A self-operating broadband spectrometer on a droplet

The paper addresses how to perform broadband spectroscopic analysis with a highly miniaturized and inexpensive device that avoids conventional moving parts. Spectral analysis is fundamental across science and industry, with instruments ranging from scanning grating monochromators and optical spectrum analyzers to Fourier transform (FT) spectrometers. Miniaturization trends, driven by integrated photonics, have yielded microscale spectrometers, but FT devices typically require mechanical scanning, challenging robust miniaturization despite MEMS advances. Non-mechanical approaches (electro/thermo-optic tuning, SWIFTS, SHS, RAFT) have emerged, mainly in the near-infrared, yet often require complex fabrication. The authors propose and demonstrate a self-operating FT spectrometer that uses the spontaneous motion of an evaporating liquid droplet on a fiber tip as the scanning mechanism. They develop an analysis strategy to extract reproducible spectral information despite non-reproducible droplet dynamics, aiming to achieve competitive resolution and bandwidth with minimal cost and complexity.

The authors survey compact spectrometers: dispersive devices (grating-based monochromators, optical spectrum analyzers, virtual image phase arrays) and FT spectrometers. FT devices offer multiplex (Fellgett) advantage but traditionally need moving parts. MEMS technologies enabled chip-scale FT interferometers, followed by non-mechanical microspectrometers exploiting electro/thermo-optic tuning, stationary-wave sampling (SWIFTS), and multiple fixed-OPD interferometers (SHS, RAFT). These integrated devices predominantly operate in the near-IR with 0.5–5 nm resolution over 100–500 nm spans. This context motivates a simpler alternative leveraging unforced physical processes (evaporation) to provide scanning without complex integration or moving parts.

Technique and setup: A small droplet placed on the ferrule of a single-mode fiber creates two partially reflecting interfaces (fiber–liquid with reflectivity R1, and liquid–air with R2). Broadband light with spectral distribution I(k) is launched through the fiber. The back-reflected intensity coupled into the guided mode is the sum of: (i) a weak constant term from fiber–liquid reflection; (ii) a slowly varying term proportional to C(L)R2T1∫I(k)dk; and (iii) a fast interferometric term proportional to C(L)T1R1R2∫I(k)cos(2kL)dk, where T1=1−R1, L is instantaneous droplet thickness, and 0<C(L)<1 is the coupling efficiency depending on droplet shape. The fast term encodes the real part of the FT of I(k), scaled by material factors and C(L). To mitigate droplet-shape dependence (C(L)), the authors record two synchronous signals: a broadband signal (det1) and a tapped narrowband reference at wavelength λref (det2). Under the assumption that C(L) is approximately constant across the source spectrum, combining the AC (interferometric) and DC components yields a normalized interferogram proportional to the FT of I(k), up to a constant response factor T1R1R2. Experimental details: At each measurement, ~2 µl of isopropanol (chosen for negligible absorption and nearly constant refractive index across the NIR) is deposited on a 2.5 mm FC-PC connector ferrule. The back-reflected light is routed by a fiber circulator to an InGaAs detector (det1); a fiber splitter plus notch filter provides a narrow spectral slice around λref to a second photodiode (det2). The droplet forms and evaporates spontaneously after deposition, driving the effective optical path difference (OPD) scan. Spectrum reconstruction: The time-domain interferograms are mapped to OPD using the monochromatic reference: each full fringe corresponds to 2nΔL=λref; the last point before full evaporation corresponds to OPD=0. The number of complete fringes from any sampled point to the end, plus the fractional fringe, provides the OPD coordinate. Because evaporation velocity is nonuniform, equal time steps map to nonuniform OPD steps; thus the OPD-domain signal is interpolated and resampled at uniform OPD spacing before applying a discrete Fourier transform. Processing includes time reversal, removal of droplet-shape factors via AC/DC normalization, OPD transposition using fringe counting, and uniform resampling. No apodization or phase correction was necessary in the demonstrations; zero filling was optionally applied. Absorption spectroscopy procedure: To measure sample absorbance, the source is sent through a 15 cm acetylene gas cell with controllable pressure; transmitted spectra are measured by the droplet spectrometer at various pressures and normalized off-line to the source spectrum recorded with an empty cell. Parallel measurements with a commercial optical spectrum analyzer (OSA) provide a benchmark.

• The droplet-based FT spectrometer accurately reproduces the spectrum of a broadband supercontinuum source. The spectrum obtained from the processed interferogram overlays well with that from a commercial OSA (ANDO AQ6317B), validating the method and assumptions.
• Achieved spectral resolution of ~11 cm⁻1 (~2.6 nm) using isopropanol on a 2.5 mm ferrule, corresponding to interferograms longer than 600 fringes at λref=1538 nm, with a recording time of about 100–120 s (reported as ~100 s or about 2 minutes depending on run).
• Bandwidth of the demonstrated setup: 6000–7000 cm⁻1 (~250 nm span) in the NIR, limited by the InGaAs detector responsivity and fiber components, not by the principle of operation.
• Resolution–time trade-off via liquid choice: Water forms a larger droplet (L0≈2 mm) enabling ~2 cm⁻1 (~0.5 nm) resolution, but requires tens of minutes to evaporate; isopropanol provides faster acquisitions with lower resolution.
• Quantitative gas absorption: Transmission spectra through a 15 cm acetylene cell at pressures from 260 to 950 Torr show clear absorption features. After normalization to the empty-cell spectrum, the retrieved absorbance curves match OSA results (slightly higher OSA resolution of 8.5 cm⁻1 noted). Integrated absorbance values from both instruments agree closely, demonstrating quantitative capability and droplet-to-droplet consistency.
• The method effectively compensates for non-reproducible droplet dynamics by using a synchronous reference and AC/DC normalization, yielding reproducible spectra without mechanical scanning hardware.
• The approach is low-cost, compact, and compatible with fiber delivery; evaporation can be accelerated with heating to reduce acquisition time.

The study shows that an evaporating droplet on a fiber ferrule functions as a self-driven scanning-arm interferometer. By leveraging a reference fringe signal to construct an OPD axis and normalizing the AC component by the DC level to suppress droplet-shape coupling effects, the method reliably recovers the source spectrum and sample absorbance despite inherently variable evaporation dynamics. The findings address the challenge of miniaturizing FT spectrometers without moving parts or complex electronics by exploiting natural fluid dynamics to generate OPD scans. The demonstrated agreement with a commercial OSA for both broadband source characterization and quantitative acetylene absorbance validates accuracy and robustness. The device’s performance—nanometer-scale resolution over a few hundred nanometers bandwidth in ~100 s—compares favorably with many integrated photonic spectrometers while dramatically simplifying fabrication and cost. The concept generalizes to an optofluidic class of analyzers where evaporation or capillary forces drive FT measurements, with liquid selection and surface engineering offering tunable resolution–time trade-offs and potential extension in wavelength range (NIR to mid-IR) using appropriate detectors and components.

The work introduces and validates a simple, low-cost, self-operating FT spectrometer based on an evaporating droplet on a fiber tip. Through a robust analysis pipeline that compensates for non-reproducible evaporation, the device reconstructs broadband spectra and quantifies gas absorbance with results matching a commercial OSA. Demonstrated performance includes ~2.6 nm resolution over ~250 nm bandwidth in ~100 s, with potential improvements or alternative operating points via different liquids (e.g., water yielding ~0.5 nm resolution at longer acquisition times). The approach paves the way for a family of optofluidic FT analyzers driven by evaporation or capillary forces. Future work could integrate microfluidics and engineered ferrule surfaces to control liquid dynamics and tailor spectral parameters, extend operation across the full NIR and into the mid-IR with suitable components, and employ controlled heating for faster scans.

• The method assumes the coupling factor C(L) is approximately constant across the source spectrum; deviations can introduce spectral distortion if not mitigated by normalization.
• Liquid selection is constrained: the liquid should have negligible absorption and a nearly flat refractive index over the band of interest; otherwise, heating and absorption can destabilize the droplet and cause fragmentation late in evaporation.
• Acquisition time and resolution are coupled to evaporation dynamics and droplet geometry; higher resolution typically requires larger, slower-evaporating droplets, increasing measurement time.
• Bandwidth in the demonstration is limited by detector responsivity and fiber component transmission (here 6000–7000 cm⁻1); extending to other bands requires different detectors and components.
• The OPD scan is nonuniform in time due to fluctuating evaporation velocity, necessitating careful interpolation and resampling; environmental factors (e.g., temperature, airflow) may influence evaporation behavior and stability.