined)
[1.0, 2.0, 3.0, 3.3].forEach(E => {
  const beta = sppWavevectorScaled(E, 9.6);
  console.log(`E=${E} eV: k_x/(ω/c) = ${isFinite(beta) ? beta.toFixed(3) : '→∞'}`);
});

// SPP wavelength compared to free-space wavelength
function confinementFactor(E, omega_p, eps_d = 1) {
  // λ_spp / λ_0 = 1 / Re(k_x/(ω/c))
  return 1 / sppWavevectorScaled(E, omega_p, eps_d);
}
console.log(`λ_spp/λ_0 at 2 eV: ${confinementFactor(2.0, 9.6).toFixed(3)}`); // < 1: shorter than light

// Grating momentum matching: which order m launches an SPP?
// k_spp = k0·sinθ + m·(2π/Λ), with k0 = 2π/λ0
function gratingMatch(lambda0_nm, period_nm, thetaDeg, betaTarget) {
  const k0 = 2 * Math.PI / lambda0_nm;
  const G  = 2 * Math.PI / period_nm;
  const kIn = k0 * Math.sin(thetaDeg * Math.PI / 180);
  const kTarget = betaTarget * k0;        // k_spp = β·k0
  return Math.round((kTarget - kIn) / G);  // nearest diffraction order
}
console.log(`Grating order: ${gratingMatch(600, 500, 10, 1.05)}`);

Where surface plasmons show up

• Biosensing. SPR instruments (Biacore and similar) measure binding kinetics, affinities, and concentrations label-free in drug discovery and diagnostics.
• Spectroscopy. SERS and tip-enhanced Raman push chemical fingerprinting to the single-molecule limit using plasmonic hot spots.
• Nanophotonics & integrated optics. Plasmonic waveguides, modulators, and antennas route and concentrate light on chips below the diffraction limit.
• Color and metamaterials. Structural color, plasmonic pixels, and negative-index/metasurface designs exploit engineered resonances.
• Energy. Light trapping in thin-film photovoltaics and plasmon-driven photocatalysis (hot-carrier chemistry).
• Medicine. Photothermal therapy, where gold nanorods absorb near-infrared light and heat tumours; and lateral-flow LSPR color readouts.
• Data storage. Heat-assisted magnetic recording uses a plasmonic near-field transducer to heat a nanoscale spot on the disk.

Common mistakes

• Trying to excite an SPP with light at normal incidence on a flat film. The momentum never matches. You need a prism, grating, or near-field coupler.
• Confusing the bulk plasma frequency with the surface plasmon frequency. The surface mode sits at ω_p/√(1+ε_d), below the bulk ω_p — about ω_p/√2 in air.
• Using s-polarized (TE) light. Only TM/p-polarization launches an SPP; the field geometry of TE light cannot satisfy the boundary conditions for a bound mode.
• Treating a nanoparticle resonance as a propagating SPP. Localized plasmons (LSPR) are standing modes set by size and shape; propagating SPPs need an extended interface and momentum matching.
• Ignoring loss. Tighter confinement means more field inside the lossy metal, so propagation length and quality factor drop fast near resonance. Confinement and loss trade off; you cannot have both.
• Forgetting the environment. The resonance depends on the dielectric on both sides. Changing the surrounding index (the basis of sensing) shifts the resonance, so an air-clad and a water-clad film behave differently.