physics2u
Tier
⌕ Search ⌘K
physics2u.com/vault/resonance-lorentzian-lineshape.html
Derivation

Power Absorption and the Lorentzian Lineshape

Statement

For a linear damped oscillator mẍ + bẋ + kx = F0cos(ωt) driven in steady state, the time-averaged power absorbed from the drive is (ω) = (F02/2m) γω2 / [(ω02ω2)2 + γ2ω2], where ω02 = k/m and γ = b/m. Near resonance this reduces to a Lorentzian of full width at half maximum Δω = γ, peak max = F02/2, so the absorption bandwidth and the quality factor are locked together by Q = ω0ω.

Why it matters

The Lorentzian is the universal fingerprint of a weakly damped resonance responding linearly to a harmonic drive. It is what a spectrometer, a swept-frequency network analyser, or a nuclear-magnetic-resonance receiver actually measures: not the amplitude curve, but the power the system removes from the drive as a function of frequency. Its width is not a fitting nuisance — it is a direct read-out of the internal loss rate.

Because the width equals γ and the centre equals ω0, a single lineshape simultaneously yields the resonant frequency and the Q. This is the operational meaning of Q: the sharpness of a resonance in the frequency domain is the same physical quantity as the slowness of its energy decay in the time domain. Every downstream idea — cavity finesse, laser linewidth, filter selectivity, the energy–time reading of a decaying state — descends from this one calculation.

Assumptions
Linear restoring and damping.If the spring stiffens or the drag becomes quadratic, superposition fails, the steady state is no longer a single sinusoid at ω, and the response develops harmonics whose shape is not Lorentzian.
Steady state has been reached.During the transient the amplitude and phase still evolve, so the instantaneous power differs from the period-averaged value and the swept lineshape is smeared by the finite settling time τ ≈ 1/γ.
Viscous (velocity-proportional) damping, single mode.Dry friction or radiation damping give a different frequency dependence of the loss; a nearby second mode adds a second Lorentzian and the two overlap if their separation is not ≫ γ.
Weak damping, γω0, and the drive is swept slowly (adiabatically) compared with γ.If γ is not small the exact curve is asymmetric and the "width = γ" Lorentzian approximation errs by O(γ/ω0); if the sweep is fast the line is dynamically broadened and can ring.
Derivation
1
x(t) = A(ω) cos(ωtδ),   A = (F0/m) / √[(ω02ω2)2 + γ2ω2],   tan δ = γω/(ω02ω2)
Steady-state particular solution imported from the driven-oscillator result; the transient has decayed. A
2
P(t) = F(t) ẋ(t) = F0cos(ωt) · [−Aω sin(ωtδ)]
Instantaneous power delivered by the external force equals force times velocity; ẋ obtained by differentiating step 1. A
3
⟨cos(ωt) sin(ωtδ)⟩ = ⟨cos ωt (sin ωt cos δ − cos ωt sin δ)⟩ = −½ sin δ
Expand and average over one period: ⟨cos ωt sin ωt⟩ = 0 and ⟨cos2 ωt⟩ = ½. Only the component of velocity in phase with the force does net work. B
4
= ⟨P⟩ = ½ F0 A ω sin δ
Substitute the average from step 3 into step 2; the two minus signs cancel. This is the general "power = ½ (force)(velocity amplitude) cos(phase between them)" statement, with sin δ the velocity–force in-phase factor. B
5
sin δ = γω / √[(ω02ω2)2 + γ2ω2]
From tan δ in step 1, build the right triangle with opposite = γω, adjacent = ω02ω2, hypotenuse = the square root. A
6
(ω) = (F02/2m) · γω2 / [(ω02ω2)2 + γ2ω2]
Insert A (step 1) and sin δ (step 5) into step 4. Both carry one factor of the square root, so the denominator becomes the bracket to the first power. This is the exact absorption curve — no approximation yet. A
7
ω02ω2 = (ω0ω)(ω0 + ω) ≈ 2ω0(ω0ω),   ωω0
Near-resonance, weak-damping expansion (γω0): keep only the leading behaviour for |ωω0| ≲ γω0. Errors are O(γ/ω0). C
8
(ω) ≈ (F02/2m) · γ / [4(ωω0)2 + γ2] = (F02/2) · (γ/2)2 / [(ωω0)2 + (γ/2)2]
Substitute step 7, set ω2ω02 in numerator and in the γ2ω2 term, then factor 4 out to expose the standard (γ/2)2 half-width form. B
9
(ω0 ± γ/2) = ½ max  ⟹  FWHM Δω = γ,   Q = ω0/γ = ω0ω
The Lorentzian falls to half its peak when (ωω0)2 = (γ/2)2; using Q = ω0/γ from the energy-decay result ties bandwidth to Q. A
Result
(ω) = max · (γ/2)2 / [(ωω0)2 + (γ/2)2],   max = F02/2,   Δω = γ = ω0/Q

Reading. The power the drive pumps into the oscillator is a symmetric bell centred on ω0, whose full width between the half-power points equals the damping rate γ. A high-Q system (small γ) presents a tall, narrow line: it absorbs strongly but only within a sliver of frequency. The peak height F02/2 rises as damping falls, yet the area under the curve, ∫ dω = πF02/4m, is independent of γ — sharpening the line trades width for height at fixed integrated absorption.

Units check. max = F02/2 has units N2/(kg·s−1) = (kg m s−2)2/(kg s−1) = kg m2 s−3 = W. The Lorentzian factor is dimensionless (each term ~ s−2), and Δω = γ carries s−1 as an angular frequency should.

Limiting cases
  • On resonance (ω = ω0): δ = π/2, velocity is exactly in phase with the force, absorption is maximal at F02/2.
  • Far off resonance (|ωω0| ≫ γ): → (F02γ/2m)/[4(ωω0)2], a Lorentzian tail falling as 1/(detuning)2.
  • Zero damping (γ → 0): the line becomes a Dirac spike of zero width and infinite height but fixed area πF02/4m — a lossless oscillator absorbs only precisely at ω0.
  • Half-power points (ω = ω0 ± γ/2): = ½max, phase δ = π/4 and 3π/4; these are the −3 dB frequencies of the resonance.
Breaks when
  • Strong damping (γω0): the factorisation ω02ω2 ≈ 2ω0(ω0ω) collapses, the exact curve is visibly asymmetric (heavier on the low-frequency side), and its peak shifts below ω0.
  • Nonlinearity (large amplitude, Duffing-type stiffness): the resonance bends, becomes hysteretic and multivalued, and the notion of a single width γ loses meaning.
  • Overlapping modes: when a second resonance lies within a few γ, the total power is a sum of Lorentzians and the individual widths cannot be read off directly.
  • Non-viscous loss (dry friction, hysteretic/structural damping): the loss per cycle no longer scales as velocity squared, so the lineshape is not Lorentzian and its width becomes amplitude-dependent.
Failure modes
  • Confusing the amplitude curve with the power curve. The amplitude A(ω) peaks slightly below ω0 (at √(ω02γ2/2)) and its FWHM is not γ; only the power peaks exactly at ω0 with width γ.
  • Dropping the phase and writing = ½F0v0. Without the sin δ factor the reactive power is counted as absorbed; off resonance most of the velocity is 90° out of phase and does zero net work.
  • Using half-width for full-width. The FWHM is γ, the half-width at half maximum is γ/2; halving this error doubles the inferred Q.
  • Time-averaging F and v separately.Fv⟩ ≠ ⟨F⟩⟨v⟩ = 0; the product must be averaged, not the factors.
  • Mixing conventions for γ. If one defines the equation with 2γẋ (amplitude decay rate γ), the width is 2γ, not γ. State whether γ is the energy or amplitude rate.
Discussion

The physical content of step 3 is that only the velocity component in phase with the driving force does net work over a cycle; the quadrature component merely sloshes energy in and out of the spring and mass without loss. On resonance the phase lag is exactly 90°, which places the velocity precisely in phase with the force (velocity leads displacement by 90°), maximising the overlap and hence the absorption. This is why the power peaks sharply at ω0 even though the amplitude peak is slightly detuned: absorption is governed by the velocity–force alignment, not by displacement.

The relation Δω = γ is the frequency-domain twin of the time-domain statement that stored energy decays as eγt. A resonance that rings for a long time (small γ, long τ = 1/γ) must, by the reciprocity of Fourier conjugates, respond over a correspondingly narrow band. The product (lifetime)(bandwidth) ≈ 1 is the classical ancestor of the energy–time relation ΔEΔt ≈ ℏ for a decaying quantum state, whose spectral line is itself a Lorentzian of width ℏ/τ — the "natural linewidth." The mathematics is identical because both are a first-order linear system relaxing at a single rate.

The scale-invariance of the integrated absorption, ∫ dω = πF02/4m, is a sum rule: it depends only on the drive strength and the inertia, never on the damping. It says that engineering a higher Q does not let a resonator absorb more total energy across the spectrum; it concentrates the same absorption into a narrower, taller line. This underlies why high-finesse cavities and sharp filters are prized for selectivity, not for raw throughput, and it is the classical form of the oscillator-strength sum rules of atomic spectroscopy.

The exact curve of step 6 is the modulus-squared of a complex susceptibility χ(ω) = 1/[m(ω02ω2 − iγω)], and the absorbed power is set by its imaginary part, = ½F02ω Im χ(ω). This is a fluctuation–dissipation statement: Im χ, the dissipative part of the response, controls both the energy the system takes from a coherent drive and the spectrum of its thermal fluctuations. Kramers–Kronig relations then force a matching dispersive (real) part — the phase lag δ and the amplitude curve — so the Lorentzian absorption line and the S-shaped dispersion curve are not independent facts but two projections of one analytic response function.

Common misconceptions. The width being γ does not mean the oscillator responds only within ±γ/2 and is dead elsewhere — the Lorentzian tails fall only as 1/(detuning)2 and carry non-negligible power far out. And "high Q" does not mean "absorbs more energy"; it means "absorbs over a narrower band," with the peak height rising to keep the integrated area fixed.

Worked examples
1
Loudspeaker cone. m = 0.020 kg, resonance f0 = 60 Hz, Q = 5, driven by force amplitude F0 = 3.0 N. Find the peak absorbed power and the −3 dB bandwidth.
Set up: ω0 = 2πf0, γ = ω0/Q, then max = F02/2. A
2
ω0 = 2π(60) = 377 rad s−1,   γ = 377/5 = 75.4 rad s−1
Symbols to numbers: damping rate from the imported Q = ω0/γ. A
3
max = (3.0)2 / [2(0.020)(75.4)] = 9.0 / 3.02 = 2.98 W
Insert into the boxed peak. Units: N2/(kg·s−1) = W. A
4
Δf = Δω/2π = γ/2π = f0/Q = 60/5 = 12 Hz
Bandwidth from Δω = γ, converted to ordinary frequency. A
max ≈ 3.0 W,   Δf = 12 Hz (from 54 to 66 Hz)

Reading. A low-Q woofer absorbs strongly but over a broad 12 Hz band — deliberately, so its response is smooth across the bass region rather than a peaky whistle.

1
Quartz tuning-fork resonator. f0 = 32.768 kHz, Q = 8.0×104. A drive is applied 1.0 Hz above resonance. What fraction of the peak power is absorbed, and what is the −3 dB bandwidth?
Set up: ratio /max = (γ/2)2/[(ωω0)2 + (γ/2)2]; work in angular units. B
2
Δf3dB = f0/Q = 32768/8.0×104 = 0.41 Hz,   γ = 2π(0.41) = 2.57 rad s−1
Bandwidth first, since γ follows directly from it. A
3
γ/2 = 1.29 rad s−1,   ωω0 = 2π(1.0) = 6.28 rad s−1
Detuning of 1.0 Hz converted to angular frequency; compare with the half-width. A
4
/max = (1.29)2 / [(6.28)2 + (1.29)2] = 1.66 / (39.4 + 1.66) = 0.040
Evaluate the Lorentzian factor; dimensionless. B
Δf3dB ≈ 0.41 Hz,   /max ≈ 0.040 (4.0%)

Reading. A detuning of only 1 Hz — 30 parts per million — already cuts the absorbed power to 4% of peak. This razor-sharp line is exactly why such a crystal makes a precise clock reference.

Problems
  1. A driven oscillator has m = 0.10 kg, ω0 = 200 rad s−1, γ = 4.0 rad s−1, F0 = 5.0 N. Find max, the FWHM in rad s−1, and Q.
    Solutionmax = F02/2 = 25/[2(0.10)(4.0)] = 25/0.80 = 31.3 W. FWHM Δω = γ = 4.0 rad s−1. Q = ω0/γ = 200/4.0 = 50.
  2. Show that the half-power points of the near-resonance Lorentzian lie at ω = ω0 ± γ/2, and confirm the full width is γ.
    SolutionSet = ½max: (γ/2)2/[(ωω0)2 + (γ/2)2] = ½. Cross-multiplying, 2(γ/2)2 = (ωω0)2 + (γ/2)2, so (ωω0)2 = (γ/2)2, giving ω = ω0 ± γ/2. The separation is γ/2 − (−γ/2) = γ.
  3. An oscillator absorbs 8.0 W at resonance. By what factor does the absorbed power drop when the drive is detuned by exactly one full linewidth, |ωω0| = γ?
    SolutionRatio = (γ/2)2/[γ2 + (γ/2)2] = (1/4)/[1 + 1/4] = (1/4)/(5/4) = 1/5 = 0.20. So = 0.20(8.0) = 1.6 W, a factor-5 drop.
  4. A tuning fork rings down: after being struck, its amplitude falls to 1/e in 3.0 s. Estimate its Q and absorption bandwidth if f0 = 440 Hz. (Amplitude decays as eγt/2.)
    SolutionAmplitude 1/e time: γ/2 = 1/3.0 s−1γ = 0.667 rad s−1. ω0 = 2π(440) = 2765 rad s−1. Q = ω0/γ = 2765/0.667 = 4.15×103. Bandwidth Δf = γ/2π = 0.106 Hz. (Equivalently Δf = f0/Q = 440/4150 = 0.106 Hz.)
  5. Starting from the exact result = (F02/2m)γω2/[(ω02ω2)2 + γ2ω2], prove the sum rule ∫0 (ω) dω = πF02/4m in the weak-damping limit, and comment on its independence from γ.
    SolutionIn the weak-damping limit the integrand is sharply peaked near ω0, so use the Lorentzian form ≈ (F02/2)(γ/2)2/[(ωω0)2 + (γ/2)2] and extend the lower limit to −∞ (negligible error). The standard integral ∫−∞ dx/(x2 + a2) = π/a with a = γ/2 gives ∫ = (F02/2)(γ/2)2(π/(γ/2)) = (F02/2)(γ/2)(π) = πF02/4m. The γ cancels: peak height ∝ 1/γ and width ∝ γ compensate, so total integrated absorption depends only on F0 and m. Sharpening a resonance concentrates, but does not increase, its total absorption.