June 1, 2015 / Vol. 40, No. 11 / OPTICS LETTERS
2603
High-performance infrared light trapping in nano-needle structured p+ SnOx (x ≤ 1)/thin film n-Ge photodiodes on Si Xiaoxin Wang,1,3 Andrew Wong,1 Stephanie Malek,1 Yan Cai,2 and Jifeng Liu1,* 1
Thayer School of Engineering, Dartmouth College, 14 Engineering Drive, Hanover, New Hampshire 03755, USA 2
Microphotonics Center, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139, USA 3
e-mail: [email protected] *Corresponding author: [email protected] Received March 27, 2015; revised May 7, 2015; accepted May 10, 2015; posted May 11, 2015 (Doc. ID 237042); published May 28, 2015 We report nano-needle structured conductive SnOx (x ≤ 1) as a self-assembled electrode for high-efficiency light trapping in thin-film infrared (IR) photonic devices, benefiting from the high scattering efficiency, high density, and low IR loss of the nano-needles. We demonstrate a 2.2× responsivity enhancement for a 1.5-μm-thick Ge absorber in a nano-needled p SnOx ∕n-Ge photodiode on Si at λ 1580 nm, in good agreement with theoretical calculation of 2.3× enhancement assuming no IR loss in the nano-needles. Such low-loss light trapping can potentially enable 15–30× absorption enhancement at λ 1600–1650 nm in the Ge layer when integrated with a perfect rear reflector. © 2015 Optical Society of America OCIS codes: (230.5160) Photodetectors; (040.3060) Infrared; (310.7005) Transparent conductive coatings; (130.0250) Optoelectronics. http://dx.doi.org/10.1364/OL.40.002603
Light trapping strategies are widely used in thin-film solar cells to increase the absorption in the thin active absorber layer [1–4]. Conventionally, light trapping is achieved by conformal deposition of the active absorber on surface-textured transparent conductive oxides (TCOs) [1]. However, this method increases the surface area of the active absorber, and correspondingly, the surface leakage [1,2]. Over the last decade, plasmonic effects [3,4] and scattering from nanostructures [5,6] have been employed for light trapping. The subwavelength nanostructures can be engineered to scatter the light preferentially towards the active absorber layer at large incidence angles, which increases the optical path length and enables lateral waveguiding in the absorbing semiconductor thin film [7,8]. Remaining challenges include optical loss in metallic structures, broad-band light trapping, and the relatively high cost of the noble metal nanoparticles for large-area device such as solar cells. Considering that TCOs are directly adjacent to the active thin film absorber, optimal results could be obtained if low-loss nanostructured TCOs could be engineered for efficient light trapping without increasing the active absorber surface area. In addition, most studies on light trapping focus on solar cell applications in the visible and near infrared regime (λ < 1100 nm). It is therefore interesting to explore the possibility of light trapping in other thin-film photonic devices in the infrared (IR) regime, such as thermophotovoltaic (TPV) cells and IR detectors, using nanostructured TCOs deposited on a planar thin-film absorber. In TPV systems, small band gap semiconductors such as bulk GaSb and Ge wafers are utilized to absorb the IR thermal radiation, yet they are very costly (10–50 × more expensive than Si wafers) [9,10]. Light-trapping strategies for thin-film Ge can play an important role in reducing the cost. On the other hand, the research on thin-film Ge-on-Si optoelectronic devices has made rapid progress over the last decade, owing to the excellent optoelectronic properties 0146-9592/15/112603-04$15.00/0
of epitaxial Ge at optical communication wavelengths and its compatibility with Si-complementary metal oxide semiconductor (CMOS) processing [11]. Thanks to the high material quality and the tensile strain in the Ge-on-Si thin films that red-shifts the direct band gap, Ge-on-Si photodetectors can respond in the entire C band (1528–1560 nm) and L band (1561–1620 nm) of optical communications. However, the responsivity at 1620 nm is still low even for 2.35-μm-thick Ge with 0.25% tensile strain, where the responsivity of 0.1 A∕W corresponds to an absorption of only 13% [12]. Therefore, light trapping in Ge-on-Si photodiodes could greatly improve their performance as IR detectors and TPV cells. In this Letter, we report nano-needle structured conductive p SnOx (x ≤ 1) [13] as a self-assembled electrode for high-efficiency, low-loss light trapping in thin-film Ge-on-Si IR photonic devices. We have recently demonstrated nano-needle structured p SnOx at a low temperature of 225°C–300°C with a high hole density of p ∼ 5 × 1021 cm−3 , a mobility of μp ∼ 2 cm2 V−1 s−1 , and a low IR optical loss of 2000 nm exceeds 95% for Si and Ge substrates due to the low optical loss of Sn in this regime, indicating that the core/shell structure is ideal for lowloss light trapping at λ > 2000 nm. At λ 1600–1900 nm the scattering efficiency is reduced to 70%–90%, mainly due to the absorption of the Sn core. On the other hand, since σ abs ∼0 at λ 1000–2500 nm for SnO, the SnO nanoneedle has ηsc ∼ 100% although the absolute σ sc is ∼1∕5 of the core/shell structure [Fig. 2(a)]. Figures 3(a)–3(d) further show polar plots of normalized angular scattering distribution (ASD) of various nano-needles/substrates at λ 1600 nm. The loss cones for Si/air and Ge/air interfaces are defined by the critical angles of θc 17° and 14°, respectively. The scattered light outside the loss cone in Si or Ge is trapped by total internal reflection after each scattering event. Clearly, the nano-needle on the high index substrates exhibits a narrower angular distribution of the scattered light compared to the case in air, together with greatly reduced backscattering that is beneficial for light trapping. Nano-needle Sn/SnO core/shell on Si
1 Sn-SnO
0.1 core/shell in air
0.01
SnO in air
(a) 500
1000
1500
2000
Wavelength (nm)
Fig. 1. SEM images of high-density, randomly oriented nanoneedles formed in a SnO0.85 thin film after annealing at 225°C for 15 min in (a) N2 and (b) air. Nano-needle thickness is the same as the film thickness, as indicated by the red arrow.
2500
Scattering Efficiency ηsc
a prefect rear reflector. Compared to conventional surface texturing, these structures can be deposited on planar semiconductor thin films without increasing the surface area or surface leakage. Compared to common plasmonic metals (Au, Ag), these nanostructured TCOs exhibit low loss in IR regime and broadband light trapping. The Sn-SnO core/shell SnOx nano-needles were prepared by RF magnetron co-sputtering of Sn and SnO2 targets on Si and 1.5-μm-thick, 0.2% tensile-strained n-type Ge-on-Si substrates, followed by annealing in N2 or air at 225°C for 15 min in a tube furnace [13]. Scanning electron microscopy (SEM) reveals that the film consists of a single layer of high-density SnOx nano-needles, as indicated by the red arrow in Fig. 1(a). Therefore, the thickness of the nano-needles is the same as that of the film, which was measured to be 155 nm by cross-sectional SEM. The average length of the nano-needles is 500 nm and the width is 100 nm. The average composition is SnO0.85 from Rutherford Back Scattering (RBS) analysis for the asdeposited thin film and the samples annealed in N2 , the latter showing n conductivity. The 1:0.85 Sn:O atomic ratio indicates that the average SnO shell thickness is 40 nm. When annealed in air, the nano-needled SnOx film [Fig. 1(b)] is close to stoichiometric SnO since most of the Sn cores have been oxidized, and the conductivity becomes p correspondingly. In this study, the scattering mechanisms for both n Sn/SnO core-shell and nearly stoichiometric p SnO nano-needles are investigated via theoretical modeling and compared to the experimental results. For the pn photodiode fabrication, air annealing was adopted to form p nano-needle structured SnO (p ∼ 1021 cm−3 ) on n-type epitaxial Ge-on-Si (n ∼ 1017 cm3 ). The transmission and reflection spectra were measured using a UV–visible–IR spectrometer with an integrating sphere, and the nominal absorptance is derived as 1-transmittance-reflectance. To understand the scattering mechanism of the nanoneedles, the theoretical scattering (σ sc ) and absorption (σ abs ) cross-section spectra as well as the scattering patterns for Sn/SnO core/shell nanostructures on different substrates were calculated numerically using finite element method (COMSOL multiphysics 4.3a, Wave Optics module), similar to our previous work [14]. The dielectric functions of the substrates (bulk Si and Ge) [15], β-Sn [16], and SnO [17]] are utilized in the simulations. To the first-order approximation, we assume that the interactions between nano-needles are negligible, and that the scattering behavior of the nano-needles as a whole is simply the aggregate of the individual
Normalized Scattering Cross Section σsc
2604
1.0
on Ge Sn/SnO core/shell
0.9 nano-needle
on Si in Air
0.8 0.7 0.6
(b) 1500
2000
2500
Wavelength (nm)
Fig. 2. (a) Normalized scattering cross-section versus wavelength for a single Sn/SnO core/shell nano-needle in air and on the Si substrate, and a SnO nano-needle (without the Sn core) in air. The light is incident normally on the thin films, and the electric field is polarized along the nano-needle. (b) Scattering efficiency versus wavelength for Sn/SnO core/ shell structures in air, on Si, and on Ge substrates.
June 1, 2015 / Vol. 40, No. 11 / OPTICS LETTERS
100
(a)
90 80
Sn/SnO core-shell nano needle on DSP Si
70 60 50
Measured
40 30
Simulated 1000
1500
2000
Wavelength (nm)
Fig. 4. Scattered electric field Ex (along the nano-needle) at λ 1600 nm for (a) Sn/SnO core/shell nano-needle, and (b) SnO nano-needle on Ge. The light is incident along z direction and polarized along the x direction. The Sn core is outlined by the yellow rectangle in (a). The outer boundaries of the nano-needles are outlined by white rectangles.
2500
1.5 μm thick Ge absorption
There is negligible difference between the ASD of SnO nano-needle and that of its core-shell counterpart [Figs. 3(c) and 3(d)], as further confirmed by the scattered electric field (E x ) distribution in Fig. 4. However, the scattered field is more concentrated on the Sn core for the core/shell structure. Therefore, while the Sn core increases the scattering cross-section [Fig. 2(a)], it also tends to increase the ohmic losses. Overall, optimizing the size of the Sn core by controlled oxidation may help decrease the metal loss and improve the scattering efficiency without affecting the scattering pattern. In order to evaluate the light-trapping effect of the nano-needles for comparison with optical measurements, we calculated the trapped fraction (A) by integrating the trapped light over the solid angles greater than θc and normalizing to the total scattered power I sc . We also calculated the fraction of inward scattering within the loss cone (D), also normalized to I sc . Therefore, the total optical coupling into the semiconductor absorber is B A D. The out-coupling loss (C) is defined as the fraction of light scattered into air. Following the multiscattering method developed in Ref. [18], we obtain
the total nominal absorption, including both the absorbed light and trapped/guided light that cannot be collected by the integrating sphere. Figure 5(a) shows that this simulated nominal absorption agrees very well with the experimental data. Referring to Fig. 2, we find that the nominal absorption below the band gap of Si at λ > 1100 nm is due to the light trapping effect of the Sn/SnO core shell structure and the absorption of Sn. Especially, the ∼45% nominal absorption at λ > 2000 nm is mainly due to the light trapping effect since the scattering efficiency is >95% [Fig. 2(b)]. Therefore, nearly half of the incident optical power at λ > 2000 nm is trapped by the core-shell nano-needle structures. Considering the high scattering efficiency of SnOx nano-needles in the IR regime, it is very interesting to investigate their applications for light trapping in Ge thin-film photonic devices on Si. Because Ge can only absorb light at λ < 1900 nm due to the band gap constraint (∼0.66 eV), low-loss SnO nano-needles (without Sn cores) are more beneficial for light trapping in this case, while core/shell structures only offer high ηsc at λ > 1900 nm [Fig. 2(b)]. The Sn/SnO core/shell structure, on the other hand, can potentially provide high-efficiency light trapping for thin-film GeSn absorbers with a direct band gap corresponding to λ ∼ 2500 nm [19]. Figure 5(b) shows the simulated absorption spectra of 1.5-μm-thick, 0.2% tensile strained Ge film on Si without light-trapping structures (bare Ge-on-Si), with SnO nano-needles alone, and with both front-side SnO nano-needles and an ideal back-reflector. We assume no IR loss in SnO nanoneedles in the simulation. The absorption coefficients of 0.2% tensile strained Ge-on-Si are taken from the experimental data in Ref. [20]. With the SnO nano-needle light trapping, the absorption at 1560–1650 nm is enhanced by 2–5 × . We also found that the relative enhancement due to reduced reflection on SnO is only 30%, so most of the enhancement indeed comes from the light-trapping effect. When an ideal back-reflector is incorporated (e.g., a distributed Bragg reflector, DBR), the enhancement at 1600–1650 nm can be further increased to 10–30 × . Such a strong enhancement is because even the scattered light within the loss cone in Fig. 3 can now be reflected back. As the back-reflected light hits the frontside nano-needle structure again, it has another chance to be partially scatter outside the loss cone and guided by total internal reflection. Thus, the back reflector prevents Nominal Absorption (%)
Fig. 3. Polar plots of normalized angular scattering distribution of a single Sn/SnO core/shell nano-needle at λ 1600 nm: (a) in air, (b) on a Si substrate, and (c) on a Ge substrate. For comparison, the case for a SnO nano-needle is shown in (d). The light is incident vertically on the nano-needle. The black curves show the results for the polarization along the nanoneedle, while the red ones correspond to the polarization perpendicular to the nano-needle. The loss cones for Si/air and Ge/air interfaces are also shown (green lines), defined by the critical angles for total internal reflection.
2605
1.0 0.8 0.6
With SnO nano-needle +ideal back reflector
With SnO nano-needle
0.4 0.2
(b)
2.3x
15x
Bare Ge-on-Si
30x
Simulation
0.0 1450
1500
1550
1600
1650
Wavelength (nm)
Fig. 5. (a) Comparison of measured (blue line) and simulated (full red circle) nominal absorption spectra of Sn/SnO core/ shell nano-needles on a double-side polished Si substrate; (b) simulated absorption spectra of 1.5-μm-thick, 0.2% tensile strained Ge film on Si for 3 cases: without light trapping structures, with SnO nano-needles alone, and with both front-side SnO nano-needles and an ideal back-reflector.
OPTICS LETTERS / Vol. 40, No. 11 / June 1, 2015 0
-3
(a)
=10 cm
I (mA)
-2
2
10
10
0
-3
10
-4
10
-5
10
-6
10
-2
-1
10 Ideality factor -1 n=1.6 10 Series -2 10 resistance -3 Rs=2.4 kΩ 10 0 1 2
Voltage (V)
0.6
10 10
-2
Responsivity (A/W)
10
3
Junction area
2
-1
J (mA/cm )
10
Laser
(b)
Probe
40 0.4
p-SnOx n-Ge Si
Probe Laser 30 p-SnOx n-Ge Si
20
0.2
EQE(%)
2606
10
Incident power =50 μW Light on SnO 0 Light on Ge
0.0 0.0
-0.5
-1.0
-1.5
-2.0
Reverse Bias Voltage (V)
Fig. 6. (a) I-V/J-V curve of p SnO nano-needles/n-type Ge photodiode. The 1.5-μm-thick n-Ge film is grown epitaxially on Si substrate. The inset shows the ideality factor and series resistance. (b) Responsivity/EQE as a function of reverse bias when the λ 1580 nm laser shines on the SnOx region (green line) and the n-Ge region (red line) of the diode, respectively. The insets show the schematics of the corresponding configurations.
the backside leakage and allows the light to hit the front-side nano-needle scatterers multiple times, thereby significantly enhancing light trapping. To verify the theoretical model, we fabricate p nanoneedle structured SnO (155 nm thick) on 1.5-μm-thick n-type epitaxial Ge thin films on Si to form a pn photodiode. Note that this type II heterojunction facilitates the collection of photogenerated carriers. The SnOx layer was patterned by HCl to define the 10−3 cm2 junction region. The device is to demonstrate the concept of light trapping by SnO nano-needles without optimization/ metallization. The low temperature fabrication of p SnO (225°C) is also fully compatible with back-end-of-line (BEOL) CMOS processing (
2603
High-performance infrared light trapping in nano-needle structured p+ SnOx (x ≤ 1)/thin film n-Ge photodiodes on Si Xiaoxin Wang,1,3 Andrew Wong,1 Stephanie Malek,1 Yan Cai,2 and Jifeng Liu1,* 1
Thayer School of Engineering, Dartmouth College, 14 Engineering Drive, Hanover, New Hampshire 03755, USA 2
Microphotonics Center, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139, USA 3
e-mail: [email protected] *Corresponding author: [email protected] Received March 27, 2015; revised May 7, 2015; accepted May 10, 2015; posted May 11, 2015 (Doc. ID 237042); published May 28, 2015 We report nano-needle structured conductive SnOx (x ≤ 1) as a self-assembled electrode for high-efficiency light trapping in thin-film infrared (IR) photonic devices, benefiting from the high scattering efficiency, high density, and low IR loss of the nano-needles. We demonstrate a 2.2× responsivity enhancement for a 1.5-μm-thick Ge absorber in a nano-needled p SnOx ∕n-Ge photodiode on Si at λ 1580 nm, in good agreement with theoretical calculation of 2.3× enhancement assuming no IR loss in the nano-needles. Such low-loss light trapping can potentially enable 15–30× absorption enhancement at λ 1600–1650 nm in the Ge layer when integrated with a perfect rear reflector. © 2015 Optical Society of America OCIS codes: (230.5160) Photodetectors; (040.3060) Infrared; (310.7005) Transparent conductive coatings; (130.0250) Optoelectronics. http://dx.doi.org/10.1364/OL.40.002603
Light trapping strategies are widely used in thin-film solar cells to increase the absorption in the thin active absorber layer [1–4]. Conventionally, light trapping is achieved by conformal deposition of the active absorber on surface-textured transparent conductive oxides (TCOs) [1]. However, this method increases the surface area of the active absorber, and correspondingly, the surface leakage [1,2]. Over the last decade, plasmonic effects [3,4] and scattering from nanostructures [5,6] have been employed for light trapping. The subwavelength nanostructures can be engineered to scatter the light preferentially towards the active absorber layer at large incidence angles, which increases the optical path length and enables lateral waveguiding in the absorbing semiconductor thin film [7,8]. Remaining challenges include optical loss in metallic structures, broad-band light trapping, and the relatively high cost of the noble metal nanoparticles for large-area device such as solar cells. Considering that TCOs are directly adjacent to the active thin film absorber, optimal results could be obtained if low-loss nanostructured TCOs could be engineered for efficient light trapping without increasing the active absorber surface area. In addition, most studies on light trapping focus on solar cell applications in the visible and near infrared regime (λ < 1100 nm). It is therefore interesting to explore the possibility of light trapping in other thin-film photonic devices in the infrared (IR) regime, such as thermophotovoltaic (TPV) cells and IR detectors, using nanostructured TCOs deposited on a planar thin-film absorber. In TPV systems, small band gap semiconductors such as bulk GaSb and Ge wafers are utilized to absorb the IR thermal radiation, yet they are very costly (10–50 × more expensive than Si wafers) [9,10]. Light-trapping strategies for thin-film Ge can play an important role in reducing the cost. On the other hand, the research on thin-film Ge-on-Si optoelectronic devices has made rapid progress over the last decade, owing to the excellent optoelectronic properties 0146-9592/15/112603-04$15.00/0
of epitaxial Ge at optical communication wavelengths and its compatibility with Si-complementary metal oxide semiconductor (CMOS) processing [11]. Thanks to the high material quality and the tensile strain in the Ge-on-Si thin films that red-shifts the direct band gap, Ge-on-Si photodetectors can respond in the entire C band (1528–1560 nm) and L band (1561–1620 nm) of optical communications. However, the responsivity at 1620 nm is still low even for 2.35-μm-thick Ge with 0.25% tensile strain, where the responsivity of 0.1 A∕W corresponds to an absorption of only 13% [12]. Therefore, light trapping in Ge-on-Si photodiodes could greatly improve their performance as IR detectors and TPV cells. In this Letter, we report nano-needle structured conductive p SnOx (x ≤ 1) [13] as a self-assembled electrode for high-efficiency, low-loss light trapping in thin-film Ge-on-Si IR photonic devices. We have recently demonstrated nano-needle structured p SnOx at a low temperature of 225°C–300°C with a high hole density of p ∼ 5 × 1021 cm−3 , a mobility of μp ∼ 2 cm2 V−1 s−1 , and a low IR optical loss of 2000 nm exceeds 95% for Si and Ge substrates due to the low optical loss of Sn in this regime, indicating that the core/shell structure is ideal for lowloss light trapping at λ > 2000 nm. At λ 1600–1900 nm the scattering efficiency is reduced to 70%–90%, mainly due to the absorption of the Sn core. On the other hand, since σ abs ∼0 at λ 1000–2500 nm for SnO, the SnO nanoneedle has ηsc ∼ 100% although the absolute σ sc is ∼1∕5 of the core/shell structure [Fig. 2(a)]. Figures 3(a)–3(d) further show polar plots of normalized angular scattering distribution (ASD) of various nano-needles/substrates at λ 1600 nm. The loss cones for Si/air and Ge/air interfaces are defined by the critical angles of θc 17° and 14°, respectively. The scattered light outside the loss cone in Si or Ge is trapped by total internal reflection after each scattering event. Clearly, the nano-needle on the high index substrates exhibits a narrower angular distribution of the scattered light compared to the case in air, together with greatly reduced backscattering that is beneficial for light trapping. Nano-needle Sn/SnO core/shell on Si
1 Sn-SnO
0.1 core/shell in air
0.01
SnO in air
(a) 500
1000
1500
2000
Wavelength (nm)
Fig. 1. SEM images of high-density, randomly oriented nanoneedles formed in a SnO0.85 thin film after annealing at 225°C for 15 min in (a) N2 and (b) air. Nano-needle thickness is the same as the film thickness, as indicated by the red arrow.
2500
Scattering Efficiency ηsc
a prefect rear reflector. Compared to conventional surface texturing, these structures can be deposited on planar semiconductor thin films without increasing the surface area or surface leakage. Compared to common plasmonic metals (Au, Ag), these nanostructured TCOs exhibit low loss in IR regime and broadband light trapping. The Sn-SnO core/shell SnOx nano-needles were prepared by RF magnetron co-sputtering of Sn and SnO2 targets on Si and 1.5-μm-thick, 0.2% tensile-strained n-type Ge-on-Si substrates, followed by annealing in N2 or air at 225°C for 15 min in a tube furnace [13]. Scanning electron microscopy (SEM) reveals that the film consists of a single layer of high-density SnOx nano-needles, as indicated by the red arrow in Fig. 1(a). Therefore, the thickness of the nano-needles is the same as that of the film, which was measured to be 155 nm by cross-sectional SEM. The average length of the nano-needles is 500 nm and the width is 100 nm. The average composition is SnO0.85 from Rutherford Back Scattering (RBS) analysis for the asdeposited thin film and the samples annealed in N2 , the latter showing n conductivity. The 1:0.85 Sn:O atomic ratio indicates that the average SnO shell thickness is 40 nm. When annealed in air, the nano-needled SnOx film [Fig. 1(b)] is close to stoichiometric SnO since most of the Sn cores have been oxidized, and the conductivity becomes p correspondingly. In this study, the scattering mechanisms for both n Sn/SnO core-shell and nearly stoichiometric p SnO nano-needles are investigated via theoretical modeling and compared to the experimental results. For the pn photodiode fabrication, air annealing was adopted to form p nano-needle structured SnO (p ∼ 1021 cm−3 ) on n-type epitaxial Ge-on-Si (n ∼ 1017 cm3 ). The transmission and reflection spectra were measured using a UV–visible–IR spectrometer with an integrating sphere, and the nominal absorptance is derived as 1-transmittance-reflectance. To understand the scattering mechanism of the nanoneedles, the theoretical scattering (σ sc ) and absorption (σ abs ) cross-section spectra as well as the scattering patterns for Sn/SnO core/shell nanostructures on different substrates were calculated numerically using finite element method (COMSOL multiphysics 4.3a, Wave Optics module), similar to our previous work [14]. The dielectric functions of the substrates (bulk Si and Ge) [15], β-Sn [16], and SnO [17]] are utilized in the simulations. To the first-order approximation, we assume that the interactions between nano-needles are negligible, and that the scattering behavior of the nano-needles as a whole is simply the aggregate of the individual
Normalized Scattering Cross Section σsc
2604
1.0
on Ge Sn/SnO core/shell
0.9 nano-needle
on Si in Air
0.8 0.7 0.6
(b) 1500
2000
2500
Wavelength (nm)
Fig. 2. (a) Normalized scattering cross-section versus wavelength for a single Sn/SnO core/shell nano-needle in air and on the Si substrate, and a SnO nano-needle (without the Sn core) in air. The light is incident normally on the thin films, and the electric field is polarized along the nano-needle. (b) Scattering efficiency versus wavelength for Sn/SnO core/ shell structures in air, on Si, and on Ge substrates.
June 1, 2015 / Vol. 40, No. 11 / OPTICS LETTERS
100
(a)
90 80
Sn/SnO core-shell nano needle on DSP Si
70 60 50
Measured
40 30
Simulated 1000
1500
2000
Wavelength (nm)
Fig. 4. Scattered electric field Ex (along the nano-needle) at λ 1600 nm for (a) Sn/SnO core/shell nano-needle, and (b) SnO nano-needle on Ge. The light is incident along z direction and polarized along the x direction. The Sn core is outlined by the yellow rectangle in (a). The outer boundaries of the nano-needles are outlined by white rectangles.
2500
1.5 μm thick Ge absorption
There is negligible difference between the ASD of SnO nano-needle and that of its core-shell counterpart [Figs. 3(c) and 3(d)], as further confirmed by the scattered electric field (E x ) distribution in Fig. 4. However, the scattered field is more concentrated on the Sn core for the core/shell structure. Therefore, while the Sn core increases the scattering cross-section [Fig. 2(a)], it also tends to increase the ohmic losses. Overall, optimizing the size of the Sn core by controlled oxidation may help decrease the metal loss and improve the scattering efficiency without affecting the scattering pattern. In order to evaluate the light-trapping effect of the nano-needles for comparison with optical measurements, we calculated the trapped fraction (A) by integrating the trapped light over the solid angles greater than θc and normalizing to the total scattered power I sc . We also calculated the fraction of inward scattering within the loss cone (D), also normalized to I sc . Therefore, the total optical coupling into the semiconductor absorber is B A D. The out-coupling loss (C) is defined as the fraction of light scattered into air. Following the multiscattering method developed in Ref. [18], we obtain
the total nominal absorption, including both the absorbed light and trapped/guided light that cannot be collected by the integrating sphere. Figure 5(a) shows that this simulated nominal absorption agrees very well with the experimental data. Referring to Fig. 2, we find that the nominal absorption below the band gap of Si at λ > 1100 nm is due to the light trapping effect of the Sn/SnO core shell structure and the absorption of Sn. Especially, the ∼45% nominal absorption at λ > 2000 nm is mainly due to the light trapping effect since the scattering efficiency is >95% [Fig. 2(b)]. Therefore, nearly half of the incident optical power at λ > 2000 nm is trapped by the core-shell nano-needle structures. Considering the high scattering efficiency of SnOx nano-needles in the IR regime, it is very interesting to investigate their applications for light trapping in Ge thin-film photonic devices on Si. Because Ge can only absorb light at λ < 1900 nm due to the band gap constraint (∼0.66 eV), low-loss SnO nano-needles (without Sn cores) are more beneficial for light trapping in this case, while core/shell structures only offer high ηsc at λ > 1900 nm [Fig. 2(b)]. The Sn/SnO core/shell structure, on the other hand, can potentially provide high-efficiency light trapping for thin-film GeSn absorbers with a direct band gap corresponding to λ ∼ 2500 nm [19]. Figure 5(b) shows the simulated absorption spectra of 1.5-μm-thick, 0.2% tensile strained Ge film on Si without light-trapping structures (bare Ge-on-Si), with SnO nano-needles alone, and with both front-side SnO nano-needles and an ideal back-reflector. We assume no IR loss in SnO nanoneedles in the simulation. The absorption coefficients of 0.2% tensile strained Ge-on-Si are taken from the experimental data in Ref. [20]. With the SnO nano-needle light trapping, the absorption at 1560–1650 nm is enhanced by 2–5 × . We also found that the relative enhancement due to reduced reflection on SnO is only 30%, so most of the enhancement indeed comes from the light-trapping effect. When an ideal back-reflector is incorporated (e.g., a distributed Bragg reflector, DBR), the enhancement at 1600–1650 nm can be further increased to 10–30 × . Such a strong enhancement is because even the scattered light within the loss cone in Fig. 3 can now be reflected back. As the back-reflected light hits the frontside nano-needle structure again, it has another chance to be partially scatter outside the loss cone and guided by total internal reflection. Thus, the back reflector prevents Nominal Absorption (%)
Fig. 3. Polar plots of normalized angular scattering distribution of a single Sn/SnO core/shell nano-needle at λ 1600 nm: (a) in air, (b) on a Si substrate, and (c) on a Ge substrate. For comparison, the case for a SnO nano-needle is shown in (d). The light is incident vertically on the nano-needle. The black curves show the results for the polarization along the nanoneedle, while the red ones correspond to the polarization perpendicular to the nano-needle. The loss cones for Si/air and Ge/air interfaces are also shown (green lines), defined by the critical angles for total internal reflection.
2605
1.0 0.8 0.6
With SnO nano-needle +ideal back reflector
With SnO nano-needle
0.4 0.2
(b)
2.3x
15x
Bare Ge-on-Si
30x
Simulation
0.0 1450
1500
1550
1600
1650
Wavelength (nm)
Fig. 5. (a) Comparison of measured (blue line) and simulated (full red circle) nominal absorption spectra of Sn/SnO core/ shell nano-needles on a double-side polished Si substrate; (b) simulated absorption spectra of 1.5-μm-thick, 0.2% tensile strained Ge film on Si for 3 cases: without light trapping structures, with SnO nano-needles alone, and with both front-side SnO nano-needles and an ideal back-reflector.
OPTICS LETTERS / Vol. 40, No. 11 / June 1, 2015 0
-3
(a)
=10 cm
I (mA)
-2
2
10
10
0
-3
10
-4
10
-5
10
-6
10
-2
-1
10 Ideality factor -1 n=1.6 10 Series -2 10 resistance -3 Rs=2.4 kΩ 10 0 1 2
Voltage (V)
0.6
10 10
-2
Responsivity (A/W)
10
3
Junction area
2
-1
J (mA/cm )
10
Laser
(b)
Probe
40 0.4
p-SnOx n-Ge Si
Probe Laser 30 p-SnOx n-Ge Si
20
0.2
EQE(%)
2606
10
Incident power =50 μW Light on SnO 0 Light on Ge
0.0 0.0
-0.5
-1.0
-1.5
-2.0
Reverse Bias Voltage (V)
Fig. 6. (a) I-V/J-V curve of p SnO nano-needles/n-type Ge photodiode. The 1.5-μm-thick n-Ge film is grown epitaxially on Si substrate. The inset shows the ideality factor and series resistance. (b) Responsivity/EQE as a function of reverse bias when the λ 1580 nm laser shines on the SnOx region (green line) and the n-Ge region (red line) of the diode, respectively. The insets show the schematics of the corresponding configurations.
the backside leakage and allows the light to hit the front-side nano-needle scatterers multiple times, thereby significantly enhancing light trapping. To verify the theoretical model, we fabricate p nanoneedle structured SnO (155 nm thick) on 1.5-μm-thick n-type epitaxial Ge thin films on Si to form a pn photodiode. Note that this type II heterojunction facilitates the collection of photogenerated carriers. The SnOx layer was patterned by HCl to define the 10−3 cm2 junction region. The device is to demonstrate the concept of light trapping by SnO nano-needles without optimization/ metallization. The low temperature fabrication of p SnO (225°C) is also fully compatible with back-end-of-line (BEOL) CMOS processing (
