Microelectronic Engineering 57–58 (2001) 737–748 www.elsevier.com / locate / mee
Thermal nano-probe a, a ,1 a ,2 b I.W. Rangelow *, T. Gotszalk , N. Abedinov, P. Grabiec , K. Edinger a
Institute of Technological Physics, IMA, University of Kassel, FB-18, Heinrich-Plett Str. 40, 34132 Kassel, Germany b University of Maryland, College Park, MD 20742, USA
Abstract The novel thermal probe presented here is based on the changes of the electrical resistivity of a nanometer-sized filament with temperature. The filament is integrated into an atomic force scanning probe piezoresistive type cantilever. Using a focused ion beam technique, the front end of the Al meander is cut through, forming an approximately 1-mm wide gap. Employing an electron beam deposition technique a sub-100 nm diameter Pt filament is deposited across the gap. The filament consists of an approximately 2-mm high loop with an additional spike deposited at the apex of the loop to improve spatial resolution. The new probe is an example on how a combination of CMOS technology, bulk and surface micromachining, focused ion beam technology and electron beam-induced deposition can be used to successfully fabricate unique nanoprobes. A spatial resolution of the order of 20 nm and a thermal resolution of 10 23 K is obtained. 2001 Elsevier Science B.V. All rights reserved. Keywords: Scanning thermal microscopy; Thermal nano-probe; Nanofabrication; Focused ion / electron beam technology
1. Introduction The progress of modern lithography has forced today’s IC devices technology to be in the range of 200 nm. In case of the electrical characterisation of electronic devices, analytical tools with highest lateral and thermal resolution have to be developed. Moreover, the thermal failure of future 70-nm node electronic devices may prove to be one of the significant problems due to the fast escalating package density and thus increasing energy dissipation. The very first thermal-sensitive probes consisted of a miniaturized thermocouple probe for both STM / SThM [1]. Owing to the fact that metal / metal contacts exhibit a rather small temperature sensitivity, a resistive thermal probe integrated with an AFM cantilever was proposed for this application [2]. In this article we present the
* Corresponding author. E-mail address:
[email protected] (I.W. Rangelow). 1 Permanent address: Institute of Microsystem Technology, Wroclaw University of Technology, ul. Janiszewskiego 11 / 17, 50-372 Wroclaw, Poland. 2 Present address: Institute of Electron Technology, 02-668 Warsaw, Al. Lotnikow 32 / 46, Poland. 0167-9317 / 01 / $ – see front matter 2001 Elsevier Science B.V. All rights reserved. PII: S0167-9317( 01 )00466-X
738
I.W. Rangelow et al. / Microelectronic Engineering 57 – 58 (2001) 737 – 748
efforts devoted to the development of a new miniaturized thermo probe sensor which belongs to the piezoresistive scanning probe sensors family developed at the University of Kassel [3–7]. All the sensors described here are fabricated based on advanced silicon micromachining and standard CMOS processing. The fabrication scenario presented in this article demonstrates the base for the production of different sensors with the same piezoresistive detection scheme. In this way we designed and fabricated, as a basic sensor, piezoresistive cantilevers for atomic force microscopy, which enables surface topography measurements with a resolution of 0.1 nm [3]. Next, by introducing a conductive tip isolated from the beam we obtained a microprobe for scanning capacitance microscopy and scanning tunnelling microscopy [4,5]. With this microprobe we measured capacitance between the microtip and the surface in the range of 10 222 F. Furthermore, a modification of the piezoresistors placement, based on the finite element method (FEM) simulation permits fabrication of the multipurpose sensor for lateral force microscopy, which enables measurements of friction forces with a resolution of 1 nN [6]. Using the same basic device idea and modified process sequence we manufactured a femtocalorimeter for the detection of heat energy in the range of 50 pJ [7]. In the following the fabrication technology and the principle of operation of a novel thermal probe sensor are presented.
2. Fabrication of the thermal nano-probe The fabrication process used in this work is a modification of a double-sidde silicon micromachining process developed for manufacturing of piezoresistive AFM sensors [1,2]. Double-sided polished k100l-oriented, 3–7-V cm silicon wafers are used as the starting material (Fig. 1). Next, standard CMOS processing like oxidation, phosphorus and boron diffusion ion implantation (Fig. 1, steps 1–5), dry and wet etching, insulator and aluminum film deposition and photolithography (Fig. 1, steps 6–8), are sequentially applied to form piezoresistors, p 1 diffusion connecting paths, contact windows, metallic connections and the connections to the micro-core at the front side of the wafer. In order to achieve deflection sensitivity of the cantilever better than 6 3 10 25 DR /R per nm, we redesigned the well-known piezoresistor formation technology. Taking into consideration that the highest stress caused by bending of the cantilever is concentrated on the surface, we employ a low-voltage (20 keV) boron implantation step and rapid thermal annealing at 8008C for 30 s. In this way, very shallow resistors are fabricated. Moreover, one has to note that the resistors are buried 50 nm under the surface. Under this condition with respect to the carrier scattering, surface states are not involved in the carrier transportation which effects improved stability of the piezoresistive sensor device and high sensitivity at the time. In the following back side processing sequence, a corner compensated membrane pattern is created by a two-sided photolithographic process and anisotropic deep etching with electrochemical etch stop of silicon in 10% TMAH solution at 708C, to create a 10-mm thick silicon membrane where the cantilever will be defined. Finally, the cantilever is defined in the membrane by a last photolithographic step at the top side of the wafer and silicon dry etching using ICP (inductively coupled plasma) with gas chopping of SF 6 and C 4 F 8 gases [8]. A 5-mm thick photoresist AZ4562 was used to mask the piezoresistive circuit and the resistive microheater during the dry etching. Fig. 1(1–8) shows the fabrication process sequences of the piezoresistive cantilever.
I.W. Rangelow et al. / Microelectronic Engineering 57 – 58 (2001) 737 – 748
Fig. 1. Fabrication process sequences of the piezoresistive cantilever.
739
740
I.W. Rangelow et al. / Microelectronic Engineering 57 – 58 (2001) 737 – 748
3. Nanoprobe integration For the fabrication of the nanoprobe a combination of focused ion beam and focused electron beam direct writing techniques were used. All fabrication steps were carried out in a combined scanning electron microscope / focused ion beam workstation (FEI 620 dualbeam). However, simultaneous use of both beams is not required and therefore the process could also be done sequentially in two separate systems. The process consists of three steps: (1) trimming of the cantilever beam; (2) cutting a gap into the aluminium leads; and (3) depositing the filament. For practical reasons the height of the deposited filament is limited to about 2–5 mm in the first step; the end of the cantilever was cut off using focused ion beam milling to increase tip clearance and allow for a larger tilt angle of the cantilever. This step brings the aluminium leads for the filament within 10 mm of the front edge of the cantilever. In this step a high ion beam current (10 nA) is used to cut through the entire thickness of the cantilever beam. Since a large amount of material needs to be removed, the process takes about 30–60 min. This time could be reduced by using a reactive gas such as chlorine during milling (gas-assisted etching [9]). While this step can be eliminated by changing the microfabrication design and moving the lead structure closer to the edge of the cantilever, it also demonstrates the flexibility of focused ion beam direct writing techniques in optimising prototype devices without the need to go through a time-consuming design cycle. In the second step a lower beam current of 100 pA is used to cut the aluminium leads (see Fig. 2). The width of the produced gap was usually in the range of 1–2 mm, but also sub-100 nm gaps can be milled using even lower beam currents (beam diameter 25 nm at 4 pA). In the third step the filament is deposited by electron beam deposition [10,11]. In this method the substrate is exposed to a flux of a precursor gas, using a thin gas feed tube placed in close
Fig. 2. Focused ion beam cut of Al-meander before electron beam induced nano-thermal probe deposition.
I.W. Rangelow et al. / Microelectronic Engineering 57 – 58 (2001) 737 – 748
741
proximity to the substrate and within the scanning field of the electron beam. The incident electron beam causes a fragmentation of the gas molecules leading to a deposit. Methylcyclopentadienyl dimethyl platinum was used as precursor gas, which leads to a conductive deposit consisting of platinum with a relatively high degree of carbon contaminants and a resistivity in the range of 900 V cm. Total deposition times for the filament structure were between 5 and 10 min. At these relatively long times the avoidance of sample drift with respect to the incident electron beam becomes extremely critical. To avoid electrical charging of the substrate during deposition a low resistivity connection of both metal leads to instrument ground was provided. Drift due to the introduction of a precursor gas flow was controlled by scanning a larger field (e.g., 10310 mm) with the electron beam and the gas valve open for about 1 min before switching to the electron beam deposition pattern. As discussed in the following, two different deposition strategies have been used to fabricate the filament structures. In both cases the final step was the electron beam deposition of a sharp needle (apex |20 nm) at the apex of the filament to increase the spatial resolution of the probe. In the first method straight deposits have been grown parallel to the direction of the electron beam by keeping the beam stationary at one spot during the process. In order to obtain a structure that bridges the gap between the two leads, the first leg of the structure was deposited at an angle between 25 and 308 by tilting the sample with respect to the beam. The second leg was then deposited at an angle of 225 to 2308 by rotating the stage by 1808 (see Fig. 3). Prior to deposition of the second leg, the location of the first leg was registered by acquiring an SEM image and the position of the second spot was placed next (|10 nm) to the apex of the first deposit. Deposition time for both legs is kept the same. In this scheme the beam has to be refocused for deposition of the second leg and obtaining a constant beam diameter for both deposits becomes crucial so that both legs grow a the same rate and eventually merge together at the apex. In cases where, due to drift or different focus, the ends of the two legs are separated by only a small distance (e.g., ,10 nm), they can be connected by scanning over the small area covering both ends with the gas flux on. Since material tends to deposit material preferentially at the ends, the additional deposit will eventually bridge the gap. The success rate for this process was found to be around 70% (see Fig. 4 for final structure). Although this is considerably lower than the second method described below, it has several advantages, which will be discussed later. In the second method, the sample remains perpendicular to the electron beam and the filament is produced by scanning the beam in small successive steps across the gap. The scanning pattern is controlled by a so-called ‘arch’, which defines the x and y position of the beam and the time the beam remains at a particular position (dwell time). Starting from two base points at each side of the gap, the beam is very slowly moved across the gap. Both sides of the legs are grown quasi simultaneously by switching position from one side to the other after each dwell time step. The shape of the structure is determined mainly by the dwell time for each point and best results have been obtained when the dwell time was decreased constantly for each successive point (e.g., from 0.3 to 0.04 s). After optimal values for the operating conditions (e.g., gas flux and beam focus) have been determined empirically, the process is very reproducible and is therefore better suited for producing a larger volume of nanoprobes (see Fig. 5 for final structure). However, in some circumstances the first method has some advantages, which need to be considered. First, since the beam is not scanned during deposition, unwanted deposition onto the substrate is limited to a small area at the base of the growing leg, reducing the risk of depositing a conducting surface layer across the gap. Second, the height and shape of the structure is more easily controlled by the deposition time and the tilt angle. Finally, because the leg is grown in successive steps, one can, in principle, deposit two different materials, thus creating a thermocouple structure.
I.W. Rangelow et al. / Microelectronic Engineering 57 – 58 (2001) 737 – 748
742
Fig. 3. Filament growth steps for stationary electron beam (method 1, see text).
4. Experimental In contrast to the thermocouple probes in our design a resistive thermal nanoprobe is integrated with a piezoresistive cantilever. The temperature of the thermal nanoprobe is observed by measuring the resistance of the probe. Generally the thermal nanoprobe can be applied in two measurement
I.W. Rangelow et al. / Microelectronic Engineering 57 – 58 (2001) 737 – 748
743
Fig. 4. Piezoresistive beam thermal nanoprobe (ThNP) fabricated by method 1.
modes. In the passive mode the probe temperature is monitored during scanning over the surface (the supply voltage of the sample is as small as possible in order to avoid self-heating of the sensor). In this technique the temperature difference between the probe and surface is observed. In the active mode the probe is heated by the current flowing through the sensor. The heat flow between the probe and the sample is influenced by the thermal conductivity of the sample and the temperature difference. The supply voltage of the thermal sensor is changed as necessary in feedback loop to maintain the sensor resistance (sensor temperature) constant. In our first experiments we measured the probe resistance at various temperatures. We observed the resistance changes from 7 up to 2 kV in the temperature range from 20 up to 808C (Fig. 6). Such a result was observed previously by Koops et al. [12]. In our measurement setup we connected the thermal nanoprobe with three passive resistors to form a bridge circuit. In this case the bridge output voltage can be calculated in the range of 2 mV/ K (in comparison to 20.05 mV/ K for the platinum resistor) (Fig. 7). The measurements of such small
744
I.W. Rangelow et al. / Microelectronic Engineering 57 – 58 (2001) 737 – 748
Fig. 5. Piezoresistive beam with thermal nanoprobe fabricated by scanned electron beam (method 2, see text).
voltages are subject to significant thermoelectric potentials from connections, the input offset voltage drifts due to the applied amplifiers and 1 /f noise. These problems may be reduced by increasing the bridge supply voltage however this is limited by the effect of self-heating of bridge resistance elements. Exciting the bridge with ac voltage and using synchronous detection eliminates the described problems. In passive mode the bridge is excited with the ac voltage with frequency off 10 kHz and amplitude of 10 mV (Fig. 8). The voltage Uth is recorded by the analogue / digital converters
I.W. Rangelow et al. / Microelectronic Engineering 57 – 58 (2001) 737 – 748
745
Fig. 6. Resistance of thermal nanoprobe versus temperature.
of the scanning probe microscope and corresponds with the difference of temperatures between the sample and probe. The feedback loop adjusts the dc voltage applied to the bridge so that the bridge remains balanced. Heat flux from the thermal probe is observed by monitoring the voltage and current applied to the circuit. Achieved spatial resolution is of the order of less then 80 nm and a thermal resolution of 10 23 K (Fig. 9).
Fig. 7. Output bridge signal of thermal nanoprobe versus temperature.
I.W. Rangelow et al. / Microelectronic Engineering 57 – 58 (2001) 737 – 748
746
Fig. 8. Measurement setup in passive mode of scanning thermal microscopy with thermal nanoprobe.
5. Summary We have developed a EBID Pt-based resistance thermal nano-probe which is integrated with piezoresistive silicon AFM-cantilever. The thermal nanoprobe can be used in two modes: passive (the temperature is monitored as it is scanned across the surface) and active (the probe induced heat flow to the sample which depends on the thermal conductivity of the substrate). In this case the thermal nanoprobe operates as a highly localised source of heat used for thermal analysis. The observed thermal contrast results from variations in thermal conductivity of the surface components or diffusivity. Extremely low thermal capacity of the probe allows performing of high frequency temperature imaging. Scanning nearfield thermal nano-probe allows observations of topography features in the range of 1 nm in the bandwidth of 1 kHz. The thermal interaction area between probe and sample determines the spatial resolution of the sensor. We achieved a spatial thermal resolution of the order of less then 80 nm and a thermal resolution of 10 23 K.
Acknowledgements The authors would like to acknowledge Hans W.P. Koops from Deutsche Telekom for helpful discussions of electron beam-assisted deposition. They also like to thank E. Lorenz for enthusiastic help in the preparation of the measurement experiments and SThM software development.
I.W. Rangelow et al. / Microelectronic Engineering 57 – 58 (2001) 737 – 748
Fig. 9. Temperature distribution on implanted resistor (scan field 2.532.5 mm).
747
I.W. Rangelow et al. / Microelectronic Engineering 57 – 58 (2001) 737 – 748
748
References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12]
C.C. Williams, H.K. Wickramasinghe, Appl. Phys. Lett. 49 (1986) 11387. R.B. Dinwiddie, R.J. Pylkki, P.E. West, Thermal Conductivity 22 (1994) 23. I.W. Rangelow, T. Gotszalk, P. Hudek, F. Shi, P.B. Grabiec, P. Dumania, SPIE 2879 (1996) 56. R. Linnemann, T. Gotszalk, I.W. Rangelow, P. Dumania, J. Vac. Sci. Technol. B 14 (1996) 856. T. Gotszalk, F. Shi, P. Grabiec, P. Dumania, P. Hudek, I.W. Rangelow, J. Vac. Sci. Technol. B 16 (6, Nov / Dec) (1998) 3948–3953. T. Gotszalk, P. Grabiec, I.W. Rangelow, Ultramicrscopy 82 (2000) 39–48. N. Abedinov, P. Grabiec, T. Gotszalk, Tz. Ivanov, J. Voigt, I.W. Rangelow, J. Vac. Sci. Technol. (A) 2001, submitted. B. Volland, F. Shi, P. Hudek, H. Heerlein, I.W. Rangelow, J. Vac. Sci. Technol. B 17 (6) (1999) 2768. R.J. Young, J.R.A. Cleaver, H. Ahmed, J. Vac. Sci. Technol. B 11 (2) (1993) 234. H.W.P. Koops, patent applied. H.W.P. Koops, J. Kretz, M. Rudolph, M. Weber, G. Dahm, K.L. Lee, Jpn. J. Appl. Phys. 33 (1994) 7099–7107. ¨ H.W.P. Koops, C. Schossler, A. Kaya, M. Weber, J. Vac. Sci. Technol. B 14 (6) (1996) 4105.
