Topographic Interferometer Overview

• Topographic interferometers are instruments that measure surface heights by encoding geometry into observables such as phase, coherence envelopes, or synthetic wavelengths across optical, radar, and quantum systems.
• They utilize diverse architectures including Michelson configurations, shearing interferometry, and SAR-based systems, each tailored for absolute height or gradient detection.
• These systems achieve nanometer-level precision and robust performance by optimizing measurement pipelines to mitigate challenges like phase noise, dispersion, and surface roughness.

Searching arXiv for recent and foundational papers on topographic interferometry and related interferometric topography methods. Search query: topographic interferometer surface topography interferometry arXiv A topographic interferometer is an interferometric instrument or architecture that measures spatially resolved surface height, height gradients, elevation, or related topographic observables by encoding geometry into phase, coherence envelope, synthetic-wavelength phase, cross-detector correlations, or Doppler-derived interferometric quantities. Across optical, radar, and quantum-inspired settings, the common structure is that a reference relation—optical path difference, synthetic phase, mutual coherence, or interferometric Doppler-rate—is converted into a topographic quantity such as $z(x,y)$, $h$, $\partial_x h$, $\partial_y h$, or elevation angle. The literature spans full-field low-coherence interferometry for noncontact surface metrology (Pernechele et al., 2018), synthetic-wavelength full-field sensing under aberrations and subsurface scattering (Kotwal et al., 2022), multi-band frequency-scanning interferometry for absolute surface profile recovery (Peca et al., 2016), passive thermal topography via cross-detector correlations and balanced homodyne detection (bao et al., 1 Jun 2026), interferometric synthetic aperture radar elevation mapping (Kabuli et al., 14 Jan 2025), polarization-based shearing interferometry for reconstructing surface deformation from gradients (Beyersdorf et al., 2012), Doppler-SAR interferometry using ultra-narrowband continuous waves (Yazici et al., 2017), and a broader classification of interferometric and non-holographic topography methods by their dominant precision limits (Häusler et al., 2021).

1. Definition and scope

In reflective optical topography, a topographic interferometer commonly maps optical path difference to height. For low-coherence interferometry, the standard relation is

$$\mathrm{OPD}(x,y) = 2\,n\,z(x,y) \quad \Rightarrow \quad z(x,y) = \frac{\mathrm{OPD}(x,y)}{2n},$$

with $n \approx 1$ in air (Pernechele et al., 2018). In phase-based reflective interferometry, the corresponding phase–height relation is

$z = \frac{\lambda}{4\pi n \cos\theta}\,\varphi,$

and axial precision follows

$$\sigma_z = \frac{\lambda}{4\pi n \cos\theta}\,\sigma_\varphi$$

(Häusler et al., 2021). In synthetic-wavelength interferometry, depth is encoded in the phase difference between two narrow optical lines, so that

$$z = \frac{\phi_{\mathrm{syn}}\lambda_{\mathrm{syn}}}{4\pi n}$$

for near-normal reflection geometry (Kotwal et al., 2022).

The term also applies beyond optical metrology. In automotive interferometric synthetic aperture radar, interferometric phase is linked to height through baseline geometry, with the paper’s vertical-baseline model

$\Delta\psi = 4\pi (D_v/\lambda) \sin \phi$

and height obtained from spherical-to-Cartesian conversion after estimating elevation $h$0 (Kabuli et al., 14 Jan 2025). In Doppler-SAR interferometry, height is recovered from the intersection of iso-Doppler, iso-Doppler-rate, and interferometric Doppler-rate surfaces rather than from conventional range difference (Yazici et al., 2017). In the quantum-inspired formulation, the “topographic interferometer” does not attempt to reconstruct the full distance profile $h$1; instead, it reads out correlations and quadratures selectively sensitive to shape or topography (bao et al., 1 Jun 2026).

This suggests that the defining property of a topographic interferometer is not a single optical layout, but the use of interferometric observables to recover a topographic quantity.

2. Measurement principles and governing relations

A broad division in the literature is between systems that encode absolute or relative height, systems that encode coherence-gated axial position, and systems that encode gradients or moments.

In low-coherence topographic interferometry, interference occurs only when the optical path difference lies within the source’s short coherence length. The interferometric intensity at a pixel is described by

$h$2

where $h$3 provides coherence gating (Pernechele et al., 2018). For a Gaussian-spectrum source, the axial resolution is written as

$h$4

(Pernechele et al., 2018). In this class, absolute height is recovered from the coherence-envelope peak rather than from phase unwrapping.

In synthetic-wavelength interferometry, two closely spaced wavelengths $h$5 and $h$6 define

$h$7

and the synthetic phase obeys

$h$8

for reflection-mode near-normal incidence (Kotwal et al., 2022). In two- and multi-wavelength rough-surface methods, the same synthetic-wavelength construction is used to extend unambiguous range, but the broader metrology discussion emphasizes that the ultimate precision on rough surfaces is limited by roughness, with

$h$9

for two-wavelength holography/interferometry on rough surfaces (Häusler et al., 2021).

In frequency-scanning interferometry, per-pixel intensity varies with optical frequency according to

$\partial_x h$0

so that

$\partial_x h$1

and, for near-normal reflection in air,

$\partial_x h$2

(Peca et al., 2016). This replaces phase-shifting hardware with frequency scans and enables absolute per-pixel phase recovery by bridging multiple narrow frequency bands.

In shearing interferometry, the measured quantity is a finite difference of wavefront phase rather than height itself. The phase difference is

$\partial_x h$3

with reconstruction of height after estimating both gradient components and integrating them (Beyersdorf et al., 2012). In slope methods more generally, the broader metrology analysis states

$\partial_x h$4

(Häusler et al., 2021).

In the quantum-inspired formulation, the crucial observable is the cross-detector correlation, with

$\partial_x h$5

and the interferometer is constructed so that balanced homodyne outputs isolate the quadrature that carries topographic dependence through $\partial_x h$6 and $\partial_x h$7 (bao et al., 1 Jun 2026). This is explicitly different from distance-centric triangulation.

3. Representative architectures

A topographic interferometer appears in several recurring architectures.

Michelson full-field systems are central in the optical metrology papers. The 2D-CMOS low-coherence instrument is a classical Michelson interferometer illuminated by a super-luminescent light-emitting diode at $\partial_x h$8 nm with reported coherence length $\partial_x h$9, a telecentric objective, and a 2D CMOS camera of $\partial_y h$0 pixels with $\partial_y h$1 pixel pitch (Pernechele et al., 2018). Swept-Angle Synthetic Wavelength Interferometry also uses a Michelson configuration, but with two distributed-Bragg-reflector lasers near $\partial_y h$2 nm, a dual-axis galvo scanning the illumination angle, and full-field CCD acquisition (Kotwal et al., 2022). Multi-band frequency-scanning interferometry uses a Michelson imaging interferometer fed by three temperature-tuned DFB diode lasers at nominal center wavelengths $\partial_y h$3 nm, $\partial_y h$4 nm, and $\partial_y h$5 nm (Peca et al., 2016).

Gradient-sensitive common-path systems are represented by polarization-based shearing interferometry. There, a negative uniaxial birefringent crystal generates lateral shear by spatial walk-off, and a z-cut LiNbO$\partial_y h$6 electro-optic modulator modulates the relative phase between polarization components (Beyersdorf et al., 2012). The crystal can be rotated by $\partial_y h$7 to measure $\partial_y h$8 and $\partial_y h$9 independently (Beyersdorf et al., 2012).

Radar topographic interferometers include both baseline-phase and Doppler-rate forms. Automotive elevation mapping with interferometric synthetic aperture radar uses two TI AWR1243BOOST sensors operating in $$\mathrm{OPD}(x,y) = 2\,n\,z(x,y) \quad \Rightarrow \quad z(x,y) = \frac{\mathrm{OPD}(x,y)}{2n},$$0–$$\mathrm{OPD}(x,y) = 2\,n\,z(x,y) \quad \Rightarrow \quad z(x,y) = \frac{\mathrm{OPD}(x,y)}{2n},$$1 GHz, a synthetic aperture of about $$\mathrm{OPD}(x,y) = 2\,n\,z(x,y) \quad \Rightarrow \quad z(x,y) = \frac{\mathrm{OPD}(x,y)}{2n},$$2 m per image frame, and vertical baselines with spacing $$\mathrm{OPD}(x,y) = 2\,n\,z(x,y) \quad \Rightarrow \quad z(x,y) = \frac{\mathrm{OPD}(x,y)}{2n},$$3 (Kabuli et al., 14 Jan 2025). Doppler-SAR interferometry instead uses two monostatic antennas configured with different velocities, and the interferometric phase is tied to a difference in Doppler-rate rather than to conventional wideband range difference (Yazici et al., 2017).

Quantum-inspired multiarm architectures reinterpret stereo paths as interferometric arms. The proposed topographic interferometer uses three input ports aligned along $$\mathrm{OPD}(x,y) = 2\,n\,z(x,y) \quad \Rightarrow \quad z(x,y) = \frac{\mathrm{OPD}(x,y)}{2n},$$4: left, central, and right. The left and right ports feed the lower beam splitter to form Mach–Zehnder arms, the central path serves as the bright local oscillator, and balanced detectors at the outputs measure differential homodyne currents (bao et al., 1 Jun 2026).

Architecture	Core observable	Representative paper
Michelson low-coherence full-field interferometer	Coherence-gated OPD peak	(Pernechele et al., 2018)
Michelson synthetic-wavelength full-field interferometer	Synthetic phase $$\mathrm{OPD}(x,y) = 2\,n\,z(x,y) \quad \Rightarrow \quad z(x,y) = \frac{\mathrm{OPD}(x,y)}{2n},$$5	(Kotwal et al., 2022)
Michelson imaging FSI interferometer	Phase slope versus frequency	(Peca et al., 2016)
Polarization-based shearing interferometer	Wavefront gradient via shear	(Beyersdorf et al., 2012)
InSAR / Doppler-SAR interferometer	Baseline phase or Doppler-rate phase	(Kabuli et al., 14 Jan 2025, Yazici et al., 2017)
Mach–Zehnder plus balanced homodyne TI	Cross-detector correlations and quadratures	(bao et al., 1 Jun 2026)

These architectures differ in what they measure directly, but each uses an interferometric reference channel to produce a topographic estimate.

4. Acquisition and reconstruction pipelines

The 2D-CMOS low-coherence system performs an axial scan of the reference arm while recording interferograms at $$\mathrm{OPD}(x,y) = 2\,n\,z(x,y) \quad \Rightarrow \quad z(x,y) = \frac{\mathrm{OPD}(x,y)}{2n},$$6 fps with an axial step of $$\mathrm{OPD}(x,y) = 2\,n\,z(x,y) \quad \Rightarrow \quad z(x,y) = \frac{\mathrm{OPD}(x,y)}{2n},$$7 per frame. For each pixel, the system assembles an intensity-versus-delay sequence, locates the coherence-envelope peak, and converts OPD to height via $$\mathrm{OPD}(x,y) = 2\,n\,z(x,y) \quad \Rightarrow \quad z(x,y) = \frac{\mathrm{OPD}(x,y)}{2n},$$8 (Pernechele et al., 2018). The paper notes that acquisition and processing are “performed almost in real time,” and the example scan covers $$\mathrm{OPD}(x,y) = 2\,n\,z(x,y) \quad \Rightarrow \quad z(x,y) = \frac{\mathrm{OPD}(x,y)}{2n},$$9 in $n \approx 1$0 steps, taking about $n \approx 1$1 s (Pernechele et al., 2018).

SA-SWI uses a two-stage phase-demodulation pipeline. First, for each synthetic bucket position, it acquires $n \approx 1$2 carrier-shifted frames and estimates the interference-free image and the envelope-squared amplitude,

$n \approx 1$3

$n \approx 1$4

Second, it retrieves the envelope phase through generalized $n \approx 1$5-shift demodulation,

$n \approx 1$6

and then converts to depth (Kotwal et al., 2022). The practical implementation uses $n \approx 1$7 shifts, or $n \approx 1$8 images per depth map (Kotwal et al., 2022).

Frequency-scanning interferometry estimates phase within each narrow wavelength band, then bridges phase between bands using

$n \approx 1$9

The prerequisite to bridge a gap without losing the integral $z = \frac{\lambda}{4\pi n \cos\theta}\,\varphi,$0 count is

$z = \frac{\lambda}{4\pi n \cos\theta}\,\varphi,$1

and, after bridging, the OPD error contracts to

$z = \frac{\lambda}{4\pi n \cos\theta}\,\varphi,$2

(Peca et al., 2016). This is the paper’s mechanism for absolute, per-pixel phase recovery without conventional phase-shifting hardware.

In polarization-based shearing interferometry, the interferogram is modulated as $z = \frac{\lambda}{4\pi n \cos\theta}\,\varphi,$3, and per-pixel Fourier analysis yields the first harmonics

$z = \frac{\lambda}{4\pi n \cos\theta}\,\varphi,$4

Phase difference is then recovered by

$z = \frac{\lambda}{4\pi n \cos\theta}\,\varphi,$5

gradients are formed from $z = \frac{\lambda}{4\pi n \cos\theta}\,\varphi,$6, and the full wavefront is reconstructed in the Fourier domain via

$z = \frac{\lambda}{4\pi n \cos\theta}\,\varphi,$7

(Beyersdorf et al., 2012).

In automotive InSAR, the pipeline comprises motion sensing and compensation, Fast Back-Projection per virtual element, interferogram formation via $z = \frac{\lambda}{4\pi n \cos\theta}\,\varphi,$8, phase-to-elevation inversion by

$z = \frac{\lambda}{4\pi n \cos\theta}\,\varphi,$9

filtering and coherence management, and 3D point generation through

$$\sigma_z = \frac{\lambda}{4\pi n \cos\theta}\,\sigma_\varphi$$0

(Kabuli et al., 14 Jan 2025).

In Doppler-SAR interferometry, image formation is followed by phase equalization, interferogram formation, phase flattening, and height inversion using the simultaneous solution of iso-Doppler, iso-Doppler-rate, and interferometric Doppler-rate equations (Yazici et al., 2017). In the quantum-inspired approach, the pipeline is formulated in terms of POVMs, homodyne samples, and moment estimation rather than pixelwise disparity or phase unwrapping (bao et al., 1 Jun 2026).

5. Performance regimes and precision limits

The reported performance of topographic interferometers depends strongly on which observable is measured.

For the 2D-CMOS low-coherence system, the reported field of view is about $$\sigma_z = \frac{\lambda}{4\pi n \cos\theta}\,\sigma_\varphi$$1 mm $$\sigma_z = \frac{\lambda}{4\pi n \cos\theta}\,\sigma_\varphi$$2 $$\sigma_z = \frac{\lambda}{4\pi n \cos\theta}\,\sigma_\varphi$$3 mm, or about $$\sigma_z = \frac{\lambda}{4\pi n \cos\theta}\,\sigma_\varphi$$4 cm$$\sigma_z = \frac{\lambda}{4\pi n \cos\theta}\,\sigma_\varphi$$5, with lateral sampling of $$\sigma_z = \frac{\lambda}{4\pi n \cos\theta}\,\sigma_\varphi$$6 (Pernechele et al., 2018). On a certified optical flat, the RMS along a $$\sigma_z = \frac{\lambda}{4\pi n \cos\theta}\,\sigma_\varphi$$7-pixel profile is $$\sigma_z = \frac{\lambda}{4\pi n \cos\theta}\,\sigma_\varphi$$8 nm for orthogonal alignment and $$\sigma_z = \frac{\lambda}{4\pi n \cos\theta}\,\sigma_\varphi$$9 nm for the tilted case with $$z = \frac{\phi_{\mathrm{syn}}\lambda_{\mathrm{syn}}}{4\pi n}$$0 fringes across the field (Pernechele et al., 2018). The measured slope limit on a calibrated sphere is $$z = \frac{\phi_{\mathrm{syn}}\lambda_{\mathrm{syn}}}{4\pi n}$$1 (Pernechele et al., 2018).

SA-SWI reports full-frame depth recovery at a lateral and axial resolution of $$z = \frac{\phi_{\mathrm{syn}}\lambda_{\mathrm{syn}}}{4\pi n}$$2 microns and frame rates of $$z = \frac{\phi_{\mathrm{syn}}\lambda_{\mathrm{syn}}}{4\pi n}$$3 Hz, even under strong ambient light (Kotwal et al., 2022). For a highly scattering “chocolate” sample moved in $$z = \frac{\phi_{\mathrm{syn}}\lambda_{\mathrm{syn}}}{4\pi n}$$4 steps, the method achieved MedAE values of about $$z = \frac{\phi_{\mathrm{syn}}\lambda_{\mathrm{syn}}}{4\pi n}$$5, $$z = \frac{\phi_{\mathrm{syn}}\lambda_{\mathrm{syn}}}{4\pi n}$$6, $$z = \frac{\phi_{\mathrm{syn}}\lambda_{\mathrm{syn}}}{4\pi n}$$7, and $$z = \frac{\phi_{\mathrm{syn}}\lambda_{\mathrm{syn}}}{4\pi n}$$8, depending on Gaussian blur kernel width, with corresponding RMSE values of about $$z = \frac{\phi_{\mathrm{syn}}\lambda_{\mathrm{syn}}}{4\pi n}$$9, $\Delta\psi = 4\pi (D_v/\lambda) \sin \phi$0, $\Delta\psi = 4\pi (D_v/\lambda) \sin \phi$1, and $\Delta\psi = 4\pi (D_v/\lambda) \sin \phi$2 (Kotwal et al., 2022). The unambiguous range in a microscopic setting is about $\Delta\psi = 4\pi (D_v/\lambda) \sin \phi$3 for suitable wavelength separation, while a macroscopic synthetic wavelength of about $\Delta\psi = 4\pi (D_v/\lambda) \sin \phi$4 mm enables centimeter-range scenes at about $\Delta\psi = 4\pi (D_v/\lambda) \sin \phi$5 depth accuracy (Kotwal et al., 2022).

Multi-band FSI demonstrates absolute, unambiguous OPD over a total range of $\Delta\psi = 4\pi (D_v/\lambda) \sin \phi$6 mm with surface height precision $\Delta\psi = 4\pi (D_v/\lambda) \sin \phi$7 nm RMS in flat, near-parallel areas (Peca et al., 2016). After joining all three bands, the refined OPD estimate reaches $\Delta\psi = 4\pi (D_v/\lambda) \sin \phi$8 RMS before final absolute extrapolation (Peca et al., 2016). Relative phase measurements on smooth slopes also yield surface roughness $\Delta\psi = 4\pi (D_v/\lambda) \sin \phi$9 nm RMS (Peca et al., 2016).

Polarization-based shearing interferometry achieved sensitivity better than $h$00, with observed thermo-elastic deformation of a gold-coated BK7 mirror yielding a reconstructed peak height of $h$01 nm before spatial-filter correction and $h$02 nm after applying the reported factor of $h$03; the analytical prediction was $h$04 nm (Beyersdorf et al., 2012).

Automotive InSAR reports centimeter-level vertical accuracy at short range in controlled tests: a mounted reflector at $h$05 cm measured at $h$06 cm, a mounted reflector at $h$07 cm measured at $h$08 cm, and a ground reflector center at $h$09 cm measured at $h$10 cm (Kabuli et al., 14 Jan 2025).

The broader metrology synthesis distinguishes four precision classes. In classical interferometry on specular surfaces, the dominant limit is photon noise and there is “principally, no lower physical bound” (Häusler et al., 2021). In triangulation, a reported precision bound is

$h$11

and for focus-search methods,

$h$12

(Häusler et al., 2021). In rough-surface interferometry, the cited ultimate limit for two-wavelength methods is again $h$13 (Häusler et al., 2021). This classification is directly relevant to topographic interferometers because it separates photon-noise-limited regimes from roughness-limited or speckle-limited regimes.

6. Error sources, ambiguities, and recurring design trade-offs

A recurring misconception is that interferometric topography is uniformly phase-limited. The cited literature shows that the dominant error term depends on modality, surface class, and geometry.

In low-coherence full-field interferometry, reported error sources include visibility variations due to surface tilt and roughness, alignment errors, objective aberrations, pixel response nonuniformity, shot and electronic noise, and environmental drift (Pernechele et al., 2018). The measured precision loss with slope is documented by the $h$14 slope limit on the calibrated sphere and by the rise from $h$15 nm to $h$16 nm RMS between orthogonal and tilted flat-surface alignments (Pernechele et al., 2018).

In SA-SWI, the major design objective is suppression of phase corruption from aberrations, stray pupils, and indirect subsurface paths. Swept-angle illumination emulates spatial incoherence so that angular compounding rejects indirect paths, suppresses aberration-induced stray-path contributions, and reduces speckle contrast (Kotwal et al., 2022). The method still retains phase wrapping: a single synthetic wavelength yields an unambiguous range of $h$17, and multi-$h$18 unwrapping is discussed as future work (Kotwal et al., 2022).

In multi-band FSI, the principal limitation on sloped regions is retrace error coupled with dispersion. The measured wavelength-dependent deviations between the $h$19 nm and $h$20 nm OPD maps are about $h$21 nm on the steeper conical slope and about $h$22 nm on the milder slope (Peca et al., 2016). The paper attributes failures of absolute recovery on steep slopes to systematic errors from dispersion and retrace error rather than to random phase noise alone (Peca et al., 2016).

In shearing interferometry, the finite-difference transfer function

$h$23

introduces spatial-frequency nulls, so reconstruction requires dual-orientation measurements and compensation or regularization near zeros of the transfer function (Beyersdorf et al., 2012). The method is robust mechanically because of its common-path configuration, but high-frequency features are attenuated when the shear is large (Beyersdorf et al., 2012).

In automotive InSAR, coherence loss arises from volume scattering, multipath, and motion, with phase noise related to coherence by

$h$24

and height error propagating as

$h$25

(Kabuli et al., 14 Jan 2025). The small baseline $h$26 avoids explicit phase unwrapping but reduces sensitivity (Kabuli et al., 14 Jan 2025).

In Doppler-SAR interferometry, the sensitivity shifts from spatial baseline to baseline velocity. The approximate interferometric phase relation

$h$27

shows that trajectory and velocity knowledge become central calibration quantities (Yazici et al., 2017).

The quantum-inspired paper introduces a different trade-off altogether: the optimal measurements for absolute distance and for topographic angle are “different and incompatible,” with the multiparameter constraint

$h$28

(bao et al., 1 Jun 2026). That formulation explicitly rejects the assumption that a topographic interferometer should simultaneously optimize distance and topography.

7. Context, comparison, and conceptual unification

The broader metrology discussion emphasizes that topographic information can be encoded in phase, coherence envelope, synthetic wavelength beat, or slope, and that the physical origin of the ultimate precision limit differs across classes (Häusler et al., 2021). Within that classification, classical interferometry on specular surfaces is photon-noise-limited, rough-surface interferometry is roughness-limited, triangulation is speckle-limited, and slope-measuring methods are characterized by the uncertainty product $h$29 (Häusler et al., 2021).

Against that backdrop, low-coherence full-field Michelson systems (Pernechele et al., 2018), synthetic-wavelength systems with swept-angle incoherence (Kotwal et al., 2022), and multi-band FSI systems (Peca et al., 2016) can be viewed as three distinct optical strategies for overcoming the same pair of problems: ambiguity in absolute height and loss of robustness under surface roughness, slope, or stray optical paths. Shearing interferometry (Beyersdorf et al., 2012) occupies a different branch, recovering topography through gradient integration rather than direct height sensing. Automotive InSAR (Kabuli et al., 14 Jan 2025) and Doppler-SAR interferometry (Yazici et al., 2017) transpose the same logic into coherent radar, replacing optical path difference by range or Doppler-derived phase. The quantum-inspired topographic interferometer (bao et al., 1 Jun 2026) extends the concept still further by treating topography as a distinct observable that should be measured through cross-detector correlations and balanced homodyne quadratures rather than through classical triangulation.

A plausible implication is that “topographic interferometer” is best understood as a family of interferometric topography instruments defined by the observable they optimize, not by a single beam-path topology. Some optimize absolute height through coherence gating (Pernechele et al., 2018); some optimize tunable range and robustness through synthetic wavelength and swept-angle incoherence (Kotwal et al., 2022); some optimize absolute per-pixel profile over centimeter OPD ranges through multi-band phase–frequency slope bridging (Peca et al., 2016); some optimize nanometric deformation sensitivity through common-path gradient measurement (Beyersdorf et al., 2012); some optimize 3D elevation mapping in radar through baseline phase or Doppler-rate differences (Kabuli et al., 14 Jan 2025, Yazici et al., 2017); and some optimize topographic Fisher information while deliberately giving up distance precision (bao et al., 1 Jun 2026).