Numerical modeling of SNSPD absorption utilizing optical conductivity with quantum corrections

M. Baránek¹, P. Neilinger^1,2, S. Kern¹, and M. Grajcar^1,2
1 Department of Experimental Physics, Comenius University, SK-842 48, Bratislava, Slovakia
² Institute of Physics, Slovak Academy of Sciences, Dúbravská cesta 9, SK-845 11, Bratislava, Slovakia

ABSTRACT. Superconducting nanowire single-photon detectors are widely used in various fields of physics and technology, due to their high efficiency and timing precision. Although, in principle, their detection mechanism offers broadband operation, their wavelength range has to be optimized by the optical cavity parameters for a specific task. We present a study of the optical absorption of a superconducting nanowire single photon detector (SNSPD) with an optical cavity. The optical properties of the niobium nitride films, measured by spectroscopic ellipsometry, were modelled using the Drude-Lorentz model with quantum corrections. The numerical simulations of the optical response of the detectors show that the wavelength range of the detector is not solely determined by its geometry, but the optical conductivity of the disordered thin metallic films contributes considerably. This contribution can be conveniently expressed by the ratio of imaginary and real parts of the optical conductivity. This knowledge can be utilized in detector design.

1 INTRODUCTION

Strengthening the security of digital communication channels by building Quantum Key Distribution (QKD) infrastructure is an ongoing effort worldwide[1, 2]. Most QKD systems are based on entangled photon pairs as quantum bits and utilize the existing, low-loss fiber optical infrastructure [3]. To implement an efficient algorithm, a single-photon detector with a high overall detection efficiency, low dark counts, and high speed is required[4]. The State-of-the-art detectors that fulfill these requirements are the superconducting nanowire single-photon detectors (SNSPD)[5]. They are superior in terms of detection efficiency and dark count rates to other types[6], such as the Avalanche Photodiodes (APD)[7], and do not require sub-kelvin temperatures, as the Transition Edge Sensors (TES)[8] do. Moreover, their superior properties make them beneficial in a broad range of applications in different areas of physics, for example, physical chemistry and spectroscopy[9], fluorescent luminescence[10], or fast space-to-ground communication[11] often requiring a lower wavelength range. The detector’s wavelength can be optimized by its design depending on the required spectral range. This is also important in the implementation of modern quantum networks[3, 12] as not just the detection efficiency of the detector, but its bandwidth is a relevant parameter.

Figure 1: Scheme of the SNSPD with coupled optical fiber. Nanowire has thickness $t$, width $w$, and spacing $s$. Optical $\lambda$λらむだ/4 resonator consists of a dielectric spacer with thickness $d_{1}$ and a gold reflector with $d_{2}$. The gap between optical fiber and nanowire is $g$. Right: Top-down view of the nanowire of diameter $2r$, with red line showing the position of the cross-sectional view.

The working principle of SNSPD is straightforward: a nanowire in the superconducting state is current-biased close to the edge of the normal metal transition. The absorption of a single photon results in the destruction of superconductivity at some place along the nanowire. Due to the current bias, the normal state region expands, which results in a voltage pulse across the nanowire. These nanowires are fabricated on 6-15 nm thick, highly disordered superconductors, such as Niobium Nitride[13], Titanium Niobium Nitride[14], Tungsten Silicide[15] and their width is in the range of 30-200 nm.

The overall system detection efficiency (SDE) of detectors is determined by the efficiency of the photon coupling to the active area of the detector $\eta_{\mathrm{coupling}}$ηいーた start_POSTSUBSCRIPT roman_coupling end_POSTSUBSCRIPT, the photon absorption efficiency of the superconductor $\eta_{\mathrm{abs}}$ηいーた start_POSTSUBSCRIPT roman_abs end_POSTSUBSCRIPT, and by the above-described efficiency of transformation of the absorbed photon to electric signal $\eta_{\mathrm{intrinsic}}$ηいーた start_POSTSUBSCRIPT roman_intrinsic end_POSTSUBSCRIPT and is given as[16]:

$$\mathrm{SDE}=\eta_{\mathrm{coupling}}\cdot\eta_{\mathrm{abs}}\cdot\eta_{\mathrm{intrinsic}},$$ηいーた start_POSTSUBSCRIPT roman_coupling end_POSTSUBSCRIPT ⋅ italic_ηいーた start_POSTSUBSCRIPT roman_abs end_POSTSUBSCRIPT ⋅ italic_ηいーた start_POSTSUBSCRIPT roman_intrinsic end_POSTSUBSCRIPT , (1)

The absorption of a nanowire (thin film) $\eta_{\mathrm{abs}}$ηいーた start_POSTSUBSCRIPT roman_abs end_POSTSUBSCRIPT on bare substrate is insufficient (below 30$\%$, Fig. 2). To increase the absorption at the desired wavelength ($\lambda=$λらむだ =1550 nm) the nanowires are commonly fabricated in an optical resonator, at the cost of limiting its bandwidth. A common choice is a $\lambda/4$λらむだ / 4 resonant cavity [17].

This resonator consists of a thin dielectric layer with thickness $d_{1}$, matching its optical length $\mathrm{<ruby><rb>\Lambda </rb><rt>らむだ</rt></ruby>}$Λらむだ\Lambdaroman_Λらむだ with $\lambda/4$λらむだ / 4 of the desired light’s wavelength. The nanowire is fabricated on one side of the dielectric and the other side is covered with a low-loss mirror metallic layer (as shown in Fig. 1). The resonator enhances the absorption of the photons in the nanowire close to 100$\%$ (Fig. 2) by creating an antinode of the electric field of TE mode in the superconductor. The detectable light is guided by an optical fiber to this structure.

To properly model the optical properties of this structure, and thus to maximize the absorption $\eta_{\mathrm{abs}}$ηいーた start_POSTSUBSCRIPT roman_abs end_POSTSUBSCRIPT in the nanowire, not only has the dielectric thickness $d_{1}$ to be optimized, but the precise optical properties of the disordered metal have to be considered. However, their optical properties in the metallic state are not trivial due to the presence of quantum corrections, which the standard Drude model fails to describe [18]. This means that, to model their optical absorption, the optical properties of the specific thin film have to be either measured in the infrared spectra, or the right optical model - or an extrapolation method - has to be chosen, if measurements in the required range are not accessible.

In this paper, we investigate the dependence of the optical absorption of NbN nanowires in $\lambda/4$λらむだ / 4 resonant cavity on film thickness and nanowire design. The optical conductivity of a set of NbN films was determined from spectroscopic ellipsometry in the visible range, and they were fitted by the modified Drude-Lorentz model, which takes the quantum corrections into account [19] and provides an excellent fit of these conductivities. The presence of quantum corrections in the IR range results in a strong wavelength and thickness dependence of the optical properties of NbN films. The obtained optical fits and thickness dependencies of its parameters are used to model the absorption spectra of nanowires. The presented approach can be applied to other disordered films, such as MoC and NbTiN. The optical properties of the latter[20] are almost identical to the properties of NbN.

Figure 2: Nanowire optical absorption $\eta_{\mathrm{abs}}$ηいーた start_POSTSUBSCRIPT roman_abs end_POSTSUBSCRIPT on bare sapphire substrate and on $\lambda/4$λらむだ / 4 resonator.

2 Optical properties of NbN films

Niobium nitride samples were deposited by means of Pulsed Laser Deposition[21] from a 99 % pure niobium target on top of single-side polished sapphire substrate in a nitrogen atmosphere with 1 % of hydrogen. The film thickness varied from 8 up to 22 nm. The optical conductivities of our films were determined from spectroscopic ellipsometry measured in wavelength range from 400 to 1000 nm at room temperature.

In Ref.[19] the quantum corrections to the optical conductivity of these NbN films were thoroughly studied. These corrections are arising from localization and interaction effects both having a square root energy dependence[22]. The optical conductivity $\tilde{\sigma}=\sigma_{1}+i\sigma_{2}$σしぐま end_ARG = italic_σしぐま start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT + italic_i italic_σしぐま start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT is described by the modified Drude-Lorentz model in the form:

$$\sigma_{1}(\omega)= \frac{\sigma_{D}}{1+(\frac{\Omega}{\Gamma})^{2}}\Big{(}1-\mathcal{Q}^{2}(1-\sqrt{\frac{\Omega}{\Gamma}})e^{-\frac{1}{2}(\frac{\Omega}{\Gamma})^{2}}\Big{)}+$$σしぐま start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT ( italic_ωおめが ) = ΩおめがΓがんま21superscript𝒬21ΩおめがΓがんまsuperscript𝑒12superscriptΩおめがΓがんま2\displaystyle\frac{\sigma_{D}}{1+(\frac{\Omega}{\Gamma})^{2}}\Big{(}1-\mathcal% {Q}^{2}(1-\sqrt{\frac{\Omega}{\Gamma}})e^{-\frac{1}{2}(\frac{\Omega}{\Gamma})^% {2}}\Big{)}+divide start_ARG italic_σしぐま start_POSTSUBSCRIPT italic_D end_POSTSUBSCRIPT end_ARG start_ARG 1 + ( divide start_ARG roman_Ωおめが end_ARG start_ARG roman_Γがんま end_ARG ) start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT end_ARG ( 1 - caligraphic_Q start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT ( 1 - square-root start_ARG divide start_ARG roman_Ωおめが end_ARG start_ARG roman_Γがんま end_ARG end_ARG ) italic_e start_POSTSUPERSCRIPT - divide start_ARG 1 end_ARG start_ARG 2 end_ARG ( divide start_ARG roman_Ωおめが end_ARG start_ARG roman_Γがんま end_ARG ) start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT end_POSTSUPERSCRIPT ) + (2)

${\displaystyle \frac{{\mathrm{<ruby><rb>\sigma </rb><rt>しぐま</rt></ruby>}}_L}{1+{\left(\frac{{\mathrm{<ruby><rb>\Omega </rb><rt>おめが</rt></ruby>}}_L^2-{\mathrm{<ruby><rb>\Omega </rb><rt>おめが</rt></ruby>}}^2}{\mathrm{<ruby><rb>\Omega </rb><rt>おめが</rt></ruby>}{\mathrm{<ruby><rb>\Gamma </rb><rt>がんま</rt></ruby>}}_L}\right)}^2}},$Ωおめが𝐿2superscriptΩおめが2ΩおめがsubscriptΓがんま𝐿2\displaystyle\frac{\sigma_{L}}{1+\left(\frac{\Omega_{L}^{2}-\Omega^{2}}{\Omega% \Gamma_{L}}\right)^{2}},divide start_ARG italic_σしぐま start_POSTSUBSCRIPT italic_L end_POSTSUBSCRIPT end_ARG start_ARG 1 + ( divide start_ARG roman_Ωおめが start_POSTSUBSCRIPT italic_L end_POSTSUBSCRIPT start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT - roman_Ωおめが start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT end_ARG start_ARG roman_Ωおめが roman_Γがんま start_POSTSUBSCRIPT italic_L end_POSTSUBSCRIPT end_ARG ) start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT end_ARG ,

$$\sigma_{2}(\omega)=\mathcal{H}[\sigma_{1}(\omega)]-(\varepsilon_{\infty}-1)\varepsilon_{0}\omega$$σしぐま start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT ( italic_ωおめが ) = caligraphic_H [ italic_σしぐま start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT ( italic_ωおめが ) ] - ( italic_εいぷしろん start_POSTSUBSCRIPT ∞ end_POSTSUBSCRIPT - 1 ) italic_εいぷしろん start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT italic_ωおめが (3)

respectively. Here, $\sigma_{D}$σしぐま start_POSTSUBSCRIPT italic_D end_POSTSUBSCRIPT is the Drude conductivity, $\mathrm{<ruby><rb>\Gamma </rb><rt>がんま</rt></ruby>}$Γがんま\Gammaroman_Γがんま is the electron relaxation rate and $\mathcal{Q}$ is the strength of the quantum corrections, also referred to as quantumness. The second term in eq. 2 describes the transition peak at ${\mathrm{<ruby><rb>\Omega </rb><rt>おめが</rt></ruby>}}_L$Ωおめが𝐿\Omega_{L}roman_Ωおめが start_POSTSUBSCRIPT italic_L end_POSTSUBSCRIPT, with strength $\sigma_{L}$σしぐま start_POSTSUBSCRIPT italic_L end_POSTSUBSCRIPT, and width ${\mathrm{<ruby><rb>\Gamma </rb><rt>がんま</rt></ruby>}}_L$Γがんま𝐿\Gamma_{L}roman_Γがんま start_POSTSUBSCRIPT italic_L end_POSTSUBSCRIPT. This formula describes the anomalous suppression in the IR range and below, which is characteristic of these films. The optical conductivity for films with different thicknesses is shown in Fig.3 The same result can be represented by $\sigma_{1}$σしぐま start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT vs. $\sigma_{2}$σしぐま start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT plots in Fig.4, where the wavelength dependence is color-coded. The solid spirals are the fits with eqs. 2,3. The fitted parameters are listed in Tab. 1.

As can be seen in Fig. 3a, the Drude conductivity $\sigma_{0}$σしぐま start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT varies weakly with thickness, but the relaxation rate $\mathrm{<ruby><rb>\Gamma </rb><rt>がんま</rt></ruby>}$Γがんま\Gammaroman_Γがんま and the quantumness $\mathcal{Q}$ are strongly thickness dependent. This reflects the commonly observed suppression of DC conductivity and, most importantly, the significant changes in the optical properties of films, especially in the IR range. By lowering the thickness, the real part of the IR range conductivity is suppressed and the imaginary part changes its sign. The dielectric function

$$\varepsilon_{r}(\omega)=\varepsilon_{1}+i\varepsilon_{2}=1+i\tilde{\sigma}(\omega)/(\epsilon_{0}\omega)$$εいぷしろん start_POSTSUBSCRIPT italic_r end_POSTSUBSCRIPT ( italic_ωおめが ) = italic_εいぷしろん start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT + italic_i italic_εいぷしろん start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT = 1 + italic_i over~ start_ARG italic_σしぐま end_ARG ( italic_ωおめが ) / ( italic_ϵ start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT italic_ωおめが ) (4)

for various film thicknesses is plotted in Fig.3b. The known thickness dependence of these parameters allows us to conveniently model the optical properties of SNSPDs. The complex refraction index can be expressed in terms of permittivity, as:

$$\tilde{n}=\sqrt{\varepsilon_{r}(\omega)},$$εいぷしろん start_POSTSUBSCRIPT italic_r end_POSTSUBSCRIPT ( italic_ωおめが ) end_ARG , (5)

$d[\textrm{nm}]$	$\sigma_{D}[MS/\textrm{m}]$σしぐま start_POSTSUBSCRIPT italic_D end_POSTSUBSCRIPT [ italic_M italic_S / m ]	$\mathrm{\hslash }\mathrm{<ruby><rb>\Gamma </rb><rt>がんま</rt></ruby>}[\text{eV}]$Γがんまdelimited-[]eV\hbar\Gamma[\textrm{eV}]roman_ℏ roman_Γがんま [ eV ]	$\mathcal{Q}$	$\sigma_{1}[MS/\textrm{m}]$σしぐま start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT [ italic_M italic_S / m ]	$\mathrm{\hslash }{\mathrm{<ruby><rb>\Gamma </rb><rt>がんま</rt></ruby>}}_1[\text{eV}]$Γがんま1delimited-[]eV\hbar\Gamma_{1}[\textrm{eV}]roman_ℏ roman_Γがんま start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT [ eV ]	$\mathrm{\hslash }{\mathrm{<ruby><rb>\Omega </rb><rt>おめが</rt></ruby>}}_1[\text{eV}]$Ωおめが1delimited-[]eV\hbar\Omega_{1}[\textrm{eV}]roman_ℏ roman_Ωおめが start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT [ eV ]
8	0.77	2.1	0.7	0.9	1.0	6.13
9	0.84	2.0	0.6	0.95	1.1	6.14
14	1.00	1.76	0.66	1.12	0.56	5.19
22	0.94	1.77	0.50	1.07	0.68	5.54

Table 1: Parameters of optical model eq.(2) obtained from the ellipsometric data fit.

Figure 3: Specific conductivity (top) and permittivity (bottom) of thin NbN films, consisting of real (solid line) and imaginary (dashed line) parts. Thick lines are the result of spectroscopic ellipsometry measurement, and thin lines result from the model described above. Dots at zero frequency are the room temperature DC conductivities.

Figure 4: Wavelength-dependent real and imaginary part of conductivity for films with various thicknesses. The arrows show the mismatch in the simple rescaling of the conductivities, easily visible by the change in the angle $\theta_{0}$θしーた start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT of arrows, which point at the conductivity at 1550 nm.

It is important to note, that not just the magnitude of the conductivity changes with the thickness, but the ratio of real to imaginary parts changes as well. This is easily visible by the change of the angle of the arrow pointing to the conductivity at $\lambda_{0}=$λらむだ start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT =1550 nm in Fig. 4. This angle is defined by the tangent of real and imaginary parts of the optical conductivity as $\tan{\theta_{0}}=\sigma_{2}(\lambda_{0})/\sigma_{1}(\lambda_{0})$θしーた start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT end_ARG ) = italic_σしぐま start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT ( italic_λらむだ start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT ) / italic_σしぐま start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT ( italic_λらむだ start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT ) and is a characteristic of the films for a given thickness - it does not depend on the fill factor of the nanowire, since scaling the conductivity scales both real and imaginary part.

3 Optical absorption simulations

We studied the effect of the film’s conductivity on the absorption spectra of SNSPD with a similar detector design to [14]. The numerical simulations were carried out in a commercial frequency-domain EM solver[23]. The air gap between the fiber and the surface of the detector leads to additional Fabry-Perót resonances superposing the absorption spectra of the $\lambda/4$λらむだ / 4 resonator. This could be expressed in the wavelength dependence of the coupling parameter $\eta_{\mathrm{coupling}}(\lambda)$ηいーた start_POSTSUBSCRIPT roman_coupling end_POSTSUBSCRIPT ( italic_λらむだ ). In our 2D simulations, the optical fiber output is modelled as a plane wave coupled to air and we assume infinite size of the detector in the plane perpendicular to the light propagation, which corresponds to the ideal coupling $\eta_{\mathrm{coupling}}=1$ηいーた start_POSTSUBSCRIPT roman_coupling end_POSTSUBSCRIPT = 1 and allows us to focus on the absorption in the nanowire $\eta_{\mathrm{abs}}$ηいーた start_POSTSUBSCRIPT roman_abs end_POSTSUBSCRIPT.

The optical absorption of the SNSPD can be described by the effective impedance model[24], which approximates the thin superconducting meander as a lumped element effective homogeneous medium, with effective sheet conductance $\tilde{G}=G_{1}+iG_{2}$:

$$\tilde{G}=\tilde{\sigma}\frac{w}{w+s}t\mathrm{\ [S/\square]},$$σしぐま end_ARG divide start_ARG italic_w end_ARG start_ARG italic_w + italic_s end_ARG italic_t [ roman_S / □ ] , (6)

where $\tilde{\sigma}(t)$σしぐま end_ARG ( italic_t ) is the optical conductivity (in units of S/m) of the NbN film with thickness $t$, and the ratio ${f={w}/{(w+s)}}$ is the fill factor. This approximation is valid for the width and the thickness of the nanowires $w,t\ll\lambda$λらむだ. As these parameters are usually $w\sim$ 30-200 nm, $t\sim$ 5-22 nm, and $s\sim$ 50-300 nm, this approximation holds well for $\lambda>1\mu\mathrm{m}$λらむだ > 1 italic_μみゅー roman_m. The validity of this approximation is demonstrated by numerical modeling of the absorption spectra of a detector with different nanowire widths and constant fill factor $f=0.5$, shown in Fig. 5. The absorption spectra are almost identical for $w\lesssim 400\mathrm{nm}$. The negligible differences in the absorption maximum ($<0.5\%$) and its wavelength ($<50$ nm) may originate in simulation mesh discretization. The $\approx$ 1% losses originate in the Au layer with $d_{2}=~{}100$ nm.

Figure 5: Simulated absorption spectra of the detector for different nanowire widths and constant fill factor $f=0.5$

In Fig. 6, we present the absorption spectra of a detector with 100 nm width NbN nanowire of different thicknesses $t=8-22$ nm on SiO₂ layer with thickness $d_{1}$= 250 nm. The absorption maximum corresponding to the optical length of an ideal $\lambda/4$λらむだ / 4 resonator is $4\mathrm{<ruby><rb>\Lambda </rb><rt>らむだ</rt></ruby>}=4n{d}_1\approx 1600$Λらむだ4𝑛subscript𝑑116004\Lambda=4nd_{1}\approx 16004 roman_Λらむだ = 4 italic_n italic_d start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT ≈ 1600 nm. For each thickness, the fill factor $f$ was tuned to maximize the absorption at $\lambda_{0}$λらむだ start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT= 1550 nm. The maximal absorptions at $\lambda_{0}$λらむだ start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT, and the wavelength of the absolute maximal absorption $\lambda$λらむだ are listed in Table  2. Note, that for this analysis, the optical properties of NbN films were extracted from ellipsometric measurements on films deposited on c-cut sapphire substrate. The optical properties of films deposited on different substrates, for example, SiO₂ or Si₃N₄, can differ. However, the presented results can be easily generalized for any combination of film and dielectric material, as we will show below.

Figure 6: Simulated absorption spectra of nanowires with different thicknesses for fill factor maximizing absorption at 1550 nm.

250 nm SiO2, $4\mathrm{<ruby><rb>\Lambda </rb><rt>らむだ</rt></ruby>}=1600$Λらむだ16004\Lambda=16004 roman_Λらむだ = 1600 nm
$d[\textrm{nm}]$	$\sigma_{2}/\sigma_{1}$σしぐま start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT / italic_σしぐま start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT @1550 nm	max($\eta_{abs}$ηいーた start_POSTSUBSCRIPT italic_a italic_b italic_s end_POSTSUBSCRIPT) $@1550~{}nm$ [%]	fill factor	$\lambda_{max}$λらむだ start_POSTSUBSCRIPT italic_m italic_a italic_x end_POSTSUBSCRIPT [nm]	max ($\eta_{abs}$ηいーた start_POSTSUBSCRIPT italic_a italic_b italic_s end_POSTSUBSCRIPT) [%]
8	0.025	99.1	0.6	1650	99.3
9	0.092	99.2	0.45	1550	99.2
14	0.242	96.8	0.3	1350	99.1
22	0.314	91.5	0.225	1250	97.9

Table 2: Simulated maximum absorption at 1550 nm and the optimized fill factor for 1550 nm, wavelength of maximum absorption, and maximum absorption value for different thickness

The absorption spectra for $t=$ 8, 9, 14, 22 nm are shown in Fig. 6. As it is visible, due to the thickness dependence of the optical conductivity (refraction index), the thickness of the film significantly shifts the position of the absorption maxima and changes the shape of the absorption spectra. The maximum of the absorption value max($\eta_{abs}$ηいーた start_POSTSUBSCRIPT italic_a italic_b italic_s end_POSTSUBSCRIPT), at its peak wavelength $\lambda_{max}$λらむだ start_POSTSUBSCRIPT italic_m italic_a italic_x end_POSTSUBSCRIPT, is also lowered for thicker films due to the increased reflection, but it is negligible compared to the drop of absorption at the desired wavelength $\lambda_{0}$λらむだ start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT = 1550 nm. To study how $G_{1}$ and $G_{2}$ separately affect the absorption spectra, we simulated the absorption of the nanowire with $w=100$ nm, $s=100$ nm ($f=0.5$), $t=10$ nm on $d_{1}=250$ nm SiO₂ dielectric and $d_{2}=100$ nm Au reflector.

The absorption spectra for varying $G_{1}$ and constant $G_{2}$ = 0 reveal that the real part of the effective sheet conductance $G_{1}$ ($\sigma_{1}$σしぐま start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT, respectively) mainly affects the value of the absorption maxima, but not their position $\lambda_{max}$λらむだ start_POSTSUBSCRIPT italic_m italic_a italic_x end_POSTSUBSCRIPT, shown in Fig.7a. On the other hand, variation of $G_{2}$($\sigma_{2}$σしぐま start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT, respectively) for constant $G_{1}$ results in the shift of absorption peak $\lambda_{max}$λらむだ start_POSTSUBSCRIPT italic_m italic_a italic_x end_POSTSUBSCRIPT, shown in Fig.7b. The simulated range of $\tan{\theta}$θしーた end_ARG ) is approx. 0 - 0.4, which is comparable to the values of our films (0 - 0.32). Thus, the resonant absorption peak of the detector is shifted by the imaginary part of the conductance towards lower values and thus, its wavelength $\lambda_{max}$λらむだ start_POSTSUBSCRIPT italic_m italic_a italic_x end_POSTSUBSCRIPT is not solely determined by the thickness of the dielectric layer, as one could simply assume. This effect, governed by $\tan{\theta}$θしーた end_ARG ), is universal and should be present in any thin film with similar values of $\tan{\theta}$θしーた end_ARG ).

Figure 7: Optical absorption in nanowire with fixed $G_{2}=0\ mS$ (left) and varying $G_{1}$, and with fixed real effective conductance $G_{1}=2.5\ mS$ (right).

The dependence of $1/\lambda_{max}$λらむだ start_POSTSUBSCRIPT italic_m italic_a italic_x end_POSTSUBSCRIPT on the conductivity of the NbN films, represented by $\tan{\theta}$θしーた end_ARG ), is linear for a broad range of $\tan{\theta}$θしーた end_ARG ), as is visible in Fig. 8, and can be approximated as:

$$\frac{1}{\lambda_{max}(\sigma_{2})}={\frac{1}{\lambda_{max}(\sigma_{2}=0)}+\frac{1}{\lambda_{c}}\tan\theta},$$λらむだ start_POSTSUBSCRIPT italic_m italic_a italic_x end_POSTSUBSCRIPT ( italic_σしぐま start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT ) end_ARG = divide start_ARG 1 end_ARG start_ARG italic_λらむだ start_POSTSUBSCRIPT italic_m italic_a italic_x end_POSTSUBSCRIPT ( italic_σしぐま start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT = 0 ) end_ARG + divide start_ARG 1 end_ARG start_ARG italic_λらむだ start_POSTSUBSCRIPT italic_c end_POSTSUBSCRIPT end_ARG roman_tan italic_θしーた , (7)

where $\lambda_{c}$λらむだ start_POSTSUBSCRIPT italic_c end_POSTSUBSCRIPT is a fitting parameter. The value $\lambda_{c}$λらむだ start_POSTSUBSCRIPT italic_c end_POSTSUBSCRIPT is governed by the optical properties of the dielectric layer and its thickness. For the studied geometry of the detector on SiO₂, $\lambda_{c}\sim 4.2$λらむだ start_POSTSUBSCRIPT italic_c end_POSTSUBSCRIPT ∼ 4.2 $\mu$μみゅーm. Here, we assumed $\tan\theta$θしーた to be constant and equal to $\tan\theta_{0}$θしーた start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT, which is a feasible approximation in the vicinity of $\lambda_{0}$λらむだ start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT, as is shown in Fig. 10.

Since $\tan{\theta}$θしーた end_ARG ) increases with decreasing wavelength, the positive feedback further lowers the resonance wavelength.

Figure 8: Dependance of wavelengths of the simulated absorption maxima on the conductivity ratio $\tan{\theta}=\sigma_{2}/\sigma_{1}=G_{2}/G_{1}$θしーた end_ARG ) = italic_σしぐま start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT / italic_σしぐま start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT = italic_G start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT / italic_G start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT, given by samples with different thicknesses. Left: On top of the 250nm SiO₂ substrate, Right: On top of the 200 nm Si₃N₄ membrane

This simple relation can be directly implemented in the detector design. The absorption maximum shift due to the imaginary part of the conductivity (contribution of the thin film), for a given dielectric layer, limits the possible range of film thicknesses.

However, using e.g. a commercially available 200 nm thick Si₃N₄ membrane, the resonance wavelength determined by the membrane is $4\mathrm{<ruby><rb>\Lambda </rb><rt>らむだ</rt></ruby>}\approx 1800$Λらむだ18004\Lambda\approx 18004 roman_Λらむだ ≈ 1800 nm and the corresponding absorption at $\lambda_{0}$λらむだ start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT is thus lower, as can be seen in Fig. 8. Using thicker NbN films, the resonance wavelength can be lowered and the absorption can be significantly increased, see Tab.3. For each film thickness, the fill factor was optimized to obtain the maximum absorption at $\lambda_{0}$λらむだ start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT. Thus, fine-tuning the absorption peak position up to 200 nm is possible, without significant loss of total optical absorption $\eta_{abs}$ηいーた start_POSTSUBSCRIPT italic_a italic_b italic_s end_POSTSUBSCRIPT.

200 nm Si3N4, $4\mathrm{<ruby><rb>\Lambda </rb><rt>らむだ</rt></ruby>}=1800$Λらむだ18004\Lambda=18004 roman_Λらむだ = 1800 nm
$d[\textrm{nm}]$	$\sigma_{2}/\sigma_{1}$σしぐま start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT / italic_σしぐま start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT @1550 nm	max($\eta_{abs}$ηいーた start_POSTSUBSCRIPT italic_a italic_b italic_s end_POSTSUBSCRIPT) $@1550~{}nm$ [%]	fill factor	$\lambda_{max}$λらむだ start_POSTSUBSCRIPT italic_m italic_a italic_x end_POSTSUBSCRIPT [nm]	max ($\eta_{abs}$ηいーた start_POSTSUBSCRIPT italic_a italic_b italic_s end_POSTSUBSCRIPT) [%]
8	0.025	93.3	0.65	1800	98.8
9	0.092	95.3	0.5	1750	98.7
14	0.242	98.6	0.3	1600	98.6
22	0.314	97.5	0.225	1450	98.1

Table 3: Table of sample parameters with simulated maximum absorption at 1550 nm, optimized fill factor for 1550 nm, the wavelength of maximum absorption, and maximum absorption value, respectively.

The fit parameter for the 200 nm Si₃N₄ membrane is $\lambda_{c}=6.3\mu$λらむだ start_POSTSUBSCRIPT italic_c end_POSTSUBSCRIPT = 6.3 italic_μみゅーm, estimated from the linear fit in Fig. 8.

3.1 Discussion

In similar studies, the properties of a single deposited film are often measured (for example by spectroscopic ellipsometry), and the determined specific conductivity is considered to be thickness-independent in the simulations. Consequently, the nanowire’s thickness and width are then tuned to maximize the optical absorption in the nanowire at the aimed wavelength $\lambda_{0}$λらむだ start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT. For example, in Ref. [25], or Ref. [26] the same refraction index was used for samples with thicknesses ranging from 4 to 10 nm, in case of Ref. [13] even from 1 to 16 nm. Rescaling the thickness of a specific film to a requested value results in the inaccuracy of the simulated absorption. This fact is acknowledged in [24], however, no further analysis is present. Rarely, the refractive index is also considered to be wavelength independent[25].

To illustrate the error arising from this approximation, we simulated the optical absorption in the above-described nanowire for the respective conductivity $\tilde{\sigma}(t)$σしぐま end_ARG ( italic_t ), and the corresponding optical indices, given by eq. 5. For example, using the conductivity of the film with $t_{meas}=$14 nm and simply rescaling it to $t=9$ nm for the simulation would result in a decrease of the simulated absorption maxima by approximately 200 nm, as can be seen in Fig.9.

Figure 9: Simulated optical absorption of 9 nm thick sample on top of the resonant cavity. Other curves show the change in optical absorption of the optical conductivity (refraction index) taken from a sample with different thicknesses.

If one assumes the optical conductivity (refraction index) of these films to be wavelength-independent, the error in the absorption spectra will accumulate with the distance from the reference wavelength, as is shown in Fig. 10.

Figure 10: Simulated absorption of a sample with wavelength dependent $n(\lambda)$λらむだ ), constant $n(\lambda=1000,1300,1550\mathrm{nm})$λらむだ = 1000 , 1300 , 1550 roman_n roman_m ) and $n$ corresponding to constant $\sigma(\lambda=1550\mathrm{nm})$σしぐま ( italic_λらむだ = 1550 roman_n roman_m ), respectively.

4 CONCLUSION

We modelled the absorption spectra of NbN nanowires in a $\lambda/4$λらむだ / 4 resonator for different film thicknesses and showed that the optical conductivity of thin films significantly affects the absorption spectra. The optical conductivity of disordered metallic films is strongly affected by the presence of quantum corrections, resulting in nontrivial wavelength and thickness dependence of the optical conductivity. The real part of the conductivity determines the amplitude of absorption of the nanowire in the $\lambda/4$λらむだ / 4 optical resonator, whereas the imaginary part contributes to the wavelength shift of the maxima. The wavelength shift of the absorption maxima can be expressed in terms of the ratio of imaginary and real parts of the optical conductivity, which is a characteristic of a film with a particular thickness. This knowledge can be utilized in detector design and optimization of its optical properties. Thus, to properly simulate and optimize the absorption of the detectors at a required wavelength, it is necessary to work with the precise optical conductivity of the particular layers. These can be obtained by direct measurements of the samples in the required range, or by utilizing a model that describes the optical conductivity of the disordered samples well, as presented in Ref. [19].

5 ACKNOWLEDGMENT

This work was supported by the Slovak Research and Development Agency under the contract APVV-20-0425, and by the project skQCI (101091548), founded by the European Union (DIGITAL) and the Recovery and Resilience Plan of the Slovak Republic.