1. Introduction

Coherent light and surface plasmon generation on the nanoscale has recently attracted significant attention due to numerous potential applications in the fields ranging from biosensing to optical interconnects. The physical dimensions of a laser are defined by its cavity which generates optical feedback based on constructive interference of propagating waves. Since the interference requires phase accumulation of at least 2π, the resonator dimensions cannot be much smaller than half a wavelength in each direction and, in fact, are limited by the classical diffraction. The smallest lasers approaching this limit are VCSELs [1].

Conceptually different approaches are required for creation of sub-diffraction coherent light generating systems. One of the promising solutions is based on incorporation of nanometric structures made of noble metals supporting plasmonic excitations [2]. Such structures can confine light far-beyond the natural limit of diffraction by coupling it into coherent oscillations of free conduction band electrons near metal surfaces [3]. The plasmonic effect can confine optical mode in one or two dimensions [4], while the facet reflections in third dimension can be used to generate sufficient feedback for lasing. Several concepts of surface plasmon polariton lasers have been recently demonstrated [5, 6]. The third dimension in laser cavities may be further reduced by employing the idea of negative phase accumulation on the reflection from negative permittivity objects, proposed in [7] and recently demonstrated in coaxial plasmonic cavities [8].

Ideologically new approach to laser resonators was proposed by D. J. Bergman and M. I. Stockman [9], who suggested replacing the interference phenomenon by the near-field feedback, introducing the concept of spaser as a coherent source of plasmonic excitations. In a case of efficient coupling of localized surface plasmons (LSPs) in far-field photons, the spaser can be considered as a nanolaser. This system was successfully demonstrated, employing a core metal nanoparticle and surrounding active dyes [10].

One of the most basic properties of a laser is its linewidth which represents the measure of temporal coherence of the output beam. The distinctive property is the narrowing of the output spectrum when the laser passes from the spontaneous emission regime through a threshold to the predominant stimulated emission and lasing. The Schawlow-Townes theory predicts the linewidth to be proportional to the square of the resonator bandwidth divided by the output power [11]. However, almost all realistic laser systems, in particular those based on semiconductor structures, cannot reach this limit due to additional noise in a resonator caused by inhomogeneities, mechanical instability and other factors. The important parameter, especially in semiconductor lasers, is a linewidth enhancement (LWE) factor (α) resulting from fluctuations of the refractive index of the laser cavity [12]. The actual line width is α² + 1 times larger than the predicted by the Schawlow-Townes theory.

The basic theoretical description of nanolasers and spasers is based on the notion of quasistatic localised surface plasmon (LSP) resonances. Unfortunately, the rigorous problem of their emission linewidth cannot be considered with the same formalism as for conventional lasers which involves time-dependent treatment of the laser cavity and absent in quasistatic approximation. The quasistatic description of LSP resonances defining the cavity of a spaser or nanolaser does not include the radiation losses, at the same time, material losses are predominant for particle diameter smaller than 20 nm [13]; moreover, the quasistatic approach does not explicitly include the time derivatives.

Here we present a theoretical framework to treat the linewidth and, in particular, its enhancement in nanolasers and spasers (Fig. 1 ). In order to investigate the linewidth properties, we have introduced the time dependence in the quasistatic treatment via the dispersion of spectral parameter which defines the LSP frequency. The LWE was estimated for bulk GaAs material as an active medium of a spaser to be of order of 3-6, strongly depending on carrier density in the active layer.

 

Fig. 1 Basic spaser/nanolaser configuration.

Download Full Size | PDF

2. Theoretical description

The basic rate equations describing spaser action have been derived in [14]. The description of a plasmonic mode in the quasistatic regime used for this purpose is given by:

Equation (1) does not include time evolution of the field in an explicit way, and this is its fundamental difference from the wave equation, which is used in the description of conventional lasers. In order to introduce the times dependence in the nanolaser action description, we will introduce the time derivatives using the dispersion properties of the material permittivities present in Eq. (1). Note that both real and imaginary parts of permettivities of metal and active dielectric have to be considered dispersive in order to satisfy Kramers-Kronig relations.

We consider a single cavity mode to be of the following form in a time domain representation:

Using the conventional relations between mode amplitude$\beta$, intensity$I$and phase$\phi$, namely, $I=\beta {\beta }^{*}$and$\phi =\ln \left(\beta /{\beta }^{*}\right)/2i$, the resulting set of differential equations, describing a spaser operation, can be obtained:

These equations, derived for a spaser, have similar structure to those, describing conventional lasers [12], but with the coefficients specific for any fluctuations in the spaser cavity.

Equations (6) may be integrated, and the phase may be shown to have a Gaussian probability distribution, which determines the spaser linewidth via the phase-dependent autocorrelation function $g^{(1)}\left(\tau \right)=\frac{〈{\beta }^{*}(t)\beta (t+\tau )e^{-i\omega \tau }〉}{〈{\beta }^{*}(t)\beta (t)〉}$. From the autocorrelation$g^{(1)}(\tau )$, the coherence time can be determined by${\tau }_{coh}={\displaystyle \underset{-\infty }{\overset{\infty }{\int }}{\left|g^{(1)}(\tau )\right|}^{2}d\tau }$. The spaser linewidth can then be found as$\Delta \nu =1/\pi {\tau }_{coh}$. The total linewidth taking into account fluctuations in the cavity can be related to the Schawlow-Townes quantity$\Delta \tilde{\nu }$ as:

In the case of a spaser with a semiconductor active medium, we can obtain the LWE from Eq. (6) following considerations in Ref [12]. as:

3. Results

To estimate the influence of the LWE on spaser operation, we have investigated the simplest model where bulk GaAs was taken to be the active medium and plasmonic resonator has been considered to be silver sphere ([19] for material parameters. In particular, ε_silver ≈-40 + i0.5). The bandgap of the active layer dictates the resonant frequency of the spaser and the LSP resonance of the metal nanostructure should be tuned to this frequency. Fortunately, plasmonic resonances may be engineered to cover entire visible and infrared frequencies using approaches such as particle-particle coupling [20], particle elongation [21], concavity tuning [22] or ultimate evolutionary methods [23]. The gain characteristics were estimated using free-carrier theory [24]; many-body effects, such as bandgap shrinkage, electron-electron scattering, and lineshape broadening may be included by more advanced modeling [25]. Here, we used the former approach, taking the material parameters of GaAs [24] (In particular, E_gap = 1.424 eV, n_background ≈3.65) and neglecting the minor free carries contribution to the index change [18].

Figure 2(a) represents the spaser central wavelength as the function of injected carrier density above the threshold. It may be seen that the increase of the carrier injection shifts the central wavelength in the short wavelength range where the gain has a maximum value, similar to the experimental observations in [8]. It should be noted that conventional lasers also exhibit short-wavelength shift of the emission peak which, however, corresponds to the decrease of the refractive index of the cavity material and, thus, the shorter wavelength in order to conserve the 2π phase accumulation per round trip [26].

 

Fig. 2 (a) Spaser central wavelength as the function of injected carrier density. (b) The linewidth enhancement factor and (c) The spaser linewidth: $\Delta \nu ~({\alpha }^2+1)/n$ dependence on the injected carrier density above the spasing threshold.

Download Full Size | PDF

Figure 2(b) shows the LWE as the function of injected carrier density taking into account the shift of the spasing wavelength as discussed above. The figure shows the characteristic decrease of LWE with the carrier density at low injection currents followed by a small increase and saturation with the increase of carrier density. Generally, this behavior is determined by several competing effects, the simplified intuitive explanation may involve the reduction of to the relative carrier’s noise with the increase of the overall carriers’ density, similarly to the situation in bulk semiconductors [25]. The further increase to the carrier’s densities in the case of spaser leads to refractive index changes in the surrounding of the resonator resulting in the shift of the emission line wavelength and, subsequently, leading to a more complicated LWE behavior. The data were recast to show the spaser linewidth dependence ($\Delta \nu ~({\alpha }^2+1)/n$ (Fig. 2(c)). The carrier density ($n$) dependent gain profile of semiconductors and dispersive nature of LSP resonances result in the dispersion of the LWE, and, as the result, the deviation of the spaser linewidth from the classical ‘one over power’ proportionality. The spaser linewidth exhibits monotonic narrowing with the injection current, similarly to the recently reported results [8]. Moreover, the shape of the curve on Fig. 2(c) has similar behavior to the linewidth, reported at [5].

The distinctive difference between plasmonic-based lasers and conventional lasers is that the quality factors of the resonators used in the former case, metal particles, are generally of order of 10-100 resulting in the very broadband spectral response (Fig. 3 ). Typical quality factors of the resonators of high-quality semiconductor laser is over 1000 with and much higher for other lasers [24]. The mode selection in classical laser system results from the interplay of sharp-resonant cavity modes competing for the gain, while the spaser operation frequency is determined by the material gain properties. Therefore, we assumed that the LSP resonance is relatively flat in the vicinity of the gain line (Fig. 3). The gain bandwidth in bulk semiconductors could exceed 100nm and reach values of ~1000 cm⁻¹. Complex plasmonic structures may, however, support number of the so-called dark modes obeying non-exponential decay laws [27] with much shaper resonances and higher quality factors, governed mainly by material loss; these will be of main importance for spasers, while conventional bright LSP modes are important for plasmonic-based nanolasers. The difference in LWE between dark and bright modes (if the far field radiation is neglected) will enter via the dispersion of the spectral parameter and will depend on specific spaser realization.

 

Fig. 3 Schematics of the gain and cavity spectral responses in the case of a conventional laser and a plasmonic nanolaser: (red solid lines) gain profile for different carrier densities, (green dashed-dotted line) LSP resonance, (blue dotted lines) laser (Fabry-Perot) cavity resonances.

Download Full Size | PDF

4. Conclusion

We have developed the theory to introduce the time dependence in the description of spaser action through the dispersion of the spectral parameter, describing the localized surface plasmon spectrum. This has allowed investigations of the spaser and plasmonic nanolaser linewidth properties. The developed theory predicts significant linewidth enhancement with an order of magnitude broadening of the output spectrum in the case of the active medium based on bulk GaAs, with the deviation of the spaser linewidth from the classical ‘one over spasing power’ proportionality. The developed theory could be of potential interest to rapidly developing nano-plasmonic devises with incorporated gain [28].