Micromachined atomic force microscopy sensor with integrated ... - AVS

Micromachined atomic force microscopy sensor with integrated piezoresistive sensor and thermal bimorph actuator for high-speed tapping-mode atomic force microscopy phase-imaging in higher eigenmodes R. Pedrak, Tzv. Ivanov, K. Ivanova, T. Gotszalk, N. Abedinov, and I. W. Rangelowa) Institute of Microstructure Technologies and Analytics (IMA), University of Kassel, 34109 Kassel, Germany

K. Edinger NaWoTec GmbH, Industriestr. 1 64380 Rossdorf, Germany

E. Tomerov NTS Ltd. 2140 Botevgrad, Bulgaria

T. Schenkel University of California, Berkeley, LBNL, 1 Cyclotron Road, Berkeley, California 94270

P. Hudek Leica Microsystems Lithography GmbH, 07745 Jena, Germany

共Received 18 July 2003; accepted 4 August 2003; published 10 December 2003兲 This article describes microprobes for noncontact scanning force microscopy that make use of a direct-oscillating thermally driven bimorph actuator with integrated piezoresistive readout sensor. The sensitivity has been increased using direct current for biasing and alternating current for exciting the thermally driven cantilever in a higher flexural mode. The cantilever operates in the phase-shift atomic force microscopy 共AFM兲 detection technique. The main advantage of phase imaging is the higher z resolution at high scan rates and much lower forces than in height imaging with contact AFM. Critical dimensions measurements illustrating the imaging capability and resolution of our new scanning proximal probe are demonstrated. © 2003 American Vacuum Society. 关DOI: 10.1116/1.1614252兴

I. INTRODUCTION Tapping mode atomic force microscopy 共TM–AFM兲 allows imaging with nanometer-scale resolution, where the tip strikes the surface with negligible force. This technique is a promising candidate for noncontact and damage free sub-50 nm critical dimension 共CD兲 measurement applications. In TM–AFM1 the cantilever is excited to oscillate close to its resonance. The topography information is collected from the phase lag between vibration excitation and response of the cantilever deflection sensor. This phase lag is related to the energy dissipated in the tip-sample interaction. Cantilever probes based on piezoresistive sensing principle provide a simple and convenient technique enabling easy access to novel applications of force microscopy. However, the scan speed of current TM–AFMs is limited to about 250 ␮m/s due to actuation time constant of the piezotube feedback loop that keeps the tapping amplitude constant. This limitation can be overcome by reducing the size of the system and consequently its inertia. In this manner it is possible to significantly increase the scanning speed. Furthermore, by increasing the sensitivity of the tapping probe the signal to noise ratio can be improved thus leading to a further reduction in the time constant needed for stable feedback loop operation. Two issues are considered as fundamentals in the realization of this idea. a兲

Electronic mail: [email protected]

3102

J. Vac. Sci. Technol. B 21„6…, NovÕDec 2003

共i兲 The integration of the piezoresistive readout and actuator provides the best solution to realize high speed TM– AFM. In 1993 Tortonese et al. first demonstrated that the minimum detectable deflection for piezoresistive cantilevers can be as low as 0.01 nm for a 1 kHz bandwidth, confirming that this detection scheme is adequate for AFM operation.2 Piezoresistive detection is an attractive technique compared to the conventional optical beam deflection technique especially in the cases where laser detection is difficult. Piezoresistive cantilevers are irreplaceable in such areas as AFM in high vacuum,3 arrays of cantilevers,4 high frequency small cantilevers,5 in applications such as microbalance or infrared radiation detection6 or cryogenic conditions.7 共ii兲 For signal conversion from the electrical to the mechanical domain, it is necessary to add conversion elements compatible with complementary metal–oxide– semiconductor 共CMOS兲 processing, using thin film technology. Signal conversion by electrostatic forces has been demonstrated.8 However, electrostatic forces are normally only significant for small separation of the plates because electrostatically driven cantilevers are based on electrostatic forces that are proportional to the square of the separation of the cantilever plates. The main advantages of electrostatically driven cantilevers are low power consumption and short actuation times. Piezoelectric actuators based on sputtered ZnO films have been successfully used to excite mechanical vibration in micromachined cantilevers for AFM

1071-1023Õ2003Õ21„6…Õ3102Õ6Õ$19.00

©2003 American Vacuum Society

3102

Pedrak et al.: Micromachined AFM sensor with integrated sensors

3103

applications.9 The microactuators described in this article are based on the so-called bimetal effect.10 The actuator consists of a sandwich of layers, namely Al, SiO2 , and Si. The aluminum layer forms the heating microresistor and is used as the driving element. The main objective of the work described in this article is to enable the construction of cantilevers with piezoresistive readout and to integrate bimorph actuators for TM–AFM as an alternative to optically detected TM–AFM devices with piezotube z actuator. II. DEVICE DESCRIPTION Many research groups have looked at different alternative techniques replacing the optically detected cantilever techniques, using different types of integrated sensors for AFM applications. These are piezoresistive resistor,2,11 MOS transistor channel,12,13 piezoelectric,14,15 and capacitive16,17 techniques. In this article we will describe the piezoresistive cantilever sensor, where the deflection of the AFM tip is measured using an externally direct current 共dc兲-biased Wheatstone bridge circuit to detect the change in the resistance of the piezoresistors when the tip interacts with the sample surface. The circuit produces a voltage signal, which is proportional to the cantilever deflection. This signal is amplified using a low noise amplifier and supplied, in place of the signal from the photodetector, to a conventional and commercially available AFM control electronics. In addition, standard AFM control electronics and software has been designed to be used with a conventional photodetector sensor but can be used to operate the AFM in contact or dynamic mode. Phase imaging is based on measuring the phase lag between the voltage drop in the piezoresistors and the driving current of the thermal bimorph actuator. The variations in sample topology result in phase shifts of the signals, which are mapped to produce phase contrast images. The important parameters in the noncontact 共tapping mode兲 AFM are the minimum detectable force and the minimum force gradient. The main factors which determine this parameters are related to the cantilever geometry and the operating conditions such as temperature and working environments 共air, vacuum兲.18 –21 The minimum detectable force and force gradient are given by F min⫽

冉

冊冉冊冉冊

4kk B TB 2 ␲ f 0Q

1/2

⫽

␦ F min 1 2kk B TB ⫽ ␦z A 2 ␲ f 0Q

III. THERMAL BIMORPH ACTUATOR Until resently, a variety of micromachined cantilevers, with deflection-sensor and some of them with an integrated actuators, have been developed.19,20 Due to the low heat capacity of the transducer element, a high temperature rise per unit input power can be achieved. The maximum value of the deflection is a function of the cantilever material, the ratio of the thicknesses of the bimorph layers, and Young’s modulus E. Cantilever built from Al–S has a largest deflection value due to the large Young’s modulus of silicon, however, with such cantilevers a smaller displacement dynamics is obtained in comparison with bimorph made cantilevers from Al–SiO2 . In the devices presented in this article we used a system which consists of three layers: 1 ␮m Al 0,6 ␮m SiO2 , and 5 ␮m Si where a 50 nm chrome film underlayer is employed to improve the thermomechanical stability and the adhesion to SiO2 . For a cantilevers 650 ␮m long and total thickness of 6.6 ␮m, a specific deflection of approximately 21 nm/K at the free end where the AFM tip is placed can be achieved. The tip displacement ⌬ of the cantilever by the actuator, at small deflection angel can be described as ⌬⫽

共 E ␳ 兲 1/4共 k B TB 兲 1/2,

1/2

,

where: l, w, and t are length, width, and thickness of the cantilever; k, Q, and f 0 are it’s spring constant, quality factor

r⫽

and resonance frequency; E, ␳ is the Young’s module and the density of the cantilever material; T is the temperature and k B ⫽1.38⫻10– 23 J/K is the Bolzman constant. The measurement bandwidth is B and amplitude of oscillations is A. In order to estimate the capability of the presented cantilever device, with respect to the minimum detectable force and force gradient it is necessary to obtain the reasonable values of these quantities. In case of air environment the quality factor is about 220. At such conditions, a 4-␮m-thick silicon cantilever, 150 ␮m width and 600 ␮m long, has a spring constant of k⫽1.89 N/m. In the bandwidth of 1 Hz the cantilever allows to detect forces in range of 13 fN 共at resonance兲. Moreover, if the amplitude of oscillation is 10 nm this cantilever has the ability to recognize the force gradient 2.68 ␮N/m. The cantilever operates at room temperature.

2 1/2

wt lQ

3103

2 l canti

2r

for l⬍⬍r with the bending radius r and length l of the cantilever. Considering assumptions from Ref. 21 that the sum of applied forces have to be zero and the generated moments have to be balanced than the radius of curvature of the multilayered cantilever beam can be expressed by the equation

共兰 t0 兰 w0 Edydz 兲共兰 t0 兰 w0 Ez 2 dydz 兲 ⫺ 共兰 t0 兰 w0 Ezdydz 兲 2 共兰 t0 兰 w0 Edydz 兲共兰 t0 兰 w0 E ␣ Tzdydz 兲 ⫺ 共兰 t0 兰 w0 E ␣ Tdydz 兲共兰 t0 兰 w0 Ezdydz 兲

JVST B - Microelectronics and Nanometer Structures

,

3104


3104

TABLE I. Physical properties of materials used in the construction of the bimorph actuator.

Material Si SiO2 Al Au Cr

Thermal expansion coefficient (10⫺6 /K)

Young’s modulus (1011 N/m2 )

Specific heat (103 J/kg K)

Thermal conductivity 共W/mK兲

Density (103 kg/m3 )

2.6 0.4 23.0 14.3 4.9

1.62 0.74 0.69 0.8 2.54

0.691 0.71 0.9 0.129 0.47

170 1.4 234 318 91

2.42 2.66 2.692 19.4 7.19

where T is temperature increase, E are the Young’s modules of the layers, ␣ are the coefficients of thermal expansion, z is the variable distance to the cantilever bottom, t is the total cantilever thickness, and w is the layer width. The physical properties of some materials of interest which have been used in the construction of the actuator are summarized in Table I. We found that the best difference in the thermal expansion coefficients using CMOS compatible metalisation, can be achieved by using a bilayer system consisting of sputtered Al with 2% Cu. IV. DEVICE FABRICATION We have fabricated piezoresistive AFM probes using two kind of tips: 共i兲 integrated Si tips which are formed at the beginning of the cantilever micromachining or 共ii兲 electron beam deposited tips, done after the fabrication of the cantilever beam which are grown directly on the Al microheater which is a part of the bimorph actuator. If the tips fabricated at the beginning, they have to be protected during subsequent processing steps, because they would be vulnerable to damage. The fabrication procedure for the cantilever beam is similar to that outlined by Rangelow et al.22 Here silicon on insulator 3 in. wafer comprising of 40-␮m-thick Si top layer 共12 ⍀ cm n type兲 of 具100典 orientation are used for the wet etching of the tips. The thickness of the top layer may be varied depending upon the required height of the Si tip. This layer is bonded to a 700-Å-thick thermal oxide layer grown on 具100典 base silicon wafer. The oxide layer is used as an etching stop layer. If the silicon tip has to be integrated on the cantilever, a thermal oxide needs to be grown and patterned to form the 8000-Å-thick mask which is subsequently used for wet etching of the Si tip. After a standard RCA clean, a 8000-Å-thick oxide is grown. This film was patterned and the resist mask over the oxide was employed as a mask for the boron contacts implantation at 1.1⫻1015 cm2 , 30 keV. This resist mask is removed using ␮ wave plasma striping and this is followed by growth of passivating thermal oxide during a 1 h anneal at 900 °C. Using again a resist mask the piezoresistors are configured in a Wheatstone bridge configuration defined in the oxide layer and boron implanted at 4⫻1014 cm2 20 keV, followed by growth of passivating thermal oxide during the 1050 °C for 30 min annealing. The cantilevers are then patterned and plasma etched to open the contact holes to the highly doped areas. Aluminum for the contacts to the J. Vac. Sci. Technol. B, Vol. 21, No. 6, NovÕDec 2003

piezoresistors and the metal layer, forming the microheater and bimorph actuators are then deposited, and annealed in a forming gas at 410 °C for 50 min. The oxide layer on the back of the wafer is patterned and a gas chopping reactive ion etching 共GChRIE兲23 combined with KOH wet step is used to release the cantilever membranes and partially dice the wafer. The buried oxide, used to stop the silicon etch, is then removed with a buffered oxide etch solution, using a mechanical wafer-chuck to protect the top side of the cantilevers. To form the cantilever beam and to cut up the single sensor chip employing GChRIE step, a thick resist mask is used. Finally, the resist mask is removed in oxygen plasma and the AFM cantilever are separated mechanically by pushing them from the support silicon frame 共Fig. 1兲.

FIG. 1. 共a兲 Micromachined TM–AFM probe with integrated bimorph actuator, piezoresistive deflection sensor, and Si tip. 共b兲 Experimental setup of the TM–AFM head. The preamplifier is visible as mirror image onto the measured optical mask.

3105


3105

FIG. 3. 共a兲 Schematic diagram of the experimental setup and PLL PID feedback electronics. 共b兲 Measured resonance frequency and phase shift 共⌬␾兲 curves in air and vacuum 共100 mTorr兲 of the thermally driven TM–AFM probe with bimorph actuator and piezoresistive deflection sensor.

FIG. 2. 共a兲 Static displacement of the cantilever vs applied heating power. 共b兲 Resonance frequency 共first order mode兲 shift vs actuator dc bias for different ambient pressures.

V. HIGH-SPEED TAPPING-MODE AFM AND PHASE IMAGING IN HIGHER EIGENMODES Employing the micromachining technique the feedback speed can be improved through integration of the actuator with the cantilever providing a feedback directly by applying an acting force to the cantilever. Then the resonance frequency will be much higher 共which is the own res. frequency of the cantilever兲 in comparison to a usually used piezotube actuator in commercial system. The magnitude of the feedback gain determines the speed by which the actuator can restore accurate feature topography. The high resonance frequency of the cantilever, feedback loop with smaller time constants, and bigger gain determine the surface scan speed by which the tip can accurately follow the topography features. The integrated piezoresistive cantilever, Si tip, and electrothermal actuator24 共bilayer effect which occurs between the Al and SiO2 layer兲 has been described in detail previously 共Fig. 1兲.25 For nanometer-scale resolution, the TM–AFMs scan speed is limited to a few tens of microns per second. At this speed a single, moderately sized 1024 ⫻1024 pixel image takes several minutes to acquire. Increasing the scan speed with a directly actuated AFM cantilever 共without pattern damage兲 the image acquisition time can be JVST B - Microelectronics and Nanometer Structures

reduced 5–10 times, which is very important for critical dimension measurements of lithographic patterns. Here, cantilever excitation is done by periodic heating which is very simple and has a high efficiency. It requires only a power-dissipating resistance on the cantilever and alternating current 共ac兲 driving voltage. The excitation heat power is a square of the driving sinusoidal voltage

FIG. 4. Force curve measured with the piezoresistor as a sensor and thermal actuator performed with an excitation of 716 kHz 共third order mode兲 and tip amplitude of 20 nm.

3106


3106

FIG. 5. 共a兲 Piezoresistor response vs drive frequency of the bimorph actuator for a free cantilever 共without contact with the surface兲. The fundamental ( f 0 ⫽19 kHz) and the next three natural modes ( f 1 ⫽162 kHz, f 2 ⫽348 kHz, and f 3 ⫽716 kHz) of the cantilever are plotted. 共b兲 A typical image of profile measurements obtained with 5 Hz for Cr features of optical mask. 共c兲 CD measurements with 10 Hz done on Cr features 共the 5-nmthick Pt islands were resolved兲.

P heater⫽ ⫽

1 R heater

共 V dc⫹V ac sin ␻ t 兲 2

1 2 sin2 ␻ t 兲 . 共 V 2 ⫹2V dcV ac sin ␻ t⫹V ac R heater dc

Figure 2 shows a plot of cantilever deflection response versus applied heating power. Static deflection of 3.6 ␮m J. Vac. Sci. Technol. B, Vol. 21, No. 6, NovÕDec 2003

was achieved by applying a heating power of 30 ␮W to the actuator. The dc biasing causes a shift in the resonance frequency. This can be explained by the fact that due to the increase of the dc bias the temperature of the cantilever will increase and this will increase the compressive stress. By heating of the bimorph actuator at temperature higher than the ambient, signal to noise ratio is improved and the stiff-

3107


ness of the cantilever increased, resulting in a resonance frequency shift 关Fig. 2共b兲兴. In our setup the cantilever is driven by thermal actuation at its resonance frequency ( f res) by an ac current with frequency ( f res/2) while z axis actuation in z axis is provided by applying a dc current. The magnitude of the dc current determinates the deflection of the cantilever, controlled by the proportional integral differential 共PID兲 feedback loop 关Fig. 3共a兲兴. The total ac driving power is below 1 ␮W.25 In this manner, a stable actuator and a soft cantilever for probing of the surface of the sample is provided. If the cantilever is operated in vacuum the quality factor increases 关Fig. 3共b兲兴. The result of phase change caused by approaching the thermally vibrating tip to the sample surface 共standard force curve兲 is shown in Fig. 4. The phase signal shift of the driving voltage and voltage drop response of the piezoresistive sensor is monotonic, ensuring a stable and reproducible control over the full z range of probe sample interaction. This phase signal offers shorter response times and should be the preferred signal for distance control. Figure 5共a兲 shows piezolever response versus drive frequency of the bimorph actuator for a free cantilever 共without contact with the surface兲. The fundamental ( f 0 ⫽19 kHz) and the next three natural modes ( f 1 ⫽162 kHz, f 2 ⫽348 kHz, and f 3 ⫽716 kHz) and the corresponding phase shift peaks of the cantilever are given. We found that the most sensitive mode of flexural vibration for an AFM cantilever is the first mode. However, the high-order vibration modes are more sensitive than the first mode when the contact stiffness is greater. These results are compatible with the theoretical results obtained by Ref. 26. Figure 5共b兲 shows the measured topography of a chromium pattern on quartz with deposited 5 nm Pt film recorded with 5 and 10 Hz. The height contrast between the top of the Cr pattern 共bright兲 and the quartz substarte between the features 共dark兲 is 40 nm, as expected. The TM–AFM image reveals a cluster structure of the deposited Pt onto the Cr pattern. From this observation, we conclude that the Cr pattern is perfectly reproduced and in particular, some residual Cr is present on the patterns. The vibration amplitude was 10 nm, which means that the tip provide a complete approach-contact retraction cycle during each period of oscillation. Figure 5共c兲 depicts the tapping mode image of a Cr grating with a period of 1 ␮m. The Cr patterns are 40 nm high. The image was taken at 10 Hz scan speed at the mechanical resonance frequency 716 kHz. Images 5b and 5c were obtained employing the phase signal for distance regulation. The magnitude of cross coupling from actuator to sensor is an essential issue, since they are closely spaced due to their integration on the cantilever. We have evaluated the signal cross-coupling from the thermalactuator input to the piesoresistive sensor output. The heat used for actuation can induce a temperature gradient at the location of the Wheatstone bridge sensor and lead to a large offset or signal drift of the sensor output. The sensor signal was measured as a function of the frequency at which the

JVST B - Microelectronics and Nanometer Structures

3107

thermal actuator was driven using a lock-in technique. At dc level, the amount of cross-coupling for a given actuator power is about 3 nV/mA. This relatively low cross-coupling is a result of the good thermal separation between sensor and actuator. VI. CONCLUSION The higher flexural vibrations of piezoresistive AFM cantilevers have been successfully employed in the CD measurement of lithographic features. Furthermore, the obtained experimental results with this cantilever show that utilizing TM–AFM in higher eigenmodes is a suitable technique for realizing high-speed topography imaging. The method we have used for a constant force imaging is based on the constant phase lag at higher eigenmodes of the cantilever. At higher eigenmodes, we can increase the bandwidth of imaging. The use of higher eigenmodes allows to obtain topographic images with high sensitivity and increases the force resolution. We found, that for reliable CD measurement and pattern imaging at higher eigenmodes the use of a phase locked loop 共PLL兲 phase shift detector is very suitable. 1

Q. Zhong, D. Imniss, K. Kjoller, and V. B. Elings, Surf. Sci. 290, L688 共1993兲. 2 M. Tortonese, R. C. Barrett, and C. F. Quate, Appl. Phys. Lett. 62, 834 共1993兲. 3 F. J. Giessibl and B. M. Trafas, Rev. Sci. Instrum. 65, 1923 共1994兲. 4 W. Barth et al., Microelectron. Eng. 57–58, 825 共2000兲. 5 R. P. Reid et al., Int. Solid-State Sens. Actuators Conf., 1997, p. 447. 6 N. Abedinov et al., J. Vac. Sci. Technol. A 19, 2884 共2001兲. 7 A. Volodin and C. Van Haesendonck, Appl. Phys. A: Mater. Sci. Process. A66, 305 共1998兲. 8 K. E. Petersen, IBM J. Res. Dev. 24, 631 共1980兲. 9 T. Itoh, T. Ohashi, and T. Suga, J. Vac. Sci. Technol. B 14, 1577 共1996兲. 10 S. Timoshenko, J. Opt. Soc. Am. 11, 233 共1925兲. 11 R. Linnemann, T. Gotszalk, L. Hadjiiski, and I. W. Rangelow, Thin Solid Films 264, 159 共1995兲. 12 T. Ivanov, T. Gotszalk, T. Sulzbach, and I. W. Rangelow, Ultramicroscopy 97, 377 共2003兲. 13 T. Ivanov, T. Gotszalk, T. Sulzbach, I. Chakarov, and I. W. Rangelow, Microelectron. Eng. 67–68C, 534 共2003兲. 14 T. Itoh, C. Lee, and T. Suga, Appl. Phys. Lett. 69, 2036 共1996兲. 15 T. Fujii, S. Watanabe, M. Suzuki, and T. Fujiu, J. Vac. Sci. Technol. B 13, 1119 共1995兲. 16 M. Despont et al., Proceedings of the MEMS 2000 Conference, Miyazaki, Japan, 2000, p. 126. 17 D. Lange et al., Proceedings of the MEMS’99 Conference, Orlando, FL, 1999, p. 447. 18 K. Y. Yasumura, T. D. Stowe, E. M. Chow, T. Pfafman, T. W. Kenny, B. C. Stipe, and D. Rugar, J. Microelectromech. Syst. 9, 117 共2000兲. 19 W. Riethmuller and W. Benecke, IEEE Trans. Electron Devices 35, 758 共1988兲. 20 Chui et al., J. Microelectromech. Syst. 7, 69 共1998兲. 21 J. Bu¨hler, J. Funk, O. Paul, F.-P. Steiner, and H. Baltes, Sens. Actuators A 46–47, 572 共1995兲. 22 I. W. Rangelow et al., Proc. SPIE 2879, 56 共1996兲. 23 I. W. Rangelow, Proc. SPIE 1392, 180 共1990兲. 24 T. Itoh and T. Suga, in Proceedings of Transducers’95, Stockholm, Sweden, Royal Swedish Academy of Engineering Science IVA, Stockholm, 1995, p. 632. 25 Tzv. Ivanov, T. Gotszalk, P. Grabiec, E. Tomerov, and I. W. Rangelow, Microelectron. Eng. 67–68C, 550 共2003兲. 26 W.-J. Chang, Nanotechnology 13, 510 共2002兲.