1. Introduction

Whispering gallery mode resonators (WGMRs) are a unique platform for fundamental studies in physics and have also allowed a large variety of technological developments. WGMR geometries range from microspheres to microbubbles, going through microtoroids and microdisks. All of them have played an important role in lasing, sensing, non-linear optics, optomechanics and ultrasound applications [1–4]. The common properties of these WGMRs are high quality factors (Q) and small mode volumes, resulting in a significant enhancement of light-matter interactions and a high cavity power build-up. Consequently, optothermal effects are ubiquitous in all kinds of WGMRs and have been extensively studied in the last decade [5–7]. Most of these studies focused on characterizing and compensating the optothermal effects of the cavity as a whole [5,6,8,9] since the thermal drift of the optical resonances could be highly detrimental in sensing applications, such as label-free sensing [10].

On the other hand, however, this sensitivity can be exploited to implement the WGMRs as highly sensitive thermometers, allowing to detect tiny temperature changes in their local environment through the shift of an optical resonance [11]. Photothermal microscopy of non-luminescent nano-objects on integrated ultra-high Q toroidal microresonators was first demonstrated on carbon nanotube [12] and then in conductive polymers [13]. By exploiting an ultra-sensitive detection scheme, optical excitation and absorption of a single gold nanorod could be also detected [14]. However, all these experiments were performed in an almost dry environment and an extremely small tuning range, which might not be ideal in key applications like biochemical or environmental sensing [15,16].

Microbubble resonators (MBRs) made from glass capillaries [17,18] allow a physical separation between the optical layer, where light is injected into the resonator through the external surface of the capillary, and the inner region, where a non-solid sample is confined. The intrinsic microfluidics through the capillaries and the ultra-high Q are unique features of the MBR system that cannot be implemented through other types of WGMRs, such as integrated microdisks or on-chip microtoroids. It has been demonstrated that MBR can be sensitive ultrasound transducers [1], versatile flow cytometers [19] and even exceptional nanoparticle sensors [20] when working in the quasi droplet regime, where the high-order optical modes are almost within the liquid core of the resonator. Recently, an MBR system has been used in combination with a double-modulation detection scheme for measuring the chemical etching dynamics of gold nanorods through polarization sensitive photothermal absorption microscopy [21].

Here we propose the MBR as an absorption spectrometer to probe non-luminescent targets, i.e., gold nanorods (GNRs), in an optically dense real-scenario matrix, such as milk at room temperature. In our system an MBR is filled with the matrix containing the nanoparticles and a wide-range laser source is filtered to illuminate the nanoparticles with a series of wavelengths. Upon resonant illumination, the particles release heat into the matrix and produce an overall temperature increase in the system. This, in turn, modifies the refractive index of the MBR walls because of the thermo-optical effect and, in the end, produces a shift of the MBR resonances. This shift, ultimately, represents an indirect measurement of the nanoparticles absorption coefficient in the case of a non-saturated sample, since all the transduction processes previous mentioned are linear. Indeed, the high quality factor of the MBR is a key element for the measurement, since it allows a clear detection of small temperature changes and therefore an high-sensitivity towards particles absorption.

Due to the thermal nature of the detection, the presented approach is insensitive towards light scattering from both the liquid hosting the nanoparticles or the scattering spectrum of the nanoparticles, and therefore grants scattering-free spectroscopy. This aspect is particularly promising for the characterization of highly scattering and highly opaque samples, since in these cases the scattering spectrum could hide the absorption spectrum in a typical light extinction measurement. This spectroscopic approach can yield to a better material characterization and/or unambiguous molecular identification, paving the way for a wide range of applications in the fields of environment, well-being, food safety and security.

2. Experiment description and results

The experimental setup implementing the detection approach previously outlined is sketched in Fig. 1 and is basically analogous to the one we have designed for the detection of photoacoustics waves through the microbubble [1,19]. An MBR fabricated with an arc-discharge technique [18] was connected through polymer tubings (Tygon R-3607, Saint-Gobain, La Défense, Courbevoie, France) to a peristaltic pump (Minipuls 3, Gilson, Middleton, WI, USA) and a reservoir to fill the MBR with a suspension containing PEGylated gold nanorods (GNRs) synthesised with the seed-mediated approach [22,23]. The MBR external diameter was 400 µm and the silica walls thickness was estimated to be 3.8 µm, following the methods discussed in [24].

 

Fig. 1. Sketch of the experimental setup. SC: supercontinuum, AO: acousto-optical, MR: 45$^\circ$ mirror, AL: aspherical lens, PD: photodiode, Pol.Con.: polarisation controller.

Download Full Size | PDF

The MBR was positioned in the focus point of an aspherical lens (C110TME-B, Thorlabs, Newton, NJ, USA; focal length 6.24 mm) to have a supercontinuum source (SuperK COMPACT, NKT Photonics, Birkerød, Denmark; spectral range 400 nm-2400 nm) excite the GNRs through an inverted microscope configuration. In particular, the supercontinuum emission was filtered through an acousto-optic modulator (SuperK SELECT, NKT Photonics, Birkerød, Denmark; spectral range 680 nm-1100 nm) to illuminate the GNRs at selected wavelengths with a repetition rate set to 20 kHz, which basically simulate the emission of a low-power CW source (approx. 20 µW). Laser emission was enabled through a square pulse provided by the built-in generator of the oscilloscope (MDO4000C, Tektronix, Beaverton, OR, USA) and a feedback signal was used as reference to trigger the oscilloscope acquisition.

Finally, an home-made tapered fiber [25] was coupled to the MBR to excite its Whispering Gallery mode (WGM) resonances through a low-noise CW infrared laser (Koheras ADJUSTICK, NKT Photonics, Birkerød, Denmark; spectral range 1550 nm - 1551 nm). In particular, a waveform generator (Keysight 33220A, Agilent Technologies, Santa Clara, CA, USA) allowed to perform a 1 GHz (8 pm) scan around the laser wavelength, a polarization controller allowed to optimise the coupling of the resonances and the system transmission was recorded through an InGaAs photodiode (PDA400, Thorlabs Newton, NJ, USA). To avoid perturbations caused by air currents, the area surrounding the MBR was shielded with transparent PMMA walls.

The first operations of the experiment were filling the MBR with the suspension containing the GNRs and then searching for a narrow high-contrast WGM resonance through a scan of the probe laser. Then, by filtering the supercontinuum emission through the acousto-optic modulator, the GNRs were illuminated with a series of excitation wavelengths $\lambda _{\mathrm {exci}}$ using a series of exposure times $\Delta t$. For each of these combinations, the new position of the WGM resonance was recorded and the shift $\Delta \nu$ with respect to the unperturbed position was computed. Since the microbubble WGMs excited in the experiment are mostly localized within the silica walls and since silica has a positive thermo-optic coefficient, the WGM shifts recorded during the experiment are always red-shifts.

Figure 2 shows a selection of these shifts versus exposure times for a series of excitation wavelengths (dots), while the inset highlights the unperturbed and the shifted position of the WGM resonance for an exemplifying wavelength-exposure combination. The time evolution of the optical shifts is compatible with a rising exponential trend (continuous curves) and shows that the MBR system reaches thermal equilibrium after a few seconds of illumination. In turn, this means that, for the experimental configuration here considered, an illumination above 6 s would not yield any improvement in the measurement, since the additional energy would be dissipated into the environment surrounding the MBR. To estimate the stability of the system, a series of acquisitions was performed before illuminating the GNRs. In particular, the resonance was acquired every 5 seconds for 25 times, finding a 5 MHz residual oscillation in its position. This value represents the stability threshold of the system and the error to be applied to each datapoint in Fig. 2 (error bars are not shown in Fig. 2 to visualize properly the datapoints for short expositions). Finally, the right vertical axis in the main plot shows the temperature shift $\Delta T$ of the system, as measured through the silica thermo-optical effect [26]. As previously mentioned, this shift is a minute fraction of the room temperature and peaks at around 120 mK for the low-power source used in this experiment (approx. 20 µW).

 

Fig. 2. Optical shifts $\Delta \nu$ of the resonance versus several exposure times $\Delta t$ for a selection of excitation wavelengths $\lambda _{\mathrm {exci}}$ (dots). The continuous curves show that a rising exponential profile gives a valid description of the evolution towards thermal equilibrium and the axis on the right shows the conversion of the optical shifts into temperature shifts $\Delta T$. The inset in the top-right corner shows the WGM resonance in its umperturbed (blue curve) and the shifted (red curve) position for the combination $\left ( \Delta t= 2 \mbox { s},\, \lambda _{\mathrm {exci}} = 750 \mbox {~nm} \right )$.

Download Full Size | PDF

To account for inhomogeneities, the raw shifts $\Delta \nu$ were normalized to the optical power of the excitation lines ($P_\mathrm {opt}$), with the resulting $\Delta \nu /P_\mathrm {opt}$ versus wavelength profiles representing the absorption spectrum of the nanoparticles. Figure 3 shows these profiles for two iterations of the experiment, where GNRs were dispersed in different environments. In particular, in the first iteration [Fig. 3(a)] the GNRs were dispersed in water to simulate a transparent environment, while in the second iteration [Fig. 3(b)] the GNRs were dispersed in an aqueous suspension of milk powder at a concentration of 1% (w/w) to simulate an opaque environment. In both iterations the GNRs concentration was 3 mM Au.

 

Fig. 3. Power-normalized optical shifts $\Delta \nu /P_\mathrm {opt}$ for an MBR filled with an aqueous suspension of GNRs at a concentration of 3 mM Au (a, dots), and with an aqueous suspension of GNRs and milk powder at a concentration of 3 mM Au and 1% (w/w) milk powder (b, dots). The panels also show the profiles obtained with an MBR filled with just the host liquid [crosses, water for (a) and water with 1% (w/w) milk powder for (b)]. In the insets the profiles are normalized to the value of the 740 nm read-out, respectively.

Download Full Size | PDF

In both cases, the profiles reconstruct the plasmonic band of the GNRs at 750 nm and are close in value when comparing analogous wavelength-exposure combinations. Consequently, these results prove that the MBR absorption measurements are insensitive towards the scattering of the host environment and demonstrate the MBR as an effective scattering-free spectrometer. To further demonstrate this point, Fig. 3 also shows the profiles obtained for the MBR filled with only the host liquids (crosses). In both cases, the contribution of the host is null for each excitation wavelength (except for little hops due to residual system instabilities), proving a very low background level for the absorption measurements. This feature is expected from the presented configuration, since water shows low absorption in the 700 nm-900 nm range. In a more general scenario, however, a subtraction between the GNRs suspension dataset and the host dataset may be necessary to highlight the absorption spectrum of the former with respect to the latter. Finally, the exposure time does not influence the shape of the absorption profiles, since these profiles collapse to a common shape once they are a normalized to a reference value. In particular, the insets of Fig. 3 show the result of this additional normalisation, where the profiles are normalized to their 740 nm values.

As a final validation test, the absorption spectrum obtained with the MBR system was compared with the one obtained with a spectrophotometer (V-560, Jasco, Tokyo, Japan). In particular, the MBR was used to profile the absorption of a suspension having a reduced particle concentration (GNRs 1.5 mM Au, same 1% w/w milk powder), while the spectrophotometer measured a 20-fold dilution of this suspension, both in terms of GNRs and milk powder. In analogy with Figs. 3 and Fig. 4(a) shows the power-normalized optical shifts $\Delta \nu /P_\mathrm {opt}$ (dots) and the background associated with the milk-only suspension (crosses), while Fig. 4(b) shows the profiles after the normalisation at the 740 nm values. Even for this reduced concentration, the profiles collapse to a common shape and a very good agreement is found with the spectrophotometer profile (cyan curve) both in terms of position and width of the GNRs plasmonic band. As shown in the inset of Fig. 4(b), the reference profile (cyan curve) was deduced by subtracting the milk background (black curve) from the raw dataset (blue curve), since the extinction produced by the Rayleigh scattering is substantial. By comparison, this subtraction is not needed for the MBR system because of its scattering-free detection principle, which makes the host contribution null [cfr. crosses in Figs. 3 and 4(a)].

 

Fig. 4. (a) power-normalized optical shifts $\Delta \nu /P_\mathrm {opt}$ for an aqueous suspension of GNRs and milk powder at a concentration of 1.5 mM Au and 1% (w/w) milk powder (dots). The panel also shows the shifts for an aqueous suspension with only the milk powder (crosses). (b) $\Delta \nu /P_\mathrm {opt}$ profiles after the normalisation to their 740 nm values, together with a reference profile from a spectrophotometer (cyan curve). The top-right inset shows the raw spectrophotometer profiles (black and blue curves) used to compute the reference one (cyan curve).

Download Full Size | PDF

Finally, Fig. 4 also allows to estimate the system limit-of-detection (LOD) for the presented configuration. In particular, Fig. 4(a) shows that the host liquid produces a baseline shift close to 0.5 MHz, which can be assumed as the background level of the system. In the same panel, the 1.5 mM Au GNRs suspension produces a peak shift of 6.5 MHz for the maximum exposition, giving therefore a signal-to-background ratio of 13. Due to the linearity of absorbance, the concentration yielding a background equivalent signal is around 150 µM Au GNRs, which can be taken as a qualitative estimate of the LOD. In turn, 150 µM Au GNRs corresponds to a particle number density around $3 \times 10^{11}$ ml$^{-1}$ [1]. It is important to highlight that this LOD can be lowered by optimising the experimental setup, with the most straightforward approach being an increase in the optical power illuminating the sample. Indeed, it also possible to implement more refined approaches based on stabilization techniques and phase sensitive detection [21], promising to increase the sensitivity level (and therefore lowering the LOD) by several order of magnitude. As a final note on Fig. 4, it is also important to mention that a proper calibration of the system, in combination with a simple thermal model [27], would also would allow to assess the absolute value of the nanoparticles absorption coefficient $\mu$.

3. Conclusions

In conclusion, a new method based on MBRs for scattering-free absorption spectroscopy of nanoparticles in strongly scattering medium has been proposed. With the proposed approach, the MBR contains the nanoparticle suspension under study and the absorption spectrum is detected through the temperature increase produced by the particle absorption upon irradiation. This approach has been validated by measuring the spectrum of an aqueous suspension of GNRs and, in particular, in a suspension stained with milk powder to increase its opaqueness. In prospective, the scattering-free nature of the measurement is a promising feature for the detection and the characterisation of nanoparticles or molecules in highly opaque liquids. In addition, the MBR system also shows interesting technical features such as the extreme compactness, the intrinsic microfluidics and the decoupling between the read-out and the excitation wavelengths. Indeed all the aforementioned features make the MBR system promising for applications in the fields of environment, well-being, food safety and security.