Whispering gallery mode (WGM) microlasers have been widely used for biological sensing both in vitro and in vivo [1,2]. Their high sensitivity to the refractive index (RI) of their surroundings has allowed measurements of target molecules with concentrations down to the single molecule level [3], while their small size makes them ideal for biointegrated sensing applications [2,4]. They have been used as intracellular lasers inside single cells [5,6] and cancer spheroid environments [7,8], for tracking cell migration [7,9,10], as well as in single cardiomyocytes for contractility measurements [4]. Furthermore, microlasers have been used for in vivo experiments in zebrafish embryos to measure the contractility of the heart [4].

To excite the microlasers, image the biological sample, and collect the laser emission, previous studies have made use of epi-fluorescence microscopy systems, where a single objective is used to focus light onto the sample and to subsequently collect the spatial and spectral information [4–6]. Others have also implemented confocal scanning microscopy for improved optical sectioning, creating detailed volumetric data of the trajectories of the microlasers [7,11]. Unfortunately, these two imaging modalities either have limited sectioning capabilities or require relatively large scanning times, respectively, while both have in common that they can cause photobleaching and phototoxicity while scanning the biological sample.

Light sheet fluorescence microscopy (LSFM) is an imaging method in which a plane of light is used to selectively illuminate whole slices of a sample at one time [12–14]. Typically, fluorescence from this slice is then collected via detection optics that are orientated orthogonally to the illumination plane. By scanning the illumination and detection planes in unison, a sample volume can be divided into several slices that are reconstructed to form a final 3D image. This method allows imaging large volume samples with lower phototoxicity rates and higher imaging speeds than what is possible with standard confocal microscopy systems. As a result, it has been implemented in studying large samples over long time periods including the embryogenesis of organisms such as Drosophila melanogaster [15] as well as full 3D imaging of the zebrafish heart [16]. With relevance to sensing with additional modalities, LSFM has been used in conjunction with second harmonic generation nanoprobes for the purpose of microangiography in live zebrafish embryos [17]. However, to the best of our knowledge, no studies have been performed on the use of an LSFM system in combination with microlaser spectroscopy.

In this study, we present the design and application of a light sheet fluorescence microscope that is coupled to a high-resolution spectrometer. We explore the use of a scanned light sheet for pumping WGM microlasers and analyze the effects of light coupling on the characteristics of the laser spectra. We then use the system for refractive index sensing of the microlaser environment in an agarose matrix and in the tail muscles of a zebrafish embryo.

Figure 1 illustrates the principle of WGM microlaser refractive index sensing on a light sheet microscope. Polystyrene microspheres doped with a fluorescent dye that act as microlasers are injected into a biological sample [Fig. 1(a)]. The sample is mounted onto the microscope and the light sheet is used to pump the microlasers [Fig. 1(b)]. The light sheet is scanned across the microlaser in steps and an image as well as a spectrum is recorded at each step. Subsequently, the images and spectra can be stacked and the variation of the peak position in the spectra can be quantified [Fig. 1(c)]. Upon determining the peak positions, we fit the laser modes to an optical model to determine the external refractive index (next) for each pump position. As a result, a 3D image of the microlaser and information about the refractive index distribution in its direct environment is obtained. Apart from the fluorescence signal from the microlaser, which typically results from the long wavelength edge of the gain material, additional fluorescent probes that stain specific biological structures can be imaged either simultaneously or via sequential scanning with different excitation lasers and suitable filters in the emission path.

The general layout of the LSFM system comprises an illumination and a detection arm that are aligned orthogonally to one another (Fig. 2). The detection arm is further divided into an imaging and a spectroscopy path. The sample is mounted inside a water-filled sample chamber and positioned with a motorized xyz stage (Zaber, X-XYZ-LSM025A).

In the illumination arm, a laser combiner (Hübner Photonics, C-FLEX) with continuous wave lasers emitting at 488 nm (Cobolt, 06-MLD) and 561 nm (Cobolt, 06-DPL), as well as a co-aligned nanosecond-pulsed diode-pumped solid state laser emitting at 532 nm (Coherent, HELIOS 532) with tunable pulse repetition frequency is used as sources for fluorescence excitation and microlaser pumping, respectively. The lasers are aligned such that they are collinear and are focused into a light sheet via a cylindrical lens (Thorlabs, LJ1567L1-A). The light sheet is formed on a galvanometer mirror system (Thorlabs, GVS102) and imaged onto the back focal plane of a 20× water immersion objective (Olympus, UMPLFLN 20× WI, numerical aperture=0.5) via a beam expander system consisting of a scan lens (Thorlabs, CLS-SL) and a tube lens (Thorlabs, AC254-150-A-ML) such that they fill the back plane along the z-axis. An iris is placed at the back focal plane of the objective to control the effective numerical aperture of the illumination arm to allow different widths of the light sheet. A dichroic beam splitter (Thorlabs, DMLP550R) is placed between the illumination objective and the tube lens, which allows the excitation lasers to pass through to the sample, but which reflects the long wavelength components of the light collected by the illumination objective. The reflected light then passes through a long pass or bandpass filter to remove the excitation light. An image is then formed by a tube lens positioned in front of a CMOS camera (CAM1, Basler, AC2040-80). This camera provides a view along the direction of the light sheet and is used to observe the position of the pump light sheet on the microlasers to assist alignment.

The detection arm uses a 40× water immersion objective (Olympus, UMPLFLN 40× WI, numerical aperture=0.8) for collecting the fluorescence and microlaser signals. A 50:50 beam splitter (Thorlabs, BSW10R) is placed behind the objective, which divides the signal between the light sheet imaging path and the spectroscopy path. The light sheet imaging path consists of a tube lens (Thorlabs, TTL180-A) that forms an image on a CMOS camera (CAM3, Hamamatsu, ORCA Flash 4.0 v3). A filter wheel (Zaber, FWR06A-E02) is used to select fluorescence channels. To ensure that the focal plane of the detection objective overlaps with the plane of the light sheet illumination, we typically moved the sample through the static pump beam to scan the entire sample. Alternatively, the detection objective can be scanned by a piezo stage (Physik Instrumente, P-628.1CD).

The spectroscopy path couples the microlaser signal to a round-to-linear fiber bundle (Andor, SR-OPT-8002) made up of 19 fiber cores each with a diameter of 100 μm via an objective (Zeiss, A-Plan 10×/0.25 M27). With this design, we directly image a part of the detection plane onto the fiber bundle. The field of view of the fiber bundle has a diameter of about 330 μm, significantly larger than what would be possible with a single fiber. However, this field of view is smaller than the image observed by the CMOS camera (CAM3). Consequently, the microlasers were positioned in the center region of the CMOS image to ensure efficient collection of the emitted light by the fiber bundle. At the other end of the fiber bundle, the image of the 19 vertically aligned fiber cores is projected onto the entrance slit of the spectrometer (Andor, Shamrock 303i) using a 4f relay system consisting of two achromats (Thorlabs, AC254-50-A-ML). Spectra are detected with a CCD camera (CAM2, Andor, Newton 971) in full vertical binning mode with 50 μm slit width and a 1800 lines/mm grating. The nominal optical resolution of the setup is about 80 pm. For sample mounting, we make use of a cobweb holder design [18] in which a thin film of mounting media is applied to the holder into which the sample is embedded. Polystyrene microspheres (Microparticles GmbH, PS-FluoRed-Fi264, 15 μm diameter) are used as microlasers due to their low lasing threshold and their previous use as microlaser sensors [4,5].

We first investigate the efficiency of the microlaser excitation in terms of signal strength as recorded with the spectrometer. For this, we apply two different excitation beams, a Gaussian beam and a light sheet. The microlaser is then scanned through the static pump beams from one end to the other [Fig. 3(a)]. Here, microlasers are embedded in a homogeneous environment, consisting of 1% agarose matrix. The microlaser and pump beam are aligned such that the pump light sheet is initially incident on the edge of the microlaser that is further away from the detection objective and the xyz stage is set to scan in the z-direction in steps of 2 μm. The spectrometer is set to trigger with each z step resulting in one spectrum per step until the entire microlaser has passed through the pump beam. This experiment is also performed without the cylindrical lens to generate a Gaussian beam while keeping the laser pulse repetition frequency (500 Hz) and pulse energy constant.

For both pump beam profiles, we observe a characteristic intensity distribution that consists of two maxima located roughly at the point where the pump beam is at the edge of the microlaser [Fig. 3(b)]. In addition, the intensity of the maximum that originates from the side that is closer to the detection optics is typically higher than the maximum at the rear edge. Importantly, we observe that pumping with either a light sheet or a Gaussian beam allows WGMs to be excited over the entire diameter of the microlaser, but that the recorded intensity is significantly reduced when the pump beam is in the center region of the microlaser. This effect can be explained by considering the coupling behavior of Gaussian beams into WGM resonators and the directionality of the emission of the excited WGMs. These characteristics indicate that the effect of light coupling of the excitation beam into the microlaser and the geometry of collecting the microlaser emission play an important role when optimizing the collection efficiency.

To obtain a detailed understanding of the free-space light coupling of a Gaussian beam into a spherical resonator we adapt a previously published theoretical model [19]. The model assumes a 2D circular cavity and gives an analytical expression for Gaussian beams that couple into cylindrical waves that resonate inside the cavity (i.e., WGMs). We use the following expression from the model to examine the coupling efficiency of the Gaussian beam: cm=eim(π2−θ0)−ikq−(kr0d−m)2k2w2+i2kq21+i2qkw2,(1)where d=r0sin(φ0+θ0) and q=r0cos(φ0+θ0). Furthermore, θ0 is the incidence angle of the Gaussian beam, φ0 is the azimuthal angle on the resonator at which the beam is coupled into the resonator, m is the mode number of the WGM being coupled into, k is the wavenumber of the pump beam, r0 is the distance of the Gaussian beam waist from the center of the resonator, and w is the width of the Gaussian beam. We calculate cm for the case of a beam of width 0.2 times the radius of the cavity (rc) intersecting the cavity at 90° and adjust r0 while holding all other values constant. We then normalize the result to examine where the maximum of coupling efficiency occurs with regard to the beam position [Fig. 3(b)]. In accordance with the experimental data we observe for the Gaussian beam, the model features pronounced maxima in coupling efficiency at the edges of the cavity. Furthermore, the coupling efficiency is strongly reduced when the beam is located in the center of the cavity.

The experimental light sheet excitation curve follows a similar trend but the intensity decay in the center of the microlaser is smaller in comparison to the Gaussian beam. This is because part of the light sheet will always intersect with the edge of the microlaser when being scanned across the cavity. However, despite the improved coupling the reduction of the signal intensity in the center is still visible. We attribute this effect to the directionality of the emission of the excited WGMs that imposes limitations on the collection efficiency. In the case when the light sheet is exciting the microlaser in the center, the WGMs are resonating and emitting in the same plane as the light sheet. However, this plane is orthogonal to the detection axis, meaning that the microlaser emission is collected with lower efficiency. Nevertheless, we can conclude that the drop in signal intensity is strongly reduced when a light sheet is used compared to a scanned Gaussian beam.

As demonstrated experimentally and theoretically, the coupling efficiency of the pump beam into WGMs is highest at the edge of the microlaser. This characteristic has important consequences for the orientation of WGMs excited under light sheet excitation. Considering that light coupling and lasing are only efficient where the light sheet overlaps with the edge of the microlasers and the fact that the oscillation plane of the WGMs has to go through the center of the microlaser, leads to the generation of two separate WGM paths that cross each other in the center of the microlaser [see the two paths indicated in Fig. 3(a)]. We will use the term “cross modes” to describe these two planes of laser oscillations inside the microlasers. Interestingly, by moving the light sheet from one end of the microlaser to the other, these cross modes will scan the entire surface of the microlaser [Fig. 4(a)]. We will use this effect later to perform independent measurements of the refractive index along these two paths.

After we have assessed the excitation properties of microlasers under light sheet excitation we now want to investigate the refractive index sensing characteristics. As before, we first measure microlasers that are embedded inside an agarose matrix. The use of this matrix is motivated by the fact that many LSFM experiments use water immersion objectives with agarose as the mounting medium to keep the sample in place while allowing refractive index matching with the aqueous environment [12–15]. Compared to more complex biological samples, the agarose matrix also provides an environment with a relatively homogeneous refractive index with no dynamic changes over time. This allows us to analyze the microlaser spectra and the observation of the cross modes without the additional complexity of a heterogenous refractive index environment.

For sample mounting we used a low melting point agarose (Fisher Bioreagents) that was diluted with deionized water to produce a 1% agarose gel. Microlasers were embedded when the gel was in a liquid state and mixed with a pipette to ensure dispersion throughout the medium. The liquid was then pipetted onto the sample holder as a thin film and left to congeal. Afterwards, the sample was mounted onto the microscope for analysis. Light sheet scanning of microlasers was performed by translation of the stage in the z-direction from one side of the microlaser to the other while collecting images and spectral data. Each scanning step was set to be 3 μm in depth with each spectrum acquired with an acquisition time of 100 ms (pump laser repetition frequency 500 Hz). The precise positions of the WGMs were extracted via fitting to a Voigt profile, which serve as input parameters for an analytical model used in our previous work [4,20,21]. This analysis reveals the microlaser size and external refractive index (next), which is the refractive index inside the evanescent field component of the WGMs.

Agarose-embedded microlasers were pumped with a light sheet and scanned along the z-direction (Fig. 4). When the light sheet pumps the microlaser at an off-center position, bright spots that indicate the presence of lasing cross modes appear on its perimeter [Fig. 4(a)]. Two of these bright spots appear at the position where the light sheet overlaps with the edge of the microlaser while another two bright spots appear on the opposite side of the microlaser. At these four points, light scattering and outcoupling from the WGMs in the direction of CAM1 is most efficient. Very faintly, the paths of the cross modes are also visible, which are on a plane that goes through the center of the spherical microlaser as explained above. As the light sheet is scanned across the microlaser, the cross modes change their oscillation plane as it is visible from the bright spots moving along the circumference of the microlaser. Consequently, it is possible to scan different paths on the surface of the sphere, each of which can probe the average external refractive index along its path. If a sufficiently small step size is used, the spectra will encode the 3D refractive index information of the local environment around the microlaser.

The multimode WGM spectrum of the microlaser inside the agarose matrix shows an almost constant mode position [Fig. 4(b)] as it is expected for a perfect sphere inside a homogeneous refractive index environment. The central peak positions were measured via a Voigt fit of the laser modes and are overlayed with the spectral data [Fig. 4(c)]. The relative change of the calculated refractive index (Δnext) over the entire microlaser is typically around 2×10−4 refractive index units or less. These small but detectable fluctuations can be attributed to inhomogeneities in the density of the agarose gel caused by the mixing process or small dust particles from the external environment that became trapped in the gel when it was in the liquid state.

To demonstrate refractive index sensing with light sheet scanning in complex tissue, we performed measurements inside the tail muscle of a fixed zebrafish embryo [Figs. 5(a) and 5(b)]. Thin muscle filaments were stained with an actin dye (Spirochrome, SPY-555 actin). Microlasers were inserted into the sample using a tapered glass capillary (Sutter Instrument, BF100-58-10) pulled with a Narishige needle puller and injected with a microinjector (Picospritzer). Post microinjection, the sample was mounted on the microscope in a layer of 1% agarose and examined under brightfield illumination to ensure correct location of the microlaser [Fig. 5(c)]. The sample was then subsequently illuminated with the 488 nm continuous wave laser and the 532 nm pulsed pump laser beam with a repetition rate set to 500 Hz. The pump was scanned across the bead in 1.66 μm steps and spectra were acquired using 100 ms acquisition time. Images of both the bead fluorescence and WGM lasing [Fig. 5(d)] were taken by switching between filters using a motorized filter wheel. The process of peak position extraction and RI fitting was performed as described above for the agarose data.

The spectra from the microlaser inside the zebrafish tail [Fig. 5(e)] are markedly different compared to those measured in agarose. First, the intensity of the laser modes inside the zebrafish is lower than in agarose, mostly because the signal must pass through a substantial depth of tissue where it can undergo scattering and absorption. However, these processes are elastic in character, resulting in no change of the photon energy. Thus, they have no effect on the WGM position measurement. It is also visible that the signal-to-noise ratio is strongly reduced, especially when the pump beam is located in the center of the microlaser. Despite the lower signal, precise mode fitting could be performed for all recorded spectra. Second, there is a significantly larger shift of the WGMs depending on the cross mode position, with some spectra showing a clear splitting of the laser modes [Fig. 5(e)], indicating a difference in the effective RI in the two cross modes.

Absolute refractive index measurements of the two cross modes were performed by peak fitting two subsequent pairs of TE and TM modes and matching these mode positions to theoretically calculated mode positions [4]. By applying a double-Voigt fit, we were able to extract two separate mode positions for the two cross modes and calculate the external refractive index for each of the two. Here, we assume that any spectral peak that has a width larger than the spectral resolution of the setup is composed of two separate modes [Fig. 5(g)]. However, at the edges of the microlaser, the intensity of the TM modes also decreases to the point that their position cannot be fitted any more. To extract the absolute refractive index next at these points, we make use of the fact that the change in refractive index scales linearly with the change in mode position [4]. Therefore, we can calibrate the sensitivity of the microlaser after which the position of only one TE mode is needed to determine next.

The corresponding next as a function of the z-position of the light sheet is shown in Fig. 5(e), which shows two separate next values, one for each of the two cross modes. At the edge of the bead (z=0) the difference between next is not as large as the cross modes probing similar paths. However, advancing the sheet by about 1.66 μm leads to a pronounced splitting of the modes [Fig. 5(f)], which translates into a refractive index difference between the two cross modes of about 0.006. Advancing towards the center of the microlaser, the longer wavelength peak almost disappears indicating next becoming more similar across the two paths as it is expected when the light sheet is centered in the middle of the microlaser. Continuing the scan towards the opposite side of the microlaser the observed peak splitting almost mirrors the characteristics from the first half of the scan. We attribute this to the symmetry of the microlaser and the fact that the cross modes will follow very similar paths when they are offset either side of the center of the microlaser, while the small differences between the left- and right-hand sides of the scan are attributed to the relatively coarse scanning that leads to slightly different paths being probed around the microlaser. The maximum measured difference in next between the two cross modes is 0.006 refractive index units, which is a factor of 10 greater than what we observed in the agarose matrix. This indicates that the refractive index fluctuations are a result of refractive index inhomogeneities in the zebrafish tail muscles that are well resolved by scanning of the microlaser.

We have designed and constructed an LSFM system with the ability to simultaneously collect high resolution spectra from WGM microlasers and collect three-dimensional imaging data. We also characterized the effects of the Gaussian light sheet on exciting WGMs and describe how the light coupling characteristics lead to the occurrence of so-called cross modes. The oscillation planes of these cross modes are dependent on the intersection of the light sheet with the microlaser by efficiently pumping WGMs at the rim of the microlaser. This newly observed effect can be used to systematically scan the entire surface of the microlaser with the two cross modes by moving the light sheet across the microlaser. Measurements were performed in agarose, an inert medium with a homogeneous refractive index, and in the muscle tissue of a fixed zebrafish embryo. Our results demonstrate that using light sheet spectroscopy for refractive index sensing can be implemented along standard imaging modalities and can be used to scan the local refractive index environment quickly and efficiently.

The here described capabilities for measuring local refractive indices with high precision are relevant in other biological settings as, for example, the determination of biomechanical properties of the tail muscle and notochord in vivo [22]. In addition, the method could also be implemented in complex organoid environments where position-specific processes could be mapped over time, allowing to extract information in very heterogeneous environments [23,24]. A further implementation could be to use our system with deformable microlasers and droplet cavities, where the light sheet selectively excites different axes inside the deformed cavities to extract geometrical information [25,26].

In the future, an increased resolution in the light sheet scanning could reveal more details in the 3D refractive index profiles and allow to correlate this information to structures in the biological environment. We have recently demonstrated spectral acquisition rates of more than 100 kHz in a fiber-coupled spectrometer [11], meaning that extremely fast scanning of large tissue segments with a swept light sheet will be possible. In this context, excitation of multiple microlasers is also feasible as the spectra can be decoupled either in the spectral domain by analyzing the characteristic, size-dependent WGM patterns or spatially, by looking at the information of separate fiber cores by reading out specific parts of the spectral CCD area detector. Alternatively, instead of fiber-coupling the light sheet excitation it should be possible to image a scanned Gaussian beam onto the entrance slit of the spectrometer by de-scanning the image and turning it 90 deg, which would strongly increase the field of view of the spectrometer and eliminate fiber-coupling losses. A further improvement that could be made would be to implement alternate light sheet layouts. For example, the current traditional two-objective LSFM design of our setup might be changed to adapt the spectroscopy part to a one-objective LSFM system, which allows for fewer constraints on the sample mounting [27,28]. Finally, by using multiphoton pumping of the biointegrated lasers [29], refractive index sensing deep inside biological tissue at a depth that is currently not accessible by light sheet microscopy could be realized.

Acknowledgment. We thank Soraya Caixeiro for providing the code for fitting Voigt profiles of WGM spectra featuring peak splitting and broadening. We also thank the workshop of the Department of Physics and Astronomy, University of St Andrews, for technical support. M.S. acknowledges funding by the Royal Society.