1. Introduction

Whispering gallery mode (WGM) resonators have recently emerged as versatile photonic devices [1]. They are efficient tools to study fundamental scientific questions, e.g., in cavity quantum electrodynamics [2] or quantum optomechanics [3]. They also find widespread application as low-threshold lasers [4], sensors [5], or comb generators [6,7] and filters [8] in optical communications. Part of this success of WGM resonators can be attributed to their ease of manufacturing and the capability to integrate quantum emitters to achieve active cavities. But most important are intrinsically very high Q-factors and the related strong confinement of light resulting from the rotational symmetry of the cavities.

This evident advantage of WGM resonators is also one of its main handicaps and makes free-space coupling to WGMs quite inefficient [9,10]. One way to enhance the energy exchange between the outside and the WGM is to use near field couplers such as tapered fibers or prisms that enable efficient evanescent field coupling [11]. The most straight-forward way to couple light directly into WGMs via free-space optics is achieved by excitation of a gain medium incorporated in a cavity. This allows remote excitation of WGMs observed as laser emission. But this method is still inefficient since the whole device and not only the modal volume of the emitting modes is optically pumped.

Quite challenging is the collection of laser emission from rotationally symmetric resonators. Typically light emitted perpendicular to the micro-disk plane has been collected using free-space optics [4,12–14]. But, ideal high-Q resonators do not emit in this direction. Observation of laser emission is here solely due to the fact that the light is scattered towards the detector by defects and impurities. So, the collection efficiency is limited since one only collects a small portion of the cavity’s emission. Further, the occurrence of imperfections is random and hardly controllable in the fabrication. Additionally, the polarization state of the emitted scattered light is not accessible and the distinction between TE and TM modes is impossible.

Since the disks are excited over and emit from the total disk area (and not only in the area of the cavity modes) one often finds the absence of any WGM signature, especially in the case of continuous wave excitation. The emission spectrum is then broad and unstructured since modal filtering is covered.

To enhance the collection of WGM emission several approaches have been applied in the past, most of them focusing on controlling the direction of the emitted light, e.g., with notches in the cavity rim [15,16], using spirals [17] or deforming the cavity geometry [18,19]. However, often an undesired reduction of the quality factor is associated with these cavity modifications. Approaching read-out tools such as fiber tips directly to the WGM region has been applied for micro-spheres in order to enhance the collection efficiency [20,21]. Wienhold et al. proposed paraboloidal micro-mirrors around micro-disks for an enhancement of the collection efficiency [22].

A determination of the mode polarization has been successfully realized in passive schemes by using fiber polarization controllers on the excitation and/or detection path [23,24], or for active micro-fibers [25,26] and micro-spheres [27] by employing a polarizer in front of the detector to distinguish TE and TM modes. The polarization is important for a full characterization of the modal spectrum. In addition its determination is essential for comparison with simulated data. Particularly important is the knowledge of the field polarization for polarization-sensitive applications, e.g., sensing [28,29]. For micro-spheres and micro-rings it has been shown in passive schemes that a combination of the sensor signal of both polarizations provides information on the orientation of adsorbed biomolecules [30] or allows the detection of conformational changes of proteins [31].

In the work presented we analyze various combinations of detection and excitation schemes to optimize the performance of WGM micro-lasers. We focus on schemes that are applicable to unmodified micro-disks which moreover do not require complex additional post-fabrication steps. In the first part of the paper we will demonstrate how localizing the pump energy to the region of the WGMs leads to a lowering of lasing thresholds and how the light collection can be greatly enhanced. In the second part, we identify the field polarization of lasing modes from active WGM micro-disks. We highlight the advantage of in-plane read-out of the laser emission for this purpose. Additionally, immersion in an aqueous environment assists the determination of the lasing mode polarization. Subsequently, we show the influence of the polarization of lasing modes for applications such as refractive index sensing.

2. Fabrication of micro-lasers and characterization methods

Polymeric micro-resonators are structured according to the fabrication scheme presented in [32]. For active devices the laser dye pyrromethene 597 (PM597) is mixed into the positive photoresist PMMA 950K at a concentration of 25 µmol/g PMMA. After stirring the solution for several hours, the dye-polymer solution is spin-coated onto a silicon substrate. To pattern disks the photo-resist is exposed by electron beam lithography. In the last step we remove the silicon beneath the disks in an isotropic etching process resulting in WGM micro-disks.

To optically characterize the micro-lasers, photoluminescence (PL) spectroscopy is applied. In order to achieve WGM lasing, the resonators are pumped with 10 ns pulses of a frequency-doubled neodymium-doped yttrium orthovanadate (Nd:YVO₄) laser at a wavelength of 532 nm and a repetition rate of 20 Hz. For the investigation of the WGM-modulated dye emission we employ a diode pumped solid state laser (DPSSL) as continuous wave laser source (532 nm).

Excitation of the micro-resonators is either performed in a free-space excitation configuration, where the laser beam is loosely focused onto the resonator or by using tapered fibers that are approached to the cavity rim enabling evanescent field coupling (fiber excitation). Similarly, for the read-out of the lasing emission on the one hand we use a microscope objective (50 ×, NA = 0.42) and additional free-space optics to image the emission onto a spectrometer equipped with a CCD camera (free-space collection). Both, collection from the top of the resonator (scattered light perpendicularly to the resonator plane) and of in-plane radiated light has been implemented. For the latter case, part of the chip adjacent to the resonators is removed in order to allow light propagation to the detector which is mounted in the plane of the disk and thus perpendicular to the excitation direction. On the other hand, collection of the emitted light is also performed by evanescently coupled tapered fibers (fiber collection).

In order to ensure efficient resonator-fiber coupling, passive transmission measurements are performed on the active cavities prior to using fibers as excitation and collection tool for fluorescence emission: Transmission spectra are recorded by coupling a tunable external cavity diode laser (New Focus, TLB 6300 Velocity) to the tapered fiber input. When the fiber is precisely aligned with the rim of the cavity sharp resonance dips in the transmission spectrum are observed, revealing efficient resonator-fiber coupling [33]. Once this state of coupling is reached, the fiber input can be connected to one of the 532 nm lasers to excite the dye-doped cavities. Tapered glass fibers are fabricated in a home-built stretching machine using flame brushing technique. For regulating the excitation power, a λ/2-waveplate in combination with a polarizer is used in the free-space measurement configuration, whereas a variable fiber attenuator is used for the fiber-based approaches.

The various excitation and collection possibilities are now combined to four different combinations:

A schematic of these different excitation and collection geometries is depicted in Fig. 1. A detailed illustration of the free-space photoluminescence setup can be found elsewhere [12].

 

Fig. 1 Excitation (grey color) and collection (red color) methods for active micro-cavities: Excitation is either performed in a free-space configuration or by evanescent field coupling using tapered fibers. Resulting WGM emission is collected with either a microscope or aligned tapered fibers and guided to a detector consisting of a grating spectrometer equipped with a CCD camera.

Download Full Size | PDF

3. Efficient excitation and detection of WGMs

In the following, we first reveal the limitations of the most straight-forward µ-PL configuration (method I) [4,12–14] as a motivation for necessary modifications of the measurement scheme. Subsequently, we discuss the performance of three alternative characterization configurations (method II, III and IV). If not stated otherwise, the presented data have been collected from identically processed dye-doped micro-lasers with a disk diameter of D = 50 µm.

3.1 Motivation

Generally, free-space excitation of micro-disks and free-space based light collection are simple and advantageous compared to complex and fragile tapered fiber coupling. However, read-out of WGM emission from the top of the micro-cavities (perpendicular to the resonator plane) is dependent on cavity impurities which scatter the light towards the detector direction. Also it suffers from a small signal-to-background fluorescence ratio, as a large part of the collected light stems from non-WGM modulated fluorescence emission from the inner part of the disk.

To motivate alternative measurement configurations, we show the limitations of method I in Fig. 2. Although the resonators shown in (a) and (b) have similar thresholds and are pumped equally above their threshold, the recorded spectra are markedly different.

 

Fig. 2 Free-space excitation of a micro-laser and free-space collection of the WGM emission perpendicular to the resonator plane. Imperfections at the resonator rim and corresponding alignment with respect to the spectrometer entrance slit greatly enhance the SNR as in case (a) compared to the case of a resonator without imperfections and missing alignment (b).

Download Full Size | PDF

The spectrum in Fig. 2(a) corresponds to a read-out where the resonator is positioned in such a way that imperfections that scatter the emission towards the detector direction are intentionally aligned on axis with the entrance slit of the spectrometer. For the presented resonator this is easily achievable as lots of imperfections at the disk rim are present [red dots at the rim in Fig. 2(a)]. In that way, a significant portion of lasing emission can be detected leading to a high signal-to-noise ratio SNR = 14.8 dB. Here, the SNR was defined as the ratio of the maximal peak intensity to the background fluorescence next to the lasing peak. For a cavity without or significantly less scattering centers and missing alignment [see Fig. 2(b)] a significantly smaller SNR of 2.2 dB is measured.

The significant reduction of the SNR can culminate in WGM resonances being buried within the spontaneous non-modulated fluorescence background. Low signal intensities are especially problematic for sensors that operate in aqueous environment where the collection efficiency is further reduced due to enhanced absorption losses. In the case of cw excitation a WGM modulation of the fluorescence spectrum is expected as a manifestation of the Purcell effect [21,34,35]. Such a modulation typically leads to a less distinctive WGM signature in the fluorescence spectrum compared to the case of WGM lasing. Due to the large collection of non-modulated spontaneous emission, the WGM signature cannot be recorded with method I. We will return to this case in more detail in section 3.3.

Without having access to the WGMs, the micro-cavities are useless for any application. The problematics with method I thus motivate alternative schemes. Moreover we are interested in measurement schemes that enable high collection efficiencies and thus signal intensities, e.g., for the determination of the spectral shifts of the WGM resonances used in sensing experiments [14,36]. White et al. showed that the uncertainty in the determination of the cavity mode is inversely proportional to the fourth-order root of the signal-to-noise ratio [37].

Additionally, low pump energy densities just slightly above the lasing threshold should be applied in order to prevent the dye molecules from photo-bleaching [14]. Also in that context methods with high collection efficiencies are beneficial as high signal intensities can be achieved even with small deposited pump energies.

On these grounds, we consider in the following three alternative measurement techniques that are independent of scattering imperfections and moreover beneficial regarding the achievable collection efficiency.

3.2 Determination of lasing threshold and collection efficiency

To compare free-space and fiber-based excitation and collection configurations we use the PL setups presented in chapter 2. The signal intensities and thus the values of collection efficiencies are comparable for each configuration as identical measurement conditions (CCD camera settings, spectrometer settings, integration time, etc.) have been used. The lasing threshold is an intrinsic property of the micro-cavity. It should depend on how the pump energy is deposited in the resonator but it should not depend on the detection scheme. Still, we need to confirm that the latter is indeed true for the collection schemes described here. Please note that the pump energy is not converted into a pump fluence but refers to the total energy deposited to the resonator area. The collection efficiency, however, has to be understood as a property of the specific measurement method and states how much light can be collected from the micro-resonator using a specific configuration. To compare the various methods we use the differential collection efficiency ξ, defined as the slope of the input-output curve above the lasing threshold [38].

The different measurement configurations are now compared to the PL setup using free-space excitation and free-space collection from the top of the micro-resonator (method I). Typical emission from a dye-doped micro-laser excited above threshold shows a complex spectrum with both, fundamental and higher-order WGMs [Fig. 3(a)]. The lasing threshold is determined from the kink in the integrated PL intensity when increasing the pump energy. To this end we evaluate the lasing mode that evolves at the lowest pump energy [see the red marked rectangle in Figs. 3(a) and 3(b)]. Several resonators from identically processed samples have been characterized leading to an average lasing threshold of E_th = (0.50 ± 0.22) nJ.

 

Fig. 3 (a): Typical PL spectrum of a PM597-doped micro-cavity excited above lasing threshold with fundamental and higher-order modes (recorded with method I). The lasing threshold is determined from the integrated PL intensity of the lasing mode arising at the lowest pump energy (marked in red color). (b): Using fiber excitation and collection results in a significant decrease of the lasing threshold and increase of the collection efficiency; note the dimension on the x-axis changed to pJ (c). Input-output curves for the different measurement configurations demonstrating significant differences in the resulting collection efficiencies (d).

Download Full Size | PDF

As already mentioned, method I relies on the existence of defects or impurities on axis with respect to the entrance slit of the spectrometer. We find an average differential collection efficiency of ξ = (1.14 ± 1.61) ∙ 10³ Counts/nJ. The fact, that the uncertainty is actually larger than the average value reflects the arbitrary nature of this detection method: Depending on the alignment of the imperfections with respect to the spectrometer entrance slit, collection efficiencies between ξ_min = 1.00 ∙ 10² Counts/nJ and ξ_max = 6.10 ∙ 10³ Counts/nJ have been measured.

We prove in the following that this arbitrariness is eliminated when the emission is either recorded in the plane of the disk using free-space optics (method IV) or by application of tapered fibers in method II or method III. First optimal coupling conditions are established for method II and III as described in chapter 2 prior to pumping the active micro-resonators with 532 nm. A resulting transmission spectrum using a tunable diode laser is depicted in Fig. 4. From the linewidth of the resonance dips we infer Q-factors as high as 6.5·10⁴.

 

Fig. 4 Transmission spectrum of a tapered glass fiber aligned to the rim of a dye-doped micro-disk. Sharp dips corresponding to the WGM resonances are observed. Q factors as high as 6.5·10⁴ have been determined from the resonance linewidths.

Download Full Size | PDF

The results of pump-energy dependent measurements taken on multiple resonators with the different configurations are summarized in Figs. 3(c) and 3(d) and listed in Table 1. From comparison to method I we can make three main statements. (i) The lasing threshold is independent of the collection method; (ii) the lasing threshold is reduced by a factor of ~100 using fiber excitation (method III); (iii) the collection efficiency is significantly increased using in-plane read-out (1-2 orders of magnitude for method IV) or fiber collection (2 orders of magnitude for method II, 4 orders of magnitude for method III).

Table 1. Averaged lasing thresholds and collection efficiencies for PM597-doped micro-lasers, investigated with different excitation and collection schemes

View Table | View all tables in this article

Statement (iii) is as expected: Obviously it is more efficient to use a phase-matched fiber for collection than relying on arbitrary scattering defects (method I). Although methods I and IV are both completely based on free-space optics on the excitation as well as the detection side, there is a crucial difference: The position of the detector with respect to the orientation of the resonator, see Fig. 1. A rotationally symmetric device actually does emit in-plane (curvature dependent radiation loss [22,39]) even in the absence of imperfections. By collecting mainly this in-plane radiated light in method IV, the collection efficiency is significantly enhanced and moreover making the method independent of random defects [compare Figs. 2(a) and 2(b)].

Also result (i) confirms expectations: The lasing threshold depends only on the laser itself and on the excitation configuration. Since the investigated resonators are identically processed we find an identical threshold in case of free-space excitation as long the collection efficiency is not too low and the onset of lasing can be clearly determined from distinct peaks.

We however find a significant reduction of the laser threshold when free-space excitation is replaced by fiber excitation (statement (ii)). We now take a closer look into the origin of this reduction. First we compare the deposited pump energy necessary for lasing with the theoretically expected value. Using [40] and following [14], the theoretically expected pump power for lasing can be expressed as

But the required pump energy to achieve lasing not only depends on sufficient concentration. It also has to be spatially localized in optimal overlap with the laser modes. In free-space excitation schemes dye molecules throughout the micro-disk are excited. However, most of these molecules do not contribute to lasing as they are located off the WGM rim region. Part of the persisting spontaneous dye fluorescence is actually directed to the spectrometer and results in a background below the WGM peaks as discussed in section 3.1.

In order to give an estimate of the energy that actually is available to the dye molecules located in the mode volume being responsible for WGM lasing, we performed FEM simulations using the eigen-mode solver JCMSuite from JCMwave. We determine the modal volume (defined according [41]) of WGM-disk geometries as used in our experiments. For fundamental WGMs in the wavelength regime between 610 nm and 640 nm we find a modal volume of V_M ~90 µm³. As rough estimate, we use the geometrical extension of a hollow cylinder with a height of 1.2 µm to determine the surface area of a WGM A_mode = 75 µm². Compared to the total surface area of a disk resonator with a diameter of D = 50 µm this corresponds to a fraction of C_mode = 0.04. Furthermore, from the absorbance measurements we performed for the determination of the absorption cross section in Eq. (1), the extinction coefficient at the pumping wavelength κ(532 nm) = 0.23 is known. Together with Lambert-Beer’s law we conclude that only a fraction of C_abs = 0.21 is absorbed from the incident pump light in free-space excitation methods. Considering these factors for the threshold pump energy E_th in free-space excitation configurations, we derive for the threshold energy that is actually available within the mode volume E_mode:

At the rim the pump energy is efficiently absorbed, especially when the pump laser is in resonance with a WGM mode. To demonstrate that this indeed is the case in our experiments we block the resulting laser emission with an appropriate filter. This allows observation of green pump light that is scattered at defects and imperfections all along the resonator rim and proves the strong overlap of pump light with the modal volume of the laser emission (see the corresponding video in the supplementary material, Visualization 1).

We can state at this point that applying fibers is a powerful tool to decrease the necessary pump energy to achieve lasing and to enhance the collection of light. But, it is also associated with a complex setup and time-consuming alignment. As an efficient alternative, in-plane read-out (method IV) enables collection of radiated light [22], also leading to an noticeable enhancement of the collection efficiency [see Table 1].

3.3 WGM-modulated fluorescence emission under cw excitation

In the following we have a closer look at the PL spectra of dye-doped resonators excited by a cw laser source. By using cw excitation instead of a pulsed laser, the lasing regime cannot be reached. However, from [21,42,43] it is clear, that a WGM-modulated fluorescence emission is expected. This modulation is explained by a modification of the spontaneous emission rate of the dyes in a cavity environment [27,34,44] and is a direct manifestation of the Purcell effect [35,45]. As we will see in the following, the possibility to observe this modulation strongly depends on the measurement scheme.

A typical fluorescence emission spectrum recorded with method I is shown in Fig. 5(a). As illustrated, no modulation through WGM resonances and a rather smooth emission is observed. Based on the results presented in section 3.2 we interpret this behavior as follows: A great portion of the emitted light originates from the non WGM-modulated inner part of the disk and is dominant to the sparse collection of WGM-modulated light from the cavity rim. Even for long integration times and a higher spectral resolution using finer diffraction gratings in the spectrometer, a WGM signature in the emission spectrum has not been observed as it is buried within the large non-modulated fluorescence.

 

Fig. 5 Fluorescence emission of dye-doped micro-disks under cw excitation: Method I in (a), IV in (b) and III in (c). Whereas in case (a) the expected WGM-modulation is not visible due to the large non-modulated spontaneous dye emission, in (b) and (c) a clear WGM-modulated dye emission with the expected free-spectral range is observed.

Download Full Size | PDF

This interpretation is reinforced by the results obtained in method IV. The corresponding emission spectrum is depicted in Fig. 5(b): Clearly, under the same free-space pumping conditions, the WGM characteristics can now be detected. Across the emission spectrum peaks separated by the free spectral range (FSR) are observed showing the micro-cavity’s influence on the dye emission; e.g., around the wavelength of 605 nm we find a FSR of 1.50 nm well coinciding with the expected FSR of 1.55 nm for a micro-disk with a diameter of D = 50 µm.

To get an even more distinct WGM signature, we employ cw excitation also in combination with fiber excitation and collection (method III). The corresponding spectrum is shown in Fig. 5(c) showing a very clear periodic modulation throughout the fluorescence emission. From the shape of the spectrum we identify related peaks and determine the FSR at 650 nm to be 1.85 nm, again in accordance with the expected value of 1.80 nm. The slight deviations can be explained with size variations of the individual resonators. By using a finer diffraction grating, additionally to the main peak, a substructure is visible showing that both fundamental and higher-order modes are excited (see insets).

Our results emphasize the necessity of appropriate measurement schemes to fully characterize and investigate the influence of WGM cavities on the fluorescence emission. The WGM signature has not been detected at all in method I. The collection of WGM emission is significantly improved in method IV (identical pumping conditions as in method I) and even further in method III when using fibers as collection tool.

4. Performance of dye-doped micro-lasers in an aqueous environment

In this chapter we optically characterize active micro-lasers in aqueous environment. Although using tapered fibers in water environment is feasible, the operation is tedious due to the fragility of the tapered fibers and is also limited to a lab environment. Therefore in aqueous environment, e.g., when used as sensing devices, often free-space schemes are employed. In the following we especially focus on measurement method IV in water environment. We will demonstrate the advantages of the aqueous environment reflected in an enhancement in collection efficiency and in the possibility to access the polarization state of the WGMs. We also address the effect of mode polarization on the sensing performance.

4.1 Collection efficiency in an aqueous environment using free-space collection configurations

In this section we compare the performance of method IV to method I in water. The accompanying reduction of the refractive index contrast results in an increase of cavity radiation losses [46,47]. As the light is radiated in-plane [22] and radiation is the dominating loss mechanism in aqueous environment [14], a further enhancement of the collection efficiency in method IV is expected. Considering that the radiation-dependent Q factor is exponentially dependent on the cavity radius (Q_rad ~exp(R) [46]), we expect method IV to be even more advantageous for smaller-sized cavities.

To check these expectations, we characterize the micro-resonators in aqueous environment by mounting the sample into a fluidic chamber. Similar as in the first chapter we performed pump-energy dependent PL measurements to determine the differential collection efficiency ξ above the lasing threshold. In Fig. 6 we present typical input-output curves for methods I (black) and IV (red) for cavities with D = 50 µm (a) and D = 20 µm (b). In Table 2 the averaged lasing thresholds and collection efficiencies are shown.

 

Fig. 6 Input-output curves for micro-lasers with diameters of D = 50 µm (a) and D = 20 µm (b) recorded in aqueous environment. In-plane (red color) collection efficiencies are enhanced compared to the collection from the top (black color). As method IV is radiation-based, it is even more advantageous for smaller sized resonators (b) with enhanced radiation losses.

Download Full Size | PDF

Table 2. Averaged lasing thresholds and collection efficiencies for PM597-doped micro-lasers in aqueous environment recorded method I and IV.

View Table | View all tables in this article

Let us concentrate on resonators with D = 50 µm first. The absolute signal intensities in aqueous environment are lowered compared to air environment owing to increased absorption losses. This leads to relatively small collection efficiencies for method I, e.g., for the case of missing alignment of a resonator’s impurity with respect to the spectrometer entrance slit we measure small collection efficiencies as low as ξ_min = 0.90 ∙ 10² Counts/nJ; even for resonators that are directly read-out at an impurity the measured collection efficiencies do not exceed ξ_max = 6.60 ∙ 10² Counts/nJ. Both values are already quite small. In sensing applications requiring pumping for long time scales, they are eventually even further reduced due to photo-bleaching of the dye molecules [14,48]. In the worst case the lasing emission cannot be recognized as distinct peaks anymore. A way to avoid this detrimental effect is to enhance the lasing collection efficiency which in turn allows using lower pump energies and thus reducing photo-bleaching.

As illustrated in Fig. 6, method IV is superior to I in terms of collection efficiency: For micro-lasers with D = 50 µm it is enhanced by one order of magnitude and independent of impurities. As illustrated in Fig. 6(b) and Table 2, the signal collection is further enhanced for size-reduced micro-lasers (D = 20 µm) where radiation from the cavities is even more pronounced. A difference in the collection efficiency of two orders of magnitude is found. The lasing threshold stays the same since the excitation geometry in both methods is identical. In-plane read-out (method IV) is thus the method of choice in water environment.

4.2 Differentiation of TE- and TM-polarized WGMs by polarization-dependent PL spectroscopy

In order to enable a more comprehensive mode characterization and differentiate between differently polarized WGMs we perform polarization-dependent PL spectroscopy. For that purpose a linear polarizer integrated into a motorized 360° rotation mount is included in the detection beam path. To prevent artefacts from polarization-dependent elements in the setup, e.g., the diffraction efficiency of grating in the spectrometer, a λ/2-wave plate is used to ensure unchanged polarization of the analyzed light.

For the read-out configuration from the top (method I) we find for both, air and water environment, peak intensities of the laser peaks that are independent of the orientation of the polarizer. This behavior is expected as for method I only scattered non polarized light is collected. A determination of the mode polarization is hereby not possible.

For method IV the total signal in the detector comprises fractions of both, scattered and radiation light. In air environment the fraction of scattered light is dominant making a clear differentiation complicated, however indications for differently polarized WGM lasing modes have been observed. By switching to aqueous surroundings, radiation is greatly enhanced due to the lowered refractive index contrast [14] enabling a clear polarization differentiation. A corresponding lasing spectrum in water environment is depicted in Fig. 7(a) showing fundamental modes that are separated by the expected FSR. Higher-order modes are suppressed due to increased cavity losses. Two combs can be identified. In the presented example, the combs are overlapping and the two different kinds of modes seem to appear as doublets. Tracking the peak intensities of the modes of such a doublet for different orientations of the polarizer results in Fig. 7(b). Clearly a 180°-periodic intensity dependence is observed for both mode types as expected from Malus´ law. One type of mode reaches its maximum when the polarizer is horizontally positioned with respect to the detection direction [θ = 90°and θ = 270°; marked blue in Fig. 7] while the other type of mode reaches its maximum when the polarizer is set vertically [θ = 0°and θ = 180°; marked red in Fig. 7].

 

Fig. 7 (a): PL spectrum of a micro-laser in aqueous environment. Fundamental modes separated by the FSR are visible. From polarization-dependent PL spectroscopy we find a 180°-periodic intensity dependence for adjacent modes (b). From the phase shift of 90° and the polarizer’s orientation we can identify TE and TM lasing peaks.

Download Full Size | PDF

The observed phase-shift of 90° between the two curves proves the existence of two orthogonal linear polarizations. Using the nomenclature of TE and TM modes, depending on whether the electric or magnetic field lies in the resonator plane [32], we assign the red (blue) marked modes to TE(TM)-like modes.

We conclude that method IV allows a clear distinction of TE and TM polarized laser modes in aqueous environment. In the following section we will demonstrate that this distinction is important in sensing applications of micro-lasers.

4.3 Influence of the mode polarization on the bulk refractive index sensitivity

One of the most important applications of micro-cavities is their usage as (bio)sensors [49]. Often the bulk refractive index sensitivity (BRIS) is used to determine the sensor performance. It is defined as the wavelength shift of a resonance Δλ for a given change of the refractive index Δn [50]. The BRIS is determined by the resonator geometry but is also found to depend on the selected mode for which the wavelength shift is measured. We will show here that it is also dependent on the mode polarization.

To determine the BRIS, a micro-laser has been immersed into solutions of different glycerol-water concentrations. The resulting lasing mode shift Δλ is plotted over Δn. The BRIS is inferred from the slope of a linear fit to the data. The detailed explanation of the procedure is found in earlier publications [14,36].

In order to obtain theoretical predictions for the polarization dependence of the BRIS, we conducted FEM simulations [14]: For the investigated resonators with a thickness of t = 1.2 µm and diameter D = 50 µm the difference of the BRIS between TE and TM modes is expected to be only 0.15 nm/RIU or 0.6% and thus typically smaller than the uncertainty of a typical BRIS measurement (≥ 0.3 nm/RIU). As a consequence the difference of TE and TM BRIS cannot be resolved. As shown in [14] decreasing the diameter is beneficial for the sensing performance on the one hand as the BRIS is increased; on the other hand it results in an increased difference between TE and TM BRIS, e.g., for D = 20 µm the simulated difference is 5.74 nm/RIU or 12.3% at 580 nm. An unambiguous and consistent interpretation of the sensor signal thus requires to consider the polarization of the modes that are evaluated to derive the mode shift $\Delta \lambda$.

To verify these results experimentally we perform BRIS measurements with micro-cavities of D = 20 µm, D = 25 µm and D = 50 µm in configuration IV. In advance to the injection of different glycerol-water solutions we determine the polarization of the lasing modes. The resulting data are presented in Fig. 8 for D = 20 µm together with the FEM simulations. The data for the other sized resonators are summarized in Table 3. For all sizes of resonators the absolute values of the measured BRIS is found to be a factor of ~0.83 smaller than the simulated one which seems to be a systematic deviation. But more importantly, the general trend of the BRIS curves is experimentally reproduced. In contrast to earlier experiments with polymeric micro-lasers based on method I [14,22], method IV allows a clear distinction of BRIS values for TE and TM lasing modes. This underlines the importance of controlling the mode polarization in sensing experiments.

 

Fig. 8 Simulated and experimentally determined BRIS of micro-resonators with D = 20 µm in method IV. Both simulated and experimental data have been fitted with straight lines. The uncertainties of the BRIS values correspond to the uncertainties of the slope of the linear regression for the plot “$\Delta \lambda$over $\Delta n$”. For the linear fits to the data the 95% Confidence Level is shown.

Download Full Size | PDF

Table 3. Measured and simulated BRIS for fundamental TE and TM modes of differently sized resonators. The values are determined at the wavelength where the linear wavelength regressions have the smallest uncertainty. Although the absolute values differ, simulation and measurements show the same trends for the dependencies on radius and polarization.

View Table | View all tables in this article

5. Summary and conclusion

In summary, in the first part we compared the performance of different excitation and collection photoluminescence schemes and characterized polymeric micro-lasers focusing on lasing threshold and collection efficiency. Using fiber excitation instead of free-space excitation reduces the lasing threshold by a factor of ~100 due to the deposition of the pump energy within the modal volume. Very small lasing thresholds on the order of ~5 pJ per pulse are measured. The read-out of lasing emission perpendicular to the disk plane depends critically on defects and imperfections scattering the emitted light towards the detector and results in rather small collection efficiencies. Such poor collection of WGM emission leads to the fact that under cw pumping conditions the expected WGM modulation of the spontaneous emission is buried within the non-modulated background. Using evanescently coupled tapered fibers as collection tool instead increases the collection efficiency by two orders of magnitude when the resonators are excited in free-space geometry and by additional two orders of magnitude when fiber excitation is applied. As a consequence, fiber-based collection enables the detection of the expected modal filtering in the fluorescence evoked by WGM resonances under cw excitation.

Less complex than fiber usage and highly advantageous concerning collection efficiency is read-out of in-plane radiated light. In the second part we verified that in-plane detection is especially beneficial in aqueous environment and for smaller-sized resonators where radiation losses are strongly enhanced. Additionally, this method enables the determination of the polarization of the WGMs since this property is sustained for in-plane radiated light. TE and TM polarization are assigned to the fundamental WGMs in water environment by performing polarization-dependent PL spectroscopy. We finally demonstrate that in-plane detection is moreover beneficial for sensing applications: In refractive index sensing the expected differences between the BRIS of TE and TM modes are experimentally reproduced. This allows consistent and unambiguous sensing as the polarization of the mode, from which the wavelength shift is derived, can be identified and monitored by the presented approach.

The presented fiber-based measurement configurations (method II and III) are beneficial for dry environments for a more comprehensive characterization of the active micro-cavities and the WGMs. In-plane read-out (method IV) is additionally well suited for wet setups, e.g. in sensor applications such as refractive index sensing.