This action might not be possible to undo. Are you sure you want to continue?
Printed in the UK
Experimental determination of the parameters of the feedback system of a scanning tunnelling microscope
˜ † and E Anguiano§ A I Oliva†, M Aguilar†‡, J L Pena
´ y de Estudios Avanzados del IPN, Unidad Merida, ´ † Centro de Investigacion ´ ´ Mexico ´ AP 73 Cordemex, 97310 Merida Yucatan, ´ § Instituto Ciencia de Materiales (CSIC), Campus Universidad Autonoma de Madrid, 28049 Madrid, Spain Received 30 August 1996, in ﬁnal form 11 November 1996, accepted for publication 21 January 1997 Abstract. The experimental determination of the main parameters of the feedback system in a scanning tunnelling microscope (STM) is discussed. Knowledge of these parameters allows one to determine accurately the region where unstable STM operation could affect measurements, and also to set the optimal working conditions to obtain high-quality images. Each parameter involved in the feedback circuitry is analysed and discussed as are their mutual interactions. Different working conditions are simulated and analysed in order to determine the parameters needed for stable operation.
1. Introduction The scanning tunnelling microscope (STM) invented by Binnig and Rohrer (1987) is today a standard instrument in surface science laboratories because of its capability to obtain images with atomic resolution. The STM utilizes the high sensitivity of the tunnel current ﬂowing through the gap formed between the sample and a sharp tip. By controlling this current and scanning the tip over the surface of the sample, the topography of the surface with atomic resolution and/or measurements of some electronic properties of the sample can be obtained. Control of the vertical movement of the tip (mounted on a piezoelectric) is possible by a feedback circuit that allows stable and precise movement. The tip movement quality depends strongly on the parameters used for control and the feedback mechanisms employed. A typical feedback system used in a STM is shown in ﬁgure 1 where each component is shown. In general, the feedback system compares a set-up reference (Vr ) with the value given by the actual system. The error signal between them is used to generate a compensation signal in order to obtain the control value in the output that compensates the system perturbations. The use of inadequate values for the control parameters can produce instabilities and imprecise movements of the tip during scanning, yielding false information or even worse, oscillations in the system. Some pioneering works describe models for STM feedback systems. The ﬁrst model was realized by Park
‡ On sabbatical leave from: Instituto Ciencia de Materiales (CSIC), Campus Universidad Aut´ onoma de Madrid, 28049 Madrid, Spain. 0957-0233/97/050501+07$19.50 c 1997 IOP Publishing Ltd
and Quate (1987). However, this model is not helpful in the laboratory because a clear way to obtain some of the parameters used was not described and this imposes severe limitations in the feedback loop. Kuk and Silverman (1989) presented a more realistic model that included the transient response. In this work, for the ﬁrst time the mechanical response of the tunnel junction as a double pole system in order to obtain instabilities was considered. Important parameters in this model are the natural frequency ω0 of the tunnel junction and the quality factor Q. More recently, an experimental method to determine the optimal feedback parameters for the critically damped response has been proposed (Jeon and Willis 1991), thus dispensing with the need to use mathematics to ﬁnd out the analytic solution. Other workers (Hammiche et al 1991, Ping and Player 1993) have used mathematical models as approximations, but they do not explain how to obtain the parameters involved experimentally. A book by Stroscio and Kaiser (1993) offers, in the chapter on instrumentation, a general account of the stability of the tunnel loop suggesting a resonant model characterized by an additional parameter: a factor α related to the phase response. However no hint as to how to locate the stability region of a particular STM is given. In this article we describe experimental work to obtain the main variables involved in a general feedback system implemented on a STM by using a theoretical model previously proposed (Oliva et al 1995). We analyse the case when the mechanical component has two poles (the most simple case) and consequently oscillations. We brieﬂy describe the model and show how to ﬁnd a zone to achieve
Fn ) as ﬁlters. This element gives a signal proportional to the error between this voltage and the reference voltage. 2. but they are expressed as a function of the unknown variables. and several RC poles (F1 . we obtained a relation between the natural or resonant frequency ω0 and the working frequency 2 ω2 = ω0 KI 2 KI + G2 τ ω0 (1) where KI is the integration constant related to the RC factor of the circuit by KI = 1/RC . the value of each parameter. characteristic of each STM system. . instabilities can be present). different elements in the direct loop are used. In the next section we show how to measure these variables. In this case this condition is given by the inverter input of G1 . 3. Control system description and feedback model Figure 1 shows the elements involved in a traditional feedback system on a STM. This value (in volts) is sent to the comparator G1 . we ﬁrst describe the possible forms of measurement of the parameters: .e. . By applying the 502 Nyquist criteria (Nyquist 1932) for stable conditions in the feedback system. The feedback system consists of a proportional–integral controller (β in the ﬁgure) whose output signal feeds the Z piezoelectric (see later). Equation (1) can be written in the usual form corresponding to a damped harmonic oscillator. in a real situation. Experimental procedure We need to know the values of the different variables involved in equation (3) to obtain the stable region of the STM.A I Oliva et al Figure 1. The particular components of a STM system such as the I –V converter. The last equation gives some restrictions on the stability regions of the STM. generating the feedback cycle. In order to clarify the following discussion. In order to minimize the error signal (after comparison). It is necessary to have an inverter element in the circuit for loop stability. ω0 and τ . integral control. As a result the z -piezo will suffer elongation. in terms of the damping factor ζ as (1 − ζ 2 ) = KI 2 KI + G2 τ ω0 . optimal values of the control system in order to avoid instabilities. such as ﬁlters. The tunnel current is detected in the tunnel junction by using a I –V converter with variable gain and its output is sent to the logarithmic ampliﬁer for linearization of the feedback system. Typical feedback loop diagram of a STM. proportional control. . assuming that the piezo has a double pole behaviour (i. the logarithmic ampliﬁer and the tunnel junction are also in the feedback loop. Other element of the circuit are the high-voltage ampliﬁer with gain A to increase the elongation of the piezos. we obtain the condition for stability G0 KI τ + G2 ζ 2 <1 2 2 KI + (1 + ζ 2 )ω0 τ (3) where G0 is the total gain of the closed loop. We then describe the experimental work to be done in order to ﬁnd. this will change the tip–sample distance (δ) and consequently will change the value of the tunnel current. (2) Taking the real part of the Nyquist diagram. etc. . ζ . In a previous paper (Oliva et al 1995) we discussed a mathematical model for the STM feedback system.
with a lock-in ampliﬁer (LA). Cambridge CB4 3NP. (ii) tunnel without feedback—the tip is approached manually to the sample until a tunnel current is obtained. after a short time. The piezoelectric scanner is available in two conﬁgurations: tripod and tube. the response of one axis of the tripod when an oscillating signal was applied to other axis (see ﬁgure 2(a)). With this arrangement we obtained. 755 Ravendale Drive. 3. UK. In this work we mainly used tripods based in different STMs and piezoelectrics with low and high sensitivities and large and short lengths. Mountain View. the tunnel current is lost. French’s Road. Experimental set-up implemented to determine: (a) the frequency response of the mechanical coupling between axes. and (c) coupling YZ . i.STM feedback system parameters Figure 2. we made the mechanical characterization of the STM when no tunnel current was detected.e. and the second letter refers to the axis from which † Chesterton Mills. The degree of rigidity of this mechanical structure is directly related to the value of the ﬁrst mechanical resonant frequency ω0 . scanning and data acquisition as well as the software to control the whole system. The STM head is the mechanical part where the tunnel junction (tip and sample) is located. Coupling between axes As a ﬁrst approximation. A scan in frequency was performed and the transfer function (TF) between axes was obtained by means of the LA. The tube can present higher mechanical resonance frequencies in the Z direction because of its compact body (only one piece). 503 . CA 94043.1. when we mention AB combination or coupling. The drawback is the relative low mechanical resonance frequencies for large scan tripods. Another STM was used with a piezo tube arrangement. small coupling between axes and similar mechanical resonant frequencies for the three axes. The tripod is formed by three elements in an orthogonal arrangement and the advantages of this system are good orthogonality (even for large scanners). This is an unstable condition because the feedback system is disconnected. (i) no tunnel—the tip is far away from the sample and no tunnel current is present. the tip was far away from the sample (condition (i) no tunnel). Figure 3 shows the results obtained with the piezo tripod of the STM from WA ˚ V−1 for the Technology. In fact. Because we developed the STM control. Two commercial STM heads ﬁtted with tripods were used: WA Technology (WA Technology Ltd†) and Hitachi model V-3000 (Hitachi Scientiﬁc Instruments Ltd‡). The next subsections describe the work done to measure the values of the main parameters (usually unknown) involved in the feedback system. However. Figure 3. (b) the ﬁrst mechanical resonance frequency of the STM under tunnel conditions and with the feedback system operating. Usually. The tripod has a sensitivity of 9 A −1 ˚ X and Y axes and of 5 A V for the Z piezo. Figures 3(a)– (c) show the TF region between 1 and 16 kHz. (b) coupling XZ . (iii) tunnel with feedback—a tunnel current is established between tip and sample while the feedback system is operative. the ﬁrst letter refers to the axis to which the signal was applied. each element involved was known to us. ‡ Nissei Sangyo America Ltd. piezo tubes yield nonlinear scans. Coupling between axes in the WA Technology STM: (a) coupling YX . Both types of scanners are used in STM equipment with different conﬁgurations. high rigidity implies a high value of ω0 . large coupling between axes and low X and Y mechanical resonant frequencies. This is the usual mode of operation of a STM. USA. For simplicity.
The experimentally obtained mechanical resonant frequency must be considered when choosing the scanning speed for imaging and setting the parameters of the feedback system. This tripod is small 10 A and compact. XZ and Y Z . at which frequencies a signal in one particular axis will perturb the other axes. Because of this. The ﬁrst peak in the three diagrams appears at 9. the Y X coupling shows little difference when compared to the results obtained from XZ and Y Z couplings for this tripod. Similar results were found when the axes were interchanged. In fact. and also with the feedback system operating. This result is consistent with the kind of tripod analysed and with the qualitative information given by the manufacturer. Further measurements were taken with the same Hitachi STM. XZ .1 were performed with the STM under condition (i) no tunnel. and (c) coupling YZ . under conditions (ii)—tunnel without feedback—and (iii)—tunnel . sensitivity and mechanical mounting made the Z axis completely different to the X and Y axes. The insets in ﬁgures 5(b) and 5(c) show a magniﬁcation of the main peak at 2.A I Oliva et al Figure 4. because the image information originates from this axis. and Y Z combinations. and 300 A ˚ V for Z (900 A piezo). However. A small peak at 9. 3. the higher peak appears near 15 kHz. i. Figures 5(a)–(c) show the results obtained where resonance at a lower frequency is 504 observed.e. has a high frequency and is thus useful for high scan rates. The most important response is that including the Z axis. Again. Coupling between axes in the Hitachi STM with the ﬁne tripod: (a) coupling YX . and (c) coupling YZ . (b) coupling XZ . A magniﬁcation of the two peaks appearing between 15 and 16 kHz is shown in the inset of ﬁgure 4(b). Figures 4(a)–(c) show the same axes combinations Y X .5 kHz appears in the Y X coupling. Coupling between axes in the Hitachi STM with the coarse tripod: (a) coupling YX .2. This piezo tripod.5 kHz and a larger peak is found near 14 kHz. usually the size. more interesting experimentation needs to be done when the STM is under the tunnel condition. the results obtained in this work indicate the opposite: it has the lowest mechanical resonance frequency. We can conclude that the average resonance frequency of this piezo is about 15 kHz—higher than that of the previous tripod. it is very important to measure the coupling response between axes. The piezo tube scanner is usually expected to have a higher mechanical resonant frequency than the tripod scanner. However. The measurements described give information about the possible coupling and transfer function between axes that can affect the topographic information. the resonant frequencies obtained yield information about the mechanical coupling between axes.75 kHz. The same method was used to characterize the ﬁne tripod of the Hitachi V-3000 STM with sensitivities of ˚ V−1 (X. This tripod has long axial lengths and consequently should have a lower mechanical resonance frequency than the previously described tripod. Similar behaviour for XZ and Y Z combinations can be seen along the scanned frequency range. with a ﬁrst peak at 9. Figure 5. Thus. the signal was measured.5 kHz. i.e. (b) coupling XZ . Y ) and 5 A ˚ V−1 (Z). but in all three cases. First mechanical resonance frequency The experiments described in section 3. Figure 3 shows the responses measured in Y X . but using a coarse tripod with a lower sensitivity ˚ V−1 for X and Y piezos.
By using equation (5) we obtain that the resonant frequency.e. Since we have already measured f0 experimentally we can obtain the value of ζ . Delay time and damping factor We have described a way to obtain ω0 . of the two tripods will be respectively f0 = 20/ζ kHz and f0 = 5/ζ kHz. Note that peaks at other frequencies disappeared and only the main peak and its harmonics appear well deﬁned. it is possible to see that the tunnel signal is perturbed at the moment of the step. Our results. Stroscio and Kaiser 1993).2 and 10. This is shown in ﬁgure 2(b). However. The damping factor ζ of the response can also be obtained directly from the experiment (without using the known value of f0 ). Afterwards. two more parameters. it can be seen that the resonance frequency for tunnel conditions with feedback is at a higher frequency than when the STM is not in the tunnel condition. the damping factor ζ will be δ ζ =√ 4π 2 + δ 2 (8) 505 . Several curves obtained by this procedure were analysed to measure the damping behaviour. by using the logarithmic decrement method. we can deduce the value of the damping factor if we know two consecutive amplitudes of the damping motion by means of the following relation: δ = ln X1 = X2 2π ζ 1 − ζ2 . by using an oscilloscope. cases with no tunnel and with tunnel current plus feedback. obviously the STM has to be under condition (iii) tunnel with feedback.e. a larger signal will appear in the X and Y piezos. by observing the signal induced in one of those piezoelectrics the tripod mechanical resonant frequencies can be obtained. To make a comparison with the no tunnel condition. Basically. They found that the resonant frequency is the same in both cases regardless of whether the feedback system was operating or not. are needed. In this case we know the combined action of the mechanical system.STM feedback system parameters with feedback. we compared conditions (i) and (iii). Because of the deﬁnition. a square signal of very low frequency is added to the control signal to simulate a step. the feedback system was prepared following the recommendations of Pohl (1986). (7) Thus. Thus. it is therefore necessary to repeat the experiment with our STM system. From the basic literature about harmonic motion. i. The experiments were realized ˚ V−1 by using the Hitachi STM with a piezo of 300 A sensitivity. The results are 0. the amplitude of the damped oscillation follows the equation (Thomson 1981) X = X0 [exp(−2π t/τ ) sin(2π t/τ + φ)] (4) where τ is the decay time of the oscillation. Because only the ﬁrst mechanical resonant frequency is known. imply that the shift in the mechanical resonance is due to the tunnel process and not to the feedback system. to use equation (3). Thus. Figure 6 shows the amplitude of the signal induced in the X axis as a function of the frequency of the oscillation applied to the Z axis. We measured this parameter by using the following method: with the STM in tunnel conditions and the feedback system operating. These values are of the same order of magnitude as others reported in the literature (Park and Quate 1987. A clear main peak at 3. f0 = ω0 /2π . i. Then. Comparisons between conditions (ii) and (iii) have been reported in the literature (van de Walle et al 1985. The measurements were made with the STM in tunnel conditions and with the feedback system operating (condition (iii)). ζ and τ .7 respectively. When the whole tripod is in resonance. In fact. For this. a scan in frequency is made. Thus. together with other reported results (van de Walle et al 1985). This is a very important result: the actual ﬁrst mechanical resonant frequency must be obtained while the STM is in the tunnel condition. The value of τ is obtained by measuring the decay time of the oscillation intensity. the experimental set-up used is similar to that shown in ﬁgure 2(b) with an oscilloscope (or an A/D converter) connected in parallel with the I –V converter. The response that appears in the scan piezoelectric is measured as in the previous case.8 kHz can be seen clearly. By comparing ﬁgures 6 and 5. Both τ and τ are related to the natural frequency of oscillation of the system ω0 and the damping factor ζ by the following relationships: 2π/t = ζ ω0 2π/t = ω0 (1 − ζ 2 )1/2 . The tunnel signal will be a damped oscillation and the recovery time will be a measurement of the delay time τ . 3.3. good values for damping.6 kHz and harmonics at 7. a sinusoidal signal is added to the control signal at the output of the control unit. The amplitude of the signal that appears in the tunnel current can also be measured: the peaks obtained correspond to a resonance. The delay time τ is the time necessary for recovery of the tunnel current signal when an instantaneous perturbation affects the stability of the tunnel junction.9 and 0. van Kempen and van de Walle 1986). the delay time of the perturbation and τ the period of the damped oscillation. the tunnel junction and the electronic elements involved in the feedback circuit. Any other method based on mechanical resonance of the STM head—without it being in tunnel conditions—will not yield the value necessary to perform feedback analysis. we devised a different experimental set-up to perform the experiment under condition (iii). the value of the resonance frequency measured in the tunnel condition is necessary for the design and use of feedback control and needs to be considered in the selection of the scan rate. This result should not be misinterpreted in the light of work published by the van de Walle group (van de Walle et al 1985) who compared the resonance frequency between conditions (ii) and (iii). (5) (6) The values of the delay time measured on the Hitachi STM were 50 ms for the ﬁne tripod and 200 ms for the coarse piezo. however.
58 2004–9 . Lex Prix Nobel en 1986 (The Nobel Foundations) pp 85–111 Hammiche A. A main peak at 3. Instrum. Spain (grant TIC95-09960-E). References Binnig G and Rohrer H 1987 Scanning Tunneling Microscopy from Birth to Adolescence. Wei Y. Sci. Acknowledgments This work was made possible by the support of CONACyT (M´ exico) by grants 211085-5-4483A and 2362P-A. In particular. and CICYT. Thus. From this work. Usually. 60 165–80 Nyquist H 1932 Bell. Instrum. We obtained the ﬁrst resonant frequency by using the experimental set-up shown in ﬁgure 2(b).3 and ω0 = 14. the resonance frequency is f0 = 2. the performance of a STM cannot be assumed by consideration of the scanner design only—the response of the complete mechanical structure must also be measured. 62 3010–21 Jeon D and Willis R F 1991 Feedback system response in a scanning tunneling microscope Rev. the STM with the piezo tube had a lower resonance frequency than all the STMs with tripods analysed in this work. Instrum. 11 126 Oliva A I. Instrum.A I Oliva et al Figure 6. Sci. Valencia M A and Pe˜ na J L 1996 A new STM design for atomic resolution Instrum. Aguilar M and Pe˜ na J L 1995 Analysis of scanning tunneling microscope feedback system Rev. i. Sci. where δ is the logarithmic decrement and X1 and X2 are two consecutive amplitudes. These results are interesting because it is usually argued that a STM built with a piezo tube scanner has a higher mechanical resonance and (as a result) higher performance than one incorporating a piezo tripod.6 kHz and its harmonics were found. Devel. Denisenko N. Conclusions In this work the parameters which must be considered for analysis of a STM feedback control system have been discussed and procedures to measure these values have been described. However. we repeated this procedure for a piezo tube scanner instead of a tripod with a STM built in our laboratory (Oliva et al 1996). This is for the design and to obtain stable working conditions. By using this method we can ensure the consistency of both measurements.3 kHz—a value lower than that in the case of the measured tripods. Instrum. we conclude that it is necessary to have a through characterization of the STM employed and control of the parameters used in the feedback control in order to obtain reliable images. we get ζ = 0. Response of the whole STM feedback loop system in tunnel conditions with the feedback system operating. Wilson I H and Webb R P 1991 The Surrey STM: construction. It should be remembered that ζ is related to the integration constant K1 and the gain G2 through equation (2). when the resonance peak is damped. development and evaluation of a scanning tunneling microscope Rev. 62 1650–1 Kuk Y and Silverman P J 1989 Scanning tunneling microscope instrumentation Rev. 66 3196–203 Oliva A I. Knowledge of the parameters involved assists the reliable and stable operation of a STM. 3 32–6 Park S and Quate C F 1987 Theories of the feedback and vibration isolation systems for the scanning tunneling microscope Rev. We checked that the result was the same for the three tripods. the bandwidth is diminished and this situation presents a comprise between stability and good scanning rate.68 rad/s. Finally. Anguiano E. Syst. Sci. Sci. Thanks are also given to the DGICYT-MEC (Spain) for partial support to the stay of M Aguilar in CINVESTAV IPN-M´ erida. Experiments performed with the STM in tunnel conditions enable the real resonance frequency for the whole feedback loop system to be 506 determined.e. From the curve obtained. 4. Corona E. we proposed an experimental method to ﬁnd coupling frequencies between axes in two types of scanners to ﬁnd their resonance frequencies. Tech.
Res. Develop. Instrum. NJ: Prentice-Hall) van Kempen H and van de Walle G F A 1986 Applications of a high-stability scanning tunneling microscope IBM J. 30 509–14 van de Walle G F A. Sci. 56 1573–6 507 .STM feedback system parameters Ping G and Player M A 1993 Control system analysis of a scanning tunneling microscope Meas. Gerritsen J W. Develop. van Kempen H and Wyder P 1985 High-stability scanning tunneling microscope Rev. Sci. Technol. Res. 30 417–27 Stroscio J A and Kaiser W J (eds) 1993 Scanning tunneling microscope Methods of Experimental Physics vol 27 (New York: Academic) Thomson W T 1981 Theory of Vibration with Applications 2nd edn (Englewood Cliffs. 4 677–81 Pohl D W 1986 Some design criteria in scanning tunneling microscopy IBM J.
This action might not be possible to undo. Are you sure you want to continue?