Observation and Manipulation of Gold Clusters with Scanning Tunneling Microscopy A Project Report Submitted to the Faculty of the Graduate School of the University of Minnesota By Dogukan Deniz In Partial Fulfillment of the Requirements for the Degree of Master of Science May 2012
Dogukan Deniz 2012
Acknowledgment I want to thank Dr. Maps for his help and patience during this project, and thanks to every other person who showed support in their own way.
Abstract In this project gold films deposited on mica are investigated. Sampling techniques, adjustment of the microscope for different observations, observation techniques and finally manipulation of gold surface with a scanning tunneling microscope (STM) tip are presented in this report. Observation and manipulation abilities of the microscope are found to be strongly dependent on the thickness of films and so on the cluster sizes. The parameters of STM are adjusted to get the best scan results by considering the trade off between the scanning features. Topographic surface pictures of gold films are obtained to show the effect of these parameters.
Table of Contents Acknowledgment Abstract I. Introduction... 1 II. Background... 4 III. The Sample... 14 a. Physical Vapor Deposition of Gold on Mica... 14 b. Surface Thickness and the Effect of Annealing.... 14 IV. Imaging and Surface Manipulation... 17 a. Manipulations... 17 b. Trade Off Between the Scan Features... 22 V. Conclusion... 25 References
I. Introduction In this project, the initial efforts have been carried out to deposit and study gold films on mica substrates with a long term goal of studying electromigration (the motion of an indention on the thin film surface in the presence of a current through the film) techniques to create recognizable features on the gold surface were also explored. However, the indentations obtained in this experiment were never small enough to observe electromigration (Figure 1). Therefore the primary goal of the project became manipulation of the surface with an STM tip to create suitable markers. 1
Figure 1. Electromigration, the motion of an indentation on a thin film surface in the presence of a current. In order to observe this motion the indentation needs to be on the order of several nanometers. A nanometer sized mark would be just a few atoms in size, hard to truly identify and track unless truly atomically flat regions are present. 2
Preparing and observing thinner metal samples on different substrates has always been a challenge for physics and chemistry experimentalists. With the spectroscopy tools developing over the last few decades, this challenge is currently carried on at the atomic level. Various vapor deposition techniques are being used for coating or doping purposes in experiments today and the quality of equipment play an important role on thickness and quality of the films. Physical vapor deposition (PVD) is used to obtain thin gold films on mica substrates in this experiment. The scanning tunneling microscope (STM) has been used to obtain atomic scale images with the help of a tunneling current, a feedback mechanism (depending on the choice of the microscope mode) and piezoelectric controllers. Since its invention [1] STM has not only been used to observe surfaces in small scales but also been used to manipulate surface atoms in atomic scale, with the help of the scanning tip. [2] Since such a manipulation changes the work function between the tip and sample and probably the tip shape, the tunneling current and so the feedback is affected quickly by such a change. When a change in feedback exceeds the limits of piezoelectric controllers, scanning after the manipulation becomes impossible for that area with that STM, since the piezoelectric voltage range is saturated. Various manipulation techniques are tried to overcome this effect and the results are presented. 3
II. Background The STM has been used as a powerful microscopy tool since it was invented by by Binnig and Rohrer. [1] Although it is a noisier setup compared to the its successor, atomic force microscope (AFM) [3], STM has still been chosen for many experiments because of its consistent and well known interaction parameters with the surfaces, and because of its ability to control the tip over rough surfaces. An STM consists of a scanning tip connected to piezoelectric positioners which are connected to computer-generated controls and a feedback circuit. Vibration isolation systems are used for all STM models. High vacuum and low temperature setups may be required for specific experiments, but they all can still be considered as parts of a complete STM setup. Here we focus on the tunneling components. When a bias voltage is applied between the tip and sample, for a small distance s, a tunneling current can be observed. Binnig and Rohrer started their project for STM with the measurement of the tunneling current from a tungsten tip to a platinum surface. [4] They observed the expected exponential behavior in tunneling resistance and this led them to use the dependence of this current on the distance between the tip and sample to plot surface topography. The distance between source (tip) and drain (sample) in Figure 2 can be considered as the distance between the tip and sample. This distance is on the order of an angstrom. 4
Figure 2.[8] The exponential decay of the electron wave function through a tunneling barrier. The source in the figure is the scanning tip and the drain is sample. When z changes by an angstrom the tunneling current changes by an order of magnitude, therefore this setup can be used to reveal atomic size structures. 5
The tunneling current density J T is given as: J T exp( 4 π ( 2m ϕ) s),(1) h and this shows the exponential dependence on tip-sample distance s. ϕ is the work function between the tip and sample and m is the free electron mass. [1] Figure 3 shows two different modes of an STM. The amplitude of tunneling current can either be used to plot the surface topography (constant height mode) or it can be used in a feedback circuit to keep the distance between the tip and sample constant (constant current mode), in this case the feedback output reflects the surface topography. 6
v Figure 3.[9] Constant current and constant height mode STMs. The distance between the tip and sample is kept constant by a feedback loop in the first figure. The tip moves at a constant height in the second figure, and the tunneling current changes depending on the distance between the tip and sample. 7
Figure 4. STM diagram used in the first STM report by Binnig and Rohrer. [1] 8
The first microscope was designed to be a constant current mode STM. In Figure 4 s denotes the distance between the tip and sample and J T shows the corresponding tunneling current density between them. In order to keep this current density in J T ±Δ J range ( Δ J J ), the feedback control unit ( CU ) adjusts the piezodrive P z by changing V p, and this change moves the tip in the z direction by Δ s, in proportion to the voltage change ( Δ V p ) applied by the feedback loop. Even though the piezoelectric design used in our experiment is quite different from the one in Figure 4, it has the same working principle and it has 600 nm range for Δ s and 200 V range for V p. This corresponds to 3 nm change in z for 1V change in V p. Considering the height s to be on the order of an angstrom and the amplitude of tunneling current to be exponentially proportional to the height s, any difference in s as tall as an atomic step (~ 3A ) during the scan can change the amplitude of tunneling current by an order of magnitude. Therefore, this setup can be used to obtain topographic plots that have atomic resolution. In Figure 4 feature A shows a step in the surface which can be detected in a step length δ for a scan in the y direction. In order to investigate the resolution limit for this structure, scanning parameters can be considered as following; for n data points taken in a scan length of l, the length of each step between the data points can be taken as l / n, the voltage applied to P y is in discrete steps. The z feedback loop tries to pull or push the tip depending on the surface structure at this point and the tip moves l / n in y direction after recording the z data. The length l /n limits the resolution of an STM image. However decreasing this value does not always increase the resolution since the data collection rate is limited by the cutoff frequency of the feedback loop, the data collection rate of the data acquisition device, and the minimum voltage step produced by voltage amplifiers for x and y piezodrives. Considering the time spent to complete one scan in y direction to be Δ t, then the time spent for each step is Δ t/ n. If the time constant of the feedback loop is smaller than this time interval, then a data point can be recorded for length l /n. However, if the period of the feedback is longer than this time interval, then 9
the structure in a step length δ can not be reliably resolved by the STM. Similarly, if the discrete voltage steps for one direction are not accurately applied to move the tip l /n further, then the data collection device over samples the data point for every point in that scan and resolution is lost for the small structures. The data collection rate is limited by the data collection speed of the data acquisition device. Therefore, the time interval of the scan along one direction needs to be adjusted depending on these parameters. The area labeled as C in Figure 4 shows a different work function region, which means when the tip is on top of this region, the amplitude of tunneling current will change compared to an area which has the same height but regular material. Binnig and Rohrer [1] called these types of areas as work function mimicked structures and in order to reduce the effect of these type of structures during the scan, they suggest modulating s faster than the feedback frequency. This means instead of using only a constant bias voltage for V T, a small amplitude sinusoidal component can be added. When the frequency of this modulation is higher than the cutoff frequency of the feedback loop, different work function areas can be neglected by the feedback loop, or at least the effect can be minimized. Modulating the bias voltage and then using a lock in amplifier to demodulate the signal is a known way to decrease the noise in STM setups, however in this experiment, since manipulation with the tip and the tip-sample interactions are also considered as a part of the experiment, modulation of the bias voltage is not used. V T, the bias voltage, is the last parameter that affects the resolution depending on the surface structures. Increasing the bias voltage is useful for getting data without approaching the surface so closely. At fixed z, it increases the amplitude of the tunneling current: V T =0.23V and I T =1.2nA, compared to V T =0.1V and I T =1nA. This increased current goes into the feedback loop and it brings V p to the saturation value faster. However, since the height difference changes at different locations, V T can be considered as a parameter which needs to be adjusted experimentally for the best resolution at the specific area. 10
Figure 4 shows the scanning process when the tip is controlled by the piezodrives. But in order to get the tip within 100 nm of the sample, a coarse approach mechanism is needed. The coarse approach mechanism consists of a motor-driven screw that is controlled by computer that can also read the tunneling current. After mounting the sample the coarse approach mechanism starts pushing the sample till the feedback loop starts detecting the tunneling current. After the tip enters the region where the tunneling current can be detected [5], the approach can be stopped. Since the scanned region is at most on the order of a micrometer, the coarse approach can be run once per an area, because starting the coarse approach almost certainly brings the tip to a different place than the previous scanning area. So it is not possible to find the same location again once the coarse approach motor is started. Figure 5 shows a saturated scan example after which the coarse approach motor needs to be started. 11
Figure 5. The saturation effect seen on top left corner of the scan. 12
A tungsten tip is used in most of the scans. Chemically etched Pt-Ir alloy tips are also used, but tungsten tips are found to be giving better results for gold samples during our experiment. The tip quality is an important factor to have a high resolution and a stable scan. Having a sharp tip edge means the tunneling occurs only at one point on the tip or at places which are very close to each other. However, having a sharp tip for the manipulation part of our experiment decreases the possibility of a successful scan after the manipulation. Therefore, instead of having chemically etched tips, tungsten wires that are sharply cut are used in this experiment. The last part of a complete STM setup is the vibration isolation mechanism. Vibration is the most influential outside parameter that can change the result of a scan. In this experiment air filled tubes are put between an aluminum block and a heavy table to reduce the effect of vibration. 13
III. The Sample a. Physical Vapor Deposition of Gold on Mica Physical vapor deposition is a technique in which the coating material is heated up to its evaporation temperature in high vacuum and the evaporated material sticks on the substrate, which is held above the source. For this experiment gold flakes are heated in a tungsten boat under 5 pascal for 45 minutes by increasing the current running through the plate up to 3A. Freshly cleaved V-4 level mica substrates are held under a mask, above the evaporating gold. Depending on the time of deposition, temperature and pressure; two sets of different gold films are obtained. Through the resistance measurements an average thickness value is estimated to be 20 nm. Compared to the gold images given so far this thickness should be considered as an average thickness of a continuous line. b. Surface Thickness and the Effect of Annealing PVD is a commonly used technique for making gold films and examples of very thin films which are obtained with this technique can be found in the literature. [6] The second step of making thin films by PVD is the annealing. Before annealing the gold atoms form clusters on the surface as seen in Figure 6, and the probability of finding atomically flat areas is quite low for this type of films. Once annealing is done the gold atoms start filling the gaps and form areas which are more planar compared to the samples that are not annealed. However, annealing is a temperature and material dependent procedure and the annealing time needs to be chosen carefully along with these parameters. [7] A propane flame is used along with a thermocouple to anneal the surface. The surface is annealed at 720 K for about 1 min. The change in the formation of clusters can be seen in Figure 7. 14
Figure 6. The formation of gold clusters before the annealing. 15
Figure 7. Gold clusters are still present between flat regions of gold after the annealing. However an offset value can be used in STM to move the tip to the flat regions. The clusters in the previous figure are quite tiny compared to the ones in this figure. The small clusters apparently have merged into larger regions after the annealing and they formed longer and planar gold clusters on the surface (the red rectangle). Annealing can also form flat regions along with long and thin clusters, these type of areas can be reached by an offset value (the yellow rectangle). The green region shows the clusters which are as big as the scanning area in Figure 6. 16
IV. Imaging and Surface Manipulation a. Manipulations The next step of the project is to use the STM to make manipulations and observations on the sample. In order to be able to make a good manipulation on the surface, firstly the samples need to be made flat or the most planar area on the surface should be found. Annealing technique is applied to achieve flat areas on the surface. After the annealing, a flat area on the surface was looked for. One of the most planar areas found during one of those scans is given in Figure 8. 17
Figure 8. A relatively flat area on the gold surface, the high scale is due to the saturation on top of the sample. Z range is ~50 nm, but the area has a uniform color in desired region, therefore the saturated part can be ignored and z range can be assumed to be much lower. 18
After finding a flat area and having stable scans on that region, like the one in Figure 8, the manipulation process is started. Our first manipulation technique was to make an indentation by touching the surface with the tip. However, the change in the work function and the tip shape, because of the transfer of Au atoms to the tip, saturated V p and coarse approach had to be carried to pull the tip back to the normal range of V p. Chemically assisted field evaporation (CAFE) technique used by Hasegawa et. al. [6] is a common technique to manipulate thin gold films on mica surfaces. It is done by increasing the bias voltage an order of magnitude for a very short time. This creates short lived indentations on the surface. Trials of the CAFE technique didn't give us any observable results since the technique is good for thin films and very planar surfaces. The last manipulation technique is to draw a line with the tip on the surface instead of trying make a hole. But in order to avoid the tip-sample contact, this process can be done in two different ways after increasing the bias voltage; one is to run the tip over the surface when the tip is fully retracted, the other is to run the tip over the surface while the feedback is kept on. The first technique did not reveal the indentation on the surface since the tip-sample interaction was not strong enough to make a manipulation on the surface. The second manipulation technique was chosen because of a shorter tip-sample distance, which means a higher interaction force. This technique needed an adjustment of the period of the signal used on the piezodrives. Because, when the bias voltage is increased by an order of magnitude (0.3V to 3 V), the z voltage V p reaches its limit very quickly, therefore there is a very short time to manipulate the surface before the saturation (before the tip-sample distance is increased by the z feedback). This short time was experimentally measured several times and an average value (6s) is assigned to the period of the piezodrive signals to draw the line. After this value is set a triangular wave with the amplitude of 13.5 V is run through x and y piezodrives. This corresponds to a line of length 21 nm along x=y direction, although the manipulation region appears a little longer than that. Every unsuccessful manipulation trial means repeating the whole procedure again starting from the coarse approach. A manipulated surface which could be 19
observed successfully after the saturation limit was avoided is given in Figure 9 which shows the manipulated surface for Figure 8. For the 50 nm x 50 nm region in Figure 8, manipulation is done along x=y line up to x=y=25 nm. The manipulation, the tip-sample interactions and the effects of the restoring surface force (long-range elastic interactions[6]) are visible in Figure 9. 20
Figure 9. Observation of manipulations along x=y line at four different times. (a) The manipulation region right after the manipulation trial. Since the work function between the tip and sample is changed, (some gold atoms possibly stuck on the tip) saturated scan lines are visible around the manipulation region. (b) After 10 minutes. The saturated scan lines are narrower, this most likely means the tip is cleaning itself during the scan. (c) After 25 minutes. Displacement of the sample with respect to the tip. This shows a possible external drift which limits the observation time (explained in the next figure). (d) After 2 hours. The restoring force is expanding the manipulation region. Possible damages to the surface along x direction can be seen around the manipulation region or saturated scan lines are still present. 21
b. Trade Off Between the Scan Features The more time spent on one pixel ( Δ t/ n ) means the more accurate the feedback output will be, but it will slow down the imaging process. This shows the trade off between the scan speed and the resolution. The main challenge to observe manipulations is to observe them before they disappear. Since a fast scan is needed to observe the manipulations, the trade off between a fast scan and a good resolution needs to be considered for each scan. Another challenge during the manipulation is to keep the feedback running while the bias voltage is an order of magnitude higher compared to its normal value. Increasing the bias voltage to this value causes the feedback output to go up to the maximum value for z piezodrive (full retraction), and this means saturation along z direction. For manipulation, it means ignoring the surface structures and so failure of a uniform manipulation along the manipulation direction. Therefore, the other trade off is between the bias voltage and the saturation of z piezodrive voltage. Combining overall, depending on the need (scan speed, resolution, less saturation limit) the parameters (feedback frequency, number of data points, settling time, bias voltage) need to be adjusted carefully. The quality of the sample (purity and uniformity) is another parameter which affects the scan for any purpose. External parameters like the temperature or vibration can affect the results of the scan dramatically. Since the samples holders are made of brass, they can be affected from a temperature change and since the scanning area is on the order of a nanometer, the thermal expansion due to a temperature change can cause a displacement of the sample with respect to the tip. A change of 1K in temperature for a 5 cm long brass causes a thermal expansion of 100 nm. In Figure 10 such a motion is shown for a graphite sample. 14 nm displacement corresponds to a 0.14 K change in temperature. When the sample is a thin metal film, the tip-sample interactions become much stronger. Eigler et. al. [2] used this interaction to move single xenon atoms on a nickel surface at cryogenic temperatures. For our case, this interaction caused the sample to be 22
manipulated slightly by the tip after each scan. Therefore, the consecutive scans over the gold sample give slightly different results. If the interaction is strong enough to move a gold atom to the tip and this might be the reason of saturated scan lines along x direction. 23
Figure 10. The motion of a graphite sample along x=y line, due to the external effects like thermal expansion. (a) to (d) Motion of the sample along x=y line in 25 minutes. 24
V. Conclusion Successful scanning tunneling microscopy of gold films deposited on mica depends on the microscope and sample features. Besides the well known speed-resolution trade off for any scan, a quick saturation of z voltage needs to be considered for thicker (non-flat and clustered) gold samples. This saturation effect might make the scanning impossible after a manipulation attempt with an STM tip. Therefore, the feedback abilities of the microscope needs to be considered and others parameters of the manipulation needs to be adjusted for the manipulation that would make a scan possible after being done. Along with the adjustment of the parameters, having a flat gold sample increases the possibility of a successful manipulation and a following successful scan. So, annealing techniques are applied to obtain flat gold samples. 25
References 1. G. Binnig, H. Rohrer, Ch. Gerber, and E. Weibel, Phys. Rev. Lett. 49, 57-61 (1982). 2. Eigler, D. M. & Schweizer, E. K. Nature 344, 524 526 (1990). 3. G. Binnig, C. F. Quate, C. Gerber, Phys. Rev. Lett. 56, 930 (1986). 4. Binnig G, Rohrer H, Gerber Ch and Weibel E Appl. Phys. Lett. 40, 178 (1982). 5. G. Kästle, H.-G. Boyen, A. Schröder, A. Plettl, and P. Ziemann, Phys. Rev. B 70, 165414 (2004). 6. Y. Hasegawa and Ph. Avouris, Science, New Series, Vol. 258, No. 5089 (Dec. 11, 1992) 7. Li Nan et al 1997 Acta Phys. Sin. (Overseas Edn) 6 531 8. http://wandlowski.dcb.unibe.ch/research/equiptment/stm/html 9. http://www-public.rz.uni-duesseldorf.de/~pc1/stm/stm.html 26