Selective vapor sensors with thin-film MoS₂-coated optical fibers

The ability to sense liquid and gaseous analytes at ultralow concentrations is of tremendous importance for a range of applications from defense to leak detection to agriculture and disease diagnosis.^1–5 Devices with semiconducting two-dimensional (2D) transition metal dichalcogenides (TMDCs) have emerged as attractive candidates for a multitude of sensing applications^6,7 owing to their semiconducting properties, visible-range bandgaps, and high surface-to-volume ratio. Of the available TMDCs, molybdenum disulfide (MoS₂), an ambient stable semiconductor with an indirect (1.29 eV) in bulk to a direct bandgap (1.9 eV) in monolayer transition, has received considerable attention within the sensor community. For example, chemresistors,^7–10 transistors,^11,12 electrochemical,^13,14 and optoelectronic^15,16 devices utilizing MoS₂ and other two-dimensional (2D) materials as the primary sensor material have pushed the limits of sensitivity to detect minute concentrations of analytes beyond many commercially available systems. Despite these appealing improvements, the ability to selectively distinguish targeted chemical analytes remains a fundamental challenge and has limited the implementation of TMDC-based sensors to practice to date.

In many cases involving semiconducting TMDC-based sensors, modification of the carrier density within the semiconducting channel can result in successfully distinguishing analytes based upon electron donor/acceptor characteristics. The modulation in carrier density within the sensor device due to interactions with the analyte manifests into a change in electronic or optical properties in the nanomaterial. Such is the case with the observed increase or decrease in MoS₂ channel resistance values upon exposure to either electron-donating NH₃ (resulting in n-doping) and electron-withdrawing NO₂ (resulting in p-doping).⁸ This has also been observed in optical-based sensors where red and blue shifts were observed in the MoS₂¹⁷ and WS₂¹⁸ photoluminescence spectra. While intrinsic thin-film MoS₂ provides little by way of selectivity, surface functionalization of 2D materials with metal nanoparticles¹⁹ and various organic molecules²⁰ have also proven to be advantageous for encouraging analyte selectivity in very specific cases.

Exploring other selectivity mechanisms intrinsic to 2D materials is vital in order to complement the ultrahigh sensitivities in a low-cost, reusable platform. Thin film–coated optical fiber sensing modalities can improve sensing selectivity through the use of broadband optical characterization. Evanescence-wave fiber-optic sensors (EWFOS) have emerged as a means to accomplish this through effectively monitoring temperature,^21–23 strain,²⁴ and contaminants in water^25,26 in a versatile, stable, modular, and low-cost platform. TMDCs have been used in conjunction with various fiber-optic based sensors in monitoring temperature,²⁷ humidity,^28–30 and human breath.³¹ To date, TMDC-coated fiber-optic sensors have only exploited single-mode optical fibers that were optimized for maximum transmission within the fiber at laser wavelengths of 1060,²⁷ 1540,²⁸ and 1550 nm.^29–31 Conversely, other fiber-based sensors have realized the importance of capturing the full broadband visible spectrum rather than a single wavelength. For example, fiber-based surface plasmon resonance sensors incorporating gold,³² silver,³³ and mixed coatings^34–36 observe plasmonic peak shifts and modulated signal intensities within a broad spectral range.

In this work, we demonstrate that the unique excitonic optical properties of thin-film MoS₂ can be further exploited by probing the anomalous optical dispersion behavior in the visible regime (400–800 nm) to selectively distinguish organic compounds in representative EWFOS systems. Changes in the measured optical transmission for a sputter-coated optical fiber are observed in response to 12 different organic vapors. As suggested, these optical sensing responses are highly correlated to exciton peak resonances of the thin-film MoS₂ at visible wavelengths. The thin-film optical properties of semiconducting MoS₂ (i.e., complex refractive index and dielectric function) are known to be dominated by the associated exciton behavior and exhibit exceptional sensitivity to the surrounding chemical and dielectric environment.^37,38 As such, we show that our sensors are particularly responsive to electron donor/acceptor organic vapor species, aromatic compounds, analytes with higher refractive indices, and known chemical intercalants such as aniline and N,N-dimethylaniline, representing a step toward selective chemical sensors with TMDC materials.

Fabrication of the thin-film MoS₂ EWFOS was accomplished starting with a 0.5 NA step-index multimode fiber with a 200 μm diameter core (ThorLabs, RP200ERT) that was trimmed to approximately 3 in. in length. The Tefzel coating was removed with fiber strippers for the entire length and the hard polymer cladding was then thinned through the use of solvent sonication to ensure that only bare glass fiber remained. Following the removal of the cladding and coating material, the optical fiber was placed into a custom-built vacuum chamber with base pressure approaching 3 × 10⁻⁹ Torr, where 10 nm of MoS₂ was coated over the top surface using pulsed DC magnetron sputtering (PDCMS). The growth of MoS₂ was performed at a substrate temperature of 400 °C, where the growth temperature was calibrated via a spot-welded thermocouple and a pyrometer for measurement during growth. PDCMS from a MoS₂ target of 99.95% purity occurred at a deposition rate of 0.4 nm/s for a total deposition thickness of 10 nm. To ensure uniform coating on both the top and the bottom surfaces of the optical fiber, MoS₂ was sputtered in two sequential steps where the bare fiber was removed from the growth chamber, flipped, and the procedure repeated to evenly coat the bottom side as well. The substrates were rotated at approximately 100 rpm while positioned 7 cm from the MoS₂ target. The synthesis chamber was initial pumped down to 10⁻⁹ Torr and was then filled with ultra-high-purity argon gas at 25 SCCM and at a pressure of 10 mTorr. A pulsed DC power of 60 W at 65 kHz was applied to the target for the sputtering process.

The devices were placed in a custom-made support rig to prevent overbending and the holder was placed in a stainless-steel chamber with 1.2 L volume. NO₂ and NH₃ were obtained from Indiana Oxygen in 10 and 1000 ppm concentrations. The other analytes were obtained by filling a bubbler and running dry nitrogen through the frit. Analyte concentrations were calculated with literature values of the Antoine equation, while temperature was controlled by air ventilation or by an ice bath as needed. Analyte flows were mixed with nitrogen to a total flow rate of 200 standard cubic centimeter per minute (SCCM) to target the desired concentration. Mass flow controllers from MKS ranged from 20 to 2000 SCCM. Final concentration values for several analytes were confirmed by a photoionization detector to verify the validity of the Antoine equation calculations (VOC-TraqII Flow Cell on Ametek Mocon).

Sensor measurements were taken using an OceanOptics (model QE65000) spectrometer. The chamber was purged with nitrogen for several hours preceding measurement. Full spectral sweeps were performed and saved every 5 s before and during testing. Custom software was written to monitor select frequencies in real time to determine when the saturation point was reached. Final Δ transmission values were taken as the difference between the spectrum at saturation (50 min) and the spectrum just before analyte was added (0 min).

The transmission of the fiber optic was modeled using COMSOL Multiphysics, which uses the Finite-Element Method in the frequency domain. The radius of the fiber optic in the simulation was 10 μm instead of the experimental 150 μm to reduce the demand on the computer. The use of a smaller radius should not affect the results since it is much larger than the wavelength of light. This assumption was confirmed by increasing the radius past 10 μm in single-wavelength simulations to determine if increasing the radius affected the transmission, and it was found that the transmission was not affected. The length of the fiber optic was 5 μm, and the only mode studied was the fundamental TE₁₀ mode. The simulations were used to determine the main mechanism of sensing for the fiber optic. The three mechanisms that were investigated were charge transfer between the solvent and MoS₂, the formation of a thin layer of the solvent on the outside of the fiber optic, and the adsorption of the solvent into MoS₂ without affecting the electronic properties of MoS₂. The refractive index values were taken from literature.

Physical vapor deposition (PVD) of a thin-film MoS₂ using PDCMS has been reported as a viable deposition technique for coating flat substrate surfaces.³⁹ Expanding upon such efforts, the deposition of MoS₂ onto conformal surfaces (e.g., cylindrical optical fibers as described in this study) represents the utility of such PVD methods to prepare controlled low-dimensional material thin films on a variety of surfaces.⁴⁰ Here, the large fraction of ionized constituents during PDCMS (further details explained in the experimental section) facilitates thin-film coating onto conformal surfaces via uniform acceleration of ions from the compositionally uniform plasma sheath around the fiber.⁴¹ A scanning electron microscopy (SEM) micrograph of a MoS₂ EWFOS is shown in Fig. 1(a). The PVD deposition of MoS₂ was performed at a substrate temperature of 400 °C in order to a 10 nm coating of crystalline 2H-MoS₂ on the bare fiber, as verified via the presence of the E_2g and A_1g Raman vibrational modes shown in Fig. 1(b). The higher processing temperature is significant as amorphous MoS₂ does not exhibit the same excitonic spectral profiles necessary for representative sensing modalities discussed later.

Following the sputtering process, optical connections were established to the end of the 2D MoS₂-coated EWFOS and the sensor was placed into a custom environmental testing chamber depicted schematically in Fig. 1(c) (further details in the supplementary material).⁶⁹ The testing setup consisted of the sensor connected to both a broadband mercury light source and an OceanOptics spectrometer on opposite ends. During sensor measurements, the analyte of interest was then continually fed into the chamber, passing over the sensor before being exhausted out the other end of the setup.

Figure 1(d) shows the optical transmission spectrum from 400–1000 nm through both a bare glass fiber and a MoS₂-coated optical fiber without the presence of any analyte. As every other variable is kept constant in this measurement, the difference between the two spectra is expected to be solely the result of the optical properties of the MoS₂ coating. The difference in transmission between the coated and the uncoated fiber is also shown in Fig. 1(d) in units of dB, which is related to the optical signal in the following expression:

where S₂₁ is the scattering parameter depicting the signal loss at a given wavelength within the MoS₂-coated optical fiber relative to the uncoated fiber. In this case, a positive dB loss value indicates a reduced transmission at that wavelength in the MoS₂-coated fiber relative to the uncoated fiber. The dB loss spectrum exhibits nine unique identifiable peaks in the spectral range of 400–800 nm, many of which are associated with strong optical excitonic phenomena for MoS₂. Note that a loss in transmission correlates to differences in (near-field) absorption or total scattering of the MoS₂ excitons. For instance, the location and intensity of these excitonic peaks within the optical properties (i.e., refractive index, n, and extinction coefficient, k) of MoS₂ have been shown to correlate to differences in crystal size, layer thickness, chemical doping, dielectric screening, and morphological properties of the nanomaterial.^37,38,42,43 Generally, three major peak signatures are observed in the spectral response labeled as the A, B, and C excitons (increasing in photon energy) and represent electron–hole Columbic interactions. The location, magnitude, and spectral characteristics of these excitons dictate the resulting optical response. As such, any change in the optical behavior can be observed with respect to any change in the optical exciton activity—significant to any optically based sensing application. The peak of the C exciton, which correlates to the optical transition at the Γ-K point, has been observed in the range of 407–424 for the extinction coefficient (k) and anywhere from 450 to 506 in the refractive index (n).³⁷ The A and B excitons, closest to the Fermi level, are directly related to the spin–orbit split valance electrons at the K and K′ points. The B exciton peak locations have been observed at a range of 606–619 nm for the extinction coefficient (k) and 620–645 nm for the refractive index (n). Similarly, A exciton peak locations have been reported at 615–676 nm for the extinction coefficient (k) and 664–689 nm for the refractive index (n).^44,Figure 1(d) shows several distinct peaks observed from the change in transmission of the MoS₂-coated optical fiber relative to the uncoated fiber. Correlating these observed peaks suggests that several are within the range of the C exciton (peak 1 at 439 nm, peak 2 at 473 nm) and the B and A excitons (peak 6 = 666 nm, peak 7 = 682 nm, and peak 8 = 695 nm) within MoS₂. While three distinct spectral exciton peaks are typically observed for MoS₂, the optical dispersion data analysis commonly necessitates 5–6 oscillators to analytically represent experimental measurements. For instance, the optical dispersion data from spectroscopic ellipsometry will incorporate exemplary peak oscillations at about 290, 390, 430, 528, 608, and 654 nm in Lorentz, Tauc–Lorentz, or Tanguy formalisms.⁴⁵ Therefore, it is not unexpected to observe more than three spectral peaks in the optical response, as has been observed in other studies, illustrating the spectral response difference in the complex refractive index and dielectric function.⁴⁵ 

As PDCMS of MoS₂ is the preferred deposition method used in this study to coat conformal optical fibers, the energetic ions emitted during the deposition of 2D MoS₂ have been shown to create sulfur vacancies within the thin films.⁴⁶ Previous studies also reported peaks within both the n and the k spectra that relate to bound electron–hole pairs near the band edge at wavelengths between 700 and 1000 nm in sputtered films and films with high defect concentration.^47,48 First-principle calculations also predict the emergence of midgap states in the presence of sulfur and molybdenum vacancies.⁴⁹ The peak in Fig. 1(d) near 750 nm, labeled peak 9, is attributed to these near-band edge defects and will be discussed below. Note that optical characteristics below the band edge for MoS₂ have also been observed in the optical constants for chemical vapor–deposited films.⁵⁰ 

Sensor tests were performed by monitoring the optical transmission through the MoS₂-coated fiber while introducing 600 ppm of a known vapor-phase analyte into the testing chamber. Each experiment was conducted by first allowing the sensor device to equilibrate in an inert nitrogen carrier gas. Once the optical signal stabilized, the analyte gas entered into the chamber and the optical transmission through the fiber was monitored for a total of 50 min, where full equilibration of the signal was reached. Twelve analytes were chosen to evaluate the selectivity of the MoS₂-coated EWFOS: aliphatic vapors NH₃, NO₂, 2,4-pentadione, butyl acetate, and proprionic acid, as well as aromatic vapors, aniline, anisole, bromobenzene, mesitylene, N,N-dimethylaniline, o-xylene, and pyridine. The resultant change in transmission from time t = 0 to t = 50 min is shown in Fig. 2(a) for all the analytes evaluated in this study. Clearly, the change in transmission across the MoS₂-coated optical fiber varies substantially with wavelength, as the interaction of the analyte and 2D material drives changes in the transmission of the broadband evanescent wave through the fiber. In all cases, a signal loss was observed within the entire viewing spectral range. Ten of the 12 analytes are clustered with the exception of N,N-dimethylaniline and aniline, which exhibit a substantially different optical response. An evaluation of the signal response of the ten nonaniline-based spectra in Fig. 2(b) reveals that the transmission losses occur very near the C (∼470 nm), B (∼630 nm), and A excitons (∼690 nm) of MoS₂ as well as the suspected midgap states near 780 nm. This suggests that the signal loss in transmission across the optical fiber is correlated to the interaction between the analytes and the excitons of the MoS₂ coating (e.g., likely charge transfer effects or dielectric field screen^38,51,52 or both⁴⁵). Additionally, the sensor device is extremely stable and reusable, as Fig. 2(c) shows the difference in sensor performance from the first NH₃ test to the same test performed after all the analytes were completely evaluated.

To more clearly observe changes within the fiber due to the interaction of the analytes with the surface of MoS₂, the change in transmission can be expressed in dB loss similar to Eq. (1). In this case, the dB loss shown in Fig. 2(d) depicts the signal loss between the initial signal of the MoS₂-coated fiber to the final equilibrium state (at t = 0 min and t = 50 min). In this case, the contributions from excited states within MoS₂ become more evident with activity surrounding 450–500 and 600–800 nm, near excited states observed in MoS₂ thin films. The sensor here is most responsive to aniline, with several peaks leading to a transmission loss exceeding 6 dB.

The steady-state transmission spectra of the fiber optic with different solvents are shown in Fig. 3(a). All analyte vapors show increased loss at wavelengths around 475 nm with a continuous redshift when going from solvents that act as electron acceptors to donors [Fig. 3(b)]. This increased transmission loss is associated with the C exciton, which typically appears around 415 nm.^53,54 Although the peak is red-shifted from the C exciton wavelength, it corresponds to the maximum in the refractive index reported to be caused by the C exciton.³⁷ This redshift between the exciton transition and n is an observed characteristic of the optical dispersion data to derive the optical properties of thin films and 2D MoS₂. For instance, the refractive index (n) represents the velocity of light as it passes through a material and is often associated with how transmissive or reflective a material is in relation to the extinction coefficient (k), which describes how absorptive the material is. Mechanistically, complementary studies show how spectral exciton characteristics change in response to controlled environmental changes. For example, the high sensitivity of MoS₂ to charge carrier injection using electrostatic gating has been studied and sensitivity was seen near the A and B excitons’ wavelengths (with minimal changes in the optical constants, Δn ≤ 0.5).⁵¹ Conversely, physisorbed chemical dopant-induced changes in the refractive index, reported by Stevenson et al.,³⁷ show much larger changes in each measured exciton along with substantial changes in the reported refractive index (i.e., up to Δn ≈ 2.2). The peak wavelength shifts and changes in peak amplitude (including the C exciton) was attributed to a combination of both carrier injection and dielectric field screening of the adsorbed chemical dopant. Stevenson et al. also observed a red (blue) shift in the C exciton for n- (p-) type dopants, which is consistent with our experimental results. Therefore, the change in the optical transmission intensity observed here is likewise assumed due to both carrier injection and dielectric screening effects from the adsorbed dopants (albeit to varying degrees with respect to the given analyte).

To further investigate the role of carriers on the optical fiber system, the transmission was simulated using the frequency domain module in COMSOL Multiphysics (see the Experiment section) and the carrier-dependent refractive index of monolayer MoS₂ from literature.⁵¹ Although the monolayer refractive index of MoS₂ is expected to differ from that of the thin-film MoS₂ used in this study, the qualitative differences in the optical dielectric function between monolayer and few-layer MoS₂ are small enough to allow qualitative analysis without significant error.^11,54,55 Furthermore, the carrier density-dependent refractive index of few-layer MoS₂ has not yet been reported to the best of the authors’ knowledge. The simulated transmission spectra are shown in Fig. 3(d). In the simulation, the fiber optic loss increases around 480 nm when either electrons or holes are injected into MoS₂. The simulations also show a higher response to electron injection than to hole injection, which agrees with the experimental result. The simulated spectra confirm the presence of a loss maximum located at the peak intensity of the refractive index due to the C exciton showing that the loss in the fiber optic is due to an increased transmission through the MoS₂ layer instead of an increased absorption by MoS₂. This conclusion is further confirmed by the electric field profile in Fig. 3(c), where the low electric field within MoS₂ shows that there is negligible absorption by MoS₂.

For all of the aromatic compounds evaluated in this study, much of the redshift in dB loss near the C exciton was observed to occur on the order of minutes over the course of the experiment rather than instantaneously. This is most likely due to the increasing vapor concentration in the chamber, as the concentration increases for the first several minutes after which it then saturates at 600 ppm (see in the supplementary material⁶⁹ for further details pertaining vapor concentration in the test chamber). Figure 4(a) depicts the progression of dB loss surrounding the C exciton for pyridine, o-xylene, and mesitylene, where each line of increasing dB loss corresponds to a unique spectrum taken every 30 s. With the case of pyridine, the peak near the C exciton of MoS₂ exudes only a minor redshift from 473 to 475 nm over the 50- min experiment. Conversely, other aromatic systems, including o-xylene and mesitylene, exhibit a substantial redshift if there is maximum dB loss over 11 and 17 nm, respectively, as depicted in the shaded region within Fig. 4(a). The complex interaction of inorganic/organic interfaces has been shown to be strongly dependent upon many factors, including charge transfer, orbital hybridization, and electronic screening.⁵⁶ A reduction of the molecular HOMO–LUMO has shown to be possible through quasiparticle excitations from the organic molecule/underlying substrate hybridization. To evaluate the contribution of the analyte HOMO–LUMO gap relative to this redshift, the HOMO–LUMO gap of aromatic compounds was calculated (further details in the supplementary material).⁶⁹ The HOMO contours of both pyridine and mesitylene are shown in Fig. 4(b), and the HOMO–LUMO value with respect to the observed redshift in max dB loss is shown in Fig. 4(c). Clearly, a strong linear trend between the HOMO–LUMO of the analyte and the redshift in the fiber signal suggests an interplay between orbital hybridization and sensor performance and potentially enabling further selectivity through monitoring the location of the C exciton over time. Furthermore, it has been shown in previous works that at the interfaces of gas/solid or solid/solid systems, the energy-level alignments play an important role in determining the strength of the interface interaction^56,57 For instance, in the system of organic–metal interfaces, the energy-level alignment between the HOMO–LUMO levels of organic molecules and the Fermi level of metals is directly linked to the efficiency of charge transfer at the interfaces.⁵⁷ Accordingly, in our system, organic compounds with different HOMO–LUMO levels can potentially have strong or weak interaction with MoS₂, thus leading to the trend observed in Fig. 4(c).

While the transmission spectra surrounding the C exciton present for several mechanisms for selective vapor sensing, an analysis of the A and B excitons may elucidate further details of the thin-film MoS₂ EWFOS. Figure 5(a) shows that the equilibrium dB loss for the ten non-aniline-based analytes near the A and B excitons and the lower energy states is centralized around 780 nm [see Fig. 1(d)]. Every dB loss transmission spectrum of the analytes tested reveal a bimodal series of peaks between 600 and 720 nm as well as a small peak between 770 and 780 nm. Figure 5(b) shows the peak dB loss intensity value of these peaks associated with the C exciton (460–490 nm), B-exciton (610–630 nm), A exciton (650–690 nm), and the unique peak between 770 and 780 nm (corresponding to potential midgap states due to defects). In general, an increase in dB loss with the analyte refractive index is observed for the peaks observed with the strongest correlation due to the suspected defect states around 780 nm.

The large-scale analyte sensor tests reveal the most unique signatures from the two aromatic amines N,N-dimethylaniline and aniline. The transmission losses within the thin-film MoS₂-coated optical fiber approach 4 and 7 dB for N,N-dimethylaniline and aniline, respectively. While the loss intensity is maximized near the C excitons, the peak locations in this region show little deviation from the other ten analytes tested [see Fig. 2(d)]. Substantial deviations from the transmission loss spectrum occur near the A and B excitons as well as near the band edge within the spectral range of 600–850 nm shown in Fig. 5(c). In this range, N,N-dimethylaniline exhibits a broad peak in dB loss, which can be further fit to three individual Gaussian peaks at 632.3, 669.3, and 724.2 nm. While the two higher energy peaks (632 and 669 nm) fit well within the expected range for the B and A excitons, respectively, the large peak at 724 nm is entirely unique to N,N-dimethylaniline and is not observed in any other vapor tested.

Of all vapor analytes tested in the study, the MoS₂-coated optical fiber sensor displays the strongest affinity to aniline. While the dB loss peak near the B exciton shown in Fig. 5(c) is within the wavelength range of other analytes (639 nm), the peak near the A exciton is the most redshifted to 708 nm. An extremely strong peak centering around 798 nm is observed in aniline, which matches the highest signal intensity from all analytes.

The emergence of two extremely intense transmission loss peaks in N,N-dimethylaniline (724 nm) and aniline (798 nm) near the band edge suggests a potentially unique coupling mechanism between the two aromatic compounds and MoS₂. Aniline and its derivatives are known to intercalate within the layers of MoS₂ via an oxidizing polymerization, where the organic molecules act as a dielectric and physically separate layers of MoS₂ due to the presence of the molecule between the van der Waals gap⁵⁸ resulting in an expansion of the crystal lattice.^59–61 In general, MoS₂ is resistant to intercalation by molecules such as pyridine and ammonia as these molecules are difficult to reduce,⁶² and the intercalation of such molecules into the van der Waals gap of MoS₂ is possible only with highly reducing agents such as alkali metals.⁶³ Previous studies investigating polyaniline/MoS₂ nanocomposites depict significant enough intercalation of the polyaniline species to observe shifts in the Raman, UV-Vis, and x-ray diffraction spectra of MoS₂.^64,65 The intercalation process of the aniline derivatives has also shown to be reversible with weak bonding between the host and the intercalant, thus allowing a reuse of the fiber sensor after purging.^59,66 Moreover, intercalants within layered van der Waals materials are strongly influenced by the presence of defects.⁶⁷ We speculate that as recent studies have shown that new plasmonic peaks can emerge in the optical spectrum of MoS₂ from an intercalation of copper atoms, these new plasmonic modes may be a result of intercalated aniline and N,N-dimethylaniline and the influence of the molecular species on defects within MoS₂ near 724 and 798 nm, respectively.⁶⁸ 

The time-dependent transmission spectrum through the MoS₂-coated optical fiber exposed to 600 ppm of aniline is shown in Fig. 6(a). The most prominent peaks within the spectra occur at 457, 492, 640, 708, and 798 nm, which all exhibit a redshift during the test, as shown in Figs. 6(b) and 6(c), presumably due to the concentration within the chamber increasing and eventually saturating at a value of 600 ppm. Figure 6(d) reveals the time-dependent sensor performance with a rapid increase in signal occurring within the first minute of exposure to aniline across all wavelengths, with the peaks associated near the C exciton (457 and 492 nm) saturating faster than others. Additionally, the detection of as low as 6 ppm of aniline was easily achieved with this fiber setup as depicted in Fig. 6(e), and further sensor device optimization is expected to be possible through improved MoS₂ deposition and strategies to reduce signal/noise within the sensor testing setup. As the Occupational Safety and Health Administration (OSHA) limit of detection for aniline is currently 5 ppm for an 8h-time window, the ability to detect at these concentrations shows real-world implications. Finally, all of the sensor studies, including those with suspected intercalant molecules, depict a complete reversibility, demonstrating that the thin-film MoS₂ EWFOS is extremely reusable.

This work describes a strategy to capitalize on the unique optical properties of thin-film low-dimensional materials (such as TMDCs including MoS₂) to increase selectivity in optical fiber sensor devices. Magnetron sputtering of MoS₂ directly onto a bare glass fiber resulted in a coating that displayed several signatures of excited states in the nanomaterial when compared with the uncoated fiber. Through evaluating the time-dependent peak locations and intensities of several of these excited states, including the C, B, and A excitons, several trends within analytes were observed. Selectivity of analytes between donor/acceptor molecules, aromatic compounds, analyte refractive index, and intercalated compounds such as aniline derivatives is proposed. This approach represents a step toward realizing highly sensitive, selective, and completely reusable gas sensors based upon 2D nanomaterials that are cost-effective and scalable.

N.R.G., L.K.B., L.D.T., and L.B. acknowledge support from the Air Force Office of Scientific Research under Grant No. FA9550-20RXCOR057. D.J. and J.L. acknowledge partial support from the Air Force Office of Scientific Research under Grant No. FA9550-21-1-0035 and AOARD Grant Nos. FA2386-20-1-4074 and FA2386-21-1-4063

The authors have no conflicts to disclose.

M.M., L.K.B., and J.L. contributed equally to this work.

The data that support the findings of this study are available from the corresponding authors upon reasonable request.

Dr. Lucas K. Beagle graduated from the Wright State University in 2005 with a B.S. in biological sciences, and Youngstown State in 2008 with a master’s in chemistry. He went on to complete his Ph.D. in Chemistry from the University of Florida in 2012 with a focus on heterocycles and medicinal chemistry. He is currently Program Manager and Lead Scientist in the Biological and Nanoscale Technologies Division at the UES Inc., where his research supports the Materials and Manufacturing Directorate at the Air Force Research Laboratory. His research areas include synthesis and processing of covalent organic frameworks (COFs) and other polymers, microwave chemistry and processing, hybrid organic/2D-inorganic systems, 2D heterostructures, and novel design of chemical and biological sensors.

Jason Lynch is currently a Ph.D. candidate in the Electrical and Systems Engineering department at the University of Pennsylvania where he was awarded with an Ashton Fellowship. He earned a B.A. in both Physics and Molecular Engineering in 2019 from the University of Chicago. He then earned a M.S. in Nanotechnology from the University of Pennsylvania in 2021 where he received a reward for Excellence in Research. His research focuses on using tunable metamaterials and 2D materials to create high-performance, optoelectronic devices.

Dr. Peter R. Stevenson received his B.S. in Chemistry from Brigham Young University–Idaho in 2012 and his Ph.D. in Physical and Materials Chemistry from the University of Utah in 2019. Prior to graduate school, he worked in the specialty chemicals industry involving applied reactive monomer and polymer technologies. In graduate school, his research interests transitioned to plasmonics, nanophotonics, and nanomaterial fabrication. He now leads the optical coatings research team within the Materials and Manufacturing Directorate of the Air Force Research Laboratory. His group’s research emphases include photonics, optical physics, emergent optical materials discovery, optical thin films, stratified optical media (e.g., multilayer stacks and metasurfaces), and optical coating design and fabrication.

Dr. Ly D. Tran graduated Valedictorian from University of Technology in Ho Chi Minh City, Vietnam with a B.S. in Chemical Engineering. She then moved to Texas and earned her Ph.D. in Organic Chemistry at University of Houston in 2013 under Dr. Olaf Daugulis’s guidance. Her research focused on C–H bond functionalization using iron, copper, and palladium catalysis. Ly continued her professional career as a post-doctoral researcher with Dr. Adam Matzger’s group at the University of Michigan in the field of Materials Science, focusing on the design and synthesis of Microporous Coordination Polymers for gas sorption and organic compound separation. After post-doctoral training, Ly started her industrial career in 2015 as a senior scientist with MilliporeSigma in Miamisburg, Ohio, where she was responsible for the synthesis of new and high-quality isotope labeled compounds. Ly is currently a scientist with UES, Inc. She is conducting research in the 2D-inorganic/organic material, a joint effort between Dr. Nicholas Glavin’s and Dr. Luke Baldwin’s teams at the Air Force Research Laboratories. Ly’s current research focuses on exploring covalent organic frameworks (COFs) and incorporating COFs with inorganic materials in sensing applications.

Dr. Luke A. Baldwin is a Research Chemist at the Air Force Research Laboratory (AFRL), Materials & Manufacturing Directorate. He received his BSc. in Chemistry from Carroll University in 2011 and his PhD in Chemistry from The Ohio State University in 2017. He joined AFRL in 2018 and is now a primary point of contact for synthetic materials at AFRL and leads the Directorate's research in continuous flow chemistry. Luke's current efforts are centered on atomistically tailoring synthetic organic materials, reprocessable elastomers, thermosets, and composites for advanced applications. More recently he has been exploring the implementation of AI active learning and optimization algorithms to more efficiently discover and synthesize materials.

Dr. Drake Austin received his Ph.D. in physics from Ohio State University in 2017 with his thesis research focusing on the surface modification of bulk, single-crystal semiconductors using mid-infrared, femtosecond laser pulses. His current research is on the modification and characterization of thin-film materials using laser light. Such modification includes localized crystallization, doping, defect formation, and induced chemical reactions with the goal of understanding the relationships between these material properties as well as the development of functional devices such as chemical sensors. His expertise is in the development and characterization of optical systems for laser-processing, the optical characterization of thin-film materials, and the interactions between light and solid materials in general.

Prof. Deep Jariwala is an Assistant Professor in Department of Electrical and Systems Engineering at the University of Pennsylvania (Penn). His research interests broadly lie at the intersection of new materials, surface science, and solid-state devices for computing, sensing, opto-electronics, and energy harvesting. Deep graduated in Metallurgical Engineering from the Indian Institute of Technology, Banaras Hindu University (2010). He pursued his Ph.D. in Materials Science and Engineering at Northwestern University (2015). He was a Resnick Prize Postdoctoral Fellow at Caltech from 2015 to 2017 in Applied Physics and Materials Science before launching his independent career at Penn in 2018. Deep’s research has earned him awards of multiple professional societies including the Russell and Sigurd Varian Award and Paul H. Holloway Award of the American Vacuum Society, The Richard L. Greene Award of the American Physical Society, The Weertman Doctoral Fellowship and the Hilliard Award at Northwestern University, the Army Research Office Young Investigator Award, Nanomaterials Journal Young Investigator Award, TMS Frontiers in Materials Award, Intel Rising Star Award, iCANX Young Scientist Award, IEEE Photonics Society Young Investigator as well as the IEEE Young Electrical Engineer of the Year Award (Philadelphia and Delaware Valley). In addition, he has been named in Forbes Magazine list of 30 scientists under 30 (2018) and is an invitee to Frontiers of Engineering conference of the National Academy of Engineering (2019). He has published over 95 journal papers with more than 13 500 citations and 7 patents.

Dr. Nicholas R. Glavin received his B.S. and M.S. degrees from the University of Dayton in 2010 and 2012 and his Ph.D. in Mechanical Engineering from Purdue University in 2016. Upon completion, he joined the Materials and Manufacturing Directorate at the Air Force Research Laboratory, where he is currently a Senior Materials Scientist focused on synthesis, processing, and characterization of 2D materials and polymers for electronics and sensors.