Plasmon resonance and perfect light absorption in subwavelength trench arrays etched in gallium-doped zinc oxide film

Near-perfect light absorption in subwavelength trench arrays etched in highly conductive gallium-doped zinc oxide films was experimentally observed in the mid infrared regime. At wavelengths corresponding to the resonant excitation of surface plasmons, up to 99% of impinging light is efficiently trapped and absorbed in the periodic trenches. Scattering cross sectional calculations reveal that each individual trench acts like a vertical split ring resonator with a broad plasmon resonance spectrum. The coupling of these individual plasmon resonators in the grating structure leads to enhanced photon absorption and significant resonant spectral linewidth narrowing. Ellipsometry measurements taken before and after device fabrication result in different permittivity values for the doped zinc oxide material, indicating that localized annealing occurred during the plasma etching process due to surface heating. Simulations, which incorporate a 50 nm annealed region at the zinc oxide surface, are in a good agreement with the experimental results.

2][3][4] Most of the previous studies made use of metals, such as silver and gold, as the plasmonic materials.More recently, efforts have been made to explore alternative plasmonic materials that potentially can overcome some of the pitfalls of traditional metal based plasmonic materials such as high ohmic loss, especially at optical and near infrared frequencies. 5In the mid infrared spectral regime, highly doped semiconductors and transparent conducting oxides are great potential candidates.][8] Previously, strong light absorption due to resonant excitation of surface plasmon polaritons has been investigated in metallic 1D subwavelength grating grove structures. 9In this work, we show that 1D subwavelength grating trenches etched in a doped semiconductor material, gallium doped zinc oxide, can near-perfectly absorb light in the mid-wave infrared spectral range.The 1D subwavelength grating etched in highly conductive gallium doped zinc oxide film can be viewed as an array of coupled split-ring resonators.This kind of structure, when made of noble metals, has been shown to trap and absorb photons at optical frequencies. 10abrication of subwavelength 1D gratings in gallium doped zinc oxide films started with the deposition of a 1300 nm thick gallium doped zinc oxide layer onto a high resistivity silicon wafer using pulsed laser deposition with a KrF excimer laser.The zinc oxide (ZnO) film, nominally Zn 0.974 Ga 0.026 O, was deposited in pure argon from a ZnO target with 3 wt.% Ga 2 O 3 .This method has been demonstrated recently for the deposition of highly conductive ZnO. 11,12The 1D grating structures were patterned using standard UV photolithography and inductively coupled plasma (ICP) etching.A negative photolithography patterning process was implemented for depositing a nickel (Ni) hard mask via electron beam evaporation followed by metal lift-off.The thickness of the Ni was chosen such that target etching depth is achieved before the Ni film is also completely etched away.For our process, the etching rate ratio of ZnO to Ni using BCl 3 based ICP etching is roughly 5:1.Following fabrication, the samples were analyzed using energy dispersive spectroscopy (EDS) to confirm the complete removal of the Ni mask.A scanning electron microscope (SEM) image of an etched 1D grating is shown in Fig. 1.
Two 1 = 4 wafers were cleaved from the same gallium doped zinc oxide substrate, each having a different Ni hard mask thickness, in order to produce intended grating etch depths, h, of 400 nm and 700 nm.On each 1 = 4 wafer, a 3 Â 3 array of devices were fabricated containing all combinations of three periods, P (5, 6, and 7 lm), and three grating trench widths, w (1, 1.5, and 2 lm).Due to imperfections in the etching process, the actual grating trench widths varied from their intended value.Additionally, the grating sidewall slopes deviated by 20 -30 from vertical creating a wider width at the top of the trench, w t , than at the bottom of the trench, w b .Table I lists the actual geometric parameters of the fabricated devices used in these experiments, as measured using SEM.
Reflectivity spectra from fabricated devices were obtained using a Fourier transform infrared spectrometer (Bruker Vertex 80V) with a coupled Hyperion microscope and a mercury cadmium telluride (MCT) photodetector, the microscope objective being removed in order to allow for normal incidence excitation.In order to determine the absolute value of the device reflectivity, all measured reflectivity spectra were normalized to a gold reference, taking into account the 3% absorption loss of gold at the wavelengths of interest.Fig. 2 shows the reflectivity spectra as a function of grating period, etch depth, and grating width.All spectra show a resonance were the reflectivity drops considerably and approaches zero, as well as a roll off in the reflectivity for wavelengths shorter than the grating period.The resonant feature is attributed to the excitation of surface plasmon polaritons which are bound to the top edges of the grating structure and, therefore, are trapped at the top of the grating trenches.Such light trapping effects have been observed in other systems such as metals at optical and near infrared wavelengths; 13 here, we show light trapping in the mid infrared regime, using the alternative plasmonic material of doped zinc oxide.For wavelengths greater than the period of the grating, the zero order diffraction is the only one which is reflected.However, for wavelengths less than the period, the 61 diffraction orders are also excited.Our experimental setup only measures the zero order reflection which explains the roll off in the reflectivity at shorter wavelengths.This is verified via simulation as shown later in this paper.
All other parameters held equal, increase in the depth of the trenches increases the peak absorption wavelength.This is due to an interference effect for plasmons travelling up and down the side wall of the grating trenches. 10Fig. 2(a) shows that for a grating with a period of P ¼ 5 lm, trench linewidth of w b ¼ 1.8 lm, and etch depth of 520 nm, the absorption exceeds 97% while for a similar grating with a deeper etch depth of 640 nm, the reflectivity is increased up to about 10% and the full width at half maximum of the resonance linewidth is nearly doubled.The increase of the resonance linewidth in deeper trenches is due to the increase of light trapping and absorption.In Fig. 2(b), it can be seen that, with trench width and etch depth held equal, change in the period of the grating results in a resonance wavelength shift, as expected from grating coupling theory of plasmon modes. 14Finally, changing only the trench width while maintaining the period and etch depth produces a small shift of the resonance wavelength; the spectra blue-shift as the width is increased, as shown in Fig. 2(c).From Fig. 2, it can be seen that the absorption peak wavelength is very sensitive to the change in the depth of the trenches and less sensitive to the change in the trench width.Of all devices, A2-P6L1 has the best performance, with a peak absorption of 99%.
Using a commercial FDTD software code from Lumerical, Inc., numerical simulations were performed for each device and the results were compared with the experimental results.The grating parameters for the simulations were extracted from SEM images.Using infrared ellipsometry (J. A. Woollam IR-VASE), the complex permittivity values of the deposited gallium doped zinc oxide layer were determined and were used in the simulations.After initial simulations failed to match the experimental spectra, an additional ellipsometry measurement was taken using a flat region of a post-fabricated device.It was found that the permittivity values of the zinc oxide layer were considerably less metallic after device fabrication, as shown in Fig. 3(a).Note that the real part of the permittivity does remain negative after fabrication and, therefore, is still capable of hosting surface plasmons. 6,157][18] The simulations were then modified so that within 50 nm of the surface the zinc oxide layer was modelled using the post-fabrication permittivity values while the rest of the zinc oxide layer was modelled using the pre-fabrication values.As seen in Fig. 3(b), after incorporating the local annealing effect, the simulated zeroth order reflectivity matches up very well with the experimentally measured reflectivity.Also shown are the simulated reflectivity spectra for the 61 diffraction orders and the  total reflection (sum of all reflected diffraction orders).It is shown that the roll off at wavelengths less than the grating period is due to the excitation of higher diffraction orders propagating in free space.The small additional roll off seen in the total reflection simulation is due to losses in the zinc oxide; the bulk plasma wavelength of zinc oxide is 1.4 lm.
To gain a better understanding of the coupling of the individual plasmon resonators, 1D scattering cross sectional simulations were carried out for device A2P6L1, which was the black reflectivity curve in Fig. 2(c).Shown in Fig. 4 are the cross sections for arrays with 1 individual trench up to 13 coupled trenches.At the fundamental plasmon resonance wavelength of roughly 6.4 lm, a scattering peak is seen, corresponding to the peak absorption or reflectivity dip in the experimental data.It is the destructive interference between the out-of-phase scattered light and the light reflected from the air/gallium doped zinc oxide interface which results in the near perfect absorption.The strength and linewidth of the scattered light were found to be dependent on the number of trenches in the array.The scattering cross sectional spectra reveal that each individual trench is essentially an independent plasmonic resonator with a very broad resonance spectrum; i.e., each trench can be viewed as an individual plasmonic split ring resonator.The resonance of each individual resonator is so broad and weak that it is barely discernible in the cross sectional spectra.As the number of coupled trenches increases, the resonance becomes stronger and the linewidth becomes narrower, indicative of a coherently coupled plasmonic system. 19The linewidth narrowing eventually saturates at 13 coupled trenches, as seen in the inset of Fig. 4, which shows the plasmon resonance quality factor versus number of coupled resonators.Note that an additional 15 coupled resonator quality-factor was used in the inset but not shown in the spectral plot for clarity.The maximum quality factor of $14 from the simulated scattering cross sections is on par with the experimentally measured quality factor value of 18.These simulations imply that, in the view of the trench array as an array of coupled split ring resonators, it is the coupling of these plasmonic resonators that leads to the near-perfect absorption observed in our experiment.This can also be thought of as the coupling of localized plasmon resonances in individual subwavelength trenches.
In order to verify that light absorption is indeed occuring when the surface plasmon mode is resonantly excited, as opposed to, say, excitation of Fabry-Perot resonances or waveguide modes, electric and magnetic field profiles from the simulation results were analyzed.Fig. 5 shows a side view of the magnitude of the electric and magnetic field for device A2-P6L1, taken at the fundemental plasmon resonant wavelength of 6.4 lm.The electric field is highly concentrated at the edges of the gallium doped zinc oxide grating trenches, consistent with light trapping.This field concentration provides an enhancement factor of roughly 13.Additionally, the magnetic field is seen to be enhanced at the surface of the grating and inside the trenches due to the electrical current of the plasmon resonance, i.e, surface plasmon current.Fundementally, from Maxwell's equations, there are only two ways to generate a magnetic field: electrical current via the movement of charges or displacement current via the change in the electric field.Here, the location of the magnetic field enhancement is not in the same location as that of the electric field enhancement; therefore, the magnetic field enhancement must be due to the surface plasmon (electric) current.
In summary, near-perfect light absorption in subwavelength periodic trenches etched in highly conductive gallium doped zinc oxide film has been demonstrated, with absorption values reaching up to 99%.The absorption resonance for the 1D grating structures was optimized for the midinfrared regime of 5-8 lm but, in principle, the wavelength regime could be extended to longer wavelengths.Scattering cross sectional spectra reveal that while each individual trench has a broad plasmonic resonance, the coupling of these individual trenches leads to enhanced light absorption and resonance linewidth narrowing.Changes in the trench depth were seen to have a more profound effect on the resonance than changes in trench width.During the dry etch fabrication process, local heating resulted in an annealing effect extending down 50 nm from the material surface.Simulations, which incorporated electric permittivity change due to the local annealing effect, reproduced the experimental results very well.

FIG. 1 .
FIG. 1.(a) Top view and (b) side cross sectional profile SEM images of a 1D grating etched in a gallium doped zinc oxide film.

TABLE I .
Geometric parameters of fabricated trench arrays extracted from SEM images.