Investigation of Al Schottky junction on n-type CdS film deposited on polymer substrate

Abstract

A systematic study has been made on the behavior of Al/n-CdS thin film junction on flexible polymer substrate (polyethylene terephthalate, PET) grown using thermal evaporation method. Temperature dependence of I−V measurements for this junction has been done which closely follow the equations of Schottky barrier junction dominated by thermionic emission mechanism. Intrinsic and contact properties such as barrier height, ideality factor and series resistance have been calculated from I−V characteristics. The barrier height of Al/n-CdS junction is found to increase with increase in temperature whereas ideality factor and series resistance decrease with increase in temperature.

1 1 Introduction

Microwave photonics have developed rapidly in recent years [1,2]. Many functions have been demonstrated, including ultra-wideband (UWB) signal generation [3], reconfigurable high-resolution radio frequency (RF) filtering [4], RF phase shifts [5], RF frequency up-conversion [6], and optical phased array beamforming [7]. Microwave photonic systems based on bulky optical components have suffered from a large volume, high-power consumption, high cost, and vulnerability to environmental disturbances. Therefore, it is highly desirable to integrate a microwave photonics system into a single chip to make it more compact, less expensive, and less power consuming.

Wideband communication dates back to the beginning of the 20th century in which spark-gap transmitters were used in World War I. However, in the 1920s, it was quickly replaced by superheterodyne radio, which can send continuous signals. Another wireless wideband technology, radar, developed rapidly, inspiring interest in signals with an ultra-wide bandwidth.

UWB signals have inherent characteristics, such as immunity to multipath fading, a wide bandwidth, and low-power spectral density [8]. One of its applications is in short-range high-throughput wireless communication (IEEE 802.15.3a) for wireless transmission of massive multimedia data without delay. It can also be used in low-speed and low-power transmission (IEEE 802.15.4a) for Internet of Things (IoT) applications, such as precise indoor positioning. Unlike WiFi- or Bluetooth-based distance estimation depending on the intensity of the signal, the UWB signal has an extremely narrow pulse width, which is similar to a radar pulse, making it possible to predict the position with an accuracy of 10 cm using the signal propagation time [9]. The Federal Communication Commission (FCC) provided UWB signal regulation in 2002 [10], and Apple Inc. has applied UWB technology in its products, such as iPod and iWatch. Many other companies, including NXP, Samsung, Bosch, Sony, and others, have formed an organization called the FiRa consortium to promote UWB application for perfect user experiences for circumstances, such as hands-free access control, location-based services, and device-to-device (peer-to-peer) services [11].

Traditional UWB pulse generation via electrical circuits needs electrical-to-optical conversion to distribute the signal over an optical fiber. Generating UWB signals directly in the optical domain with integrated photonics has many merits, including its light weight, small size, large tunability, and immunity to electromagnetic interference [3]. Several optical methods have been reported to produce UWB pulses, the most common of which is to implement phase-to-intensity modulation via dispersive devices [12] or frequency discriminators [13], such as a bandpass filter [14] or an fiber-Bragg-grating (FBG) [3,15]. Some others have obtained UWB pulses of different orders from the derivative of Gaussian pulses [16], and factors related to power efficiency were analyzed [17]. Furthermore, many studies have focused on coherent [13,16,18–20] and incoherent [14] summation of UWB signals of low orders to generate UWB pulses of higher orders. However, these approaches all require complicated bulky systems to generate stable signals. Linear summation of modified doublet pulses can be utilized for high-power efficiency UWB generation [21]. Other approaches to generate UWB signals have also been proposed by using nonlinear optical loop mirrors [22] or two-photon absorption in a silicon waveguide [23]. However, these approaches all require complicated bulky systems to generate stable signals. As for the integrated method, Wang and Yao proposed a simple method to generate UWB doublet signals using an electro-optic intensity modulator (EOM) [24]. Many have reported integrated schemes to generate UWB pulses that utilize cross-phase modulation in semiconductor optical amplifiers (SOAs) [25–27], which involve a complex system and costly device fabrication. The same shortcomings also exist for electroabsorption modulator (EAM)-based methods [28,29]. For simplicity, UWB signals can be generated using a single micro-ring resonator (MRR) [30,31]. However, the essential principle is to transform phase modulation to intensity modulation (PM-IM), and the integrated MRR only plays the role of IM conversion rather than complete key function. There are also some other special methods [32,33], such as splitting the light-wave and then combining them after certain processing to produce UWB signals; however, the waveform generated is limited [32]. Other schemes [12,33] need two lasers, which increase the system volume and cost. The effect of free-carrier dispersion (FCD) and free-carrier absorption (FCA) in silicon on generated UWB signals have also been studied [34]. Monocycle pulses have been produced on a silicon photonic chip [30]. These methods are either unstable or impractical owing to the bulky system or are monotonous because only limited waveforms are generated.

The obvious feature of UWB signals is its asymmetric line shape in the time domain. It is thus straightforward to use the asymmetric Fano resonance to generate UWB signals. Fano resonance is a ubiquitous physics phenomenon in nature, which was first observed by Beutler as spectral atomic lines that exhibit sharp asymmetric profiles in absorption. Later, Miroshnichenko et al. suggested the first theoretical explanation for this effect and suggested a formula that predicts the shape of spectral lines [35]. Since then, many researchers have studied it, and Fano resonances in various systems have been discovered and analyzed. As for photonic nanostructures, devices, such as dual-bus waveguide coupled MRR [36], the bent waveguide-based Fabry–Perot resonator [37], the waveguide micro-ring Fano resonator [38], the silicon Bragg reflector [39], MRR with a feedback coupled waveguide [40], add-drop MRR [41,42], a nanobeam cavity [43,44], coupled whispering-gallery-mode resonators [44,45], and the plasmonic resonator [46], have been proposed to produce Fano resonance. These have many applications in switching [43], sensing [39], photonic thermometers [38], instantaneous microwave frequency measurement [47], and others owing to the sharp asymmetric resonance line shape.

In this paper, we present a simple method to generate UWB pulses of different orders by utilizing a reconfigurable micro-ring resonator-coupled Mach–Zehnder interferometer (RC-MZI). The spectrum of the RC-MZI exhibits different line shapes when the phase difference between two arms of the MZI varies. Modulation is performed by the MRR when the device works at the Fano resonance with an asymmetric resonance line shape. UWB monocycle and doublet pulses are generated by using square and Gaussian electrical pulses, respectively. When the amplitude of the driving signal is further enlarged, UWB pulses of higher orders can also be generated.

2 2 Device structure and Fano resonance

2.1 2.1 Device structure

Figure 1(a) shows a schematic structure of the RC-MZI. The coupling between the MRR and the MZI (parent MZI) is enabled by a small child MZI coupler with two micro-heaters integrated for coupling tuning (indicated by the dashed box). A PN junction is integrated in the racetrack MRR for high-speed modulation. The other arm of the parent MZI is integrated with a thermo–optic phase shifter and a PIN diode-based variable optical attenuator (VOA) to adjust the phase and amplitude of the light traveling in this path, respectively. The device is highly reconfigurable, and the resonance spectrum can be flexibly tailored.

The phase shifter is enabled by a silicon resistive micro-heater that is made of a N⁺⁺-doped silicon slab, as shown in the inset of Fig. 1(a). After the current injection, the temperature of the adjacent silicon waveguide increases, leading to an increase in the effective refractive index and thus the phase of the optical beam. The phase difference between the two arms of the parent MZI determines the resonance line shape at a certain wavelength [41,48]. To obtain a high modulation efficiency, the PN junction has an L-shape cross section to maximize the overlap between the optical mode and the depletion region of the PN junction [49].

Figure 1(b) shows a microscope image of the device. The device was fabricated using CMOS-compatible processes. Direct current (DC) and RF signals were applied to the device to set the operation point and perform the modulation, respectively. The PN junction, thermo–optic phase shifter, and VOA are 500, 200, and 300 μm in length, respectively.

Fig.1 (a) Schematic structure of the micro-ring resonator-coupled Mach–Zehnder interferometer. Insets: structures of the phase shifter and the PN junction. (b) Microscope image of the fabricated device

Full size|PPT slide

2.2 2.2 Modeling

We used the transfer matrix method to model the device. The optical field transmission through the child MZI coupler can be described as

(1)$\left[\begin{array}{l}{b}_{\mathrm{1}}\hfill \\ b_{\mathrm{2}}\hfill \end{array}\right]\hspace{1em}=\hspace{1em}{\mathit{M}}_{\mathrm{DC}}·{\mathit{M}}_{\mathrm{MZI}}·{\mathit{M}}_{\mathrm{DC}}·\left[\begin{array}{l}{a}_{\mathrm{1}}\hfill \\ a_{\mathrm{2}}\hfill \end{array}\right]\hspace{0.17em}=\hspace{1em}\left[\begin{array}{ll}t\hfill & \mathrm{i}\kappa \hfill \\ \mathrm{i}\kappa \hfill & t\hfill \end{array}\right]\left[\begin{array}{ll}{\mathrm{e}}^{\mathrm{i}{\phi }_{\mathrm{1}}}\hfill & \mathrm{0}\hfill \\ \mathrm{0}\hfill & {\mathrm{e}}^{\mathrm{i}{\phi }_{\mathrm{2}}}\hfill \end{array}\right]\left[\begin{array}{ll}t\hfill & \mathrm{i}\kappa \hfill \\ \mathrm{i}\kappa \hfill & t\hfill \end{array}\right]·\left[\begin{array}{l}{a}_{\mathrm{1}}\hfill \\ a_{\mathrm{2}}\hfill \end{array}\right] \hspace{0.17em}\hspace{0.17em}=\hspace{1em}\left[\begin{array}{ll}{t}_{\mathrm{11}}\hfill & k_{\mathrm{12}}\hfill \\ k_{\mathrm{21}}\hfill & t_{\mathrm{22}}\hfill \end{array}\right]·\left[\begin{array}{l}{a}_{\mathrm{1}}\hfill \\ a_{\mathrm{2}}\hfill \end{array}\right],$

where a_i and b_i (i = 1, 2) represent the light fields at the input and output ports, respectively, k and t are the coupling and transmission coefficients of the input splitter and output combiner (k² + t² = 1 for lossless coupling) of the child MZI, and ϕ_i is the phase of the MZI arm.

Fields a₂ and b₂ are related to the racetrack ring resonator; therefore, we have

(2)$a_{\mathrm{2}}=b_{\mathrm{2}}·a_{\mathrm{Ring}}·{\mathrm{e}}^{\mathrm{i}{\phi }_{\mathrm{Ring}}}\hspace{1em},$

where a_Ring and ϕ_Ring are the loss factor and the accumulated phase when light passes the feedback ring waveguide, respectively. From Eqs. (1) and (2), the optical field transmission through the resonance arm of the parent MZI can be written as

(3)$t_{\mathrm{b}}={\displaystyle \frac{b_{\mathrm{1}}}{a_{\mathrm{1}}}}=t_{\mathrm{11}}+{\displaystyle \frac{k_{\mathrm{12}}·k_{\mathrm{21}}·a_{\mathrm{Ring}}·{\mathrm{e}}^{\mathrm{i}{\phi }_{\mathrm{Ring}}}}{\mathrm{1}-t_{\mathrm{22}}·a_{\mathrm{Ring}}·{\mathrm{e}}^{\mathrm{i}{\phi }_{\mathrm{Ring}}}}}.$

The optical field traveling through the reference arm of the parent MZI is given by

(4)$t_{\mathrm{u}}=a_{\mathrm{VOA}}·a_{\mathrm{wg}}·{\mathrm{e}}^{\mathrm{i}({\phi }_{\mathrm{ps}}+{\phi }_{\mathrm{VOA}}+{\phi }_{\mathrm{wg}})},$

where a_VOA and a_wg represent the loss factor associated with the VOA and the waveguide, respectively, and ϕ_ps, ϕ_VOA, and ϕ_wg are the phases of the phase shifter, the VOA, and the waveguide, respectively. The optical field transfer function of the entire device can then be given by

(5)${\displaystyle \frac{E_{\mathrm{out}}}{E_{\mathrm{in}}}}={\displaystyle \frac{\mathrm{1}}{\mathrm{2}}}(t_{\mathrm{u}}+t_{\mathrm{b}}).$

2.3 2.3 Fano resonance and modulation

Depending on the phase difference between the two arms of the parent MZI, the device can exhibit various resonance line shapes. In particular, Fano resonances are generated when the phase of the resonance path differs by π/2 or 3π/2 from the other reference path, that is, |ϕ₁− ϕ₂| = π/2 or 3π/2, where ϕ₁ and ϕ₂ are the arm phases of the parent MZI. At the Fano resonances, the output spectrum presents asymmetric resonance line shapes [41,48].

Fig.2 Working principle illustration for generation of (a) and (b) UWB monocycle pulses in two polarities, (c) and (d) UWB doublet pulses in two polarities

Full size|PPT slide

After applying an RF driving signal on the PN junction, UWB pulses can be generated by modulating at the Fano resonances. As depicted in Figs. 2(a) and 2(b), the Fano resonance spectrum shifts back and forth in response to the electrical driving signal. Effectively, the operation point moves along the Fano spectrum on the rising/falling edge of the square-wave driving signal. Thus, following the Fano resonance line shape, the output optical power varies correspondingly to produce a pair of monocycle UWB pulses of opposite polarities in the time domain.

When the RF driving signal is a Gaussian pulse, the device generates a UWB doublet signal, as shown in Fig. 2(c). The symmetric-looking UWB doublet pulses are produced when the halves of the two monocycle UWB pulses in opposite polarities are combined. The upside-down inverted UWB doublet can also be generated by modulating the Fano resonance with a difference in the π-phase (see Fig. 2(d)).

3 3 Measurement and results

3.1 3.1 Transmission spectrum

We first measured the device transmission spectrum, as shown in Fig. 3. Because of the arm length difference of the parent MZI, the resonance line shape varies with wavelength to cover one Fano resonance evolution cycle from 1551 to 1563 nm. The free spectral range of the MRR is approximately 0.4 nm, whereas that of the parent MZI is 11 nm. Then, owing to the interference between the MRR and the parent MZI, the device exhibits distinct resonance line shapes when the phase difference (Δϕ) between the MRR and the MZI changes. Specifically, when Δϕ≈ π/2, the typical asymmetric Fano resonance line shape is observed as shown in the upper left graph of Fig. 3. The asymmetric Fano resonance has a peak-to-valley extinction ratio of above 20 dB. When Δϕ≈ 3π/2, it exhibits the reversed asymmetric Fano line shape, as shown in the upper right graph of Fig. 3. When Δϕ≈ π, the spectrum shows an inverted Lorentzian line shape in the bottom left graph of Fig. 3. When Δϕ ≈ 0, it becomes a Lorentzian line shape as shown in the bottom right graph of Fig. 3. More detailed explanations about the Fano resonance can be found in Ref. [48].

Fig.3 Measured transmission spectrum of the RC-MZI in one Fano resonance evolution cycle. The insets illustrate the magnified resonance spectra at different wavelengths. The modeled spectra are also shown

Full size|PPT slide

By applying a DC voltage, V_ps, on the phase shifter, the phase difference between the arms of the parent MZI changes, generating different resonance line shapes in the spectrum, as shown in Fig. 4. Because the MRR is sensitive to temperature, the device was placed on a thermoelectric cooler during measurement to maintain the temperature at 22°C±0.05°C.

Fig.4 Resonance spectra at four voltages (Vps) on the phase shifter

Full size|PPT slide

Figure 5(a) shows that the Fano resonance spectrum shifted when a DC voltage was applied onto the PN junction. When the PN junction was set under the forward bias regime (V_d>0), it provided a higher modulation efficiency but a larger loss than that under the reverse bias. We extracted the phase shift from Fig. 5(a), taking the 0 V curve as the reference. Then, we calculated the variation in the effective refractive index using the formula: Δn_eff = (l/2π)·Δϕ. Figure 5(b) shows the effective refractive index change as a function of voltage. It has a slope of −2 × 10⁻⁵ and −2.2 × 10⁻⁴ V⁻¹ in the reverse and positive bias regimes, respectively.

Fig.5 (a) Fano resonance spectral shifts with different voltages (Vd) applied to the PN junction. (b) Extracted waveguide effective refractive index variation as a function of voltage on the PN junction

Full size|PPT slide

3.2 3.2 UWB pulse generation

According to the working principle described above, we set up an experimental system to generate UWB pulses, as shown in Fig. 6. Light from a tunable continuous-wave laser (EXFO, T100S-HP, Canada) passes through a polarization controller (PC) before being coupled to the device. Square-wave or Gaussian RF pulses are generated by an arbitrary waveform generator (AWG, SHF, 19120 B, Germany), amplified by an RF amplifier (SHF 810, broadband amplifier, Germany) and then combined with the DC bias voltage through a bias-tee before they are applied to the travelling-wave electrode (TWE) of the device via a 40-GHz microwave probe for electro-optic modulation. The other end of the TWE is terminated by a 50-Ω resistor for impedance matching. The modulated optical signal output from the device is then amplified by an erbium-doped fiber amplifier (EDFA) to compensate for device insertion loss, followed by a 3-nm bandwidth optical filter to suppress the amplified spontaneous emission noise. Finally, the optical signal is received by a photodetector and measured using an oscilloscope (Keysight, Infinium UXR0804A, USA) and an electrical spectrum analyzer (ESA, R&S, FSUP 50, Germany).

Fig.6 Experimental setup for UWB signal generation and characterization. PC: polarization controller, EDFA: erbium-doped fiber amplifier, PD: photodetector, AMP: RF amplifier, AWG: arbitrary waveform generator, ESA: electrical spectrum analyzer

Full size|PPT slide

Figure 7(a) shows the UWB monocycles generated at the wavelength of 1548.93 nm. The RF square-wave driving signal has a frequency of 200 MHz and a duty cycle of 35%. The voltage swing is from −0.94 to 2.76 V. The PN junction works mainly in the forward bias regime to utilize its high modulation efficiency. The UWB monocycle signal has a pulse width of approximately 150 ps at the rising edge of the RF signal. Because the falling edge of the driving signal is slower than the rising edge owing to the waveform distortion from the AWG, it generates a broader UWB monocycle pulse in the opposite polarity. Figure 7(c) shows the resulting UWB monocycle pulses when the device was modulated at the wavelength of 1553.74 nm. The pulse width was approximately 98 ps. Figures 7(b) and 7(d) show the RF spectra of the UWB waveforms. The discrete frequency lines have a spacing of 200 MHz, which is equal to the repetition rate of the monocycle pulses.

Fig.7 Generation of UWB monocycle pulses. (a) Monocycle pulses at the Fano resonance wavelength of 1548.93 nm and (b) its electrical spectrum. (c) Monocycle pulses at the Fano resonance wavelength of 1553.74 nm and (d) its electrical spectrum

Full size|PPT slide

When the RF driving signal is altered to Gaussian pulses, UWB doublet pulses can be generated. The Gaussian pulses have a frequency of 200 MHz and a duty cycle of 15% with a voltage swing from −1 to 3.7 V. Figure 8 shows the UWB doublet signals and the corresponding RF spectra.

Fig.8 Generation of UWB doublet pulses. (a) Doublet pulses at the Fano resonance wavelength of 1548.93 nm and (b) its electrical spectrum. (c) Doublet pulses at the Fano resonance wavelength of 1553.74 nm with the opposite polarity and (d) its electrical spectrum

Full size|PPT slide

If we still use a square wave as the electrical driving signal and increase its voltage swing, higher-order UWB pulses can be generated, as shown in Fig. 9. When the driving signal voltage swing is from −1.1 to 3.9 V, the device generates UWB doublet pulses (Fig. 9(a)). When the voltage swing is from −0.7 to 4.3 V, it generates UWB triplet pulses (Fig. 9(c)). When the driving voltage is further enlarged with a swing from −1.16 to 5.16 V, UWB quadruplet pulses are generated (Fig. 9(e)). The frequency spacing in the electrical spectrum is 100 MHz, which corresponds to the repetition rate of the RF driving signal. Higher-order UWB pulses have better potential for communication applications because the interference between UWB signals and other wireless signals decreases [18].

Fig.9 Generation of high-order UWB pulses using a square-wave driving signal. (a) and (b) UWB doublet pulses and the electrical spectrum. (c) and (d) UWB triplet pulses and the electrical spectrum. (e) and (f) UWB quadruplet pulses and the electrical spectrum

Full size|PPT slide

Table 1 illustrates the extracted RF central frequencies, 10-dB bandwidths, and fractional bandwidths of the generated UWB signals. The 10-dB bandwidths of the generated UWB signals are all wider than 500 MHz, and the fractional bandwidths are over 20%, which qualifies as the basic definition of a UWB signal. The RF central frequencies are approximately 1 GHz, which is out of the bandwidth of a typical power-efficient UWB signal (from 3.1 to 10.6 GHz). Perhaps the low central frequencies of the generated UWB signals are caused by the limited electro–optic (EO) bandwidth of the modulator. Particularly, as the doublet generated by Gaussian pulses is essentially a combination of two monocycles, it is possible to produce FCC-compliant UWB pulses by tuning the full width at half maximum (FWHM) of the Gaussian pulses for the UWB doublet pulses [21,50]. Of note, there is infringement in the global positioning system band (from 0.96 to 1.61 GHz). These unwanted RF frequencies could be filtered by an UWB antenna.

Tab.1 Extracted key performance specifications of the generated UWB pulses

electrical driving signal	200-MHz square wave		200-MHz Gaussian wave		100-MHz square wave
UWB signal	UWB monocycle	UWB monocycle	UWB doublet	UWB doublet	UWB doublet	UWB triplet	UWB quadruplet
optical wavelength/nm	1548.93	1553.74	1548.93	1553.74	1548.93	1548.93	1548.93
central frequency/GHz	1.2	0.8	0.8	1.2	1	1.1	1.3
10-dB bandwidth/GHz	2.2	1.4	3	1.4	1.2	0.8	1
fractional bandwidth/%	183	175	375	117	120	72.7	76.9

4 4 Conclusions

We have realized a reconfigurable RC-MZI device on a silicon photonics platform. It can generate Fano resonances with asymmetric line shapes when there is a proper phase difference between the two arms of the parent MZI. The Fano resonance is modulated when a RF signal drives the PN junction in the MRR, producing optical UWB monocycle and doublet signals. UWB signals in opposite polarity can be easily obtained by modulating at the other asymmetric Fano resonance. Even higher-order UWB pulses can be produced when the driving signal is enhanced with a larger voltage swing. UWB technology has superior performances in terms of accuracy, power consumption, robustness in wireless connectivity, and security. It is a promising technology in the era of 5G IoT, whose applications include industrial automation, sensing network, home/office automation, and accurate indoor positioning.