Wait, what? A photodiode that only measures optical power can tell you whether a laser is slightly too high or slightly too low in frequency relative to an optical cavity—even when the ordinary reflected power looks almost symmetric on either side of resonance.
That is the central trick of Pound–Drever–Hall (PDH) laser frequency stabilisation. The method turns a tiny optical phase change on reflection from a resonator into a signed electronic error signal. A feedback controller then drives the laser so that the error approaches zero.
PDH is best learned as a measurement-and-control chain: laser frequency → phase modulation → cavity-dependent reflected phase → photodetection and demodulation → signed detuning estimate → feedback → reduced frequency noise.
Quick Answer
A laser is phase-modulated before it reaches a high-finesse optical cavity. The modulation creates a carrier plus frequency sidebands. Near a cavity resonance, the carrier experiences a rapidly changing reflection phase while suitably chosen sidebands remain mostly off resonance and act as phase references. The reflected light is detected and electronically mixed with the original modulation reference. Near resonance, the demodulated signal changes sign with laser–cavity detuning. That signed signal is fed back to a laser actuator. The lock can narrow the laser relative to the cavity over the controller’s useful bandwidth, but the final stability cannot be better than the reference cavity, detector, modulation and feedback system allow.
Learning Ladder: Beginner to Professional
| Stage | What the learner should be able to do |
|---|---|
| Beginner | Explain that a feedback lock compares a changing laser frequency with a reference and applies a correction. |
| Secondary | Connect interference, resonance, phase, frequency and feedback without treating “frequency” and “power” as the same observable. |
| JC / A-Level | Use cavity free spectral range, linewidth and phase modulation to explain why sidebands can provide a reference. |
| Undergraduate | Derive the qualitative PDH error signal, distinguish open-loop from closed-loop response and identify major technical noise sources. |
| Advanced / Professional | Read noise spectra and loop transfer functions, diagnose residual amplitude modulation and cavity drift, and separate laser-noise suppression from the reference cavity’s own thermal and mechanical limits. |
1. Start With the Cavity, Not the Electronics
A two-mirror Fabry–Pérot cavity supports resonances when the round-trip optical phase is an integer multiple of 2π. For a simple cavity of optical length nL, the free spectral range is approximately
FSR ≈ c / (2nL).
If the finesse is F, the resonance linewidth is roughly
Δνcav ≈ FSR / F.
A high-finesse cavity therefore provides a narrow frequency discriminator. But a narrow resonance alone does not solve the whole control problem. A useful lock needs a signal that says not only “you are off resonance” but also which side of resonance the laser is on.
2. Why Reflected Power Alone Is Not Enough
If you scan a laser through a cavity resonance, transmitted power has a peak and reflected power has a dip. The trouble is symmetry: the same reflected power can occur at a positive or negative detuning. A controller built only from that intensity would not know which direction to correct.
What changes sign across resonance is the phase response of the reflected optical field. PDH converts that phase information into an intensity beat that an ordinary fast photodiode can measure.
3. Phase Modulation Creates a Carrier and Sidebands
An electro-optic phase modulator can drive the optical phase sinusoidally:
E(t) = E₀ exp[i(ωt + β sin Ωt)].
The result can be expanded into a carrier at angular frequency ω and sidebands at ω ± Ω, ω ± 2Ω and so on, with amplitudes governed by Bessel functions. In a practical introductory PDH picture, the carrier and first sidebands dominate.
It is tempting to say the modulator “adds two little lasers”. It does not. The sidebands are phase-coherent components of the same modulated optical field.
4. The Sidebands Act as Phase References
Choose the modulation frequency Ω so that the sidebands lie well outside the narrow cavity resonance while the carrier can approach resonance. Then the sidebands are mostly reflected without the rapid resonant phase swing experienced by the carrier.
The reflected field therefore contains components that have experienced different cavity phase responses. Their interference on the photodiode carries information about the carrier’s detuning.
5. A Photodiode Measures Beats, Not Optical Phase Directly
A square-law photodiode responds to optical intensity. When the carrier and sidebands arrive together, cross terms in the detected power oscillate at the modulation frequency Ω. The phase and amplitude of that radio-frequency beat depend on the cavity reflection coefficients seen by the optical components.
This is the bridge from optical phase to electronic measurement: phase information becomes a beat-note quadrature.
6. Demodulation Produces the Error Signal
The photodiode signal is mixed with an electronic reference derived from the same oscillator that drove the phase modulator. A low-pass filter removes the high-frequency mixer products. The resulting baseband signal is the PDH error signal.
Near resonance, and for a properly chosen demodulation phase, the signal is approximately linear in small detuning:
Verr ≈ KPDH δν.
The coefficient KPDH is a frequency-discriminator slope measured in volts per hertz. Its exact value depends on optical power, modulation depth, cavity parameters, photodetector gain, mixer gain and demodulation phase.
7. Zero Crossing Is a Control Target, Not Absolute Truth
In the ideal model, zero error corresponds to exact cavity resonance. Real systems can shift that zero. Residual amplitude modulation, etalons, electronic offsets and demodulation-phase errors can make the controller faithfully lock to a point that is not exactly the physical resonance.
A stable zero is not automatically an accurate zero.
8. Residual Amplitude Modulation Is a Major Systematic Error
An ideal phase modulator changes phase only. Real electro-optic modulators can also create residual amplitude modulation (RAM), for example through crystal birefringence, imperfect polarisation, temperature drift or parasitic reflections. RAM can add an offset to the PDH error signal and thereby create an apparent frequency shift.
APS analysis has shown how birefringence and parasitic etalons can convert RAM into frequency instability, which is why precision systems may actively measure and suppress RAM rather than treating it as an invisible nuisance.
9. Feedback Changes the Laser, Not the Cavity Physics
The PDH error signal is sent to a controller. The controller may drive laser injection current, a piezoelectric element, an intracavity electro-optic actuator, an acousto-optic frequency shifter or some combination of fast and slow actuators.
A common architecture uses a fast actuator for high-frequency noise and a slow actuator for large-range drift. The exact implementation is engineering; the Physics is closed-loop suppression of disturbances inside a bandwidth where gain is sufficiently high and phase margin remains stable.
10. Read a Servo as a Dynamical System
If the open-loop transfer function is G(f), a disturbance appearing at the controlled variable is approximately suppressed by the sensitivity function
S(f) = 1 / [1 + G(f)].
Large loop gain can strongly suppress noise, but delay and actuator resonances add phase lag. Increasing gain without checking stability can turn a lock into an oscillator. Professional work therefore treats the controller, actuator, photodetector and cavity discriminator as one transfer chain.
11. “Locked” Does Not Mean Noise-Free
A locked laser still has residual frequency noise. Some frequencies are well suppressed, others lie outside the servo bandwidth, and some noise is injected by the measurement itself. A time trace that appears visually flat can hide important phase noise at frequencies that matter for spectroscopy, interferometry or clocks.
The stronger diagnostic is a frequency-noise or phase-noise spectrum, together with the loop transfer function and a comparison against the free-running laser.
12. The Reference Cavity Eventually Becomes the Limit
PDH can make the laser follow a cavity extremely closely. It cannot make the cavity an immutable ruler. The resonance frequency changes if the optical length changes. Thermal expansion, thermo-refractive fluctuations, Brownian motion of mirror coatings, vibration, residual gas effects and long-term material drift can all move the reference.
NIST chip-scale cavity work demonstrates the modern consequence: once technical laser noise is sufficiently suppressed, intrinsic thermodynamic fluctuations of the resonator can become the floor.
13. Shot Noise Sets a Measurement Floor
Photodetection is quantised. Even a perfectly stable average optical power produces photon-counting fluctuations. When converted through the PDH discriminator slope, detector shot noise corresponds to an equivalent frequency-noise floor.
More optical power can improve shot-noise-limited sensing, but only until other effects—detector saturation, RAM, cavity heating, photothermal shifts or technical intensity noise—become important. “More power is better” is therefore not a universal rule.
14. The Modulation Frequency Is a Physical Design Choice
The modulation frequency should generally exceed the cavity linewidth so that the first sidebands do not share the carrier’s resonant phase response, while remaining compatible with the electro-optic modulator, detector and radio-frequency electronics. It also must avoid accidental overlap with neighbouring cavity modes or unwanted technical resonances.
This is a good example of boundary-condition reasoning: the familiar PDH cartoon is valid only in a regime where the carrier and sidebands interact with the cavity in the intended way.
15. Lock Acquisition and Lock Maintenance Are Different Problems
The PDH error signal is most useful in a neighbourhood around resonance. Far away, the controller may not see a monotonic discriminator. A practical system therefore needs an acquisition strategy that brings the laser into the capture region before high-gain locking takes over.
Once locked, the relevant problem changes to disturbance rejection and drift tracking. A controller designed only for a beautiful small-signal transfer function can still be frustrating if it cannot acquire or recover from interruptions.
16. PDH Is a Relative Frequency Lock
The lock makes the laser follow a cavity resonance. It does not by itself tell you the absolute optical frequency in SI hertz. Absolute or traceable frequency requires another reference chain, such as an atomic transition or an optical frequency comb linked to a time standard.
This distinction is central in metrology: stability, accuracy and traceability are different properties.
17. Optical Frequency Combs Can Verify What PDH Has Stabilised
A frequency comb can compare a cavity-stabilised laser with another optical or microwave reference. NIST spectroscopy work has used PDH cavity locking together with optical-comb calibration to produce high-resolution frequency axes. The comb does not replace the PDH discriminator; it answers a different measurement question.
18. Modern Alternatives Show the Model Has Limits
PDH remains foundational, but it is not the unique solution to ultrastable laser locking. NIST reported in 2025 a compact optoelectronic locking approach that, in its demonstrated system, provided greater suppression of some laser noise than the estimated PDH limit for that architecture while still reaching the cavity thermal-noise floor.
The correct lesson is not “PDH is obsolete”. It is that every frequency lock is an engineered measurement channel with finite signal slope, finite noise, delay and actuator authority. New architectures can trade complexity, integration, bandwidth and noise differently.
Observation Versus Inference
| What is directly observed? | What is inferred through a model? |
|---|---|
| Photodiode voltage versus time | Laser–cavity frequency detuning |
| RF beat amplitude and phase | Cavity reflection phase response |
| Closed-loop noise spectrum | Frequency-noise suppression relative to the reference |
| Error-signal zero crossing | Resonance frequency, only after systematic offsets are bounded |
| Beat note to another reference | Relative frequency stability between the two systems |
Evidence: What Makes a PDH Claim Strong?
- Measured free-running and locked frequency-noise spectra.
- A measured discriminator slope in V/Hz rather than an assumed one.
- Open-loop or closed-loop transfer-function measurements.
- Residual-amplitude-modulation characterisation and long-term drift data.
- Independent beat measurement against another stable laser or frequency comb.
- Actuator calibration and known servo bandwidth.
- Cavity temperature, vibration and environmental records.
- Demonstration that the claimed floor follows the reference-cavity noise rather than detector or electronic noise.
Misconceptions Worth Hunting
- “PDH measures frequency directly.” It measures an optical beat whose calibrated quadrature estimates cavity detuning.
- “Zero error means exact resonance.” Offsets such as RAM can move the zero.
- “A lock removes all laser noise.” Suppression is frequency-dependent and limited by loop gain, delay and measurement noise.
- “The sidebands are separate independent lasers.” They are coherent components of a phase-modulated field.
- “Higher finesse is always better.” It sharpens discrimination but can narrow capture range and increase sensitivity to cavity disturbances.
- “A stable cavity is an absolute frequency standard.” A cavity can be extraordinarily stable while drifting in absolute frequency.
- “More loop gain always improves the lock.” Excess gain with insufficient phase margin can destabilise the system.
- “PDH is the only serious ultrastable-locking method.” Modern alternatives exist and can outperform it in particular architectures.
Transfer Checks
- The error signal stays near zero but an independent frequency comb shows a slow drift. What is the first suspect? The cavity reference itself may be drifting.
- A lock becomes noisier when optical power is increased. Does this contradict shot-noise scaling? No. Technical noise, RAM, heating or detector nonlinearity may now dominate.
- The reflected optical power is the same at +100 kHz and −100 kHz detuning. Can a simple intensity servo determine the correction direction? Not from that symmetric observable alone.
- The modulation sidebands begin to overlap neighbouring cavity modes. Can the standard PDH approximation be trusted unchanged? No. The sideband–cavity interaction must be re-modelled.
- A servo has enormous low-frequency gain but rings strongly after a disturbance. What Physics is missing from a “more gain is better” model? Loop phase, delay and actuator resonances.
- A lock has a narrow instantaneous linewidth but poor day-to-day reproducibility. Which quantity is strong and which may be weak? Short-term stability may be excellent while long-term accuracy or drift is poor.
How We Know the Learning Has Held
A learner should be able to draw the full chain from electro-optic phase modulation to carrier/sidebands, cavity reflection, RF beat, mixer, low-pass error signal, controller and actuator. They should be able to explain why the error changes sign, why a high-finesse cavity is useful, why the lock is relative rather than absolute, and why RAM, detector noise, cavity thermal noise and loop stability place distinct limits on performance.
Model Limits
The clean textbook PDH derivation usually assumes small modulation depth, a simple cavity, well-separated sidebands, ideal phase modulation, linear photodetection and small detuning. Real cavities may be birefringent, multimode or lossy; optical paths may contain parasitic etalons; photodiodes and mixers have finite bandwidth; actuators have resonances; and the cavity reference has its own thermomechanical dynamics. At the professional level, PDH is therefore not one equation but an inverse problem embedded inside a control system.
Research Foundations and Freshness Check
- NIST — Chip-scale ultraprecise laser technologies and cavity locking.
- NIST / Review of Scientific Instruments (2024) — quadrature amplitude modulation for electronic-sideband and PDH locking.
- NIST / Optica (2025) — modern comparison and alternative ultrastable locking architecture.
- Physical Review A — systematic analysis of residual amplitude modulation in PDH stabilisation.
- NIST — PDH-locked frequency-stabilised cavity ring-down spectroscopy.
- LIGO 40 m technical work — practical digital frequency-stabilisation control and actuator limits.
Connect This to the eduKate Physics Estate
This page owns the narrow Physics job of PDH cavity-based laser frequency stabilisation. The broader subject owner remains the Physics hub. For instrument reasoning, continue to Scientific Instrumentation, Imaging & Measurement and Scientific Method, Evidence & Measurement. The existing Lasers and Photonics article retains the broad laser/photonics overview; this page does not replace it.
The Quiet Ending
The beginner sees a laser being “held steady”. The developing physicist sees resonance and sidebands. The advanced learner sees a phase-sensitive discriminator embedded in a feedback loop.
The professional asks the harder question: which part of the measured stability belongs to the laser, which part belongs to the cavity, which part belongs to the detector and servo, and what independent measurement can separate them?
