1. Introduction

Transmission and conversion of the electromagnetic wave is an important way for optical detection and efficient utilization of green energy and, thus, has attracted great interest all over the world. As a major device component of the electromagnetic wave transmission and conversion, rectenna is widely used in high altitude vehicle, space solar power satellite, micro robot, radio frequency identification, sensor network, and the like [1–5]. Previously rectenna usually worked in microwave range. Nowadays it can be extended to optical range after more than 40 years' development [6–12].

Recently, the absorption mechanism of noble-metal based optical rectenna is considered to be antenna resonance and plasmon resonance. Since the size of near-infrared optical antenna is at the order of hundreds of nanometers, the fabrication process has a very important impact on the device performance [13–21]. Low-epsilon material shows similar properties to the epsilon-near-zero (ENZ) material, which is a new type of electromagnetic material with real part of epsilon (ε₁) near to zero and limited imaginary part of epsilon (ε₂) [22,23]. Displacement current plays a key role in noble metal optical antenna, while it is difficult for it to flow in the low-epsilon material. This unique property can effectively inhibit the plasmon resonance and surmount the fabrication difficulty caused by the antenna size reduction [24–28]. Specifically, the materials with low-epsilon characteristics are expected to show notable response to infrared light, whose wavelength is slightly shorter than that of ENZ point [23].

In addition, a tunneling diode made of two metal electrodes with different work function can be used for optical antenna rectifying without bias [29–34]. The structure of Ti-Au antenna coupled with the Ti/TiO₂/Au tunneling diode is simple, which can work without bias [35]. However, a real rectenna may exhibit a low rectifying ratio, even in the case with an extremely small fabrication error during the preparation. Thus, the photocurrent cannot be detected in this system. One possible reason may be the uncontrollable quality of Ti and TiO₂ films. Ti metal film can be readily oxidized, which can influence permittivity notably without an obvious change in the conductivity and, thereby, rectifying property of the devices can be affected [36]. Therefore, here a TiO_x/TiO₂/Au tunneling diode coupled with TiO_x/Au antenna was constructed. The influence of oxygen content in TiO_x on the film permittivity and device rectifying property was studied thoroughly. The low-epsilon TiO_x films were obtained by controlling the deposition conditions very carefully. The TiO_x/TiO₂/Au tunneling diode was built using high permittivity TiO₂ prepared via in situ oxidation. Eventually, the infrared responsive rectenna was obtained.

2. Low-epsilon rectenna

The scale of optical antenna should be around hundreds of nanometers because its effective length is larger than the physical length due to the non-zero imaginary displacement current at both ends. In contrast, TiO_x low-epsilon material shows a low imaginary part of displacement current, giving rise to a weak plasmon resonance. The diagram of a rectenna with silica substrate (ε₁ ~2.5, ε₂ ~0) is shown in Fig. 1(a). The antenna length (L), triangular length (W), lead wire width (W₁), film thickness of Au and TiO_x (d), and vertex angle (θ) are 3.7 μm, 2.3 μm, 100 nm, 100 nm, and 60°, respectively. The diagram of metal-insulator-metal (MIM) tunneling diode with 3-nm thick TiO₂ as the insulator is shown in Fig. 1(b). All the parameters used here were chosen according to the optimized parameters of a rectenna model reported previously [36,37], except the permittivity of TiO_x.

 

Fig. 1 Diagram of (a) rectenna and (b) MIM tunneling diode.

Download Full Size | PDF

2.1 Low-epsilon materials characteristics

Au and TiO_x films were deposited in a vacuum chamber with a base pressure of 5.0E-7 Torr. An adhesive layer (5-nm thick Cr) was deposited, followed by the 95-nm thick Au film. Three different TiO_x samples (named respectively as TiO_x-1, TiO_x-2 and TiO_x-3) were obtained via sputtering at room temperature by decreasing the oxygen flow from 0.05 to 0.02, and, finally 0.00 sccm. SEM cross-section images (N-430, FEI, USA) in Figs. 2(a)-2(d) of the obtained TiO_x and Au films indicate that the thickness is about 100 nm. Figure 2(e) shows the XRD patterns of a typical TiO_x film as well as the standard samples. The TiO_x film is amorphous in nature and, thus, exhibits no obvious diffraction peaks of Ti (PDF #44-1294) or TiO₂ (PDF #21-1276 and PDF #21-1272). The major diffuse scattering broadband pattern between 10° and 50° indicates the presence of amorphous SiO₂ and TiO_x films on the crystalline silicon substrate with (400) diffraction peak (PDF #27-1402), which is consistent with the growth direction of silicon (100).

 

Fig. 2 SEM cross-section images of (a) TiO_x-1, (b) TiO_x-2, (c) TiO_x-3 and (d) Au films. (e) XRD parrterns of TiO_x compared with standard Si, Ti, rutile TiO₂ and anatase TiO₂ from MDI Jade 6.

Download Full Size | PDF

The component analysis is done by Rutherford backscattering analysis (RBS, General Ionex Corporation, GIC4117, USA) using a 2 MeV He⁺ beam with a scattering angle of 165° in 60° incidence. The SIMNRA 6.06 software is employed to analyze the RBS data. Inset (I) of Fig. 3(a) shows the fitting structure by tuning the thickness of TiO_x film and the content of Ti and O, and 1-μm silica is a substrate thick enough, as the incident energy of He⁺ becomes very low upon decay in the silica. Inset (II) of Fig. 3(a) is the zoom-in RBS spectra in oxygen range. In Fig. 3(a), the oxygen content is about 40%, 37% and 32% for TiO_x-1, TiO_x-2 and TiO_x-3, respectively. Thus, the oxygen content of TiO_x films decreases and the Ti content increases in the sequence of TiO_x-1, TiO_x-2 and TiO_x-3. TiO₂ films are prepared via in situ oxidization of three lead wire with different TiO_x using oxygen plasma processor (PJ-2, AST Products, USA). Inset (I) of Fig. 3(b) shows the fitting structure by tuning the thickness and content of Ti and O in the TiO₂/TiO_x film. An enhanced obvious shoulder is observed at high energy of O in the inset (II) of Fig. 3(b), indicating that a TiO₂ film is formed with the oxygen content of 67%, in which the oxygen content is notably higher than that of the corresponding TiO_x films shown in Fig. 3(a).

 

Fig. 3 RBS spectra of (a) TiO_x films and (b) TiO₂/TiO_x films on silica. (c) Raw and fitting Psi curves of TiO_x films. (d) Raw and fitting Psi curves of TiO₂/TiO_x films. (e) Raw and fitting Delta curves of TiO_x films. (f) Raw and fitting Delta curves of TiO₂/TiO_x films. (g) Real and (i) Imaginary part of permittivity of TiO_x-1, TiO_x-2, TiO_x-3 and Ti-Kirillova. (h) Real and (j) Imaginary part of permittivity of TiO₂-1, TiO₂-2, TiO₂-3 and rutile TiO₂-Palik. Insets of (I) in (a), (I) in (b), (c) and (d) are the corresponding fitting structure. Insets of (II) in (a) and (b) are the zoom-in RBS spectra.

Download Full Size | PDF

Ellipsometric spectroscopy (SE850, Sentech, Germany) with SpectraRay3 software is used to study the thicknesses and optical properties of the obtained TiO_x and TiO₂ films. The obtained ellipsometric spectra of TiO_x, TiO₂ and silica are fitted respectively by the model of Drude-Lorentz, Tauc-Lorentz and Cauthy. The inset of Fig. 3(c) presents the fitting structure with four layers of air/TiO_x-n/SiO₂/Si (n = 1, 2 and 3). The raw and fitting curves of Psi-Delta values of TiO_x-n (n = 1, 2 and 3) films on the SiO₂/Si substrate are shown in Figs. 3(c) and 3(e). The thickness of TiO_x films is around 100 ± 3 nm. The permittivity of TiO_x-n (n = 1, 2 and 3) is shown in Figs. 3(g) and 3(i). The real part of permittivity (ε₁) decreases as the oxygen content decreasing in the range of 750 ~2000 nm (Fig. 3(g)). Obviously, the obtained TiO_x films are typical low-epsilon materials compared to Ti-Kirillova [38], which is also measured by ellipsometric spectroscopy. The TiO_x_3 has a similar permittivity to Ti film measured by Perkin-Elmer spectrophotometer previously [39]. Moreover, the ε₁ of TiO_x-3 is between −0.4 and 0 in 750 ~1070 nm (λ_ENZ), indicating the interaction of free electrons and photons. So TiO_x-3 is a low-epsilon material while still a metal film [27]. The inset of Fig. 3(d) shows the fitting structure with five layers of air/TiO₂-n/TiO_x-n/SiO₂/Si (n = 1, 2 and 3). The raw and fitting curves of Psi-Delta values of TiO₂/TiO_x-n (n = 1, 2 and 3) films on the SiO₂/Si substrate are shown in Figs. 3(d) and 3(f). An obvious absorption is observed in the UV range compared with pure TiO_x film, indicating the presence of TiO₂ film. The thickness of TiO₂ films is around 3 ± 0.2 nm, suggesting a very thin TiO₂ insulating film is obtained atop the relatively thick TiO_x film. The permittivity of TiO₂-1, TiO₂-2 and TiO₂-3 prepared respectively via in situ oxidization of TiO_x-1, TiO_x-2 and TiO_x-3 is shown in Figs. 3(h) and 3(j). The ε₁ of TiO₂-3 is a little smaller than rutile TiO₂-Palik [40] but much larger than that of TiO₂-1 and TiO₂-2, while the ε₂ of TiO₂-3 is very close to rutile TiO₂-Palik but much smaller than that of TiO₂-1 and TiO₂-2 from 850 to 1200 nm. This phenomenon contributes to the strongest dielectric property of TiO₂-3 among the obtained three amorphous samples.

2.2 Low-epsilon rectennas simulation

The difference in permittivity of TiO_x films leads to the different displacement current. Displacement field $D={\epsilon }_r{\epsilon }_{0}E$, where ε₀ is the permittivity of free space, ε_r is the relative permittivity with ${\epsilon }_r={\epsilon }_1+j{\epsilon }_2$, where ε₂ is the imaginary part of the permittivity. A noble metal seldom behaves as a perfect conductor in the optical antenna. Furthermore, displacement current plays a more dominant role than the conduction current and, thus, is the primary one in radio frequency antenna [41].

When electric field E equals to $E=E_0\exp (-j\omega t)$, the displacement current can be written from Maxwell equation as Eq. (1).

The imaginary part of displacement current $j\omega {\epsilon }_1{\epsilon }_{0}E$ means the reciprocal transformation between the electric field and magnetic field. The real part $\omega {\epsilon }_2{\epsilon }_{0}E$ exhibits the ability to absorb electromagnetic wave [42]. Since the ε₁ of low-epsilon TiO_x approaches to 0, the imaginary part $j\omega {\epsilon }_1{\epsilon }_{0}E$ also approaches to 0. Although the ε₂ equals to 0 in an ideal ENZ material, a real low-epsilon material shows non-ignorable loss due to the non-zero ε₂ [23,28]. The displacement current coefficient of the obtained TiO_x films and Ti-Kirillova with a dimension of S/m are shown in Figs. 4(a) and 4(b). The imaginary part of displacement current coefficient of TiO_x-2 and TiO_x-3 approaches 0, while the real part is pretty large because of the high ε₂, which means strong absorbability [43].

 

Fig. 4 Displacement current coefficient of (a) TiO_x-1, TiO_x-2, TiO_x-3, and (b) Ti-Kirillova. Return loss spectra of four rectenna models in (c) near-infrared band and (d) middle-infrared band. Model-m (m = 1, 2, 3, and 4) represents simulated structure made of TiO_x-n (n = 1, 2 and 3) and Ti-Kirillova, respectively.

Download Full Size | PDF

Return loss of the four rectenna models is simulated by applying external bias voltages on the MIM diode (Figs. 4(c) and 4(d)). TiO _x-n/Au antenna coupled with the TiO_x-n/TiO₂-n/Au diode is denoted as Model-m (m = 1, 2 and 3), and Ti-Kirillova/Au antenna coupled with the Ti-Kirillova/TiO₂-3/Au diode is named as Model-4. The return loss peaks in near-infrared band appear at 998 (−10.8 dB), 994 (−11.5 dB), 994 (−11.9 dB) and 940 nm (−10.1 dB) for Model-1, Model-2, Model-3 and Model-4, respectively; while those in the middle-infrared band are at 4284 (−7.8 dB), 4161 (−8.5 dB), 5027 (−8.3 dB) and 5916 nm (−15.7 dB) for Model-1, Model-2, Model-3 and Model-4, respectively. The Model-4 exhibits the highest middle-infrared response due to the notable plasmon resonance of Ti-Kirillova-based antenna. Interestingly, Model-3 has the highest near-infrared response, although the return-loss modulus are almost the same in these four models. This agrees well with the minimum imaginary part of displacement current for TiO_x-3 among the films under test in the near-infrared band. The plasmon resonance can be very weak when the imaginary part of displacement current is close to zero.

2.3 Low-epsilon rectenna photodetectors

The rectenna patterns were defined with a bilayer resist using electron beam lithography (Vistec EBPG 5000plus ES, USA), followed by film deposition (TiO_x and Cr/Au) and lift-off process. First, the primer resist (Rohm & Haas, LOL2000, USA) was spin-coated on the substrate, followed by positive resist (ZEON Corporation, ZEP520A, Japan). Three layers of TiO_x antenna bow, TiO₂ insulator box, and Cr/Au antenna bow were fabricated sequentially using the localizer for alignment by repeating thrice of the above process. TiO₂ films were prepared via in situ oxidization of the three different TiO_x lead wires using oxygen plasma. Finally, rectenna was fabricated using the obtained TiO_x-n/Au antenna coupled with the TiO_x-n/TiO ₂-n/Au diode, denoted as Device-n (n = 1, 2 and 3). In the experimental setup, the cascaded grating acts as a wavelength filter after a tungsten strip lamp. Two groups of off-axis paraboloidal mirrors are used as beam adjustment system to focus the partially polarized monochromatic light on photodetector. The photocurrent signal is detected with a lock-in amplifier (SR830, Stanford) by setting the chopping frequency of 2 kHz. The responsivity of Device-3 is calibrated by a standard pyroelectric detector (DPe22, Zolix). It is found that the experimental photocurrent responsivity of Device-3 in the near-infrared band is about 13.6 nA/W at around 900 nm (Fig. 5(a)), which is close to the simulated near-infrared resonance wavelength. All of the four models contribute almost the same return-loss modulus in near-infrared range. Although Model-4 presents a better fit with the spectral location of the maximum photoresponse compared with Model-3, the Model-4 has a major response in the mid-infrared range, and its return-loss peak in the near-infrared has a high order oscillation and a weak absorption. Considering the effectiveness of spectroscopic ellipsometry fitting, moreover, Device-3 agrees well with Model-3 in the near-infrared band. The polarization dependence for Device-3 excited by the partially polarized light at 900 nm is shown in Fig. 5(b). Device-3 has a higher photoresponsivity when it is placed at 0° and 180°, implying the antenna response. However, there are no detectable photocurrent for the Device-1 and Device-2, possibly because of their bad rectifying property [29].

 

Fig. 5 (a) Photocurrent responsivity spectrum of Device-3, inset is SEM image of a typical rectenna. (b) Polarization depencence for Device-3 excited by the partially polarized light at 900 nm. (c) The semilog plot of current modulus from I-V curve, inset is the schematic energy band structure of an MIM tunneling diode. (d) Efficiency of different devices. Device-n (n = 1, 2 and 3) represents the TiO_x-n/Au antenna coupled with TiO_x-n/TiO₂-n/Au diode.

Download Full Size | PDF

Band diagram of the obtained MIM diode is shown in the inset of Fig. 5(c). For the unbiased structure, a built-in electric field is present due to the difference in the work function of Au and TiO_x, which is also closely related to the thickness of TiO₂ insulating layer (here, actually the width of tunneling barrier). Given Au is grounded, the barrier decreases when the potential applied on TiO_x increases, and the tunneling current increases quickly. The rectifying characteristics of the obtained devices is manifested by the semilog plot of current modulus (Fig. 5(c)) from the I-V curve measured by Keithley Sourse Meter (2636B). Device-3 possesses the maximum rectifying ratio of 11 at ± 1 V bias, while it is 1.1 for both Device-1 and Device-2. Obviously, Device-3 has the best rectifying property among them, as the ε₁ of TiO_x-3 is negative, which can promote the rectifying property of an MIM diode [27]. This also agrees well with the fact that TiO₂-3 exhibits the best dielectric property. Moreover, Fig. 5(d) shows the beta non-linear coefficient as calculated by Eq. (2) [33].

It is found that the β of Device-3 is about 1.3 A/W without bias (Fig. 5(d)), which is large enough to contribute a good rectification compared with the reported value of around 0.5 A/W [34]. On the contrary, it is about 0.07 A/W for both the Device-1 and Device-2, which is almost an order of magnitude lower than the reported value. This may explain why the Device-1 and Device-2 have no detectable photocurrent. As for its essence, TiO_x-3 has the lowest imaginary displacement current, and TiO₂-3 has the strongest permittivity. This makes Device-3 have the highest antenna impedance and the best tunneling diode rectification among these three samples.

3. Conclusion

In conclusion, the fabrication recipe has a significant impact on the TiO_x-based rectenna. Low-epsilon TiO_x film is successfully prepared by optimizing the film deposition parameters, which exhibits a negative permittivity and considerable metallic property. TiO₂ prepared via in situ oxidization of the TiO_x shows strong dielectric property. Coupled with 3.7 μm length of TiO_x/Au antenna, the MIM diode can be used to achieve photoelectric response at around 900 nm. Moreover, the performance of accurately prepared devices can be enhanced using the protocol proposed in this work, which may offer a new approach for the infrared photodetector meeting the requirements of practical applications.