1. Introduction

In the last decade plasmonics grew into the powerful branch of science, allowing to manipulate light on the subwavelength scale and get impressive results in various applications including biosensors and molecular spectroscopy [1,2], optoelectronic devices [3–5] such as solar cells, semiconductor lasers, light-emitting diodes, and photodetectors. Recently, the development of quantum dot (QD) based infrared (IR) photodetectors (QDIPs) has been substantial [6, 7]. A limitation in conventional QDIPs is that the absorbance of mid-IR radiation is weak, which results from the low density of states coupled to the dots and from the limited QD absorption thickness. Several groups have reported multiple photocurrent (PC) enhancement of mid-IR InAs/(In)GaAs [8–13] and Ge/Si [14,15] QDIPs using surface plasmon wave (SPW) coupling structures and demonstrated that the plasmonic enhancement scheme represents one more efficient route to achieve the ultrahigh responsivity in compact photodetectors [16]. The excitation of SPW modes offers an effective surface light trapping, enhancement of local field intensities, and thus interaction with the optically thin device active region.

The key component in plasmonically enhanced QDIPs is a plasmonic metasurface which is employed to convert the incident electromagnetic radiation into SPWs and to improve the PC generation. Metallic films perforated with two-dimensional subwavelength hole arrays (2DHAs) are often used as the plasmonic couplers [8–15]. Alternately, their localized counterparts, two-dimensional metal disk arrays (2DDAs), can also collect and confine light in a small volume and are able to enhance the interaction between the light and matter [17–19]. The numerical calculations have shown that SPW-like features are also presented in arrays of metallic nanoparticles [20]. However, in contrast to the 2DHA, the high damping of these plasma-like waves is expected [20]. In this paper, a comparison between ordered gold 2DHA and 2DDA plasmonic structures integrated on top of the same Ge/Si QDIP platform is analyzed in terms of detector responsivity, detectivity, and quantum efficiency. The results show that the holey films have advantages over the disk arrays in QDIP performance.

2. Methods

Figure 1(a) shows schematically the structure of the detectors discussed in this paper. The Ge/Si QD samples are grown using a Riber SIVA21 molecular beam epitaxy (MBE) system. A 0.5 µm boron-doped (p+) Si contact layer (p = 2 × 10¹⁸ cm⁻³) is first grown on a Si (100) wafer [Fig. 1(a)]. The active region of QDIPs is composed of ten stacks of Ge quantum dots separated by 40-nm Si barriers and is sandwiched in between the 200-nm-thick undoped buffer and 120-nm-thick cap Si layers. The p-type remote doping of the dots is achieved with a boron δ-doping layer inserted 5 nm above each dot layer. The areal doping density is 6 × 10¹¹ cm⁻². Finally, a boron doped 100-nm-thick p ⁺-Si top contact layer (1 × 10¹⁹ cm⁻³) is grown. After the MBE growth, the wafers are processed into 1.4 mm diameter×circular mesa-shaped QDIPs with top and bottom gold electrodes [Fig. 1(b)]. On top of the QDIPs, we fabricate metallic plasmonic structures by the deposition of a 50-nm-thick Au film and formation of a periodic square lattice of circular disks or holes using the optical lithography, e-beam metal deposition and lift-off techniques [Figs. 1(c) and 1(d)]. Both the 2DDA and 2DHA have the square lattice symmetry with lattice constant a = 1.2 µm and hole/disk diameter of 0.7 µm. A reference QDIP without the surface plasmonic structure is also fabricated for comparison of device performance. The plasmonic enhanced and the bare QDIPs are taken from a single die of the same wafer right next to each other. In this work, all measurements are performed at a temperature of 78 K. The incident IR light illuminates detectors from their substrate side.

 

Fig. 1 (a) Layer sequence of the 10-period Ge/Si QDIP coupled with the top metallic 2D hole (2DHA) or disk (2DDA) arrays. (b) Optical image of a Ge/Si photodetector integrated with 2DHA or 2DDA plasmonic structure (top view). A 1D periodicity along the vertical axis is an artifact of the image. Zoom-in scanning electron microscopy images of the square lattice of circular gold disks (c) and holes (d) in the Au film. The lattice periodicity is 1.2 µm, the hole/disk diameter is 0.7 µm. (e) Schematic image of the fragment of the valence band profile of Ge/Si heterostructures along the growth axis z showing the optical absorption η₀ involved and the carrier escape mechanism p_e, mentioned in the text.

Download Full Size | PDF

The normal incidence photoresponse spectra were obtained using a Bruker Vertex 70 Fourier transform infrared spectrometer with a spectral resolution of 10 cm⁻¹ along with a SR570 low noise current preamplifier. The devices were mounted in a cold finger inside a Specac cryostat with ZnSe windows.

3. Results and discussion

Generation of SPWs is allowed when their momentum matches the momentum of the incident photon and the reciprocal lattice vectors characterizing the periodic modulation of the metasurface. For a square array of apertures at normal incidence, the wavelength of the lowest order SPW resonance is given in a first approximation by the grating-coupling equation

 

Fig. 2 (a) Photocurrent spectral response of the Ge/Si QDIPs with the gold 2DDA and 2DHA plasmonic structures compared to the bare QDIP. The PC enhancement at ~4.2 µm is due to excitation of the resonant fundamental surface plasmon mode. (b) Photocurrent enhancement spectra.

Download Full Size | PDF

The data evaluation procedure was the folowing. First, the absolute responsivity spectra were measured and calibrated with a deuterated L-alanine doped triglycine sulfate (DLaTGS). The dark current was tested as a function of bias voltage between U_b = −1.2 V and U_b = +1.2 V using a Keithley 6430 SubFemtoamp Remote SourceMeter. For dark current measurements, the samples were surrounded with a cold shield. Then, the noise characteristics were measured with an SR770 fast Fourier transform analyzer and the white noise region of the spectra (usually, f > 100 Hz) was used to determine the gain and detectivity. The sample noise was obtained by subtracting the preamplifier-limited noise level from the experimental data. The specific detectivity is given by $D^{*}=R\sqrt{A\cdot \Delta f}/i_n$, where R is the peak responsivity, A is the device area, i_n is the noise current, and Δ f is the bandwidth. The photoconductive (optical) gain g and noise i_n relationship of a QDIP can be expressed as [22]$g=i_n^2/(4e{I}_d\Delta f)$, where e is the charge of an electron and I_d is the dark current. For the evaluation of the gain, we have subtracted the thermal (Johnson) noise from the measured noise in order to have the pure generation-recombination contribution. The Johnson noise was calculated as $i_J=\sqrt{4k_{\mathrm{B}}T\Delta f/\rho }$, where k_B is the Boltzmann’s constant, T is the temperature, and ρ is the differential resistance, which is extracted from the dark current measurements. Finally, the total net quantum efficiency η was found using the relation: $R=(e{\lambda }_{\text{peak}}/hc)\eta (1-ℛ)g$, where R is the responsivity, c is the speed of light, λ_peak is the peak photon wavelength, $ℛ$ is the measured reflectivity of the Si substrate.

Figure 3 compares performance parameters of the plasmonic and bare devices. All QDIPs demonstrate photovoltaic behavior caused by a built-in electric field provided by charge distribution between Ge QDs and delta-doping planes in the Si barriers [23]. The resulting asymmetry of the valence band profile leads to a preferrable direction of motion of photogenerated holes even in the absence of a bias applied to the sample, when the detector noise is minimal and limited to thermal noise. Measurements demonstrate that the gain (not shown here) is essentially the same for samples with and without gold metasurfaces. The QDIP enhanced by 2DHA plasmonic structure shows higher peak responsivity, detectivity, and quantum efficiency over the entire bias range. The responsivities of the 2DHA QDIP at reverse (−1.2 V) and forward (+1.2 V) U_b correspond to 0.42 and 0.12 A/W. It appears that the highest detector detectivity is achieved in the photovoltaic mode (U_b = 0), where D ^∗ = 4.5 × 10¹² cm·Hz^1/2/W at a wavelength of 4.2 µm and a field of angle of 53°. For this bias, the detectivity ratios of the 2DHA QDIP to the 2DDA and bare devices are about 2 and 3, respectively. The measured D ^∗ is much larger than that observed previously in n -type InAs/GaAs mid-IR QDIPs [8–10] due to the higher hole effective mass in the Si valence band and hence the lower dark current.

 

Fig. 3 (a) Peak responsivity (λ_peak = 4.2 µm), (b) detectivity, and (c) quantum efficiency of the Ge/Si QDIPs with the gold 2DDA and 2DHA plasmonic structures compared with the reference QDIP at different biases U_b.

Download Full Size | PDF

Let us discuss the quantum efficiency as a function of bias voltage U_b. The net quantum efficiency η is the result of two different physical mechanisms. One is the optical absorption quantum efficiency η₀. The other component is the probability p_e that a photoexcited hole will escape out of the dot region and contribute to the PC, rather than being recaptured by the original dot. An applied negative external bias compensates the internal electric field, thus leading to a drastic drop of hole escape probability and to formation of a dip in η(U_b) curves around −0.2 V as observed in Fig. 3(c). At large U_b, the photoexcited holes can easily overcome the built-in potential barrier and escape probability starts to increase and then saturates reaching the maximum values of p_e ≃ 1. In this case, the saturation values of the total quantum efficiency would be equal to η₀. In all devices, the quantum efficiency is much higher under negative bias than it is under positive bias and reaches to η₀ ≈ 2% at U_b = −1.2 V in plasmonic QDIP coupled with the 2DHA grating [Fig. 3(c))]. This value is a factor of 3 and 6 larger than η measured in 2DDA and bare QDIPs, respectively. An asymmetry of η with respect to zero bias has been previously observed in remotely doped Ge/SI QDIPs with various dot and doping densities and attriuted to the bias-dependent mechanisms of hole escape out of the dots [23].

Finally, we would like to address the issue of why the 2DHA coupled with the Ge/Si QDIP gives a more efficient IR detection as compared with the array of gold disks. Despite the fact that 3D confinement of the QD structures enables absorption of light with any polarization by modifying the polarization selection rules, the bound-to-continuum transitions of holes in Ge/Si QDs have larger oscillator strength for the electromagnetic radiation polarized along the growth axis z [23, 24]. The SPWs propogating along a planar dielectric-metal interface are classified into a tranverse magnetic mode as the magnetic field of SPWs is parallel to the plane of the interface [25]. The electric field of SPWs has two components, one is directed normally to the surface (E_z component), the other (in-plane) component is aligned with the SPW propogation direction. Numerical simulations of the near-field component distribution have demonstrated that the E_z electric field component dominates in the 2DHA plasmonic structure [14,26], while the in-plane component of the near-field vector becomes the major component in the 2DDA grating [17]. Therefore, the appearance of the large E_z field in the surface plasmon waves generated by the 2DHA plasmonic device would enable better improvement of the Ge/Si QDIP performance.

4. Conclusion

In this paper, the performance of Ge/Si mid-IR QDIPs integrated with gold 2DHA and 2DDA structures are compared in terms of peak responsivity, detectivity, and quantum efficiency. The metallic metasurfaces are employed as plasmonic couplers to convert the incident electromagnetic radiation into surface plasmonic waves. Both devices show the improvement of the detection characteristics as compared to a bare Ge/Si QDIP without the gold gratings. The results demonstrate that regular arrays of subwavelength holes perforated in a gold film implemented on the top of QDIP will be more efficient for detector enhancement than arrays of gold disks. In a 2DHA QDIP, the quantum efficency of about 2% and photovoltaic peak detectivity of 4.5 × 10¹² cm·Hz^1/2/W are determined at a wavelength of 4.2 µm at 78 K. The unequal enhancement ratios of 2DHA and 2DDA structures are attributed to the quantum selection rules in the intraband transitions and the dominant near-field vector components induced by different gold metasurfaces. Additional functionality can be achieved by adjusting the 2DHA periodicity, diameter and the shapes of holes. The information acquired from the study is valuable for feasible device applications.