Interferometry — Comprehensive Guide

Interferometry exploits the wave nature of light to measure physical quantities with sub-wavelength precision. When two coherent beams are superposed, their electric fields add according to their relative phase — producing intensity patterns that encode path-length differences as small as a fraction of a nanometer. This sensitivity makes interferometry the backbone of precision metrology, from testing telescope mirrors to detecting gravitational waves [1, 3].

The principle dates to Thomas Young's double-slit experiment in 1801, which provided the first definitive evidence for the wave theory of light. Albert Michelson extended the concept into a practical measurement tool in the 1880s, using his eponymous interferometer to disprove the luminiferous aether and later to define the meter in terms of optical wavelength [1, 2]. The Fabry–Pérot étalon, introduced in 1899, brought interferometry to spectroscopy by exploiting multiple-beam interference to achieve spectral resolving powers exceeding 10⁶ [2, 9].

Modern interferometry spans an enormous range of applications. Optical shop testing routinely measures surface figure to λ/20 or better across full-aperture optics [4]. Phase-shifting techniques extract quantitative surface height maps with sub-nanometer repeatability [6]. Fiber-optic interferometers sense strain, temperature, and rotation in environments inaccessible to free-space systems [9]. At the extreme, the Laser Interferometer Gravitational-Wave Observatory (LIGO) achieves displacement sensitivity below 10⁻¹⁹ m — a fraction of a proton diameter — by combining kilometer-scale Michelson interferometry with Fabry–Pérot cavity enhancement and quantum noise reduction [3].

This guide covers the physics of two-beam and multiple-beam interference, the major interferometer architectures, fringe analysis and phase measurement techniques, error sources, and practical guidance for selecting and specifying an interferometric system.

2.1Superposition and the Interference Condition

Interference arises when two or more coherent electromagnetic waves overlap in space. Consider two monochromatic plane waves with the same frequency ω and amplitudes E₁ and E₂ arriving at a point P with a phase difference δ. The resultant intensity at P is given by [1, 2]:

Where: I₁, I₂ = individual beam intensities (W/m²), δ = phase difference (rad).

For equal-intensity beams (I₁ = I₂ = I₀), this simplifies to:

The intensity oscillates between a maximum of 4I₀ (constructive interference, δ = 2mπ) and zero (destructive interference, δ = (2m+1)π), where m is an integer called the fringe order.

2.2Optical Path Difference

The phase difference δ between two beams relates to the optical path difference (OPD) through [1, 3]:

Where: λ = vacuum wavelength (m), OPD = optical path difference (m), n₁, n₂ = refractive indices along each path, d₁, d₂ = geometric path lengths (m).

Constructive interference (bright fringes) occurs when OPD = mλ. Destructive interference (dark fringes) occurs when OPD = (m + ½)λ. The OPD concept is central to all interferometric measurement: every architecture encodes the measurand — displacement, surface height, refractive index change, wavelength — as an OPD between two arms [3, 4].

2.3Fringe Visibility and Contrast

Real interferometers rarely achieve perfect constructive and destructive interference. The fringe visibility (or contrast) quantifies how well-modulated the fringe pattern is [1, 2]:

Where: V = visibility (dimensionless, 0 to 1), I_max = maximum fringe intensity, I_min = minimum fringe intensity.

For two equal-intensity, perfectly coherent beams, V = 1 and fringes have maximum contrast. Visibility degrades when beam intensities are unequal, when the source has finite coherence, or when wavefront aberrations scramble the phase relationship across the observation plane. Practical interferometers typically require V > 0.2 for reliable fringe detection and V > 0.5 for quantitative phase measurement [3, 6].

2.4Coherence Requirements

Interference fringes are only observable when the beams maintain a stable phase relationship. This imposes two coherence requirements [1, 5]:

Temporal coherence limits the maximum OPD over which fringes remain visible. The coherence length is set by the source linewidth:

Where: l_c = coherence length (m), Δλ = spectral bandwidth (m), Δν = frequency bandwidth (Hz), c = speed of light (m/s).

A HeNe laser with Δν ≈ 1.5 GHz has l_c ≈ 20 cm, sufficient for most laboratory interferometers. A frequency-stabilized HeNe (Δν ≈ 1 MHz) extends this to hundreds of meters [9]. A white-light source with Δλ ≈ 300 nm has l_c ≈ 1 μm, which is the basis for white-light interferometry and optical coherence tomography — the extremely short coherence length provides axial sectioning capability [7].

Spatial coherence requires that the source subtend a sufficiently small angle as seen from the interference region. Extended or poorly collimated sources reduce fringe visibility because different source points produce shifted fringe patterns that wash out when superposed on the detector.

3.1Michelson Interferometer

The Michelson interferometer is the most widely used two-beam configuration. A beam splitter divides an input beam into two arms: one reflects from a test surface (or mirror), the other from a reference surface. The returning beams recombine at the beam splitter, and the detector records the resulting interference pattern [1, 3].

The round-trip OPD between arms is:

Where: d₁, d₂ = one-way geometric path lengths in arm 1 and arm 2 (m).

The factor of 2 arises because each beam traverses its arm twice (out and back). This double-pass geometry doubles the sensitivity to mirror displacement compared to a single-pass configuration, making the Michelson interferometer particularly effective for displacement measurement and surface testing [4]. The Twyman–Green interferometer is a variant of the Michelson specifically configured for testing optical components: it uses a point source at the focus of a collimating lens, producing a plane-wave input that simplifies fringe interpretation [4].

3.2Mach–Zehnder Interferometer

The Mach–Zehnder interferometer uses two beam splitters and two mirrors arranged so that each beam traverses its arm only once. The input beam is split at BS1; the two beams follow separate paths (typically enclosing a rectangular area) and recombine at BS2. Both output ports carry complementary interference signals [1, 9].

The single-pass geometry means:

Where: n_sample, n_ref = refractive indices along sample and reference arms, d_sample, d_ref = geometric path lengths (m).

Because each beam passes through the sample region only once, the Mach–Zehnder is preferred for studying refractive index changes in gases, plasmas, and fluid flows where double-pass would complicate interpretation. It is also the standard architecture for fiber-optic interferometric sensors and integrated photonic circuits [9].

3.3Fabry–Pérot Interferometer

The Fabry–Pérot interferometer (also called an étalon when the mirror spacing is fixed) consists of two highly reflective, parallel surfaces separated by a gap d. Unlike the Michelson and Mach–Zehnder, it operates on multiple-beam interference: light bounces between the surfaces many times, and the multiply-reflected beams interfere upon transmission [1, 2, 9].

The transmitted intensity follows the Airy function:

Where: I_t = transmitted intensity, I_i = incident intensity, F = coefficient of finesse = 4R/(1−R)², R = mirror reflectance, δ = round-trip phase = 4πnd/λ.

The transmission peaks occur when δ = 2mπ, giving sharp resonances whose width decreases as R increases. The finesse quantifies the sharpness of these resonances [2, 9]:

Where: ℱ = finesse (dimensionless), R = mirror reflectance.

The free spectral range (FSR) — the frequency interval between adjacent transmission peaks — is:

Where: FSR = free spectral range (Hz), c = speed of light (m/s), n = refractive index of the gap medium, d = mirror separation (m).

A Fabry–Pérot with R = 0.97 has ℱ ≈ 100, meaning its transmission peaks are 100× narrower than the FSR. This makes the Fabry–Pérot the instrument of choice for high-resolution spectroscopy, laser cavity design, and wavelength filtering [9].

3.4Sagnac Interferometer

The Sagnac interferometer sends two beams around a closed loop in opposite directions. In the absence of rotation, both beams traverse identical paths and arrive at the detector with zero OPD. Rotation of the loop about an axis perpendicular to its plane introduces a differential path length proportional to the angular velocity [1, 3]:

Where: Δφ = phase difference (rad), A = enclosed area of the loop (m²), λ = wavelength (m), c = speed of light (m/s), Ω = angular velocity (rad/s).

Fiber-optic gyroscopes (FOGs) exploit this effect by winding hundreds of meters of fiber into a coil, multiplying the effective area A by the number of turns. FOGs achieve angular rate sensitivities below 0.001°/hr and are standard in aircraft inertial navigation [9].

3.5Fizeau Interferometer

The Fizeau interferometer places a reference flat in close proximity to the test surface, with a small air gap between them. Collimated light enters from above, reflects from both the reference and test surfaces, and the two reflected beams interfere. The resulting fringe pattern maps the gap variation — and therefore the surface figure of the test optic relative to the reference [4].

Fizeau interferometers are the workhorses of optical shop testing. Because the reference and test surfaces share a common optical path (the input beam), the configuration is inherently insensitive to vibration and source instability — environmental disturbances affect both reflections equally and cancel in the OPD. Commercial Fizeau instruments from Zygo, 4D Technology, and others routinely measure surface figure to λ/20 peak-to-valley over apertures from 25 mm to 800 mm [4, 8].

4.1Fringe Pattern Formation

The fringe pattern observed in an interferometer is a map of the OPD distribution across the observation plane. Each bright fringe connects points of equal OPD — they are contour lines of the optical path difference, analogous to elevation contours on a topographic map. The spacing and curvature of fringes encode information about the relative alignment, surface shape, or refractive index variation between the two arms [3, 4].

Three factors determine fringe geometry: the relative tilt between the interfering wavefronts (producing straight, equally spaced fringes), the relative curvature (producing circular or curved fringes), and higher-order aberrations (producing irregular fringe distortions). A skilled metrologist can diagnose surface errors by visual inspection of the fringe pattern — tilt produces parallel straight fringes, defocus produces concentric rings, astigmatism produces elliptical fringes, and coma produces asymmetric S-shaped distortions [4].

4.2Fringe Spacing from Tilt

When two flat wavefronts intersect at an angle α, the fringe spacing across the detector is [1, 4]:

Where: Λ = fringe spacing (m), λ = wavelength (m), α = tilt angle between wavefronts (rad). The approximation holds for small angles (α « 1 rad).

Each adjacent fringe represents a change of one wavelength in OPD (or λ/2 in surface height for a reflection geometry). Counting fringes across the aperture gives the total tilt in waves; the deviation of fringes from straightness reveals surface figure error [4].

4.3Circular and Localized Fringes

When the two arms produce wavefronts with different curvatures — as occurs when testing a spherical surface against a flat reference — the fringe pattern consists of concentric rings. The ring radii relate to the surface curvature mismatch. Newton's rings, formed between a convex surface and a flat, are a classic example: the radius of the mth bright ring is [1]:

Where: r_m = radius of the mth bright ring (m), R = radius of curvature of the convex surface (m), m = fringe order (integer).

Localized fringes occur in extended-source interferometry (e.g., Fizeau with an extended illuminator) where fringes appear localized at or near the surface being tested. This localization is a coherence phenomenon: fringes are visible only where the OPD is within the coherence length of the source. For broadband or white-light sources, this confines the fringes to a narrow depth range — the principle behind white-light interferometry and vertical scanning interferometry [4, 7].

4.4Reading Surface Figure from Fringes

In optical shop testing, the deviation of fringes from their ideal shape (straight lines for flats, concentric circles for spheres) quantifies surface figure error. The convention is [4, 8]:

Peak-to-valley (PV): The maximum fringe deviation expressed in fractions of a fringe. One fringe of deviation = λ/2 surface error (in reflection).

RMS figure: The root-mean-square surface deviation, typically 3–5× smaller than PV for random figure errors.

A surface specified as “λ/10 PV at 632.8 nm” has a maximum departure of 63.3 nm from the ideal form. In a Fizeau test, this would appear as fringes deviating from straightness by no more than 1/5 of a fringe spacing. Phase-shifting interferometry (Section 5) replaces this visual fringe reading with quantitative phase extraction, improving surface height resolution from ~λ/20 (visual) to ~λ/1000 (PSI) [6, 8].

5.1Phase-Shifting Interferometry (PSI)

Phase-shifting interferometry introduces known, controlled phase steps between successive intensity measurements, enabling the interferometer to extract the phase δ(x, y) at every pixel — not just at fringe locations [6, 4]. This transforms the interferometer from a qualitative fringe viewer into a quantitative surface profiler.

The intensity at any pixel (x, y) for a phase shift of αₖ is [6]:

Where: Iₖ = intensity at the kth measurement, I₀ = mean intensity (DC term), V = fringe visibility, δ = unknown phase to be measured, αₖ = applied phase shift for frame k.

With three or more intensity frames at known phase shifts, the three unknowns (I₀, V, δ) can be solved at every pixel. Phase-shifting is typically implemented by translating the reference mirror with a piezoelectric (PZT) actuator in precise steps — commonly λ/4 (90°) — between camera frames [6, 8].

5.2Common PSI Algorithms

Several algorithms trade off sensitivity, speed, and robustness to calibration error [6]:

Three-step algorithm: Uses frames at α = 0, π/2, π. The phase is:

This is the simplest algorithm but is sensitive to phase-step calibration errors and nonlinear detector response.

Four-step algorithm: Uses frames at α = 0, π/2, π, 3π/2:

The symmetry of the four-step algorithm cancels first-order errors in phase-step calibration [6].

Five-step Hariharan algorithm: Uses frames at α = 0, π/2, π, 3π/2, 2π:

The five-step algorithm is insensitive to both linear phase-step miscalibration and second-harmonic detector errors. It is the most widely used PSI algorithm in commercial instruments [6, 8].

5.3Phase Unwrapping

All PSI algorithms return phase modulo 2π (the wrapped phase). For surfaces with more than λ/2 of height variation, the wrapped phase map contains discontinuities that must be resolved. Phase unwrapping algorithms scan the phase map and add or subtract 2π at each discontinuity to reconstruct the continuous phase surface [6, 4].

Reliable unwrapping requires that the true phase difference between adjacent pixels be less than π — the Nyquist sampling condition applied to interferometry. Surfaces with steep slopes, discontinuities (steps), or isolated defects can violate this condition and cause unwrapping errors. Advanced algorithms use quality-guided unwrapping paths, starting from high-visibility pixels and propagating outward, to minimize error propagation [6].

5.4Dynamic and Single-Shot Interferometry

Conventional temporal PSI requires multiple sequential frames, making it vulnerable to vibration between frames. Several techniques address this limitation:

Simultaneous phase-shifting uses polarization-based beam splitting to capture all required phase-shifted images on a single detector array in a single exposure. The pixelated phase-mask camera (e.g., 4D Technology PhaseCam) places a micro-polarizer array over the detector pixels, routing four phase-shifted channels to interleaved pixel groups [8].

Spatial carrier interferometry introduces a large tilt to create a high-frequency carrier fringe pattern. A single image contains enough fringe data to extract the phase via Fourier-transform analysis — no phase stepping required [4, 6].

White-light interferometry (WLI) uses a broadband source to produce fringes localized at the zero-OPD position. By scanning the reference mirror through the focal range and recording the fringe envelope at each pixel, WLI maps surface height with nanometer vertical resolution and no 2π ambiguity — the coherence envelope provides an absolute height reference. Vertical scanning interferometry (VSI) is the standard implementation for rough surfaces and step heights [7, 8].

6.1Source Selection Criteria

The source determines both the measurement range and the measurement type of an interferometer. Three parameters govern source selection [1, 5, 9]:

Coherence length sets the maximum usable OPD. For equal-path interferometers (Fizeau, Sagnac), coherence requirements are relaxed because both arms have nearly equal path lengths. For unequal-path configurations (Michelson with large scan range, Mach–Zehnder with thick samples), the source must have a coherence length exceeding the maximum OPD.

Wavelength stability determines measurement repeatability. A wavelength drift of Δλ/λ causes a proportional error in all OPD measurements. Stabilized HeNe lasers (Δλ/λ < 10⁻⁸) are the standard for high-accuracy interferometry [4, 8].

Power and beam quality affect fringe visibility and detector signal-to-noise ratio. Gaussian TEM₀₀ beams produce uniform illumination across the test aperture after expansion. Multimode lasers introduce speckle that degrades fringe visibility.

6.2Laser Sources for Interferometry

The HeNe laser at 632.8 nm remains the dominant interferometric source due to its excellent coherence (l_c > 20 cm for single longitudinal mode), wavelength stability, Gaussian beam profile, and low cost [9]. Frequency-stabilized HeNe lasers locked to an iodine hyperfine transition achieve Δν < 1 MHz (l_c > 300 m), enabling long-path interferometry such as large-aperture testing and geodetic measurement [4].

Diode lasers offer compact size and wavelength tunability but typically have shorter coherence lengths (1–10 mm for multimode, up to several meters for external-cavity single-mode designs). Wavelength tuning enables frequency-scanning interferometry, where the phase shift is introduced by sweeping the laser wavelength rather than moving a mirror [9].

Nd:YAG lasers at 1064 nm (and their 532 nm doubled output) are used where high power or specific wavelengths are needed, such as in gravitational-wave detectors (LIGO uses 1064 nm Nd:YAG with output power > 200 W after amplification) [3].

6.3Broadband and White-Light Sources

Broadband sources — LEDs, superluminescent diodes (SLDs), halogen lamps — have coherence lengths from 1 μm (halogen) to 30 μm (SLD). This short coherence is a feature, not a limitation: fringes appear only when the two arm lengths are matched to within the coherence length, providing absolute position sensing without 2π ambiguity [7].

White-light interferometry is the basis for vertical scanning interferometry (VSI) and optical coherence tomography (OCT). VSI uses a broadband source in a Mirau or Linnik microscope objective to profile surfaces with nanometer vertical resolution over measurement ranges of millimeters — far exceeding the λ/2 unambiguous range of monochromatic PSI [7, 8].

OCT extends the same principle to biomedical imaging, using SLD or swept-source lasers to perform depth-resolved imaging of tissue microstructure with 1–15 μm axial resolution. The axial resolution is determined by the source coherence length, not the focusing optics — a key advantage for imaging through scattering media [9].

6.4Coherence-Gated Techniques

The ability to selectively detect light from a specific depth (determined by path-length matching) is called coherence gating. Low-coherence interferometry (LCI) uses this principle for absolute distance measurement, thin-film thickness measurement, and depth-resolved imaging [7].

In a Michelson LCI system, the reference mirror is scanned, and the fringe envelope is recorded. The peak of the envelope locates the zero-OPD position with precision limited by the coherence length. For an SLD with Δλ = 40 nm centered at 840 nm, the coherence length is approximately 7.8 μm — setting the axial resolution of the measurement.

7.1Environmental Errors

Environmental disturbances are the dominant error source in most interferometric measurements. Three mechanisms contribute [4, 7, 8]:

Vibration displaces optical components between phase-measurement frames, introducing random OPD errors. A vibration amplitude of just 10 nm at the HeNe wavelength corresponds to a phase error of ~0.1 rad — sufficient to corrupt PSI measurements. Passive vibration isolation (pneumatic tables) attenuates floor vibration above the table's natural frequency (typically 1–2 Hz), but acoustic disturbances and building sway require additional measures: short measurement times, simultaneous phase capture, or active vibration cancellation [4].

Air turbulence causes local refractive index fluctuations along the beam path. The refractive index of air at standard conditions (20°C, 101.325 kPa) is approximately 1.000272 at 632.8 nm. A 1°C temperature gradient across a 100 mm beam path changes the optical path by [4, 8]:

Where: d = beam path length in air (m), dn/dT ≈ −0.93 × 10⁻⁶ /°C (for air near 20°C at 632.8 nm), ΔT = temperature change (°C).

For d = 100 mm and ΔT = 1°C, ΔOPD ≈ 93 nm — roughly λ/7 at HeNe. Enclosing the interferometer in a thermal housing, purging with dry nitrogen, or averaging over turbulence time scales (tens of seconds) mitigates this error.

Thermal drift of the mechanical structure changes the cavity length over time. Invar or Zerodur cavity spacers (CTE < 0.05 × 10⁻⁶ /°C) minimize structural drift in high-precision instruments. Even so, a 0.1°C change over a 200 mm Invar spacer produces ΔOPD ≈ 1 nm [4].

7.2Systematic Errors

Systematic errors persist even in a perfectly stable environment [4, 6, 8]:

Retrace error occurs when the test and reference beams travel different paths through the interferometer optics (beam splitter, transmission flat, imaging lens). Aberrations in these shared optics affect the two beams differently depending on their angle of incidence, introducing a repeatable but measurable wavefront error. High-quality transmission flats with surface figure < λ/20 minimize retrace error.

Reference surface error sets a floor for absolute accuracy. A Fizeau reference flat with λ/20 figure error limits the absolute accuracy to λ/20 even if the measurement repeatability is λ/1000. Calibrating out the reference surface using three-flat testing or absolute measurement techniques removes this limitation [4].

Ghost reflections from uncoated surfaces in the optical path create parasitic fringes that corrupt the primary interference pattern. Anti-reflection coatings on all non-functional surfaces and careful optical design suppress ghosts to acceptable levels.

PZT calibration error in phase-shifting interferometers produces systematic phase errors if the actual step size deviates from the assumed value. The Hariharan 5-step algorithm (Section 5.2) suppresses this error to second order; less robust algorithms may require real-time step monitoring with a reference interferometer [6].

7.3Detector and Sampling Limitations

Nyquist sampling requires at least two pixels per fringe period. If the fringe density exceeds the detector's spatial sampling rate, aliasing corrupts the phase measurement. For a detector with pixel pitch p, the maximum measurable fringe frequency is 1/(2p), setting a limit on the tilt and surface slope that can be measured [6].

Detector nonlinearity (deviation from a linear intensity-to-signal response) introduces harmonics into the PSI measurement. For CCD cameras with < 1% nonlinearity, the resulting phase error is typically < 0.01 rad. For critical applications, lookup-table correction or algorithms inherently robust to harmonics (Hariharan 5-step) are employed [6].

Quantization noise from the analog-to-digital converter contributes phase noise of approximately π/(√6 × 2^N) for an N-bit digitizer. A 12-bit camera contributes ~0.0004 rad of phase noise per pixel — negligible in practice [6].

8.1Beam Splitters

The beam splitter is the defining optical element of most interferometers. Cube beam splitters provide convenient mounting but introduce differential dispersion between the reflected and transmitted paths (the reflected beam traverses the hypotenuse coating once; the transmitted beam passes through the full glass volume). Plate beam splitters minimize glass path but require a compensator plate in the reference arm (as in the Michelson interferometer) to equalize dispersion. Pellicle beam splitters eliminate glass-path effects entirely but are fragile and limited in aperture [1, 9].

For interferometric applications, the splitting ratio should be close to 50:50 (R:T) at the operating wavelength to maximize fringe visibility. A 60:40 imbalance reduces visibility from 1.0 to 0.98 — negligible. A 90:10 imbalance reduces visibility to 0.60, which is still usable but degrades PSI measurement noise [3].

8.2Reference Optics

Reference flats and reference spheres define the wavefront that the test surface is compared against. Surface figure quality directly limits measurement accuracy:

• λ/20 PV — standard commercial grade (Zygo, Edmund Optics)

• λ/40 to λ/100 PV — metrology grade, individually certified with interferometric test reports

• λ/200+ — ultra-precision, available from specialized manufacturers (Zygo VeriFire, SIOS)

Reference optics must be made from low-expansion materials (fused silica, Zerodur) to minimize thermal figure distortion. The clear aperture should exceed the test aperture by at least 10% to avoid edge diffraction effects [4, 8].

8.3PZT Actuators and Phase Modulators

Piezoelectric actuators translate the reference mirror for phase-shifting. Key specifications include:

• Travel range: 1–2 μm minimum (several wavelengths of phase shift)

• Resolution: < 0.1 nm (sub-nanometer positioning for precise phase steps)

• Linearity: < 1% over the operating range (or closed-loop correction)

• Bandwidth: > 100 Hz for vibration-insensitive rapid acquisition

• Hysteresis: < 2% open-loop; closed-loop PZTs with capacitive or strain-gauge feedback achieve < 0.1% [8]

Ring PZTs mounted behind the reference flat provide axial translation without tip/tilt coupling. Tubular PZTs offer higher bandwidth for dynamic measurements.

8.4Detectors and Cameras

Interferometric cameras must provide [6, 8]:

• Adequate pixel count: 512 × 512 minimum for surface mapping; 1024 × 1024 or higher for detailed figure analysis

• Bit depth: 10–14 bits for sufficient intensity quantization

• Frame rate: 30–60 fps for temporal PSI; > 1000 fps for dynamic/vibration-insensitive acquisition

• Low noise: Read noise < 10 electrons for high-visibility fringes

• Global shutter: Essential for simultaneous PSI; rolling shutter introduces row-dependent phase errors

CCD cameras remain common in metrology instruments for their excellent linearity. CMOS cameras offer higher frame rates and are increasingly used in dynamic interferometers [8].

8.5Environmental Enclosures

For measurements approaching λ/100 accuracy, the interferometer must be enclosed to suppress air turbulence and acoustic vibration. Enclosures range from simple acrylic covers (reducing convective airflow) to sealed chambers with active temperature control (±0.01°C stability). High-precision systems purge with dry nitrogen or helium to reduce the refractive index of the gas path and eliminate humidity effects [4, 8].

9.1Surface Figure Testing

Surface figure measurement is the most widespread application of interferometry in optics manufacturing. Fizeau and Twyman–Green interferometers measure the departure of a polished surface from its intended form — flat, spherical, or aspheric — with resolution from λ/20 (visual fringe counting) to λ/1000 (PSI). Results are reported as peak-to-valley (PV) and RMS surface figure in waves or nanometers [4, 8].

Standard practice: a Fizeau interferometer with a HeNe source, λ/20 reference flat, megapixel camera, and Hariharan PSI algorithm. The instrument measures the full aperture in a single acquisition, producing a surface height map from which PV, RMS, power, astigmatism, and Zernike decomposition are computed automatically [8].

9.2Thin-Film Thickness Measurement

Thin transparent films produce interference between reflections from the top and bottom surfaces of the film. The reflectance spectrum oscillates with wavelength, and the period of oscillation encodes the optical thickness [9]:

Where: n_f = film refractive index, d_f = film thickness (m), m = fringe order (integer), λ_m = wavelength of the mth reflectance extremum.

By measuring the spacing between adjacent reflectance peaks (Δ(1/λ) = 1/(2n_f d_f)), the optical thickness is determined without knowing the absolute fringe order. Spectrophotometric interferometry applies this principle using a broadband source and spectrometer to measure films from ~100 nm to hundreds of micrometers [9].

9.3Displacement and Distance Measurement

Heterodyne laser interferometers measure displacement by counting fringes as a mirror moves along the beam axis. Each fringe corresponds to λ/2 of displacement (double-pass). With electronic fringe interpolation, resolution reaches λ/1000 (~0.3 nm at 632.8 nm) [4, 9].

Displacement interferometers are the position feedback sensors in semiconductor lithography stages, coordinate measuring machines (CMMs), and precision machine tools. Agilent/Keysight and Renishaw produce commercial systems with measurement ranges up to meters and velocities up to 4 m/s, achieving accuracy of ±0.1 ppm (0.1 μm per meter) with environmental compensation [9].

9.4Fiber-Optic Interferometric Sensors

Fiber-optic interferometers replace free-space beams with optical fibers, enabling sensing in confined or harsh environments. Key configurations include [9]:

• Fiber Bragg gratings (FBGs): Periodic refractive-index modulations in the fiber core reflect a narrow wavelength band. Strain or temperature shifts the Bragg wavelength, detected by spectral interrogation.

• Fiber Fabry–Pérot sensors: A cavity formed between two fiber end-faces (or between a fiber tip and a reflective surface) senses displacement, pressure, or temperature via the cavity-length change.

• Fiber Sagnac gyroscopes (FOGs): Coiled fiber loops detect rotation via the Sagnac effect. Military-grade FOGs achieve bias stability < 0.001°/hr.

9.5Gravitational Wave Detection

LIGO operates as a modified Michelson interferometer with 4 km arms, Fabry–Pérot cavities in each arm (increasing effective path length to ~1000 km), power recycling, and signal recycling mirrors. The target displacement sensitivity is ~10⁻¹⁹ m/√Hz — achieved through a combination of high laser power (200 W), quantum-noise reduction (squeezed-light injection), multi-stage seismic isolation, and ultra-low-loss mirror coatings [3].

9.6Optical Coherence Tomography (OCT)

OCT uses low-coherence interferometry to produce depth-resolved images of semi-transparent samples (primarily biological tissue). Time-domain OCT scans the reference arm; spectral-domain OCT (SD-OCT) uses a fixed reference and analyzes the spectral interference pattern with a spectrometer, achieving faster acquisition. Swept-source OCT (SS-OCT) tunes the laser wavelength rapidly, offering the highest imaging speeds (> 100 kHz A-scan rate). Axial resolution is determined by the source bandwidth: Δz ≈ 0.44λ²/Δλ [7, 9].

10.1Selection Decision Workflow

Selecting the appropriate interferometric technique requires matching the measurement need to the architecture's strengths. The decision begins with four questions [4, 8]:

1. What is being measured? Surface figure, displacement, film thickness, refractive index, rotation rate, spectral content.

2. What precision is required? λ/20 PV (visual fringe reading), λ/100 (PSI), sub-nanometer (heterodyne displacement), 10⁻¹⁹ m (gravitational waves).

3. What is the sample geometry? Flat, spherical, aspheric, rough, transparent, reflective, in situ.

4. What are the environmental constraints? Vibration level, temperature stability, access limitations, measurement speed requirements.

10.2Architecture Matching

The answers to these questions map to architecture choices:

• Surface figure of polished flats and spheres → Fizeau interferometer with PSI (standard optical shop tool)

• Testing lenses, prisms, optical components → Twyman–Green interferometer

• Displacement or position feedback → Heterodyne Michelson interferometer (homodyne for lower cost)

• Refractive index change or gas-flow visualization → Mach–Zehnder interferometer

• Thin-film thickness → Spectrophotometric interferometry or white-light interferometry

• Surface roughness or step heights on rough surfaces → Vertical scanning interferometry (WLI)

• Rotation sensing → Fiber Sagnac interferometer

• High-resolution spectroscopy → Fabry–Pérot étalon or scanning Fabry–Pérot

• Depth-resolved biological imaging → OCT (SD-OCT or SS-OCT)

• High-vibration environments → Simultaneous PSI (pixelated-mask camera) or spatial-carrier single-shot

10.3Specification Checklist

When specifying an interferometric system, the following parameters must be defined:

• Aperture: Clear aperture matching the largest test piece, plus 10% margin

• Wavelength: 632.8 nm (standard), 1064 nm (high power, low scatter), 532 nm (higher resolution), broadband (WLI/OCT)

• Reference quality: λ/20 (routine), λ/50 (precision), λ/100+ (calibration-grade)

• Phase measurement: Temporal PSI (stable environment), simultaneous PSI (vibration), WLI (rough surfaces, steps)

• Camera: Pixel count, frame rate, bit depth per the analysis requirements

• Environmental control: Enclosure, temperature stability, vibration isolation, purge gas

• Software: Zernike decomposition, surface statistics (PV, RMS, power spectral density), data export formats