- Email: [email protected]
Micro force sensor with piezoresistive amorphous carbon strain gauge
Sensors and Actuators A 130–131 (2006) 75–82
Micro force sensor with piezoresistive amorphous carbon strain gauge E. Peiner a,∗ , A. Tibrewala a , R. Bandorf b , S. Biehl b , H. L¨uthje b , L. Doering c a
Technical University Carolo-Wilhelmina at Braunschweig, Institute for Semiconductor Technology, Hans-Sommer-Str. 66, D-38106 Braunschweig, Germany b Fraunhofer Institute for Thin Film and Surface Engineering, Bienroder Weg 54e, D-38108 Braunschweig, Germany c Physikalisch-Technische Bundesanstalt (PTB), Nano- and Micrometrology, Bundesallee 100, D-38116 Braunschweig, Germany Received 2 June 2005; received in revised form 16 September 2005; accepted 22 November 2005 Available online 20 January 2006
Abstract In this contribution we report for the first time on the successful integration of amorphous carbon (a-C) as a piezoresistive strain gauge into a silicon micro cantilever force sensor. Sputter-deposited a-C layers showing excellent tribological properties contain a percentage of nearly 20% of tetrahedral sp3 carbon bonds as observed by optical absorption and Raman spectroscopy. Temperature-dependent transport measurements revealed hopping conduction between conducting sp2 carbon sites embedded in the insulating skeletal matrix of sp3 bonds. Changing their distance by strain a change of resistivity could be expected, which was investigated with layers sputter-deposited on a silicon membrane and structured by the lift-off technique using photo resist. Cantilevers comprising a-C strain gauges were etched out of this membrane using tetra methyl ammonium hydroxide (TMAH) and potassium hydroxide (KOH) solutions in a bulk silicon micromachining process. Realised prototypes were tested by applying a variable load to the cantilever free end. We found linear characteristics of the strain gauge resistance versus the applied force in the range of 0 to ±600 N revealing piezoresistive gauge factors of a–C within 36–46. © 2005 Elsevier B.V. All rights reserved. Keywords: Amorphous carbon; Hopping conduction; Piezoresistive gauge factor; Bulk silicon micromachining; Micro force sensor
1. Introduction Diamond and hard carbon coatings are widely used to improve the micro tribological properties of micro electro mechanical systems (MEMS) [1-3]. Due its hardness, wear resistance as well as robustness to harsh environment diamond is an attractive material for many MEMS applications, e.g. micro grippers and atomic force microscope (AFM) probes. These micro components could be realised by selective deposition of polycrystalline diamond on silicon substrates and molds, respectively, using SiO2 as a sacrificial layer [4]. Furthermore, diamond is one of the most promising materials for highfrequency nano electro mechanical systems (NEMS) or surface acoustic wave (SAW) devices [5] since it exhibits the highest sound velocity of all semiconductors with 1.8 × 104 m/s as estimated by the square root of the Young’s modulus-to-densityratio.
∗
Corresponding author. Tel.: +49 531 3913761; fax: +49 531 3915844. E-mail address: [email protected] (E. Peiner).
0924-4247/$ – see front matter © 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.sna.2005.11.059
Piezoresistivity is an important property of materials to be used, e.g. for strain gauges in mechanical sensors. Boron-doped polycrystalline diamond (poly-C) grown on commercial polyC substrates exhibits a considerable piezoresistive effect with gauge factors K of a few hundreds and more than 4000 under inter- and intra-grain probing conditions, respectively [6]. In a recent study a best value of the piezoresistive gauge factor of 50 was reported for boron-doped poly-C on oxidised silicon substrates [7]. A poly-C position sensor integrated on a siliconbased cochlear implant probe has a gauge factor of 28 [8]. K depends on grain size and resistivity of the poly-C. Maximum values of 70–80 were obtained at around 102 cm [9]. Unfortunately, chemical vapour deposition (CVD) which is conventionally used for the growth of poly-C requires a substrate temperature of 500–900 ◦ C [7,8] which is much too high for substrates alternative to diamond or silicon like large-area plastics. Furthermore, it is not favourable for the deposition on micro-structured silicon. Correspondingly, low-temperature deposition, e.g. plasma-enhanced or -assisted CVD, of amorphous (a-Si:H) and microcrystalline (c-Si:H) silicon at low substrate temperature (<150 ◦ C) has recently found much atten-
76
E. Peiner et al. / Sensors and Actuators A 130–131 (2006) 75–82
Fig. 1. Schematic of micro force sensor (a) with integrated probing tip at the cantilever bottom side in an enlarged representation (b). Optical microphotographs of the amorphous carbon reference and strain gauge resistors are shown in (c) and (d).
tion [10,11]. An investigation of the piezoresistive properties of low-temperature c-Si:H on plastic substrates revealed gauge factors in the range of 10–40 [11]. For sputtered amorphous carbon (a-C) investigations on piezoresistivity are lacking. Therefore, in this contribution we report on the structural and electrical transport properties of sputtered a-C as a material for piezoresistive strain gauges in MEMS. Amorphous carbon is superior to heteroepitaxial and poly-C in terms of production cost and process requirements: high substrate temperatures and additional doping sources are not necessary. Compared to a-Si and c-Si:H it provides better tribological properties and robustness to harsh environment. In this study the piezoresistive properties of amorphous carbon strain gauges integrated on a bulk micromachined silicon cantilever (Fig. 1) are addressed for the first time. 2. Fabrication and characterization of amorphous carbon films Amorphous carbon films were deposited on silicon and SiO2 /silicon substrates by rf magnetron sputtering of a graphite target using a Balzers BAS 450 sputter plant. The substrate temperature during deposition was less than 150 ◦ C. A substrate bias of −100 V was maintained leading to superior tribological properties (high hardness, low wear and friction) of 0.25 m thick a-C layers, e.g. a hardness of 50 GPa [12]. Hardness in the range of a few tens of GPa can be expected for sputtered amorphous carbon with a content of tetrahedral-bonded carbon (sp3 ) of around 25% [13]. The hydrogen content in the films determined by secondary ion mass spectroscopy (SIMS) was less than 2%. More details on the a-C deposition process and the tribological testing were reported elsewhere [12].
The structural properties of 0.5 m thick a-C layers on silicon were investigated by spectroscopic ellipsometry and Raman spectroscopy. Fig. 2 shows the optical absorption coefficient α of a-C measured between 0.8 and 5 eV which can be used to determine the optical band gap. Constructing a plot of (αhν)1/2 versus the photon energy hν the band gap can be determined from the intercept ET of the extrapolated linear fit of Tauc’s relation to the measured data (Fig. 2, Tauc gap, [14]). Alternatively, the photon energy at which the absorption coefficient amounts to 104 cm−1 is often considered as a measure of the optical gap (E04 , [15]). From Fig. 2 we obtain values of 0.65 and 0.8 eV for the Tauc and the E04 gap, respectively. By comparison with collected data [15] we can estimate a content of conductive sp2 bonds of 80–90%. The Raman spectrum in Fig. 3 exhibits the typical G- and D-bands which can be assigned to the E2g in-plane stretching mode of graphite and to the A1g breathing mode of aromatic rings
Fig. 2. Absorption coefficient and Tauc plot [14], i.e. (αhν)1/2 in dependence on photon energy hν for a 0.5-m-thick a-C layer.
E. Peiner et al. / Sensors and Actuators A 130–131 (2006) 75–82
77
and σ(T ) = aT m ,
Fig. 3. Raman spectrum of a 0.5-m-thick a-C layer. The superimposed solid line corresponds to a fit to the measurement using an asymmetric Breit-WignerFano line shape for the G-band (dotted curve), a Lorentz function for the D-band (dashed curve) and a linear background (dash-dotted line) [17].
(graphitic oligomers) predicted at 1580 and 1360 cm−1 , respectively being in agreement with the expectation for sputtered a-C layers [16]. The solid line in Fig. 3 is a fit of a Breit-WignerFano line shape for the G-band (dotted curve), a Lorentz function for the D-band (dashed curve) and a linear background (dashdotted line) [17], revealing the respective mode wave numbers at 1600 ± 1 cm−1 (G-band) and 1388 ± 3 cm−1 (D-band). Note that due to the asymmetric Breit-Wigner-Fano line shape of the G-band the fitted value of the mode wave number does not correspond to the maximum peak position. From the G-band position a sp2 content of 83% is derived [17] being in accordance with the ellipsometry result as well as the excellent tribological properties. For amorphous carbon deposited on silicon by rf magnetron sputtering characterised by the van der Pauw/Hall technique p-type conduction is reported at a hole concentration of a few 1018 cm−3 [18]. In this study we performed temperaturedependent resistivity measurements within 20 and 300 K with a-C deposited on p-type, n-type and oxidised silicon substrates. Current blocking between layer and substrate is observed for a-C on p-silicon as well as on oxidised silicon. However, Ohmic behaviour was observed between a-C and n-silicon above 225 K while current blocking only occurred below 190 K. From the absence of diode behaviour we conclude that our a-C layers are of n-type conductivity. The hetero junction barrier between the a-C film and the silicon substrate is only effective at low temperatures. Conduction or phonon-assisted tunnelling between localised states are among the carrier transport mechanisms which are expected in sputtered a-C rather than thermally activated propagation through extended states [19]. Single- and multi-phononassisted hopping models describe the temperature dependence of conductivity in a wide range according to √ T0 1/4 VRH σ(T ) T = σ0 exp − T
(1)
(2)
respectively. To confirm this expectation conductivity measurements on a-C films on oxidised silicon were performed on square-shape samples with Ohmic contacts formed in a van der Pauw configuration. Fig. 4a and b show the measured temperature dependence of resistivity of an a-C film on oxidised silicon in representations suitable for either Eq. (1) (Fig. 4a) or Eq. (2) (Fig. 4b). At low probing current both models (straight lines) describe the experiment in the temperature range close to room temperature. A wider range of agreement is visible for the multi-phonon hopping model (Fig. 4b, Eq. (2)). As a result of the fitting we obtained σ0VRH = (1.4 ± 0.1) × 105 K1/2 /(cm) 1/4 and T0 = 26.3 ± 0.3 K1/4 from Fig. 4a and Eq. (1). For the multi-phonon hopping model (Eq. (2)) we found a = 0.012 ± 0.001 K1.25 /(cm) and m = 1.25 ± 0.01 from Fig. 4b. These values compare very well with the data reported for a-C fabricated at elevated substrate temperatures using carbon ions from a discharge in fullerene vapour [19]. We
Fig. 4. Conductivity of amorphous carbon between 20 and 300 K in representations exhibiting variable range hopping (VRH, Eq. (1)) and multi-phononassisted tunneling (Eq. (2)) between 120 and 300 K (a) and 60 and 300 K (b), respectively.
78
E. Peiner et al. / Sensors and Actuators A 130–131 (2006) 75–82
conclude that conduction in sputtered a-C is well described by the multi-phonon hopping process between localised states. Hopping conduction between conductive grains embedded in an insulating matrix is among the most representative processes of charge transport in piezoresistive composite material [20]. Considering a-C as a material which is composed of conductive clusters of sp2 bonds and a surrounding insulating sp3 matrix we can adopt the proposed theoretical model for the present study. If charge transport in a-C is governed by tunnelling between adjacent localised states which are separated by a mean distance d we can calculate the piezoresistive gauge factor, i.e. the ratio of relative resistivity change to strain using: 1 − χ/(1 − ν) 2d (1 + χ) ξ 1 − ν/(1 − ν) 2d Ktr = χ (1 + χ) ξ Kl =
(3)
where ξ is the carrier localisation length and ν denotes Poisson’s ratio. Longitudinal (Kl ) and transversal (Ktr ) gauge factors are considered via a phenomenological parameter χ, which takes into account a finite percentage of hops propagating not along but perpendicular to the direction of the applied electrical field. Isotropic behaviour is obtained at χ = 0.5. For the derivation of Eq. (3) rectangular-shaped resistors which are uniaxially strained with a contribution of the Poisson effect only in one perpendicular direction are assumed [20]. This configuration corresponds to the case of the a-C resistors integrated on a silicon micro cantilever depicted in Fig. 1. The -shaped a-C strain gauge is oriented with its both legs parallel to the cantilever axis. It is electrically contacted from the bottom side via heavily doped n-type lines (Fig. 1d). A deflection of the cantilever generates strain in the resistors R12 , R34 and R14 given by the resistances between the contacts 1 and 2, 3 and 4 and 1 and 4, respectively. R12 and R34 correspond to the legs of the -shaped strain gauge thus being under longitudinal strain. The perpendicularly oriented resistor R14 is under transversal strain.
were then deposited by e-beam evaporation and structured by lift-off. Finally, the cantilever was released wet chemically using potassium hydroxide solution (KOH, 30%, 60 ◦ C). We prefer KOH for this last etching step due to its lower mask undercut compared with TMAH. A protection of the metallisation was not necessary. Special protection of the a-C resistors was not required during the membrane as well as the final cantilever structuring since a-C was found not to be attacked by either TMAH or KOH solution. To confirm the excellent etching selectivity of a-C with respect to silicon a test structure consisting of 0.5 m thick aC was exposed to a typical 1 h etch attack by KOH solution (30%, 60 ◦ C). The scanning electron microphotographs in Fig. 5 show that the a-C layer was almost completely released from the silicon substrate without an indication of damage. Furthermore, a-C shows strong adhesion and low stress on silicon as well as on oxidised silicon, i.e. undercut of the integrated strain gauge resistors was not observed. Finally, the realised device was mounted into a connector. Voltage supply and current read-out was done by pressed contacts. Wire bonding was not necessary [21,22]. Ohmic characteristics were obtained for both the strain gauge and a reference resistor (Fig. 6a) which were connected via n+ lines and directly by the Au/Cr metallisation, respectively. The resulting resistances of 5.4 k for the strain gauge resistor and 80 k for the reference resistor are in close agreement with the design. We conclude that below 1 V the n+ connecting lines do not affect the resistance measurement. I–V characteristics of the strain gauge
3. Sensor fabrication and characterisation The tactile force sensor reported here is based on a silicon micro cantilever with length, width and thickness designed to 5 mm, 0.25 mm and 50 m, respectively. Sensor prototypes were realised using a bulk silicon micromachining process based on standard photo-lithography, thermal oxidation and wet etching [21,22]. The process was started by etching of a p-type silicon wafer using tetra methyl ammonium hydroxide solution (TMAH, 20%, 80 ◦ C) through a mask of thermal oxide to obtain a membrane structure. By diffusion of phosphorus from a spinon silica emulsion (Merck, Siodop) n+ -doped contact lines were fabricated. Subsequently, a field oxide was grown and contact holes were opened. Amorphous carbon (0.5 m) was deposited by magnetron sputtering and structured into resistors between the n+ contact lines by lift-off using photo resist. In a subsequent second TMAH etching step the silicon membrane was thinned to its final thickness. Contact holes were opened to the n+ -doped regions and gold/chromium (300 nm/30 nm) connecting lines
Fig. 5. Amorphous carbon test structure on silicon substrate illustrating the excellent etching selectivity of a-C with respect to silicon using KOH solution.
E. Peiner et al. / Sensors and Actuators A 130–131 (2006) 75–82
79
Fig. 7. Resistance change of the longitudinally strained gauge resistor R34 (Fig. 1d) of a cantilever-type micro force sensor (#1,Table 1) upon gradual increase of forces applied to the cantilever free end at the top (−F) and the bottom surface (+F), i.e. R34 is under tensile and compressive stress, respectively.
For the determination of the piezoresistive gauge factor K we consider the cantilever along its axis and assume that the force is applied at its free end, i.e. at the distance leff from the position of the strain gauge resistor:
Fig. 6. Current–voltage characteristics of amorphous carbon strain gauge and reference resistors at 300 K (a) and in the temperature range between room temperature and 150 ◦ C (b). The temperature dependence of conduction can be modelled by an Arrhenius law as depicted in the inset.
resistor were measured in the temperature range from room temperature to 150 ◦ C (Fig. 6b). A field-dependent increase of current is visible above around 1 V while below Ohmic behaviour was found. In the considered temperature range conductivity exhibits a slight temperature dependence which can be described by an Arrhenius model (see inset). By fitting we find a small activation energy of Ea = 31 meV. The sputter-deposited a-C strain gauges on silicon micro cantilevers were investigated by deflecting the cantilever and measuring the resistance change. In numerous calibration runs the cantilevers were incrementally moved with its free end against a force measuring probe mounted on the stamp of a high-resolution compensation balance (Sartorius SC2) [21,22]. Movement of the cantilever was performed at a resolution of 1 nm and a reproducibility of 5 nm. The compensation balance offers a resolution of 1 nN and a reproducibility of 2.5 nN. For the complete setup a temperature drift of less than 10 mK/h was maintained. The resistance was measured using a digital multimeter (Prema DMM 5017) at a current of 10 and 100 A, respectively. In Fig. 7a typical load–resistance curve is shown. Longitudinal compressive and tensile strain is generated in the strain gauge resistor R34 by probing the bottom and the top surfaces of the cantilever, respectively (Fig. 8). We can assume nearly uniform strain across the strain gauge resistor which is designed with its length (100 m) and height (0.5 m) being small compared with the cantilever length (5 mm) and height (50 m), respectively.
R 3 K leff =− F R 2 E weff h2
(4)
with the relative change of resistance R/R, the cantilever height h and the applied force F. Owing to the sidewalls of the cantilever [(1 1 1) crystal planes] which are inclined to the surface [(0 0 1) plane] by 54.7◦ an effective width weff = w + h/21/2 of the cantilever is taken into account. Furthermore, the undercut of the frame below the cantilever clamping leading to a not ideally fixed clamping is considered by an effective length leff exceeding its nominal value [22]. The cantilever thickness can be controlled
Fig. 8. Schematic of the cantilever-type force sensor under application of forces to the cantilever free end from bottom (+F) and top (−F) directions generating compressive and tensile strain, respectively, at the cantilever surface, i.e. in the gauge resistors.
80
E. Peiner et al. / Sensors and Actuators A 130–131 (2006) 75–82
Table 1 Characteristics of three sensor prototypes and calculated gauge factors K of integrated a-C strain gauges h (m)
Resistor/strain
R/R/F (N−1 )
K
R/R/T (K−1 )
9.7 ± 0.1
52 ± 2
R34 /tensile R34 /compressive
3.06 ± 0.05 3.13 ± 0.05
45.1 ± 3.4 46.2 ± 3.4
– –
#2A
10.7 ± 0.1
54 ± 2
R34 /compressive R14 /tensile R14 /compressive
2.66 ± 0.19 2.4 ± 0.05 2.29 ± 0.06
42.2 ± 3.2 37.9 ± 2.8 36.2 ± 2.7
8.1 × 10−3 7.9 × 10−3
#2B
13.1 ± 0.1
57 ± 2
R12 /tensile R12 /compressive R34 /tensile R34 /compressive R14 /tensile R14 /compressive
2.41 2.48 2.31 2.33 2.21 2.20
± ± ± ± ± ±
6.5 × 10−3
Sensor #1
k (N/m)
± ± ± ± ± ±
0.08 0.08 0.09 0.08 0.06 0.05
44.3 45.5 42.5 42.7 40.6 40.4
3.2 3.4 3.2 3.2 3.1 3.0
6.5 × 10−3 6.0 × 10−3
Rij denotes the resistors between the contacts i and j (Fig. 1d), i.e. R12 and R34 are under longitudinal strain and R14 is under transversal strain. By probing the cantilever from the bottom and top sides uniform compressive and tensile strain, respectively, is generated in the resistors. Each of the figures of sensitivity and gauge factor corresponds to the average of at least 40 measurements. The strain gauges of sensors #1 and #2 (A and B) were analysed at probing currents of 100 and 10 A, respectively.
using the measured stiffness k of the cantilever according to: 4k h = leff 3 (5) weff E At leff = 5.4 ± 0.1 mm, weff = 260 ± 10 m, k = 9.7 ± 0.1 N/m and E = 170 GPa we find h = 52 ± 2 N/m for the cantilever considered in Fig. 7. We observe linear resistance–force characteristics as shown in Fig. 7 yielding a piezoresistive gauge factor of K = 44.7 using Eq. (4) and the given cantilever geometry. In Fig. 9 the probability distribution of the gauge factor is displayed obtained for the above micro force sensor prototype after collecting the results of 195 measurements under compressive strain. We find an average value of −0.01727 ± 0.00026 /N corresponding to a gauge factor of K = 46.2 ± 3.4. Under tensile strain of the a-C resistor we obtain a sensitivity of −0.01685 ± 0.00025 /N corresponding to a gauge factor of K = 45.1 ± 3.4. We assign the scatter given by a Gaussian fit to the measured probability distribution mainly to the temperature drift of our calibration set-up. We find temperature coefficients of
resistivity of (6–8) × 10−3 K−1 leading to a force drift of the a-C strain gauges of 2.6–3.5 mN/K. Signal-to-noise ratio is found to be around 105 at a probing current of 100 A corresponding to a minimum detectable force in the range of few N. Full-bridge strain gauge designs which are possible on surfaces with tensile and compressive strain regions, e.g. membranes, will offer an additional increase of sensitivity (factor-of-four), improved drift stability and a sensor output convenient to standard bridge amplifiers. The results of numerous calibration runs with three prototypes of force sensors are collected in Table 1. The cantilever thicknesses determined by load-deflection measurements and Eq. (5) are within 50–60 m. Force sensitivity measured for the resistors R12 , R14 and R34 arranged between the contacts 1–2, 1–4 and 3–4, respectively (cf. Fig. 1), ranges from 2.2 to 3.1 N−1 . Using these values we calculate K-factors of 36–46. These values compare well with published values of 10–80 obtained for boron-doped polycrystalline diamond on silicon [7–9]. A slight difference between the longitudinally strained resistors (R12 , R34 ) and the resistor R14 which is under transversal strain was detected. Taking the average of all measured values for Kl and Ktr we find χ = 0.484 (Eq. (3)) assuming a Poisson ratio of ν = 0.3. The ratio of the hopping distance between localised conductive sites to the localisation length d/ξ of carriers can be estimated to be around 100. 4. Conclusions
Fig. 9. Probability distribution of sensitivity of a cantilever-type micro force sensor comprising an a-C strain gauge resistor (#1, Table 1, R34 ) under longitudinal compressive strain. The results of 195 measurements are collected.
Amorphous carbon was investigated as a functional material for MEMS technology, i.e. for piezoresistive strain gauges of bulk micromachined silicon cantilever sensors. We found that amorphous carbon is a promising material since it combines favourable mechanical properties like adhesion, hardness and wear resistance with a large piezoresistive effect. It could be deposited on thin silicon membranes maintaining low substrate temperatures (<150 ◦ C). As a consequence it could be easily structured by lift-off using photo resist. Furthermore, amorphous carbon was found to be inert in anisotropic wet etchants like
E. Peiner et al. / Sensors and Actuators A 130–131 (2006) 75–82
KOH and TMAH. Therefore, special protection of amorphous carbon on silicon and silicon oxide resistors during membrane and cantilever etching is not necessary. Extended calibration runs with cantilever prototypes yielded linear characteristics between the applied load and the amorphous carbon strain gauge resistances. Piezoresistive gauge factors were determined at compressive and tensile strain as well as under longitudinal and transversal probing conditions. We found values of K between 36 and 46 almost independent of strain and probing directions. Acknowledgements The authors are grateful to Doris R¨ummler for valuable assistance during the sensor fabrication and SEM characterisation as well as to Georgi Ginev and Wolfgang Limmer for ellipsometry and Raman measurements, respectively. Funding by the German Federal Ministry of Economics and Labour (BMWA) is gratefully acknowledged.
81
[16] N.A. Hastas, C.A. Dimitriadis, Y. Panayiotatos, D.H. Tassis, S. Logothetidis, D. Papadimitriou, G. Roupakas, G. Adamopoulos, High-field transport and noise properties of sputter-deposited amorphous carbonsilicon heterostructures, Semicond. Sci. Tech. 17 (2002) 662–667. [17] Y. Perera, J. Gottmann, A. Husmann, T. Klotzb¨ucher, E.W. Kreutz, R. Poprawe, PLD of hard ceramic coatings, Laser Applications in Microelectronic and Optoelectronic Manufacturing VI, Proc. SPIE 4274 (2001) 66–77. [18] N.A. Hastas, C.A. Dimitriadis, Y. Panayiotatos, D.H. Tassis, P. Patsalas, S. Logothetidis, Noise characterization of sputtered amorphous carbon films, J. Appl. Phys. 88 (2000) 5482–5484. [19] E.B. Maiken, T. Taborek, Amorphous carbon films deposited from carbon ions extracted from a discharge in fullerene vapor, J. Appl. Phys. 87 (2000) 4223–4229. [20] C. Grimaldi, P. Ryser, S. Str¨assler, Gauge factor of thick film resistors: outcomes of the variable range hopping model, J. Appl. Phys. 88 (2000) 4164–4169. [21] I. Behrens, L. Doering, E. Peiner, Piezoresistive cantilever as portable micro force calibration standard, J. Micromech. Microeng. 13 (2003) 171–177. [22] E. Peiner, L. Doering, Force calibration of stylus instruments using silicon microcantilevers, Sensor. Actuat. A 123–124 (2005) 137–145.
Biographies References [1] W. Wang, Y. Wang, H. Bao, B. Xiong, M. Bao, Friction an wear properties in MEMS, Sensor. Actuat. A 97–98 (2002) 486–491. [2] I.S. Forbes, J.I.B. Wilson, Diamond and hard carbon films for microelectromechanical systems (MEMS)—a nanotribological study, Thin Solid Films 420–421 (2002) 508–514. [3] M.R. Werner, W.R. Fahrner, Review on materials, microsensors, systems and devices for high-temperature and harsh-environment applications, IEEE T. Ind. Electron. 48 (2001) 249–257. [4] T. Shibata, Y. Kitamoto, K. Unno, E. Makino, Micromachining of diamond films for MEMS applications, J. Microelectromech. S. 9 (2000) 47–51. [5] W.I. Milne, Electronic devices from diamond-like carbon, Semicond. Sci. Tech. 18 (2003) S81–S85. [6] S. Sahli, D.M. Aslam, Ultra-high sensitivity intra-grain poly-diamond piezoresistors, Sensor. Actuat. A 71 (1998) 193–197. [7] A. Yamamoto, T. Tsutsumoto, Piezoresistive effect of CVD polycrystalline diamond films, Diam. Relat. Mater. 13 (2004) 863–866. [8] Yuxing Tang, Dean.M. Aslam, Jianbai Wang, Kensall.D. Wise, Technology and integration of poly-crystalline diamond piezoresistive position sensor for cochlear implant probe, in: Proceedings of the 13th International Conference on Solid-State Sensors, Actuators and Microsystems (Transducers ’05), Seoul, Korea, 2005 June 5–9, pp. 543–546. [9] M. Werner, R. Locher, Growth and application of undoped and doped diamond films, Rep. Prog. Phys. 61 (1998) 1665–1710. [10] G. de Cesare, M. Gavesi, F. Palma, B. Ricco, A novel a-Si:H mechanical stress sensor, Thin Solid Films 427 (2003) 191–195. [11] P. Alpuim, V. Chu, J.P. Conde, Piezoresistive sensors on plastic substrates using doped microcrystalline silicon, IEEE Sensor. J. 2 (2002) 336–341. [12] R. Bandorf, H. L¨uthje, C. Henke, J.-H. Sick, R. K¨uster, Tribological behaviour of thin a-C and a-C:H films with different topographic structure under rotating and oscillating motion for dry lubrication, Surf. Coat. Tech. 188–189 (2004) 530–533. [13] W. Lu, K. Komvopoulos, Effect of stress-induced phase transformation on nanomechanical properties of sputtered amorphous carbon films, Appl. Phys. Lett. 82 (2003) 2437–2439. [14] J. Tauc, R. Grigorovici, A. Vancu, Optical properties and electronic structure of amorphous germanium, Phys. Status Solidi 15 (1966) 627–637. [15] J. Robertson, Electronic and atomic structure of diamond-like carbon, Semicond. Sci. Tech. 18 (2003) S12–S19.
Erwin Peiner, (Priv.-Doz.) born in Bad M¨unstereifel, Germany, in 1960, received the Diplom-Physiker and PhD degrees, both in applied physics, from the University of Bonn, Germany in 1985 and 1988, respectively. In 1989 he joined the Institut f¨ur Halbleitertechnik (IHT) of the Technical University Braunschweig, where he was involved in the monolithic integration of III/V compound semiconductors on silicon for MOEMS. In 2000, he received the venia legendi for semiconductor technology from the Faculty of Mechanical and Electrical Engineering of the Technical University Braunschweig owing to a habilitation thesis on silicon sensors for condition monitoring of machines and plants. Presently, he is in charge of the semiconductor sensors group at the IHT. His main interests are in the field of micromachined sensors for industrial applications combining silicon with novel materials (like DLC, ZnO, SiC). Arti Tibrewala, (MSc.) born in Aurangabad, India, in 1979, received the degree of Industrial Engineering, from Marathwada University, Aurangabad, India in 2000. She got her Master of Science in Electrical Engineering from University of South Florida, Tampa, USA in 2002. She joined Institut f¨ur Halbleitertechnik of the Technical University Braunschweig in 2003, where she is involved with characterisation of diamond like carbon films. Ralf Bandorf, (Dr.-Ing. Dipl.-Phys.) was born in 1973 and studied at the Friedrich-Alexander-University Erlangen-Nurenberg. His doctoral thesis at the Fraunhofer Institute for Surface Engineering and Thin Films was focused on ultrathin tribological systems for electromagnetic microactuators. Since December 2001 he is employed as a project leader at the IST in the micro and sensor technologies group. His work is dealing with smart coatings, integrated sensors, microtribological films, and electrical and magnetic functional coatings. Saskia Biehl, (Dipl.-Ing.) was born in 1971. She received the degree of a graduate engineer in Material Science of the University of Saarbr¨ucken in 1999. In 2000 she worked as an engineer for quality assurance at a glass coating company in Germany. Since 2001, she is a project manager at the Fraunhofer Institute for Thin Film and Surface Engineering working in the field of temperature, force and biological micro sensors as well as micro structuring. ¨ Holger Luthje, (Dipl.-Ing.) was born in 1944 and studied physical technology in L¨ubeck. During 21 years industrial work in a multinational company in the field of electronics he was in-charge of different projects dealing with electrical and magnetic coatings, sensors, as well as sub-micron patterning techniques for mega-chips. Included were also technologies for the manufacturing of micromechanical components. Since 1991 he is working at the
82
E. Peiner et al. / Sensors and Actuators A 130–131 (2006) 75–82
Fraunhofer Institute for Surface Engineering and Thin Films in Braunschweig, where he is in-charge of the group micro and sensor technology. Actual work focuses on the research and development of smart coatings with integrated sensors, microtribological coatings for MEMS applications, microstructuring and molding technologies, as well as different electrical coatings. Further main topics are the development of magnetic coatings, especially thin film sensors for industrial and automotive applications.
Lutz Doering, (Dr.-Ing.) born in Salzwedel, Germany, in 1960, works in the PTB in Braunschweig since September 2001 in the work group of microsystem metrology. His graduation in the area of microelectronics technology, precision instrument technology and computer science he reached 1990 at the Technical University of Dresden. In the centre of his presently work are measuring procedures for the characterisation of micro-technically manufactured micro force sensors.
Micro force sensor with piezoresistive amorphous carbon strain gauge E. Peiner a,∗ , A. Tibrewala a , R. Bandorf b , S. Biehl b , H. L¨uthje b , L. Doering c a
Technical University Carolo-Wilhelmina at Braunschweig, Institute for Semiconductor Technology, Hans-Sommer-Str. 66, D-38106 Braunschweig, Germany b Fraunhofer Institute for Thin Film and Surface Engineering, Bienroder Weg 54e, D-38108 Braunschweig, Germany c Physikalisch-Technische Bundesanstalt (PTB), Nano- and Micrometrology, Bundesallee 100, D-38116 Braunschweig, Germany Received 2 June 2005; received in revised form 16 September 2005; accepted 22 November 2005 Available online 20 January 2006
Abstract In this contribution we report for the first time on the successful integration of amorphous carbon (a-C) as a piezoresistive strain gauge into a silicon micro cantilever force sensor. Sputter-deposited a-C layers showing excellent tribological properties contain a percentage of nearly 20% of tetrahedral sp3 carbon bonds as observed by optical absorption and Raman spectroscopy. Temperature-dependent transport measurements revealed hopping conduction between conducting sp2 carbon sites embedded in the insulating skeletal matrix of sp3 bonds. Changing their distance by strain a change of resistivity could be expected, which was investigated with layers sputter-deposited on a silicon membrane and structured by the lift-off technique using photo resist. Cantilevers comprising a-C strain gauges were etched out of this membrane using tetra methyl ammonium hydroxide (TMAH) and potassium hydroxide (KOH) solutions in a bulk silicon micromachining process. Realised prototypes were tested by applying a variable load to the cantilever free end. We found linear characteristics of the strain gauge resistance versus the applied force in the range of 0 to ±600 N revealing piezoresistive gauge factors of a–C within 36–46. © 2005 Elsevier B.V. All rights reserved. Keywords: Amorphous carbon; Hopping conduction; Piezoresistive gauge factor; Bulk silicon micromachining; Micro force sensor
1. Introduction Diamond and hard carbon coatings are widely used to improve the micro tribological properties of micro electro mechanical systems (MEMS) [1-3]. Due its hardness, wear resistance as well as robustness to harsh environment diamond is an attractive material for many MEMS applications, e.g. micro grippers and atomic force microscope (AFM) probes. These micro components could be realised by selective deposition of polycrystalline diamond on silicon substrates and molds, respectively, using SiO2 as a sacrificial layer [4]. Furthermore, diamond is one of the most promising materials for highfrequency nano electro mechanical systems (NEMS) or surface acoustic wave (SAW) devices [5] since it exhibits the highest sound velocity of all semiconductors with 1.8 × 104 m/s as estimated by the square root of the Young’s modulus-to-densityratio.
∗
Corresponding author. Tel.: +49 531 3913761; fax: +49 531 3915844. E-mail address: [email protected] (E. Peiner).
0924-4247/$ – see front matter © 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.sna.2005.11.059
Piezoresistivity is an important property of materials to be used, e.g. for strain gauges in mechanical sensors. Boron-doped polycrystalline diamond (poly-C) grown on commercial polyC substrates exhibits a considerable piezoresistive effect with gauge factors K of a few hundreds and more than 4000 under inter- and intra-grain probing conditions, respectively [6]. In a recent study a best value of the piezoresistive gauge factor of 50 was reported for boron-doped poly-C on oxidised silicon substrates [7]. A poly-C position sensor integrated on a siliconbased cochlear implant probe has a gauge factor of 28 [8]. K depends on grain size and resistivity of the poly-C. Maximum values of 70–80 were obtained at around 102 cm [9]. Unfortunately, chemical vapour deposition (CVD) which is conventionally used for the growth of poly-C requires a substrate temperature of 500–900 ◦ C [7,8] which is much too high for substrates alternative to diamond or silicon like large-area plastics. Furthermore, it is not favourable for the deposition on micro-structured silicon. Correspondingly, low-temperature deposition, e.g. plasma-enhanced or -assisted CVD, of amorphous (a-Si:H) and microcrystalline (c-Si:H) silicon at low substrate temperature (<150 ◦ C) has recently found much atten-
76
E. Peiner et al. / Sensors and Actuators A 130–131 (2006) 75–82
Fig. 1. Schematic of micro force sensor (a) with integrated probing tip at the cantilever bottom side in an enlarged representation (b). Optical microphotographs of the amorphous carbon reference and strain gauge resistors are shown in (c) and (d).
tion [10,11]. An investigation of the piezoresistive properties of low-temperature c-Si:H on plastic substrates revealed gauge factors in the range of 10–40 [11]. For sputtered amorphous carbon (a-C) investigations on piezoresistivity are lacking. Therefore, in this contribution we report on the structural and electrical transport properties of sputtered a-C as a material for piezoresistive strain gauges in MEMS. Amorphous carbon is superior to heteroepitaxial and poly-C in terms of production cost and process requirements: high substrate temperatures and additional doping sources are not necessary. Compared to a-Si and c-Si:H it provides better tribological properties and robustness to harsh environment. In this study the piezoresistive properties of amorphous carbon strain gauges integrated on a bulk micromachined silicon cantilever (Fig. 1) are addressed for the first time. 2. Fabrication and characterization of amorphous carbon films Amorphous carbon films were deposited on silicon and SiO2 /silicon substrates by rf magnetron sputtering of a graphite target using a Balzers BAS 450 sputter plant. The substrate temperature during deposition was less than 150 ◦ C. A substrate bias of −100 V was maintained leading to superior tribological properties (high hardness, low wear and friction) of 0.25 m thick a-C layers, e.g. a hardness of 50 GPa [12]. Hardness in the range of a few tens of GPa can be expected for sputtered amorphous carbon with a content of tetrahedral-bonded carbon (sp3 ) of around 25% [13]. The hydrogen content in the films determined by secondary ion mass spectroscopy (SIMS) was less than 2%. More details on the a-C deposition process and the tribological testing were reported elsewhere [12].
The structural properties of 0.5 m thick a-C layers on silicon were investigated by spectroscopic ellipsometry and Raman spectroscopy. Fig. 2 shows the optical absorption coefficient α of a-C measured between 0.8 and 5 eV which can be used to determine the optical band gap. Constructing a plot of (αhν)1/2 versus the photon energy hν the band gap can be determined from the intercept ET of the extrapolated linear fit of Tauc’s relation to the measured data (Fig. 2, Tauc gap, [14]). Alternatively, the photon energy at which the absorption coefficient amounts to 104 cm−1 is often considered as a measure of the optical gap (E04 , [15]). From Fig. 2 we obtain values of 0.65 and 0.8 eV for the Tauc and the E04 gap, respectively. By comparison with collected data [15] we can estimate a content of conductive sp2 bonds of 80–90%. The Raman spectrum in Fig. 3 exhibits the typical G- and D-bands which can be assigned to the E2g in-plane stretching mode of graphite and to the A1g breathing mode of aromatic rings
Fig. 2. Absorption coefficient and Tauc plot [14], i.e. (αhν)1/2 in dependence on photon energy hν for a 0.5-m-thick a-C layer.
E. Peiner et al. / Sensors and Actuators A 130–131 (2006) 75–82
77
and σ(T ) = aT m ,
Fig. 3. Raman spectrum of a 0.5-m-thick a-C layer. The superimposed solid line corresponds to a fit to the measurement using an asymmetric Breit-WignerFano line shape for the G-band (dotted curve), a Lorentz function for the D-band (dashed curve) and a linear background (dash-dotted line) [17].
(graphitic oligomers) predicted at 1580 and 1360 cm−1 , respectively being in agreement with the expectation for sputtered a-C layers [16]. The solid line in Fig. 3 is a fit of a Breit-WignerFano line shape for the G-band (dotted curve), a Lorentz function for the D-band (dashed curve) and a linear background (dashdotted line) [17], revealing the respective mode wave numbers at 1600 ± 1 cm−1 (G-band) and 1388 ± 3 cm−1 (D-band). Note that due to the asymmetric Breit-Wigner-Fano line shape of the G-band the fitted value of the mode wave number does not correspond to the maximum peak position. From the G-band position a sp2 content of 83% is derived [17] being in accordance with the ellipsometry result as well as the excellent tribological properties. For amorphous carbon deposited on silicon by rf magnetron sputtering characterised by the van der Pauw/Hall technique p-type conduction is reported at a hole concentration of a few 1018 cm−3 [18]. In this study we performed temperaturedependent resistivity measurements within 20 and 300 K with a-C deposited on p-type, n-type and oxidised silicon substrates. Current blocking between layer and substrate is observed for a-C on p-silicon as well as on oxidised silicon. However, Ohmic behaviour was observed between a-C and n-silicon above 225 K while current blocking only occurred below 190 K. From the absence of diode behaviour we conclude that our a-C layers are of n-type conductivity. The hetero junction barrier between the a-C film and the silicon substrate is only effective at low temperatures. Conduction or phonon-assisted tunnelling between localised states are among the carrier transport mechanisms which are expected in sputtered a-C rather than thermally activated propagation through extended states [19]. Single- and multi-phononassisted hopping models describe the temperature dependence of conductivity in a wide range according to √ T0 1/4 VRH σ(T ) T = σ0 exp − T
(1)
(2)
respectively. To confirm this expectation conductivity measurements on a-C films on oxidised silicon were performed on square-shape samples with Ohmic contacts formed in a van der Pauw configuration. Fig. 4a and b show the measured temperature dependence of resistivity of an a-C film on oxidised silicon in representations suitable for either Eq. (1) (Fig. 4a) or Eq. (2) (Fig. 4b). At low probing current both models (straight lines) describe the experiment in the temperature range close to room temperature. A wider range of agreement is visible for the multi-phonon hopping model (Fig. 4b, Eq. (2)). As a result of the fitting we obtained σ0VRH = (1.4 ± 0.1) × 105 K1/2 /(cm) 1/4 and T0 = 26.3 ± 0.3 K1/4 from Fig. 4a and Eq. (1). For the multi-phonon hopping model (Eq. (2)) we found a = 0.012 ± 0.001 K1.25 /(cm) and m = 1.25 ± 0.01 from Fig. 4b. These values compare very well with the data reported for a-C fabricated at elevated substrate temperatures using carbon ions from a discharge in fullerene vapour [19]. We
Fig. 4. Conductivity of amorphous carbon between 20 and 300 K in representations exhibiting variable range hopping (VRH, Eq. (1)) and multi-phononassisted tunneling (Eq. (2)) between 120 and 300 K (a) and 60 and 300 K (b), respectively.
78
E. Peiner et al. / Sensors and Actuators A 130–131 (2006) 75–82
conclude that conduction in sputtered a-C is well described by the multi-phonon hopping process between localised states. Hopping conduction between conductive grains embedded in an insulating matrix is among the most representative processes of charge transport in piezoresistive composite material [20]. Considering a-C as a material which is composed of conductive clusters of sp2 bonds and a surrounding insulating sp3 matrix we can adopt the proposed theoretical model for the present study. If charge transport in a-C is governed by tunnelling between adjacent localised states which are separated by a mean distance d we can calculate the piezoresistive gauge factor, i.e. the ratio of relative resistivity change to strain using: 1 − χ/(1 − ν) 2d (1 + χ) ξ 1 − ν/(1 − ν) 2d Ktr = χ (1 + χ) ξ Kl =
(3)
where ξ is the carrier localisation length and ν denotes Poisson’s ratio. Longitudinal (Kl ) and transversal (Ktr ) gauge factors are considered via a phenomenological parameter χ, which takes into account a finite percentage of hops propagating not along but perpendicular to the direction of the applied electrical field. Isotropic behaviour is obtained at χ = 0.5. For the derivation of Eq. (3) rectangular-shaped resistors which are uniaxially strained with a contribution of the Poisson effect only in one perpendicular direction are assumed [20]. This configuration corresponds to the case of the a-C resistors integrated on a silicon micro cantilever depicted in Fig. 1. The -shaped a-C strain gauge is oriented with its both legs parallel to the cantilever axis. It is electrically contacted from the bottom side via heavily doped n-type lines (Fig. 1d). A deflection of the cantilever generates strain in the resistors R12 , R34 and R14 given by the resistances between the contacts 1 and 2, 3 and 4 and 1 and 4, respectively. R12 and R34 correspond to the legs of the -shaped strain gauge thus being under longitudinal strain. The perpendicularly oriented resistor R14 is under transversal strain.
were then deposited by e-beam evaporation and structured by lift-off. Finally, the cantilever was released wet chemically using potassium hydroxide solution (KOH, 30%, 60 ◦ C). We prefer KOH for this last etching step due to its lower mask undercut compared with TMAH. A protection of the metallisation was not necessary. Special protection of the a-C resistors was not required during the membrane as well as the final cantilever structuring since a-C was found not to be attacked by either TMAH or KOH solution. To confirm the excellent etching selectivity of a-C with respect to silicon a test structure consisting of 0.5 m thick aC was exposed to a typical 1 h etch attack by KOH solution (30%, 60 ◦ C). The scanning electron microphotographs in Fig. 5 show that the a-C layer was almost completely released from the silicon substrate without an indication of damage. Furthermore, a-C shows strong adhesion and low stress on silicon as well as on oxidised silicon, i.e. undercut of the integrated strain gauge resistors was not observed. Finally, the realised device was mounted into a connector. Voltage supply and current read-out was done by pressed contacts. Wire bonding was not necessary [21,22]. Ohmic characteristics were obtained for both the strain gauge and a reference resistor (Fig. 6a) which were connected via n+ lines and directly by the Au/Cr metallisation, respectively. The resulting resistances of 5.4 k for the strain gauge resistor and 80 k for the reference resistor are in close agreement with the design. We conclude that below 1 V the n+ connecting lines do not affect the resistance measurement. I–V characteristics of the strain gauge
3. Sensor fabrication and characterisation The tactile force sensor reported here is based on a silicon micro cantilever with length, width and thickness designed to 5 mm, 0.25 mm and 50 m, respectively. Sensor prototypes were realised using a bulk silicon micromachining process based on standard photo-lithography, thermal oxidation and wet etching [21,22]. The process was started by etching of a p-type silicon wafer using tetra methyl ammonium hydroxide solution (TMAH, 20%, 80 ◦ C) through a mask of thermal oxide to obtain a membrane structure. By diffusion of phosphorus from a spinon silica emulsion (Merck, Siodop) n+ -doped contact lines were fabricated. Subsequently, a field oxide was grown and contact holes were opened. Amorphous carbon (0.5 m) was deposited by magnetron sputtering and structured into resistors between the n+ contact lines by lift-off using photo resist. In a subsequent second TMAH etching step the silicon membrane was thinned to its final thickness. Contact holes were opened to the n+ -doped regions and gold/chromium (300 nm/30 nm) connecting lines
Fig. 5. Amorphous carbon test structure on silicon substrate illustrating the excellent etching selectivity of a-C with respect to silicon using KOH solution.
E. Peiner et al. / Sensors and Actuators A 130–131 (2006) 75–82
79
Fig. 7. Resistance change of the longitudinally strained gauge resistor R34 (Fig. 1d) of a cantilever-type micro force sensor (#1,Table 1) upon gradual increase of forces applied to the cantilever free end at the top (−F) and the bottom surface (+F), i.e. R34 is under tensile and compressive stress, respectively.
For the determination of the piezoresistive gauge factor K we consider the cantilever along its axis and assume that the force is applied at its free end, i.e. at the distance leff from the position of the strain gauge resistor:
Fig. 6. Current–voltage characteristics of amorphous carbon strain gauge and reference resistors at 300 K (a) and in the temperature range between room temperature and 150 ◦ C (b). The temperature dependence of conduction can be modelled by an Arrhenius law as depicted in the inset.
resistor were measured in the temperature range from room temperature to 150 ◦ C (Fig. 6b). A field-dependent increase of current is visible above around 1 V while below Ohmic behaviour was found. In the considered temperature range conductivity exhibits a slight temperature dependence which can be described by an Arrhenius model (see inset). By fitting we find a small activation energy of Ea = 31 meV. The sputter-deposited a-C strain gauges on silicon micro cantilevers were investigated by deflecting the cantilever and measuring the resistance change. In numerous calibration runs the cantilevers were incrementally moved with its free end against a force measuring probe mounted on the stamp of a high-resolution compensation balance (Sartorius SC2) [21,22]. Movement of the cantilever was performed at a resolution of 1 nm and a reproducibility of 5 nm. The compensation balance offers a resolution of 1 nN and a reproducibility of 2.5 nN. For the complete setup a temperature drift of less than 10 mK/h was maintained. The resistance was measured using a digital multimeter (Prema DMM 5017) at a current of 10 and 100 A, respectively. In Fig. 7a typical load–resistance curve is shown. Longitudinal compressive and tensile strain is generated in the strain gauge resistor R34 by probing the bottom and the top surfaces of the cantilever, respectively (Fig. 8). We can assume nearly uniform strain across the strain gauge resistor which is designed with its length (100 m) and height (0.5 m) being small compared with the cantilever length (5 mm) and height (50 m), respectively.
R 3 K leff =− F R 2 E weff h2
(4)
with the relative change of resistance R/R, the cantilever height h and the applied force F. Owing to the sidewalls of the cantilever [(1 1 1) crystal planes] which are inclined to the surface [(0 0 1) plane] by 54.7◦ an effective width weff = w + h/21/2 of the cantilever is taken into account. Furthermore, the undercut of the frame below the cantilever clamping leading to a not ideally fixed clamping is considered by an effective length leff exceeding its nominal value [22]. The cantilever thickness can be controlled
Fig. 8. Schematic of the cantilever-type force sensor under application of forces to the cantilever free end from bottom (+F) and top (−F) directions generating compressive and tensile strain, respectively, at the cantilever surface, i.e. in the gauge resistors.
80
E. Peiner et al. / Sensors and Actuators A 130–131 (2006) 75–82
Table 1 Characteristics of three sensor prototypes and calculated gauge factors K of integrated a-C strain gauges h (m)
Resistor/strain
R/R/F (N−1 )
K
R/R/T (K−1 )
9.7 ± 0.1
52 ± 2
R34 /tensile R34 /compressive
3.06 ± 0.05 3.13 ± 0.05
45.1 ± 3.4 46.2 ± 3.4
– –
#2A
10.7 ± 0.1
54 ± 2
R34 /compressive R14 /tensile R14 /compressive
2.66 ± 0.19 2.4 ± 0.05 2.29 ± 0.06
42.2 ± 3.2 37.9 ± 2.8 36.2 ± 2.7
8.1 × 10−3 7.9 × 10−3
#2B
13.1 ± 0.1
57 ± 2
R12 /tensile R12 /compressive R34 /tensile R34 /compressive R14 /tensile R14 /compressive
2.41 2.48 2.31 2.33 2.21 2.20
± ± ± ± ± ±
6.5 × 10−3
Sensor #1
k (N/m)
± ± ± ± ± ±
0.08 0.08 0.09 0.08 0.06 0.05
44.3 45.5 42.5 42.7 40.6 40.4
3.2 3.4 3.2 3.2 3.1 3.0
6.5 × 10−3 6.0 × 10−3
Rij denotes the resistors between the contacts i and j (Fig. 1d), i.e. R12 and R34 are under longitudinal strain and R14 is under transversal strain. By probing the cantilever from the bottom and top sides uniform compressive and tensile strain, respectively, is generated in the resistors. Each of the figures of sensitivity and gauge factor corresponds to the average of at least 40 measurements. The strain gauges of sensors #1 and #2 (A and B) were analysed at probing currents of 100 and 10 A, respectively.
using the measured stiffness k of the cantilever according to: 4k h = leff 3 (5) weff E At leff = 5.4 ± 0.1 mm, weff = 260 ± 10 m, k = 9.7 ± 0.1 N/m and E = 170 GPa we find h = 52 ± 2 N/m for the cantilever considered in Fig. 7. We observe linear resistance–force characteristics as shown in Fig. 7 yielding a piezoresistive gauge factor of K = 44.7 using Eq. (4) and the given cantilever geometry. In Fig. 9 the probability distribution of the gauge factor is displayed obtained for the above micro force sensor prototype after collecting the results of 195 measurements under compressive strain. We find an average value of −0.01727 ± 0.00026 /N corresponding to a gauge factor of K = 46.2 ± 3.4. Under tensile strain of the a-C resistor we obtain a sensitivity of −0.01685 ± 0.00025 /N corresponding to a gauge factor of K = 45.1 ± 3.4. We assign the scatter given by a Gaussian fit to the measured probability distribution mainly to the temperature drift of our calibration set-up. We find temperature coefficients of
resistivity of (6–8) × 10−3 K−1 leading to a force drift of the a-C strain gauges of 2.6–3.5 mN/K. Signal-to-noise ratio is found to be around 105 at a probing current of 100 A corresponding to a minimum detectable force in the range of few N. Full-bridge strain gauge designs which are possible on surfaces with tensile and compressive strain regions, e.g. membranes, will offer an additional increase of sensitivity (factor-of-four), improved drift stability and a sensor output convenient to standard bridge amplifiers. The results of numerous calibration runs with three prototypes of force sensors are collected in Table 1. The cantilever thicknesses determined by load-deflection measurements and Eq. (5) are within 50–60 m. Force sensitivity measured for the resistors R12 , R14 and R34 arranged between the contacts 1–2, 1–4 and 3–4, respectively (cf. Fig. 1), ranges from 2.2 to 3.1 N−1 . Using these values we calculate K-factors of 36–46. These values compare well with published values of 10–80 obtained for boron-doped polycrystalline diamond on silicon [7–9]. A slight difference between the longitudinally strained resistors (R12 , R34 ) and the resistor R14 which is under transversal strain was detected. Taking the average of all measured values for Kl and Ktr we find χ = 0.484 (Eq. (3)) assuming a Poisson ratio of ν = 0.3. The ratio of the hopping distance between localised conductive sites to the localisation length d/ξ of carriers can be estimated to be around 100. 4. Conclusions
Fig. 9. Probability distribution of sensitivity of a cantilever-type micro force sensor comprising an a-C strain gauge resistor (#1, Table 1, R34 ) under longitudinal compressive strain. The results of 195 measurements are collected.
Amorphous carbon was investigated as a functional material for MEMS technology, i.e. for piezoresistive strain gauges of bulk micromachined silicon cantilever sensors. We found that amorphous carbon is a promising material since it combines favourable mechanical properties like adhesion, hardness and wear resistance with a large piezoresistive effect. It could be deposited on thin silicon membranes maintaining low substrate temperatures (<150 ◦ C). As a consequence it could be easily structured by lift-off using photo resist. Furthermore, amorphous carbon was found to be inert in anisotropic wet etchants like
E. Peiner et al. / Sensors and Actuators A 130–131 (2006) 75–82
KOH and TMAH. Therefore, special protection of amorphous carbon on silicon and silicon oxide resistors during membrane and cantilever etching is not necessary. Extended calibration runs with cantilever prototypes yielded linear characteristics between the applied load and the amorphous carbon strain gauge resistances. Piezoresistive gauge factors were determined at compressive and tensile strain as well as under longitudinal and transversal probing conditions. We found values of K between 36 and 46 almost independent of strain and probing directions. Acknowledgements The authors are grateful to Doris R¨ummler for valuable assistance during the sensor fabrication and SEM characterisation as well as to Georgi Ginev and Wolfgang Limmer for ellipsometry and Raman measurements, respectively. Funding by the German Federal Ministry of Economics and Labour (BMWA) is gratefully acknowledged.
81
[16] N.A. Hastas, C.A. Dimitriadis, Y. Panayiotatos, D.H. Tassis, S. Logothetidis, D. Papadimitriou, G. Roupakas, G. Adamopoulos, High-field transport and noise properties of sputter-deposited amorphous carbonsilicon heterostructures, Semicond. Sci. Tech. 17 (2002) 662–667. [17] Y. Perera, J. Gottmann, A. Husmann, T. Klotzb¨ucher, E.W. Kreutz, R. Poprawe, PLD of hard ceramic coatings, Laser Applications in Microelectronic and Optoelectronic Manufacturing VI, Proc. SPIE 4274 (2001) 66–77. [18] N.A. Hastas, C.A. Dimitriadis, Y. Panayiotatos, D.H. Tassis, P. Patsalas, S. Logothetidis, Noise characterization of sputtered amorphous carbon films, J. Appl. Phys. 88 (2000) 5482–5484. [19] E.B. Maiken, T. Taborek, Amorphous carbon films deposited from carbon ions extracted from a discharge in fullerene vapor, J. Appl. Phys. 87 (2000) 4223–4229. [20] C. Grimaldi, P. Ryser, S. Str¨assler, Gauge factor of thick film resistors: outcomes of the variable range hopping model, J. Appl. Phys. 88 (2000) 4164–4169. [21] I. Behrens, L. Doering, E. Peiner, Piezoresistive cantilever as portable micro force calibration standard, J. Micromech. Microeng. 13 (2003) 171–177. [22] E. Peiner, L. Doering, Force calibration of stylus instruments using silicon microcantilevers, Sensor. Actuat. A 123–124 (2005) 137–145.
Biographies References [1] W. Wang, Y. Wang, H. Bao, B. Xiong, M. Bao, Friction an wear properties in MEMS, Sensor. Actuat. A 97–98 (2002) 486–491. [2] I.S. Forbes, J.I.B. Wilson, Diamond and hard carbon films for microelectromechanical systems (MEMS)—a nanotribological study, Thin Solid Films 420–421 (2002) 508–514. [3] M.R. Werner, W.R. Fahrner, Review on materials, microsensors, systems and devices for high-temperature and harsh-environment applications, IEEE T. Ind. Electron. 48 (2001) 249–257. [4] T. Shibata, Y. Kitamoto, K. Unno, E. Makino, Micromachining of diamond films for MEMS applications, J. Microelectromech. S. 9 (2000) 47–51. [5] W.I. Milne, Electronic devices from diamond-like carbon, Semicond. Sci. Tech. 18 (2003) S81–S85. [6] S. Sahli, D.M. Aslam, Ultra-high sensitivity intra-grain poly-diamond piezoresistors, Sensor. Actuat. A 71 (1998) 193–197. [7] A. Yamamoto, T. Tsutsumoto, Piezoresistive effect of CVD polycrystalline diamond films, Diam. Relat. Mater. 13 (2004) 863–866. [8] Yuxing Tang, Dean.M. Aslam, Jianbai Wang, Kensall.D. Wise, Technology and integration of poly-crystalline diamond piezoresistive position sensor for cochlear implant probe, in: Proceedings of the 13th International Conference on Solid-State Sensors, Actuators and Microsystems (Transducers ’05), Seoul, Korea, 2005 June 5–9, pp. 543–546. [9] M. Werner, R. Locher, Growth and application of undoped and doped diamond films, Rep. Prog. Phys. 61 (1998) 1665–1710. [10] G. de Cesare, M. Gavesi, F. Palma, B. Ricco, A novel a-Si:H mechanical stress sensor, Thin Solid Films 427 (2003) 191–195. [11] P. Alpuim, V. Chu, J.P. Conde, Piezoresistive sensors on plastic substrates using doped microcrystalline silicon, IEEE Sensor. J. 2 (2002) 336–341. [12] R. Bandorf, H. L¨uthje, C. Henke, J.-H. Sick, R. K¨uster, Tribological behaviour of thin a-C and a-C:H films with different topographic structure under rotating and oscillating motion for dry lubrication, Surf. Coat. Tech. 188–189 (2004) 530–533. [13] W. Lu, K. Komvopoulos, Effect of stress-induced phase transformation on nanomechanical properties of sputtered amorphous carbon films, Appl. Phys. Lett. 82 (2003) 2437–2439. [14] J. Tauc, R. Grigorovici, A. Vancu, Optical properties and electronic structure of amorphous germanium, Phys. Status Solidi 15 (1966) 627–637. [15] J. Robertson, Electronic and atomic structure of diamond-like carbon, Semicond. Sci. Tech. 18 (2003) S12–S19.
Erwin Peiner, (Priv.-Doz.) born in Bad M¨unstereifel, Germany, in 1960, received the Diplom-Physiker and PhD degrees, both in applied physics, from the University of Bonn, Germany in 1985 and 1988, respectively. In 1989 he joined the Institut f¨ur Halbleitertechnik (IHT) of the Technical University Braunschweig, where he was involved in the monolithic integration of III/V compound semiconductors on silicon for MOEMS. In 2000, he received the venia legendi for semiconductor technology from the Faculty of Mechanical and Electrical Engineering of the Technical University Braunschweig owing to a habilitation thesis on silicon sensors for condition monitoring of machines and plants. Presently, he is in charge of the semiconductor sensors group at the IHT. His main interests are in the field of micromachined sensors for industrial applications combining silicon with novel materials (like DLC, ZnO, SiC). Arti Tibrewala, (MSc.) born in Aurangabad, India, in 1979, received the degree of Industrial Engineering, from Marathwada University, Aurangabad, India in 2000. She got her Master of Science in Electrical Engineering from University of South Florida, Tampa, USA in 2002. She joined Institut f¨ur Halbleitertechnik of the Technical University Braunschweig in 2003, where she is involved with characterisation of diamond like carbon films. Ralf Bandorf, (Dr.-Ing. Dipl.-Phys.) was born in 1973 and studied at the Friedrich-Alexander-University Erlangen-Nurenberg. His doctoral thesis at the Fraunhofer Institute for Surface Engineering and Thin Films was focused on ultrathin tribological systems for electromagnetic microactuators. Since December 2001 he is employed as a project leader at the IST in the micro and sensor technologies group. His work is dealing with smart coatings, integrated sensors, microtribological films, and electrical and magnetic functional coatings. Saskia Biehl, (Dipl.-Ing.) was born in 1971. She received the degree of a graduate engineer in Material Science of the University of Saarbr¨ucken in 1999. In 2000 she worked as an engineer for quality assurance at a glass coating company in Germany. Since 2001, she is a project manager at the Fraunhofer Institute for Thin Film and Surface Engineering working in the field of temperature, force and biological micro sensors as well as micro structuring. ¨ Holger Luthje, (Dipl.-Ing.) was born in 1944 and studied physical technology in L¨ubeck. During 21 years industrial work in a multinational company in the field of electronics he was in-charge of different projects dealing with electrical and magnetic coatings, sensors, as well as sub-micron patterning techniques for mega-chips. Included were also technologies for the manufacturing of micromechanical components. Since 1991 he is working at the
82
E. Peiner et al. / Sensors and Actuators A 130–131 (2006) 75–82
Fraunhofer Institute for Surface Engineering and Thin Films in Braunschweig, where he is in-charge of the group micro and sensor technology. Actual work focuses on the research and development of smart coatings with integrated sensors, microtribological coatings for MEMS applications, microstructuring and molding technologies, as well as different electrical coatings. Further main topics are the development of magnetic coatings, especially thin film sensors for industrial and automotive applications.
Lutz Doering, (Dr.-Ing.) born in Salzwedel, Germany, in 1960, works in the PTB in Braunschweig since September 2001 in the work group of microsystem metrology. His graduation in the area of microelectronics technology, precision instrument technology and computer science he reached 1990 at the Technical University of Dresden. In the centre of his presently work are measuring procedures for the characterisation of micro-technically manufactured micro force sensors.