In Preparation for Submission to Thin Solid Films Date: 1-7-2010

Ultrathin TaOx Film Based Photovoltaic Device Pawan Tyagi, Bruce J. Hinds* Department of Chemical and Materials Engineering, University of Kentucky, Lexington, Kentucky-40506, USA Email: [email protected] Abstract: Application of the economical metal oxide thin-film photovoltaic devices is hindered by the poor energy efficiency. In this paper we investigated the photovoltaic effect with an ultrathin tantalum oxide (TaOx) tunnel barrier formed by the plasma oxidation of a pre-deposited tantalum (Ta) film. These ~3 nm TaOx tunnel junctions showed approximately 160 mV open circuit voltage and as high as 10% energy efficiency. Our ultrathin TaOx (~3 nm) could absorb approximately 12% of the incident light in 4001000nm wavelength range; this strong light absorbing capability was found to be associated with the dramatically large extinction coefficient. Spectroscopic ellipsometry revealed that extinction coefficient of 3 nm TaOx was ~0.2, two orders higher than that of stochiometeric Ta2O5. Interestingly, refractive index of this TaOx was comparable with that of stochiometeric Ta2O5. It was also found that heating and prolonged high-intensity light exposure deteriorated the photovoltaic effect in TaOx junctions. This study provides the basis to explore the photovoltaic effect in a highly economical and easily processable ultrathin metal oxide tunnel barrier or analogous systems. Key words: Photovoltaic cell; tantalum oxide; tunnel junction; atomic defects; p-n junction

1

Introduction The realization of thin film photovoltaic devices offer a lucrative option for the replacement of costly silicon based solar cells [1]. A thin film solar cell should be able to absorb sufficient radiation. More importantly, it should be endowed with a mechanism to separate the photo-energy created electron-hole pairs [1], akin to built-in potential [2, 3] in the p-n junction solar cells. Thin film metal oxide based photovoltaic cells [4] are promising candidate for the production of economical harvesting of solar energy [5]. Devices based on ~100 nm thick semiconducting cuprous oxide (Cu2O) exhibited a photovoltaic effect [5, 6]. The mechanism of charge separation in this case was attributed to the dissimilar metal contacts with different work functions [6] and the presence of atomic defects [7, 8]. The maximum ~2% solar cell efficiency was achieved with Cu2O based solar cells, which is well below the theoretically predicted value of 20% for the same system [9] . In order to enhance the light absorption generally a thick Cu2O (~100 nm) [8] was utilized. However, the thick Cu2O also possesses large population of atomic defects [7]; a higher defect density can annihilate the photo-generated electron-hole pairs before they are separated and transferred to the opposite electrodes[9].

A better approach to make the metal oxide based solar cells is to reduce the metal oxide thickness to few nm, yet managing to keep the light absorbance reasonably high. Gratzel cells utilized only a molecular monolayer to produce ~8 % energy efficiency [10], just a monolayer of dye molecules absorb required incident radiation. This observation provides the rational for attempting an ultrathin metal oxide layer; we hypothesized that when the number of photoactive defects within tunnel barrier become comparable with the number

2

of dye molecules in a Gratzel cell a photovoltaic effect can be observed. However, in the case of photovoltaic effect in tunnel junctions the mechanism of separating photogenerated electron–hole pair has to emerge either from the defect affected band structure [1, 11]. The tunnel junctions utilizing aluminum (Al) electrodes and 3 nm alumina (AlOx) have exhibited ~0.9 eV difference in the barrier heights at two Al/AlOx interfaces [1]. This study suggested that an inhomogeneous oxidation can alone produce significant band banding, similar to the difference between the two barrier heights observed in p-n junction solar cells [1, 11]. In our knowledge, no photovoltaic effect was reported with the Al/AlOx/Al system.

A tantalum oxide (TaOx) tunnel barrier is particularly promising to observe photovoltaic effect. X-ray photoelectron emission (XPS) studies of TaOx, grown by the thermal oxidation of Ta metal, exhibited the graded distribution of Tantalum (Ta) ions impurities [12]. The gradient of atomic defects within a TaOx is consistent with the oxygen diffusion model [13]. We conjectured that a graded profile of atomic defects will also appear from the plasma oxidation of pre-deposited Ta. Photoactive defects level [14] can produce a photovoltaic effect.

Plasma oxidation is easily controllable and produces ultrathin insulators at room temperature. We hypothesized that a high density of atomic defects at the optimum spatial location [14] within TaOx can absorb significant light radiation. While the gradient in defects density [15-17] can produce the difference in the barrier height between the two Ta/TaOx interfaces. Eventually, these unequal barrier heights will produce built-in electric

3

field for the separation of electron–hole pair; similar to the case of p-n junctions [1]. In this paper, primarily we investigated the photovoltaic effect in the ultrathin TaOx tunnel junctions. Secondly, we explored the properties of TaOx barrier using electrical and spectroscopic measurement as well. Finally, the device life and reliability of TaOx based photovoltaic cells were studied.

Experimental details: In this paper we have investigated a photovoltaic effect in TaOx tunnel junctions formed by the plasma oxidation of a pre-deposited Ta film [16], followed by the deposition of a thin Ta film as the top electrode (Fig.1). Thin top Ta film thickness was kept ≤12 nm to allow significant transmission of incident light; metallic film of comparable film thickness [18, 19] has been investigated as a transparent electrodes [20] in solar cells. The Ta/TaOx/Ta junction’s fabrication steps and their characterization are illustrated in Fig.1. Throughout this study a Si wafer with thermally grown 100 nm silicon oxide was utilized. Wafer pieces were sequentially washed in acetone, isopropyl alcohol and DI water for 2 min each and then were dried in the nitrogen flow. Photolithography was done to define the geometry of bottom Ta electrode (Fig. 1A). During photolithography following steps were performed in the given order: spin coating of Shipley 1813 positive photoresist at 3000 rpm for 30 sec, soft backing at 100 ºC for 1 min, exposure for 20 sec through photomask by Karl-Suss mask aligner, developing of photo resist with MF-319 developer solution for 1 min, and rinsing of the samples in DI water 5 times followed by drying in the nitrogen gas flow. The samples were then transferred to the load lock chamber of the sputtering machine (AJA Multitarget Sputtering System) using clean sample enclosures.

4

The sputtering of ~12 nm Ta was performed in the main chamber with 2x10-7 torr base pressure at 100 W DC gun power and 1 mtorr Ar (99.99% purity) pressure at RT. Within 5 min of the sample removal from the sputtering machine, it was soaked in Shipley-1165 resist remover solution for 30 min for liftoff, followed by rinsing in DI water and drying in the nitrogen flow.

The second photolithography step, following photolithography protocol used for the patterning of bottom Ta layer, (Fig. 1B) was performed. Next, samples were transferred to the sputtering machine. Duration for which the first Ta film remained outside of the sputtering system was 2 hrs (for 90 min in contact with various electronic grade solvents and for 30 min in the clean room environment).

The exposed segment of the bottom electrode through a photoresist layer was plasma oxidized to produce a ~3 nm TaOx insulating layer (Fig. 1C). Thickness of TaOx was confirmed by the spectroscopic ellipsometry on simultaneously processed TaOx sample. Plasma oxidation was performed at 20 W substrate bias and 60 mTorr pressure of 1:1 argon(Ar):oxygen(O) mixture for 60 seconds. After the oxidation step, purging was performed with Ar for 1 hr to replace the remaining oxygen from the main chamber. Then top ~12 nm thick Ta electrode was deposited using the above mentioned Ta deposition parameters, through the same photoresist window (Fig. 1D). Next, the liftoff step with Shipley 1165 resist remover produced a Ta/TaOx/Ta tunnel junction (Fig. 1E). The AFM study of tunnel junction showed that, Ta electrodes are 12±0.5 nm thick and possessed 0.2±0.1nm Rq roughness. We typically utilized TJs with ~25 µm2 area. Area smaller than

5

this was not reproducibly possible with our photolithography facility; larger area Ta/TaOx/Ta junctions, possessing higher defect density [21], turned out to be unstable[22]. In previous study [23], the defect induced instability disabled the scaling of TaOx electrical properties with area.

Results and discussion: As prepared Ta/TaOx/Ta tunnel junctions (TJs) exhibited an asymmetric tunneling characteristic (Fig. 2A), consistent with prior studies[12], and a photovoltaic effect (Fig. 2B). An asymmetric transport can be attributed to the inhomogeneous composition [24] of the TaOx tunnel barrier [12]. A typical tunnel junction showed a breakdown voltage of 1.8±0.2 V (Fig. 2C), consistent with earlier study [25].

Tunnel barriers produced by the plasma oxidation of a metal film are known to have a depleting amount of oxygen or an increasing amount of unoxidized metal atoms on traversing the thickness starting from the top side of TaOx. Depth wise XPS study of the TaOx [17] suggested the presence of stoichiometric Ta2O5 on the top surface but Ta+ rich suboxides closer to the bottom of TaOx. In the secondary mass ion spectroscopy analysis of an electrochemically grown TaOx, [15] inhomogeneous distribution of carbon impurities was detected. The present work aims to show the photovoltaic effect due to atomic defects without delving into defect specific details. However, Ta+ ions (n-type) [15] and O- ions (p-type) defects are expected to play the key role in the plasma oxidation produced TaOx [11]. Because of the presence of defects, previous transport studies of TaOx based tunnel junctions were found to be thermally activated [15, 26]. Pool-Frenkel

6

type thermally activated transport that generally occurred via defect induced energy levels in TaOx, has been widely observed [11]. A temperature dependent transport study of our tunnel junctions indicated a thermally activated process (Fig. 2D). It clearly indicates that the defects play an active role in barrier properties and transport [15]. Thermal activation energy calculated by using Arrhenius equation was found to be ~0.22 V.

To further establish the presence of atomic defects within the barrier capacitancevoltage (C-V) measurements were performed [27, 28]. The capacitance of TJs was found to be varying with frequency (Fig. 3). For instance, capacitance of a typical

tunnel

junction at 10 kHz and 1 MHz frequency was ~140 pF and ~38 pF, respectively; capacitance of a defect free tunnel junctions are invariant of frequency [29]. This study clearly suggests that TaOx barrier possesses defect states but do not point to the position of defects within the barrier [27]. To locate the position of defects we refer to previous XPS studies on analogous system [27]. It is also noteworthy that for stoichiometric Ta2O5 capacitor, with a 3 nm thickness and a 25 µm2 junction area, calculated capacitance is 3 pF. However, in the present case capacitance of TaOx is higher and can be attributed to the defect induced states. To investigate the photo-activity of defect states C-V was performed under light radiation. The capacitance values of our tunnel junction at 10 kHz in dark and in white light irradiation are, respectively, 140 pF and 180 pF. This study suggests that TaOx barrier with atomic defects is photoactive as observed in the case of solar cells [28]

To analyze the optical properties of ~3 nm TaOx spectroscopic ellipsometry, using GA Woolam ellipsometer, was performed. Extinction coefficient for the ~3 nm thick TaOx was

7

found to be 0.15 to 0.2 for a range of 400 to 1000 nm wavelength radiation (Fig. 4A); experimentally observed k for ~1 µm thick stoichiometric Ta2O5 was 1.73-0.11x10-3 [30]. It is noteworthy that stoichiometric Ta2O5 is transparent for a significant regime of visible range [31] (Fig. 4A). In our case, a two order higher extinction coefficient affirms TaOx’s strong tendency to absorb light. It is interesting to note that in our case plasma oxidation produced TaOx showed a spike in extinction coefficient around 400 nm and this agrees with the similar spike observed with stochimetric Ta2O5 [31]. Interestingly, the magnitude of refractive index for TaOx at ~500 nm wavelength is 2.3 which is in good agreement with the refractive index of stochiometeric Ta2O5 [30, 31] (Fig. 4B).

We also directly

studied absorbance on a ~3 nm TaOx deposited on the glass substrate (inset of Fig. 4C). Ultrathin TaOx absorbed ~ 12 % of the incident light.

Photoactive Ta/TaOx/Ta with rather asymmetric transport motivated us to probe the photovoltaic effect by conducting a transport study under the varying white light intensity. A calibrated white light source (MicroLite FL 3000) was kept at 5 inch above the Ta/TaOx/Ta junctions. A significant amount of photocurrent and as high as 160 mV open circuit voltage (Vo) was obtained, (Fig. 5A). Up to 10% energy efficiency was observed with Ta/TaOx/Ta tunnel junctions. The defects state in our tunnel junction is analogous to molecular monolayer in the dye sensitized solar cell or Gratzel. Dye molecules in Gratzel could yield ~8% energy efficiency [10, 30]. To enhance the numbers of dye molecule’s aerial density highly rough surface are utilized. However, in our case high density of atomic size defects can be accommodated within 3 nm TaOx film volume.

8

To calculate the energy efficiency, product of half of the Vo and half of the photocurrent density was divided by the incident light energy per junction area. It is clear from the data that high junction resistance produces a small fill factor [1]. We argue that photo-activity of tunnel junction is due to the planar area only and not due to any impurity or residue sitting along the perimeter of the junction. Exposed area along the perimeter is almost three orders smaller than the junction area. Using perimeter area in the efficiency calculation produced unreasonably large efficiency; energy produced turn out to be more than incidence energy. We also observed a tunnel barrier lost photo-activity after electrical breakdown; if photocurrent was due any other reason or artifact than tunnel barrier breakdown could not cease it. It was observed that the magnitude of energy efficiency did not remain constant with the change in tunneling area. An increase in tunnel junction area was concomitant with a decrease in energy efficiency. This decrease is presumably due to the enhancement of the interaction between atomic defects, presumably quenching of the photo-generated electron-hole pairs. Consistent with our observation, in the previous study also TaOx electrical properties were not observed to scale with area [23]. The efficiency of our tunnel junction also decreased with increasing light intensity (Fig. 5B). The open circuit voltage (Vo) steeply decreased with increasing temperature (Fig. 5C), and saturated with increasing light intensity (Fig. 5D). The mechanism behind the photovoltaic effect and associated observations (Fig. 5 B-D) is discussed elsewhere in this paper.

An elucidation of photovoltaic effect requires the understanding of TaOx band diagram [25]. Since TaOx is inherently a tunneling barrier we attempted to calculate barrier

9

properties by fitting the experimental transport data with Brinkman tunneling model [24]. Calculated barrier height and barrier thickness were found to be 2.7±1.2 nm and 1.2±1.3 V, respectively. These numbers are comparable with the barrier properties obtained from independent experiments and published literature. According to our ellipsometry and AFM studies the magnitude of barrier thickness is 3.0 nm. Barrier height of the ultrathin TaOx was found to be 1.7 V [25]. In our case large scattered in the barrier properties affirms the presence of defects within the TaOx barrier [24]; however, Brinkman tunneling model [24] do not yield a clear understanding of accurate band diagram because it does not incorporate impurities’ effect.

To explain photovoltaic effect with TaOx we tender following qualitative model (Fig. 6). After the plasma oxidation of bottom Ta, two types of defects, presumably O- and Ta+, are present. Due to the diffusion controlled movement of oxygen [13] in Ta, a gradient in the density of O- and Ta+ defects will exist. This will result in a similar decreasing density profile of O- and Ta+ ions, but in opposite directions (Fig. 6B). The XPS study of a thermally oxidized Ta film, higher density of Ta+ ions was found near bottom of the resulting oxide [17]. We surmise that oxidation produced TaOx barrier is akin to space charge region of a p-n junction [1]. The O- doped (p-type impurities) and Ta+ (n-type impurities) [22] produces p and n regions within TaOx. Due to a gradient in the density of p- and n- type dopants the ideal tunnel barrier band diagram of TaOx (Fig. 6C) will get modified to acquire band banding (Fig. 6D). The conduction band (C.B.) and valance band (V.B.) of O- and T+ rich regions of junction shifts to have same Fermi level through the barrier. Shifting in C.B. and V.B. produces a net built-in potential [1]. Due to the presence

10

of built-in potential, similar to popular p-n junction solar cells [1], light radiation generated electron-hole pair will be separated to opposite electrodes to produce net photovoltaic effect in Ta/TaOx/Ta junctions.

The efficiency of our TaOx based solar cells were found to be decreasing with increasing intensity of incident light radiation (Fig. 5B). The increasing light intensity failed to excite a large number of electron-hole pairs in the proportion of incident flux due to the limited photoactive defect states within the TaOx barrier. Consequently, the ratio of energy produced and energy incident on junction kept decreasing after threshold intensity. The open circuit voltage (Vo) was also observed to decrease with increasing temperature (Fig. 5C). To elucidate this observation it is necessary to recall that the transport through Ta/TaOx/Ta junction was found to be thermally activated [25]. The charge transport rate is an exponential functional of the temperature [15]. As temperature increases thermal energy will reduce the built-in potential barrier height. Thermal energy will work as the external forward bias at a p-n junction [1]. We also observed that Vo saturated once white light intensity increased beyond a threshold of ~4.0 mW/cm2, (Fig. 5D). A TaOx barrier is endowed with a definite density of photo-active defect states. The Vo is governed by the electric field due to the ionized defect states in the space charge regime; Vo will saturate with increasing radiation intensity once all the defect states are ionized.

The operational life of Ta/TaOx/Ta based thin film solar cell is strongly influenced by the heat and prolonged exposure to intense light irradiation. A Ta/TaOx/Ta tunnel junction started showing instabilities soon after heating beyond 70 ºC. Long exposure to intense

11

white light is expected to produce effect similar to heating; the broad band white light source also contained radiation in infrared range.

Conclusion In summary, ultrathin TaOx tunnel barrier exhibited photovoltaic effect; as high as 10% energy conversion efficiency and ~160 mV open circuit voltage was realized. Atomic defects within ultrathin TaOx were enhanced the light absorption capability. In-depth characterization of atomic defects within TaOx is recommended for the clear understanding of mechanism. Since ultrathin tunnel barrier are inherently delicate. Application of such system would need significant efforts to improve the stability and reliability. Photovoltaic effect in our TaOx deteriorated upon heating and after exposing to intense light radiation. One judicious approach may be to use thicker tunnel barrier and optimize the doping strategy to yield net photovoltaic effect and reasonably high fill factor. This work fosters the prospects of exploring other metal oxide to obtain photovoltaic effect as well.

Acknowledgements: The authors would like to thank the Air Force Office of Scientific Research (DEPSCoR) under agreement number F49620-02-1-0225, Kentucky Science and Engineering Foundation (KSEF-621-RDE-006 and KSEF-992-RDE-008). We thank Todd Hastings for the help with ellipsometry.

12

References:

[1] S.M. Sze, Physics of Semiconductor Devices, John Wiley & Sons, 1991. [2] B. Zimmermann, M. Glatthaar, M. Niggemann, M. Riede, A. Hinsch, Thin Solid Films 493/1-2 (2005) 170. [3] J.C. Bernede, J. Chil. Chem. Soc. 53/3 (2008) 1549. [4] A.O. Musa, T. Akomolafe, M.J. Carter, Sol. Energy Mater. Sol. Cells 51/3-4 (1998) 305. [5] L.O. Grondahl, Rev. of Mod. Phys. 5/2 (1933) 141. [6] B.P. Rai, Solar Cells 25/3 (1988) 265. [7] M. Nolan, S.D. Elliott, Phys. Chem. Chem. Phys. 8/45 (2006) 5350. [8] L.C. Olsen, F.W. Addis, W. Miller, Solar Cells 7/3 (1982) 247. [9] A. Mittiga, E. Salza, F. Sarto, M. Tucci, R. Vasanthi, Appl. Phys. Lett. 88/16 (2006) 2. [10] M. Toivola, J. Halme, K. Miettunen, K. Aitola, P.D. Lund, Int. J. Energy Res. 33/13 (2009) 1145. [11] P. Rottländer, M. Hehn, A. Schuhl, Phys. Rev. B 65/5 (2002) 054422. [12] E. Atanassova, G. Tyuliev, A. Paskaleva, D. Spassov, K. Kostov, App. Surf. Sci. 225/1-4 (2004) 86. [13] A.T.A. Wee, S.X. Wang, K. Sin, IEEE Trans. Magnetics 35/5 (1999) 2949. [14] A.M. Glass, D.V.D. Linde, D.H. Auston, T.J. Negran, J. Electron. Mater. 4/5 (1975) 915. [15] T. Hirai, H. Morita, M. Tasaka, Jpn. J. Appl. Phys. Part 1 - Regul. Pap. Short Notes Rev. Pap. 39/2A (2000) 657. [16] P. Rottlander, M. Hehn, O. Lenoble, A. Schuhl, Appl. Phys. Lett. 78/21 (2001) 3274. [17] E. Atanassova, D. Spassov, App. Surf. Sci. 135/1-4 (1998) 71. [18] B. O'Connor, C. Haughn, K.H. An, K.P. Pipe, M. Shtein, Appl. Phys. Lett. 93/22 (2008) 223304 [19] H.J. Lewerenz, K. Skorupska, M. Aggour, T. Stempel, J. Grzanna, J. Solid State Electrochemistry 13/2 (2007) 185. [20] J.C. Bernede, L. Cattin, M. Morsli, Y. Berredjem, Sol. Energy Mater. Sol. Cells 92/11 (2008) 1508. [21] C. Wang, L. Fang, G. Zhang, D.M. Zhuang, M.S. Wu, Thin Solid Films 458/1-2 (2004) 246. [22] J.D. Akerman, M.; Slaughter, J.M.; Dave, R.; Sun, J.J.; Martin, J.T.; Tehrani, S, IEEE Transaction on Magnetics 42/10 (2006) 26661. [23] P.H.P. Koller, W.J.M. de Jonge, R. Coehoorn, J. App. Phys. 97/8 (2005) 083913. [24] W.F. Brinkman, R.C. Dynes, J.M. Rowell, J. App. Phys. 41/5 (1970) 1915. [25] Y. Jeliazova, M. Kayser, B. Mildner, A.W. Hassel, D. Diesing, Thin Solid Films 500/1-2 (2006) 330. [26] E. Suzuki, D.K. Schroder, Y. Hayashi, J. Appl. Phys. 60/10 (1986) 3616. [27] N. Novkovski, Semicond. Sci. Technol. 21/7 (2006) 945. [28] V.P. Singh, J.C. McClure, Sol. Energy Mater. Sol. Cells 76/3 (2001) 369. [29] H. Shimizu, F. Sugeno, S. Nishimura, H. Endo, M. Hondac, Electrochemistry 72/11 (2004) 737. [30] C.C. Lee, D.J. Jan, Thin Solid Films 483/1-2 (2005) 130. [31] "GA Woolam Spectroscopic Ellipsometer Data Base."

13

Figure captions Fig. 1. Fabrication and characterization of Ta/TaOx/Ta junction: (A) Bottom Ta layer deposition; (B) after photolithography for the oxidation of controlled area of bottom Ta and top electrode deposition; (C) plasma oxidation of Ta layer followed by (D) deposition of the top Ta electrode; (E) schematic of the complete junction after liftoff; (F) SEM micrograph of a typical Ta/TaOx/Ta junction. Fig. 2. Transport characteristic of Ta/TaOx/Ta tunnel junction. (A) Asymmetric I-V (B) photovoltaic effect, (C) electrical breakdown and (D) thermally activated transport of Ta/TaOx/Ta junction. For graph (D) current were recorded at 100mV. Fig. 3. Capacitance versus voltage study of Ta/TaOx/Ta tunnel at 10 kHz and 1 MHz frequencies in dark and light at 295K. Fig. 4. Optical properties of ~3 nm TaOx: Comparison between (A) extinction coefficients (k) and (B) refractive index (n) between ~3 nm TaOx and stoichiometric Ta2O5; (C) % light absorbance of ~3 nm TaOx, deposited on quartz. Inset of (C) contain photograph of a glass slide covered with 3nm TaOx. Fig. 5. Photovoltaic effect at Ta/TaOx/Ta junction (A) Light intensity versus photo current and photovoltage (B) energy conversion efficiency versus light intensity. (C) Open circuit potential (Vo) versus temperature at 2mW/cm2 light intensity. (D) Vo versus intensity of incident light. Fig. 6. Band diagram of photo-active TaOx barrier: Schematic of TaOx barrier with O- and Ta+ impurities (A). Density gradient for atomic defects between two interfaces (B). Modification of ideal tunnel barrier profile (C) into p-n junction type band diagram due to the presence of p-type(O-) and n-type (Ta+) defects within TaOx barrier (D). Schematic presentation of photovoltaic mechanism in TaOx barrier (E).

14

Fig. 1

15

Fig. 2

16

Fig. 3.

17

Fig. 4.

18

Fig. 5.

19

Fig. 6.

20