Intracavity dynamics-based gain-assisted sensing with microtubule Raman microlaser

Photonic sensors with an ability to monitor nanoscale material processes and dynamic biochemical reactions are essential for a wide range of applications, such as early-stage disease diagnosis, environmental monitoring, and medicine analysis.^1–3 With capabilities of transducing an analyte signal to a quantifiable optical signal with improved efficiency and higher throughput, optofluidic microsensors have aroused great interest.^3–5 Over the past decades, a variety of miniaturized and highly sensitive approaches have been developed with microfluidics, such as surface plasmon resonance (SPR) devices,^6,7 photonic crystal fibers,^8,9 and whispering gallery mode (WGM) microcavities.^10–12 Among them, ultrasensitive microsensors based on WGM microcavities are renowned for their strong light–matter interactions and small detection limit down to the single nanoparticle level.^3,13 In particular, they can be easily integrated with an optofluidic channel to form intriguing platforms for analyte delivery at a micro- or nano-liter scale, such as microbubble, microcapillary, and microtubule.^10,14,15 The quasi-droplet modes inside the microfluidic channel are beyond the evanescent distance limit,¹¹ which could enable ultrahigh sensitivity in aqueous environment sensing.

In microcavity sensing regimes, the active microlaser detection has a number of advantages over that based on passive microcavities due to the fact that it has a narrower spectral linewidth, has a rapid detection capability, and does not need external reference lasers.^16–18 The most common types of active microlasers include rare-earth-ion doped microcavity laser and microcavity Raman laser. Among them, microcavity Raman laser extends the wavelength of existing lasers due to neither a requirement of the complicated doping process of gain materials nor a limitation in operation frequencies,^19–21 which shows excellent performance in various sensing applications.^22,23 The inherent biocompatibility of silica can also be retained by the dopant-free scheme. However, its relatively weak power may be easily submerged in system noises and disturbed by environmental fluctuations, rendering it ineffective at detecting small sensing signals. Thus, laser gain amplification and sensitivity improvement are urgently needed.

For this purpose, the laser heterodyne feedback (LHF) technique is demonstrated in a microlaser-integrated microtubule to achieve the optical gain assistance. LHF occurs when part of the laser returns back to the resonator and modifies the properties (power, polarization, dynamic behavior, etc.) of the laser output.^24–26 In particular, when the feedback light is frequency-shifted and interferes with the original light field, the interference signal can resonate with the laser relaxation oscillation (LRO) and be enhanced spontaneously.^25,26 In this case, the laser signal in the optical domain obtains gain amplification and is converted into an RF response of LRO with higher sensitivity. The remarkable enhancement makes it effective for increasing intensity changes caused by the analytes. Small signals that would otherwise be buried in system noises and go undetected can be more easily resolved.

Based on the proposed Raman lasing heterodyne feedback (RLHF) microtubule, a measurement resolution with the nanoparticle detection limit down to 50 nm radius and a refractive index noise-limited resolution of 8.18 × 10⁻⁷ refractive index unit (RIU) are demonstrated, which is better than most of the microcavity active sensors. In addition, the dynamic phase transition of thermosensitive hydrogel is further investigated as a validation of its rapid detection capability. As a crucial phase transition material in biomedical science, the hydrogel has wide applications in the areas of stimulated actuators, drug release, and organ tissue engineering.^27,28 It is responsive to almost all stimuli in nature, including water, pH value, temperature, light, and electricity. The structural and chemical properties of the hydrogel depend heavily on its gelation conditions and reaction kinetics.^29,30 Therefore, monitoring the phase transition dynamics of hydrogel and revealing its structure–property relationships are important for fundamental studies and practical applications.

So far, various methods have been proposed to investigate the dynamics of hydrogel phase transition, including nuclear magnetic resonance (NMR), calorimetry, and rheology. However, the NMR method requires specialized equipment and has low resolution limitation.³¹ Calorimetry is hampered by the low sensitivity and slow response.³² Rheology cannot be easily implemented to study the rapid gelling dynamics or mechanically weak materials.^33,34 Compared with these methods, the proposed RLHF microtubule can achieve rapid monitoring of hydrogel transition dynamics with better performance. It does not apply force on materials and can be used to study mechanically weak hydrogels. The intracavity dynamics-based gain amplification makes it more sensitive to weak signals than conventional microcavity sensors. Due to these advantages, detailed transition behaviors of [poly (lactic-co-glycolic acid), PLGA] materials between hydrophilic and hydrophobic states are clearly revealed, which may help analyze the structure–property relationships of the matter state.

As a promising solution for microscale and highly sensitive biochemical analysis, the proposed RLHF microtubule brings one step closer to practicality in real-time dynamic monitoring, with ultrasmall sample consumption down to 50 pl. The gain assistance provided by LHF yields an easier-to-resolve microsensor. With merits of fast label-free detection, high sensitivity, and further miniaturization capability, the approach demonstrated here will enlighten the development of optofluidic microsensors in exploring diverse biochemical processes.

The microlaser-integrated microtubule is constituted of an inner air core and an outer spherical shell, as shown in Fig. 1(a), which provides an inherent microfluidic channel for analyte delivery. Its fabrication is based on the fuse-and-blow method.¹⁰ (See supplementary material Ⅰ for sample fabrication.) The Raman gain results from the interaction between the photons and photo-phonons of the material,³⁵ as shown in Figs. 1(b) and 1(c). The ultrahigh quality factor and extremely small mode volume of the microcavity allow for resonant buildup of high circulating power, which significantly reduces the Raman lasing threshold.¹³ 

Sensing with microcavity lasers often encounters challenges due to the weak perturbation between the analyte and evanescent field. Due to the inherent microfluidic channel, the hollow microtubule morphology supports quasi-droplet modes and allows laser modes to move beyond the evanescent reaction limitation, resulting in a substantial improvement in detection sensitivity.¹¹ The finite element simulation in Fig. 2 shows that the higher the radial order, the further the location of the mode maximum from the inner surface. In particular, the maximum interaction field of the third-order mode enters the microfluidic channel and becomes core mode. Nearly all of the sensing field is presented inside the dielectric interface so that the analyte can interact with the maximum intensity of the sensing field.

The schematic setup of the RLHF microsensor system is illustrated in Fig. 3. When pump light is coupled into the microcavity through a tapered fiber, Raman laser is generated in the microtubule with a radius of about 60 µm and a wall thickness of about 8 µm. For biochemical sensing, the sample solution enters the microtubule using a syringe connected to the microfluidic channel. For a solution-filled microtubule, the quality factor still remains 1.8 × 10⁷. When Raman laser is established in the microcavity, it will be coupled out to the same tapered fiber and then transmit. The forward propagating laser is frequency-shift-modulated using a pair of AOMs, and then, the modulated light returns back to the resonator through a circulator and modifies the laser properties. (See supplementary material Ⅱ for more details about the system construction.)

In the experiment, we tune the pump in resonance to the microcavity laser mode and generate Raman laser at 1568 nm. The LRO peak and the feedback interference signal can be clearly observed on the spectrum, as shown in Fig. 4(a). The variation in the sample properties (e.g., effective index change) inside the microtubule may result in laser relaxation shifts³⁶ and output intensity changes.

The resulting modulated laser is detected via PD₂ through the back-propagating path, carrying sensing information at the differential frequency-shift feedback frequency Ω. A lock-in amplifier (LIA) is used to demodulate these signal changes. For this purpose, a pair of signal generators provide sinusoidal reference signals via an electric mixer to satisfy the LIA bandwidth in the sub-MHz level. Then, the input signal and the reference signal are simultaneously sent to LIA for demodulation, as shown in Fig. 4. In general, the response stability of the system depends on two main aspects. One is the laser stability, which entails considering the frequency shifting stability of AOMs. The other is the microresonator coupling stability, including the coupling state of air disturbances and the stability impact of liquid flow within the microfluidic channel. A series of operations are implemented to minimize signal fluctuations and enhance the measurement stability of the system; please refer to supplementary material Ⅴ for further details.

In the experiment, the laser relaxation frequency is 2.87 MHz and the shift frequency is set to be 3.4 MHz to meet the laser relaxation band, as shown in Fig. 4(b). (See supplementary material Ⅱ for the differential frequency shift modulation process.) In this case, the sensing signal in the optical domain obtains gain amplification and is down-converted into an RF response of laser relaxation. A larger gain coefficient G as well as higher sensitivity can be obtained in the RLHF microsensor, as shown in the working principle diagram in Fig. 5(a). Notably, in contrast to our previous study where the feedback signal is utilized to narrow the laser relaxation linewidth on laser itself,³⁶ here we take advantage of the gain amplification effect provided by LRO on the frequency-shifted feedback signal. These two effects complement each other, resulting in enhanced performance.

The intracavity dynamics-based gain amplification is experimentally verified in Fig. 5(b). Locked by the reference signal, the demodulated feedback intensity increases with the enhancement of laser relaxation resonance. The closer the shift frequency is to the laser relaxation frequency, the larger the gain coefficient is, with a 12 dB enhancement from A to D. The result indicates that the signal intensity is, indeed, amplified by the LHF mechanism. It could improve the response change of the microsensor caused by a minor analyte variation, and the optical-gain-assisted regime can offer a better resolution. Different from the direct spectral analysis in the RF domain, the LHF technique does not require a large spectral measurement range, which, instead, extracts a signal change (ΔI) that is directly related to the reference frequency in a narrow band. It reduces the bandwidth limit and offers a way to observe the fast dynamic response. Moreover, the detected signal is modulated to a higher frequency, which could also effectively reduce the low frequency noise and improve the measurement signal-to-noise ratio.

Here, we first verify the performance improvement in the RLHF microsensor and test its single-particle detection capability. The ability to detect individual nanoparticles with a high resolution holds great potential in the recognition of molecular motion and conformational changes. A particle entering the lasing mode volume of the resonator induces a net change in polarizability and optical path, which perturbs the microlaser properties.¹⁷ In our system, these property changes appear as a slight frequency shift of LRO and an intensity variation in the heterodyne feedback signal. (See supplementary material Ⅳ for more details about the detection mechanism.) Through signal demodulation, particle information can be obtained.

We compare the single-particle detection performance with and without LHF under the same experimental conditions. A U-shaped fiber taper is used to make particles attach and detach from the microsensor. The comparative result shown in Fig. 6(a) indicates that only under gain amplification of the LHF technique can the square response signal be distinguished. The black line shows that when the system noise is strong enough to submerge the laser response signal, particle information cannot be extracted at this time. However, based on the enhancement of LHF, the corresponding response changes caused by individual particles can be clearly resolved by demodulating the heterodyne feedback signal [red line in Fig. 6(a)]. In this case, particles attaching or detaching from the microsensor bring step-like discrete changes. Every time one particle is transferred to disturb the microcavity lasing field, an obvious response change is observed; once the particle is taken off, the response intensity goes back to its original value, which further verifies the detection repeatability and reliability. The discrete response heights of individual particle binding steps obviously exceed the system noise level, indicating that the LHF mechanism possesses better detection performance than the conventional microlaser sensing method.

To further explore the detection capability of the microsensor, 50-nm-radius polystyrene (PS) particles are successively deposited on the surface of the microcavity at the same location. The result is shown in Fig. 6(b), in which the five discrete response changes indicate five individual particle binding events. A cross correlation method is then used to extract the particle transition points, as the red line shows, which has a high signal-to-noise ratio, and the variation heights are almost consistent. A comparison of the proposed microsensor with other microcavity measurement methods is made in Table Ⅰ.

We also analyze that the response changes vary with different particle sizes. The particles are controlled within the mode distribution at the same position on the resonator to ensure a consistent response condition. Figure 6(c) shows that the larger the particle size, the greater the slope of the response intensity change. Note that the field distribution in different laser modes and their spatial overlap with particles can also affect the detection efficiency and signal response;¹⁷ please see supplementary material Ⅳ for relevant discussions. In summary, as a new label-free sensing method with single-particle detection capability, the proposed RLHF microsensor is expected to provide guidance in molecular dynamics and biomolecular recognition.

As a further validation of rapid detection capability, the dynamic phase transition of a thermo-responsive hydrogel PLGA is investigated using the RLHF microsensor. The sample solution enters the microtubule via a syringe and is reacted inside the microfluidic channel. As shown in Fig. 8(a), an 808 nm laser diode is used as the irradiation source to optically modulate the phase transition of the material. Below the lower critical solution temperature (LCST) of the material, the hydrophilic amide group is connected with water via a hydrogen bond and exists as a fully hydrated random coil.⁴⁶ As temperature exceeds LCST, the solution experiences a deswelling phase transition and transforms from the hydrophilic to hydrophobic state [bottom of Fig. 8(a)], mainly accompanied by heat transfer and RI variation. Typically, monomers polymerize due to an increase in the hydrogel mass density, resulting in an increase in RI. Furthermore, the chemical reaction may involve heat transfer, which also affects the temperature and RI of the material.³⁰ In this process, the RLHF microsensor is used to monitor the subtle changes in temperature and RI in real time, which could illustrate the dynamic evolution of hydrogel.

The measurement result is shown in Fig. 8(c). The hydrogel solution absorbs energy, and then, both the thermo-optic effect and RI change simultaneously. The laser heterodyne feedback signal provides unambiguous evidence of the PLGA phase transition dynamics, which can be clarified as the following stages: (Ⅰ) Pure hydrophilic state: At first, hydrogel is in the hydrophilic state. When the induction light turns on, the solution absorbs light and a sharp response increase occurs mainly due to the thermo-optic effect of silica. (Ⅱ) Hydrophobic transition: The solution starts to phase-transform and the hydrophobic transition absorbs heat from the environment, resulting in a small temperature decrease and a slight signal intensity drop. At this point, the contribution of the temperature decrease dominates that of the RI increase. (Ⅲ) Subtransition: The phase transition has not been fully induced, and the signal change slope is small, which is mainly attributed to the slight change in the physical properties of the material, e.g., volume shrink. (Ⅳ) Gelation state: Due to irradiation energy accumulation, LCST is reached and the rapid gelation process is accelerated. The significant increase in RI leads to a remarkable change in response intensity, causing the hydrogel to transition into the hydrophobic state. This process typically takes around ten seconds, and then, the gelation dynamic completes as indicated by the nearly constant response intensity. The initial liquid sample turns into a jelly-like substance and its liquidity becomes poor, as shown in the microscope images in Fig. 8(b). (Ⅴ) Pure hydrophobic state: Upon turning off the induction, there is a rapid downward shift in response due to the instantaneous thermal decrease. The hydrogel undergoes a swelling transition with a decrease in its intrinsic RI. Notably, the signal deviates from its original value after a phase transition cycle, as indicated by the blue circle in Fig. 8(c). This difference mainly reflects the RI change of the sample volume, which is because the molecular unwinding of the gel group is relatively slow during the inversion process so that the response intensity does not return to its original value within a short period.

To indicate the exclusive impact of hydrogel phase transition, a calibration curve detected with deionized water is made for comparison, as shown as the black line in Fig. 8(c). It is intuitive that in cases where hydrogel reaction is absent, only the response transients caused by thermal changes are observed. This further strengthens the reliability and validity of the obtained results, emphasizing the distinct effect of hydrogel phase transition on the measured signals.

To validate the repeatability and accuracy of the measurement, a hydrogel phase transition experiment is conducted multiple times. As shown in Fig. 9(a), the response trend of the multiple tests is similar and the specific response changes at different reaction stages (e.g., the endothermic intensity changes by 0.018 V, the gelation changes by 0.025 V, and the signal deviation after a phase transition cycle is 0.013 V), which demonstrates the good repeatability of the results. We extract the response changes that reflect the bulk RI variation after one gelation cycle. These response changes are then plotted at different induction levels. The result in Fig. 9(b) shows that higher irradiance results in larger RI and the response intensity changes. Notably, the response signal may reach saturation when the induction level surpasses a specific value, indicating that the hydrogel has transformed into a fully hydrophobic state. Furthermore, a standard refractometer is used to calibrate the RI change for a comparative analysis. As Fig. 9(b) shows, the measurement trend detected by the RLHF microsensor aligns with the RI change obtained by the refractometer. For two different types of hydrogels, phase response curves are both consistent with the calibration results, which further validates the measurement accuracy. The slight deviation may come from the change of factors, such as chemical bond formation, gel density, and additional strain introduction.⁴⁷ 

This section demonstrates the rapid response, shear stress-free operation, and excellent repeatability of the RLHF microsensor, which serve as valuable guidance for the application of hydrogel materials. The inherent microfluidic channel enables sample consumption as low as 50 pl. Besides real-time monitoring of hydrogel phase transition, the proposed microsensor could pave the way for many other label-free biochemical detection and provide portable devices for medical diagnosis.

By employing the laser differential frequency-shift feedback technique in a microlaser-integrated microtubule, the sensing signal in the optical domain is down-converted into an RF response with a higher intensity when the feedback signal resonates with LRO. It contains two kinds of enhancement factors, namely the strong light–matter interaction provided by the microcavity and the intracavity dynamics-based gain amplification provided by the LHF effect. It does not require an external reference laser, which also avoids the limitations of tracking the resonance of a particular laser mode. The optical gain assistance of LHF enables the microsensor to extract tiny intensity variations of the sensing signal, which helps offer a better resolution. In addition, the proposed microsensor reduces the spectral measurement range and offers a way to observe the fast dynamic response.

Due to this regime, a measurement resolution with the nanoparticle detection limit down to 50 nm radius and a noise-limited refractive index resolution of 8.18 × 10⁻⁷ RIU are demonstrated. This is achieved without exploiting the plasmonic effects, external references, or active stabilization. The dynamic phase transition of thermosensitive hydrogel is further investigated as a validation of its rapid detection capability. Notably, specific packing schemes can be further demonstrated to reduce environmental perturbations and improve the stability. Exploring encapsulation with elastic polymers may be a practical solution.³⁴ Moreover, the microsensor can be further biomodified and combined with other techniques to increase the application scenarios, such as spectral fingerprint determination and specific sensing.^2,48 As a compact, small sample consuming, and label-free technique with better sensing performance, the platform demonstrated here may be considered as a forerunner in the field of reactive biochemical sensing.

See the supplementary material for supporting content, including (I) Materials and reagents, (II) System construction, (Ⅲ) Theoretical analysis and simulation, (Ⅳ) Supplementary for particle detection, and (Ⅴ) Noise and stability analysis.

This work was supported in part by the National Science Fund for Excellent Young Scholars of China (Grant No. 51722506), the National Natural Science Foundation of China (Grant No. 51961130387), and the Royal Society-Newton Advanced Fellowship (Grant No. 191072).

The authors have no conflicts to disclose.

M.L. and Y.T. conceived the experiments. M.L. and Z.D. built up the experimental system. M.L. and M.T. conceived the data analysis. Y.T. provided assistance in discussing and analyzing the experimental results. M.L. and Y.T. wrote this manuscript, which was revised by all authors.

Mingfang Li: Conceptualization (equal); Investigation (equal); Validation (equal); Writing – original draft (equal). Zongren Dai: Formal analysis (equal); Investigation (equal); Writing – review & editing (equal). Mingwang Tian: Formal analysis (equal); Investigation (equal); Writing – review & editing (equal). Yidong Tan: Conceptualization (equal); Formal analysis (equal); Investigation (equal); Project administration (equal); Supervision (equal); Writing – review & editing (equal).

The data that support the findings of this study are available within the article and its supplementary material.