Atomic Force Microscope nanolithography on titanium: Influence of the anodic voltage waveform on the formation of oxide nanodots

Atomic Force Microscope nanolithography on titanium: Influence of the anodic voltage waveform on the formation of oxide nanodots

Superlattices and Microstructures 44 (2008) 670–676 www.elsevier.com/locate/superlattices Atomic Force Microscope nanolithography on titanium: Influe...

805KB Sizes 0 Downloads 43 Views

Superlattices and Microstructures 44 (2008) 670–676 www.elsevier.com/locate/superlattices

Atomic Force Microscope nanolithography on titanium: Influence of the anodic voltage waveform on the formation of oxide nanodots Tae Young Kim a , Ermanno Di Zitti a , Davide Ricci a,b,∗ , Silvano Cincotti a a Dipartimento di Ingegneria Biofisica ed Elettronica, Universit`a di Genova, via Opera Pia 11a, 16145 Genova, Italy b Istituto Italiano di Tecnologia (IIT), via Morego 30, 16163 Genova, Italy

Available online 21 April 2008

Abstract We investigate the effect of different voltage waveforms on the growth of titanium oxide nanodots using Atomic Force Microscope (AFM) nanolithography. The resulting oxide features are compared by taking into account the current data detected during oxidation under the application of constant and linear ramp voltages. The experimental analysis of current waveforms during oxidation upon a constant bias voltage gives quantitative criteria to reduce space charge effects. The use of ramp voltages gives higher flexibility on the control of volume and aspect ratio of oxide features by varying the duty cycle. c 2008 Elsevier Ltd. All rights reserved.

Keywords: Atomic force microscope (AFM); Local anodic oxidation; Nanolithography; Titanium oxide nanodots

1. Introduction Since the application of Atomic Force Microscope (AFM) local anodic oxidation in patterning nanometer-scale oxide features on silicon and metals, efforts have been made to control the oxide growth by tuning the operational parameters on nano-oxidation such as relative humidity, exposure time, amplitude of bias voltage and AFM operation mode [1–5]. This technique has enabled fabrication of a variety of prototype nanodevices [6–9]. All these works pointed out that ∗ Corresponding author at: Dipartimento di Ingegneria Biofisica ed Elettronica, Universit`a di Genova, via Opera Pia 11a, 16145 Genova, Italy. Tel.: +39 3483053863. E-mail address: [email protected] (D. Ricci).

c 2008 Elsevier Ltd. All rights reserved. 0749-6036/$ - see front matter doi:10.1016/j.spmi.2008.02.012

T.Y. Kim et al. / Superlattices and Microstructures 44 (2008) 670–676

671

the buildup of space charge causes self-limiting growth, therefore this effect should be minimized to obtain better performance of the oxidation process and higher aspect ratio (height/width) of oxide features [10–14]. For this purpose, a voltage modulation technique was proposed, leading to significant enhancement of the growth rate and improvements in the aspect ratio of oxide features by changing voltage-pulse parameters such as modulation frequency, reset time and reset voltage [15–17]. The fact that oxidation kinetics is influenced by shaping the waveform voltage opens up the possibility of controlling the oxide growth by using other waveform voltages such as linear ramps. With this aim in mind, in this paper we consider a conductive AFM tip-induced oxidation of a Ti film deposited on SiO2 , under the application of different waveforms. We report on the main features of the fabricated TiO2 nanodots, as it is known from the literature [3,18] that Ti is anodized to TiO2 . We show how the use of different waveform voltages affects process control during oxide growth, providing higher flexibility in the oxide shaping. 2. Experimental All experiments of titanium oxide growth were carried out on samples prepared by evaporating a thin Ti film on a commercial p-type silicon wafer (1–10  cm) covered with ˚ of thermal SiO2 . The measured thickness of the Ti layer was 7 nm on average with a 1000 A 0.5 nm rms roughness and was obtained by using a quartz-crystal thickness and AFM imaging. All the AFM patterning and measurements were performed in contact mode at room temperature (26–28 ◦ C) over the range of 40%–44% of relative humidity, using a commercial AFM system (PSIA, XE-100). Tips used in our experiments were silicon cantilevers coated with 20–30 nm of TiN. The height of the tip was in the range of 10–15 µm with a typical curvature radius of 35 nm. The sample was mounted onto the sample holder and a conductive bridge was formed between the Ti surface and the sample holder with Ag paste. The tip and sample holder set up were electrically isolated from the AFM unit. The AFM feedback circuit was always activated in order to maintain a constant tip-sample distance during imaging and patterning. A Keithley 236 source measure unit, capable of generating various voltage waveforms with the desired timing and shape, was integrated with the AFM system for tip biasing and simultaneous collection of current versus time (I − t) data. The I − t signals were transferred to and processed on a PC using the LABVIEW program. Two types of AFM oxidations were performed according to the chosen waveform voltage, a constant bias voltage and a linear ramp voltage, as shown in Fig. 1(a) and (b), respectively. Once specific positions were selected within the scan area, current data before pulling down the tip to the sample surface and after the tip approach were recorded in order to calculate the oxidation current by subtracting the detected current data before oxidation from those recorded during oxidation, as described in our previous work [19]. 3. Results and discussion 3.1. Case 1: Constant bias voltage Bias voltages in the range from −8 to −10 V were applied in this experiment, since no appreciable morphology change was observed if the sample bias value was above −5 V. The value of the exposure time, Td = 500 ms, was chosen by looking at the time evolution of the oxidation current, in order to let the oxidation process reach the steady-state condition. In fact,

672

T.Y. Kim et al. / Superlattices and Microstructures 44 (2008) 670–676

Fig. 1. (a) Constant bias and (b) linear ramp voltage waveforms used in AFM oxidation experiments. Td : exposure time. V p : amplitude of bias voltage. R f : V p /T f (forward sweep rate), Rr : V p /Tr (reverse sweep rate), T f : forward sweep R duration, Tr : reverse sweep duration. TI (time integration) = V (t)dt. Duty cycle = T f /Td .

Fig. 2. (a) Detected current data during oxidation performed under constant bias voltage ranging from −8 to −10 V for Td = 500 ms. The regions where a different dynamical regime takes place are also indicated. (b) Height growth as a function of the exposure time. Dashed lines are best-fit logarithmic curves. (c) AFM images of titanium nanodots obtained for Td = 800 ms.

as displayed in Fig. 2(a), three distinct current regimes are observed during oxidation, namely an initial nonlinear regime followed by an intermediate transition and a final linear regime. The nonlinear regime can be associated with the first stage of the oxide growth. In this stage, both the electric field formation and the oxide growth resulting from the transport of the oxyanions to the surface take place simultaneously. During the transition regime, the rising current across the tip and the sample surface is suppressed to a large extent as the space charge builds up, thus limiting the growth of oxides. It can be seen from Fig. 2(a) that this regime develops after approximately 50 ms and tends to end up earlier as the bias voltage increases, reaching a steady-state condition. In this condition the oxidation current exhibits a linear behaviour. Therefore it is assumed that the last stage of the growth mechanism is associated with such a linear regime. The height growth with time for the same voltage amplitudes is shown in Fig. 2(b). A logarithmic law supports these kinetics, in agreement with what was reported for similar operating conditions [20]. However, a different consideration holds for the lateral oxide growth. As previously observed [19], in the steady-state condition of the oxidation process, the lateral growth is enhanced, thus leading to decrease of the aspect ratio of the oxidized structures, as can be seen in Fig. 2(c). The current data is allowed to establish the variation range of the exposure time in which the lateral growth is limited, thus obtaining an effective control parameter of the

T.Y. Kim et al. / Superlattices and Microstructures 44 (2008) 670–676

673

oxide growth under constant bias voltage. Moreover, this knowledge allows optimization of the use of the voltage modulation, which consists of repeated application of short voltage pulses so as to produce high aspect ratio oxide features [15–17]. The optimal duration of each pulse can be determined from the second stage of the current profile during dot oxidation [21]. 3.2. Case 2: Linear ramp bias voltage Various symmetric ramp types of bias voltages were applied by varying the exposure time Td from 50 ms to 1500 ms and the peak voltage V p from −10 to −15 V. Fig. 3(a) shows two captured current waveforms during oxidation as a function of time and sweep voltage. These current waveforms obtained with different exposure times are similar in shape but exhibit different features. During the forward sweep, the onset of the oxidation current in nanoampere range (such as the one previously observed in Fig. 2(a)), takes place when the sweep voltage reaches a certain value which is defined as threshold voltage. This value is close to the one found in the constant bias voltage case for long Td . At shorter Td , which corresponds to faster voltage sweep, a higher threshold voltage is observed and a lower peak current is detected. The Td dependency is also found on examining the temporal position of the peak current with respect to the voltage sweep. As Td becomes shorter, the temporal position of the current peak shifts from forward to reverse sweep region. This means that, at shorter Td , the current continues to increase for some time even in the reverse sweep where the voltage falls linearly. Evidence of this behaviour is given in Fig. 3(b), where a negative differential conductance dI /dV occurs for Td = 400 ms, while dI /dV is always constant for Td = 1500 ms. Fig. 3(c) shows the height and volume growth with time of oxide dots as a function of Td . It is found that the use of longer Td , which corresponds to slower sweep rates, produced dots with increased height and larger lateral thickness, resulting in greater volume features. As a first approximation, a logarithmic law supports these data, as previously found for the constant bias voltage case. As expected, using longer Td accounts for the increased growth of the fabricated nanodots (see Fig. 3(d)), which can also be related to the higher differential conductance with longer Td exhibited in Fig. 3(b). Hence, the dependency of the threshold voltage on Td , the temporal position of the current peak and the dI /dV behaviour are associated with the oxidation kinetics which led to the fabrication of the oxide dots. Interesting results were found by making use of asymmetric waveforms, as shown in Fig. 4. Even though both signals have the same TI and Td , different morphological oxide features were produced. Ramp voltages with faster forward sweep rate R f and slower reverse sweep rate Rr produced more voluminous oxide features with lower aspect ratio, while ramp voltages with faster Rr and slower R f produced less voluminous oxides with higher aspect ratio. This result can be explained in the following way. In the former case, the faster R f leads to high threshold voltage which, in turn, induces the shift of the current peak towards the reverse sweep region, as shown in Fig. 4 for type 1 waveforms, and the longer Rr provides sufficient time for oxidation so as to produce voluminous oxide features with lower aspect ratio. In the latter case, less voluminous oxide features with higher aspect ratio are obtained owing to the reduced duration of the oxidation time by the shorter Rr . It is worth noting that these data are aimed at showing the different behaviour of these waveforms; it is expected that the slight differences detected would become more remarkable if pulsed waveforms were used. 3.3. Estimation of waveform performance We estimated the waveform performance for growing oxide nanodots in terms of % volume and aspect ratio by changing the duty cycle defined as T f /Td . (see Fig. 1(b)). The % volume is

674

T.Y. Kim et al. / Superlattices and Microstructures 44 (2008) 670–676

Fig. 3. (a) Current versus time during oxidation performed under ramp voltage with V = −12 V for two exposure times (400 ms and 1500 ms), T f = Tr = Td /2, R f = Rr = V /T f . Solid lines represent the corresponding bias voltage waveform. (b) Differential conductance dI /dV plot of (a). (c) Height and volume of the oxide dots obtained with different exposure times. (d) AFM images of (c).

Fig. 4. (left) Current versus time curves during oxidation performed with two asymmetric ramp waveforms of bias voltage. (right) Volume and aspect ratio (height/width) of obtained oxide dots.

defined as the fractional percent of the volume of oxide dots fabricated using a ramp waveform with respect to the one obtained at constant bias voltage with the same TI and V p . As shown in Fig. 5, the highest % volume (70.7%) was obtained using the asymmetrical type 2 waveform, while the symmetric waveform, type 3, produced 63.7% of volume. The aspect ratio obtained using the type 4 waveform was higher than the one of type 6 (constant bias) and of type 3 (symmetric linear ramp) by 10% and 20%, respectively. Type 1 and type 5 waveforms are extreme cases with duty cycles equal to 0 and 1, respectively. In this way, the change of duty cycle which leads to the complementary change of R f and Rr with the same TI results in a ‘tuning’ effect for the control of oxide volume and aspect ratio. A bias voltage with a higher duty cycle can be tuned to fabricate higher aspect ratio oxides, while a lower duty cycle results in features with

T.Y. Kim et al. / Superlattices and Microstructures 44 (2008) 670–676

675

Fig. 5. Performance estimation of various voltage waveforms on oxide growth. (left) Aspect ratio and volume of oxide dots as a function of the applied waveforms whose parameters are reported in the table. (right) Characteristic parameters of the waveforms used in this experiment and fractional percent of the volume of oxide dots fabricated by a ramp waveform (type 1–type 5) with respect to the one obtained at constant bias (type 6).

higher volume. Nevertheless, it should be considered that the aspect ratio enhancement obtained with the type 1 waveform appears overestimated owing to the high standard deviation. Finally, it is worth mentioning that the use of a voltage modulation technique with pulsed asymmetrical waveforms might turn out to be a flexible nanolithographic strategy. If a high duty cycle waveform were to be used in a pulsed way, it may be expected to obtain oxide dots with higher volume. If instead, waveforms of low duty cycle were used, fabrication of oxide dots with high aspect ratio is expected. 4. Conclusions We investigated the influence of voltage waveform in AFM anodic oxidation on titanium by analyzing the current behaviour and the morphology of fabricated oxide nanodot features. In the case of constant bias voltage, three distinct current regimes are detected during oxidation, also making it possible to associate the time evolution of the morphology features of the growing oxide with the space charge buildup. Accordingly the exposure time, that is, the effective parameter that controls the oxide growth, can be chosen to minimize such space effects inside the grown oxide and make it possible to optimize the duration of single pulses in the application of voltage modulation. The application of the ramp bias voltage allows higher control of oxide nanodot features in terms of volume and height. Under the same time integration parameter TI, ramp waveforms with high duty cycles produce more voluminous oxide dots while ramp waveforms with low duty cycles produce oxide dots with enhanced aspect ratio. References [1] [2] [3] [4] [5] [6] [7] [8] [9]

D. Wang, L. Tsau, K.L. Wang, P. Chow, Appl. Phys. Lett. 67 (1995) 1295. P. Avouris, T. Hertel, R. Martel, Appl. Phys. Lett. 71 (1997) 285. C. Huh, S.J. Park, J. Vac. Sci. Technol. B 18 (2000) 55. H. Kuramochi, K. Ando, H. Yokoyama, Surf.Sci. 542 (2003) 56. N. Clement, D. Tonneau, B. Gely, H. Dallaporta, V. Safarov, J. Gautier, J. Vac. Sci. Technol. B 21( (2003) 2348. P.M. Campbell, E.S. Snow, P.J. McMarr, Appl. Phys. Lett. 66 (1995) 1388. U.F. Keyser, H.W. Schumacher, U. Zeitler, R.J. Haug, Appl. Phys. Lett 76 (2000) 457. K. Matsumoto, Y. Gotoh, T. Maeda, Appl. Phys. Lett. 76 (2000) 239. F.H. Chiu, F.C. Chiu, S.K. Fan, K.C. Tai, J.Y. Lee, Appl. Phys. Lett. 87 (2005) 243506.

676

T.Y. Kim et al. / Superlattices and Microstructures 44 (2008) 670–676

[10] J.A. Dagata, J. Itoh, K. Matsumoto, H. Yokoyama, J. Appl. Phys. 84 (1998) 6891. [11] E. Dubois, J.L. Bubendorff, J. Appl. Phys. 87 (2000) 8148. [12] J.A. Dagata, F. Perez-Murano, G. Abadal, K. Morimoto, T. Inoue, J. Itoh, H. Yokoyama, Appl. Phys. Lett. 76 (2000) 2710. [13] J.A. Dagata, et al., J. Appl. Phys 96 (2004) 2386. [14] J.A. Dagata, et al., J. Appl. Phys 96 (2004) 2393. [15] F.P. Murano, K. Birkelund, K. Morimoto, J.A. Dagata, Appl. Phys. Lett. 75 (1999) 199. [16] K. Unal, B. Aronsson, Y. Mugnier, P. Descouts, Surf. Interface Anal. 34 (2002) 490. [17] J.F. Lin, C.K. Tai, S.L. Lin, J. Appl. Phys. 99 (2006) 054312–1. [18] H. Sugimura, T. Uchida, N. Kitamura, H. Masuhara, J. Vac. Sci. Technol. B 12 (1994) 2884. [19] T.Y. Kim, D. Ricci, E. Di Zitti, S. Cincotti, Surf. Sci. 601 (2007) 4910. [20] Z. Shen, S. Hou, H. Sun, X. Zhao, Z. Xue, J. Phys. D: Appl. Phys. 37 (2005) 1357. [21] T.Y. Kim, D. Ricci, E. Di Zitti, S. Cincotti, Physica E (2007), doi:10.1016/j.physe.2007.08.085.