Nanotechnology 10 (1999) 458–463. Printed in the UK
Atomic force microscopy lithography as a nanodevice development technique∗ A Notargiacomo†, V Foglietti‡, E Cianci§, G Capellinik, M Adami†, P Faraci†, F Evangelisti‡k and C Nicolini¶ † Polo Nazionale Bioelettronica, Via Roma 28, 57030 Marciana (Li), Italy ‡ Istituto di Elettronica dello Stato Solido (IESS), CNR, Via Cineto Romano 42, 00156 Roma, Italy § Osservatorio Astronomico di Roma, Via Frascati 33, 00040 Monteporzio Catone, Italy k Unit`a INFM, Dipartimento di Fisica ‘E Amaldi’, Universit`a di Roma TRE, Via Vasca Navale 84, 00146 Roma, Italy ¶ Institute of Biophysics, University of Genoa, Corso Europa 30, 16132 Genova, Italy E-mail: [email protected]
Received 19 April 1999, in final form 9 August 1999 Abstract. Nanoscale science and technology is today mainly focused on the fabrication of
nanodevices. Our approach makes use of lithography processes to build the desired nanostructures directly. The fabrication process involves an electron-beam lithography technique to define metallic microstructures onto which nanometre scale patterning is performed using an atomic force microscope (AFM) as a mechanical modification tool. Both direct material removal and AFM-assisted mask patterning are applied in order to achieve the smallest possible separation between electrode pairs. The sample preparation involves a polymer deposition process that results in conformal growth and in surface roughness comparable to that of the substrate. The results of the application of this technique show that the process is reproducible and exhibits a good operation control during the lithographic steps, both ensured by the imaging facilities of the AFM. The nanolithography technique has been used to fabricate nanogap electrodes to be used for molecular devices. The study reported here can be considered as a reliable starting point for the development of more complex nanodevices, such as single-electron transistors.
Recently, much effort has been made in the field of nanotechnology for the development of nanodevices, due to interests in both fundamental physics (low-dimension structures, single charge effects, etc) and applied research (ultralarge-scale integration, high-density memory storage, etc). The next generation of new devices will require a circuit patterning resolution and a positioning accuracy ranging beyond the limits of the present fabrication processes. However we must consider that industrial processing and analysis techniques, such as optical and electron-beam lithography and scanning probe microscopy (SPM), will not easily be replaced by completely new solutions. A practical way consists in an optimization of the above-mentioned techniques to take advantage of their peculiar features. In particular, the integration of conventional lithography processes with SPM-based techniques allows the definition of circuit patterning down to a nanometre scale. This represents a promising way towards the fabrication of new concept ∗ This paper is based on work presented at the First ELBA Foresight Conference on Molecular Nanotechnology (14–16 April 1999, Rome, Italy).
© 1999 IOP Publishing Ltd
devices for electronic application and the improvement of conductivity measurement techniques, even on a molecular scale. Applying different principles, several attempts were made using SPM to induce modification of oxides, semiconductors and metals on the nanometre scale: lowenergy exposure of resists , thermo-mechanical writing , local oxidation [3, 4], mechanical modification  and nanomanipulation . The atomic force microscope (AFM) nanolithography technique takes advantage of the imaging facility and the ability of moving a probe over the sample surface in a controllable way. In the next paragraph we report on the investigation of direct writing, where the AFM probe ‘scratches’ a metal stripe. For ‘scratching’ we mean the mechanical action of the tip that is used as a sharply pointed tool in order to produce fine grooves. The direct scratching is possible with high precision but low-quality results are obtained due to probe wear during the lithographic process. Another solution is the addition of a soft resist polymer, in general spun PMMA, as a mask for the etching processes, thus reducing tip damage but precluding an accurate alignment to the structures underneath.
Atomic force microscopy lithography
Following as a starting point the work of Hu and coworkers [7, 8], a further investigation has been pursued on a two-layer mask (polymer plus thin metal film) deposited on the sample. The AFM is used to scratch these layers, thereby forming a pattern which is transferred to the underlying sample by a subsequent etching process. A suitable soft polymer, as a part of the multilayer structure, ensures an increased probe lifetime. It is possible to combine a low tip wear with the exact location of the structures underneath, onto which modifications are desired. For this purpose we used a polymer plasma-deposition technique that resulted in a conformal growth and in a surface roughness comparable to that of the substrate. The results of the application of this technique show that the process is reproducible and exhibits a good operation control during the lithographic steps, both reproducibility and control ensured by the imaging facilities of the AFM. Our goal is the application of this technique to obtain very narrow gaps on metal stripes that can be considered as electrodes covered by a thin native oxide layer forming a natural tunnelling barrier. A future step will be the electrical characterization of the nanometre-size gap after the deposition of organic islands.
2. Nanometre-size gaps by direct material scratching
All the nanolithographic processes and imaging in this paper were performed in ambient air using a commercial AFM, a Digital Instruments ‘Dimension 3100’ driven by a Nanoscope III controller, and silicon cantilevers. As mentioned previously, the first nanolithography process investigated was the direct modification induced by scratching a sample. The samples were prepared by standard electron-beam lithography on 200 oxidized silicon wafers. On these insulating substrates we fabricated metal stripes ending in square pads for electric contacts. A scanning electron micrograph showing a sample consisting of an aluminium stripe with electrical contact is reported in figure 1(a), while an AFM topography shows in figure 1(b) a detail of the very middle region of the sample. When cut by AFM lithography such stripes will act as electrodes separated by a narrow gap. Indeed it is possible to remove material from a metal stripe by applying an amount of force of several µN . Direct material removal was performed in contact-mode using a 125 µm long silicon cantilever with a high value of the spring constant ranging from 20 to 100 N m−1 . The pyramidal silicon tip had a radius ∼10 nm and a sidewall angle of 35¡. An accurate control of the scanning parameters also allowed the use of the same cantilever for imaging before engraving without significantly wearing the sample. Although the topography is not perfectly defined it is good enough for a proper alignment to the scratching site. The scratches were made at a rate of 0.1 µm s−1 . Different rates were tried but resulted in a bad cut-profile and/or material removal. An exact evaluation of the force
Figure 1. Scanning electron micrograph of a metal stripe with electrical contact pads (a). The whole structure is about 1 mm long. The three-dimensional view (b) shows a 6 × 6 µm2 detail of the middle region obtained by an AFM scan.
applied is not possible because the cantilever spring constant is not well known. Moreover, for a large amount of cantilever bending the −k · (z − z0 ) description for the applied force is not accurate. However, a rough estimate using a typical spring constant value of about 40 N m−1 for that type of cantilever leads to a value of ∼25 µN as the force necessary to scratch an aluminium stripe, a value very similar to that proposed by Bruckl et al . The regions of the stripes where engraving was performed were about 10 nm thick and 80 nm wide. To obtain a good material removal it was necessary to pass several times over the same scratching site resulting in a widening of the gap. A series of six writing passes produced nanogaps of the order of 60 nm. In figure 2 it is shown a three-dimensional view of a metal stripe before (a) and after (b) scratching, together with a longitudinal section profile of the processed sample (c). A contribution that increases the gap dimension is that of the wear of the tip during each scratching pass. Therefore narrower gaps are difficult to obtain. Such damaging of the probe results in a poor accuracy in further operations performed with the same tip. The image acquired in this condition will lose feature definition. A new probe is then required for an accurate quantitative analysis of the results even after just one complete series of scratches. Stripes thicker and wider than those discussed were also considered but required many scratching passes resulting in wider gaps. Furthermore, the material removal was often unsatisfactory. Therefore we can conclude that direct scratching is not a versatile technique. 459
A Notargiacomo et al
Figure 2. Three-dimensional view of an aluminium stripe before (a) and after (b) directly removing the metal in the thinnest region. The longitudinal section profile (c) shows a 60 nm wide gap obtained after material removal.
3. Thin metal layer mask patterning by AFM lithography
3.1. Sample preparation In order to obtain nanometre-size gaps an alternative technique was tried which avoided a direct interaction between the tip and the metal to be patterned. In this case the samples were covered by two sacrificial layers [7, 8]. A scheme showing the multilayer structure and the processing steps is depicted in figure 3 and will be discussed in detail in the following. An advantage of this procedure is the reduced wear of the tip, which remains available for further processes. The samples to be processed were metal stripes deposited on 200 oxidized silicon wafers, similar to those illustrated in the previous section and shown in figure 1. The only difference consisted in a change of metal: a titanium layer was used. As a first step in the process, the samples were covered by a 60 nm thick polymeric film. The polymer deposition was carried out in a plasma system consisting of a conventional parallel-plate reactor where the plasma is sustained by a 13.5 MHz source. Plasma polymerization using fluorinated gases is a well known process  and has been popular in recent years in microelectronic technology. Plasma deposition of this kind of polymers is influenced by the effective fluorine-to-carbon ratio of the chemically reactive species. In the ‘effective ratio’ only the elements participating actively in the plasma process must be taken into account. An effective fluorine/carbon ratio smaller than three produces high crosslinked polymers with useful properties. In our case we deposited the polymer from CHF3 . The effective fluorine/carbon ratio of this gas is comprised between two and three because hydrogen decreases the effective ratio by consuming part of the active fluorine by production of HF. We found plasma conditions which 460
Figure 3. Layout of the sample and the process steps: (a) sample multilayer structure; (b) thin metal mask patterning; (c) polymer removal in plasma oxygen; (d) titanium stripe etching; and (e) resulting electrodes after sacrificial layers removal.
induced polymerization at controllable rate resulting in a good quality film. The polymer film properties which affect the lithographic process (i.e. deposition rate, etching rate and adhesion, etc) were very reproducible from deposition to deposition. Using a pressure of 500 mTorr and a power density of 0.25 W cm−2 , a constant deposition rate of 10 nm min−1 was obtained. We did not analyse the chemical composition of the polymer but the Teflon-like polymer nature of our film is typical of the fluorinated gas used, as can be found in recent literature [11, 12]. The polymer films were very good from the point of view of the morphological smoothness; indeed, roughness values measured by AFM were very small. We measured a mean roughness Rpolymer = 0.4 nm on a silicon substrate with Rsilicon < 0.1 nm. A further advantage of this type of deposition is that the polymer covered the sample conformally so that the underlying structures were readily located with the AFM, resulting in a precise alignment and positioning of the cut on the metal stripes. Subsequently, the polymer layer was thermally treated at 180 ◦ C for 30 min in air and exposed to UV broadband light. These post-deposition treatments usually produce crosslinking and branching of molecules, which cohesively strengthen the polymer. As a matter of fact, we found that this post-deposition treatment was very effective for improving the mechanical properties of the polymer and its resistance to wet chemical etchers. Furthermore, the treatment improves the adhesion of the sacrificial metal layer deposited on top. A 5 nm thin aluminium sacrificial layer (see figure 3(a)) was deposited by sputtering, obtaining a smooth metal surface. The low value of the surface roughness, RAl = 0.6 nm, measured by the AFM indicates that the aluminium
Atomic force microscopy lithography
Figure 5. Typical silicon cantilever with pyramidal tip: (a) upper view; (b) lateral view showing the 10◦ angle formed with the substrate surface; (c) cantilever bending; and (d) torsion.
Figure 4. AFM surface topography of the multilayer structure: (a) oxidized silicon substrate; (b) sandwiched polymer film; and (c) aluminium top layer.
layer completely wetted the polymer. This conclusion is consistent with previous results on sputtered thin aluminium films . Figure 4 shows the AFM surface topography of the layers used: oxidized silicon substrate, sandwiched polymer film and aluminium top layer. 3.2. Results of the AFM lithography An accurate study was performed on the samples described above in order to find the optimum patterning conditions for the aluminium mask. We found that an applied force of about 1 µN is required to pattern the aluminium mask. Therefore a suitable cantilever spring constant must be in the range of 1–5 N m−1 . Tapping ModeTM and standard contact mode cantilevers have typical values over 20 N m−1 and below 1 N m−1 , respectively and, therefore, were rejected. Silicon cantilevers with 2–3 N m−1 spring constant were considered with two different kinds of tip: the first one is pyramidal (35◦ sidewall angle) with a ∼10 nm tip radius whereas the second is a conical oxide sharpened tip (20◦ sidewall angle) with a radius of about 5–10 nm. We found that the latter kind of tips tend to sink uncontrollably in the soft polymer, resulting in a pattern shape not well defined and often in the tip break. Therefore, the first kind of tips was subsequently employed. The following features were found for the lithographic process. The cuts made in the aluminium mask with the same tip and force applied exhibit different shapes if the lines are patterned in directions perpendicular or parallel to the cantilever. Indeed, the cantilever tends to twist if a lateral translation is made while a longitudinal movement results in a
bending (see figures 5(c) and (d)). As a result the tip interacts with the substrate in different ways. Moreover, an up or down bending occurs when the movement is made in the forward or reverse longitudinal directions, respectively. Since an angle of about 10◦ is typically set between the cantilever and the substrate (see figure 5(b)) this bending results in a further different tip–substrate interaction that makes the tip sink or float over the soft substrate, producing scratches with different geometrical size or even completely preventing their occurrence. In figure 6 a typical result of the first step of the lithographic sequence i.e. the engraving of the top aluminium sacrificial layer, is reported. The AFM topography shows a series of lines patterned on the aluminium mask with a different amount of force applied. Several features are worth mentioning. An increase of material pile up on both sides of the engraved lines is apparent upon increasing the applied force. The profile section along the lines (see figure 6(d)) exhibits a well-defined depth with minimal roughness. As the applied force increases, curved cuts (‘tails’) become visible at the ends of the lines. This effect is due to the cantilever bending at the beginning of the patterning process in order to reach the desired force and to the bending relaxation at the end of the process. It suggests the use of high spring constant cantilevers in order to apply the desired amount of force without a large bending. However, in this situation, the image acquisition is not controllable at small forces and the sample could be damaged. Writing lines in other directions results in the compounding of the effects mentioned above. Moreover, a high value of the surface roughness could produce unwanted features and inhomogeneous results during patterning. It is therefore important that each layer comprising the sample has a very smooth surface. The polymer and thin aluminium film we deposited satisfy this condition well. We conclude that a single line can be patterned in a readily controllable way whereas patterns that extend in two dimensions result in an inhomogeneous line depth and 461
A Notargiacomo et al
Figure 6. AFM topography (a) and three-dimensional view (b) of a series of lines patterned on the aluminium mask with different amount of force applied. The section profile along the aa 0 direction (c) shows the increasing average depth of the grooves. In the section profile along the bb0 direction (d) a well-defined depth with small roughness is the main feature of AFM lithographed lines.
width and require an ad hoc adjustment of the parameters during the lithography process. More studies on these effects and suitably fabricated or modified cantilevers and tips are planned to overcome the limits of this technique. For the time being, it was found that a silicon cantilever with a pyramidal tip was able to ensure the best results in terms of reproducibility, probe lifetime and homogeneity of depth and width. A low tip wear allows a calibration procedure before cutting the desired scratch in order to establish the best lithography parameters. The need of a correct calibration procedure has already been reported in the literature  and is fundamental for a good yield. Indeed, even by pushing two different tips with the same force or by using the same tip at different times, the sample surface could experience two distinct values of pressure. Since the amount of pressure depends on the tip shape and is not measurable it is necessary to make a preventive calibration on the very sample we are about to pattern. The mask patterning technique discussed above was applied to open a narrow gap in the metallic samples shown in figure 1 according to the procedure schematically depicted in figures 3(b)–(e). The result of patterning the upper sacrificial layer is shown as a thin furrow in figure 7(a). In a first attempt, the subsequent metal removal was performed by wet etching. This procedure did not allow a good control of the gap size which was always larger than the scratch width and never smaller than 180 nm. A remarkable 462
improvement in size and reproducibility was obtained by dry etching the titanium stripes with the procedure described in the following. After the AFM lithography, the gaps opened in the top aluminium layer were transferred to the polymer layer underneath by means of a plasma etching process using an oxygen plasma. This aluminium/polymer bilayer acted as a mask for the subsequent titanium etching by a second RIE process using a SF6 –CHF3 –O2 mixture. This three-gas mixture can produce very anisotropic profiles even at very low power density . The CHF3 gas acts as a scavenger gas for improving the etch uniformity of the titanium film. With this mixture we have not observed parasitic effects which are usually present when aluminium masking is used and which degrade the level of anisotropy of the etching . We used an extremely low power density of 0.15 W cm−2 which led to a low self-voltage bias of ∼40 V. The very thin top metallic layer was able to sustain the whole etching process without being appreciably backsputtered. Finally, aluminium and the residual polymer film were removed respectively by wet etching and oxygen plasma. We have obtained gaps as narrow as 40 nm (see figure 7(b)). In the section profile, reported in figure 7(c), the gap width is taken at the base of the stripe that is located at a depth of 30 nm. The exceeding depth in figure 7(c) is due to an overetching performed in order to ensure the gap opening.
Atomic force microscopy lithography
order to use these structures for electronic device fabrication and current characterization down to molecular scale. Acknowledgments
We would like to thank Mario Acciarini for his skilled technical assistance. This work was partially supported by ‘Progetto finalizzato MADESS II’. References
Figure 7. Thin aluminium mask patterned by the AFM tip (a): the three-dimensional view shows a thin furrow across the titanium stripe. The final result of the entire process is a 40 nm wide gap: (b) and (c) show the three-dimensional view and the section profile of the nanogap obtained on a titanium stripe by dry-etching technique. The gap width is taken at the base of the stripe that is located at a depth of 30 nm. The large depth is due to an over-etching.
AFM lithography is a suitable method for patterning circuit elements with good reproducibility and positioning accuracy. Its use allowed us to obtain nanometre size gaps as small as 60 nm by direct scratching of metal stripes. However, reliable and smaller gap sizes down to 40 nm required a more complex set-up such as the use of sacrificial layers and dry-etching processes. An accurate sample preparation was found to be fundamental to achieve good quality results. The preliminary results obtained are very encouraging but further optimization of these techniques is required in
 Wilder K and Quate C F 1998 Noncontact nanolithography using the atomic force microscope Appl. Phys. Lett. 73 2527–9  Mamin H J 1996 Thermal writing using a heated atomic force tip Appl. Phys. Lett. 69 433–5  Campbell P M, Snow E S and McMarr P J 1995 Fabrication of nanometre-scale side-gated silicon field effect transistors with atomic force microscope J. Appl. Phys. 84 1776–81  Fontaine P A, Dubois E and Stievenard D 1998 Characterization of scanning tunnelling microscopy and atomic force microscopy based techniques for nanolithography on hydrogen-passivated silicon Appl. Phys. Lett. 73 2527–9  Magno R and Bennett B R 1997 Nanostructure patterns written in III–V semiconductors by an atomic force microscope Appl. Phys. Lett. 70 1855–7  Baur C, Bugacov A, Koel B E, Makhukar A, Montoya N, Ramachandran T R, Requicha A A G, Resh R and Will P 1998 Nanoparticle manipulation by mechanical pushing: underlying phenomena and real-time monitoring Nanotechnology 9 1–5  Hu S, Altmeyer S, Hamidi A, Spangenberg B and Kurz H 1998 Novel approach to atomic force lithography J. Vac. Sci. Technol. B 16 (4) 1983–6  Hu S, Hamidi A, Altmeyer S, Koster T, Spangenberg B and Kurz H 1998 Fabrication of silicon and metal nanowires and dots using mechanical atomic force lithography J. Vac. Sci. Technol. B 16 2822–4  Bruckl H, Rank R, Vinzelberg H, Monch I, Kretz L and Reiss G 1997 Observation of Coulomb blockade effects in AFM-machined tunnel junctions Surf. Interface Anal. 25 611–13  Coburn J W and Winters H 1979 Plasma etching–a discussion of mechanism J. Vac. Sci. Technol. 16 391–403  Gr¨oning P, Schneuwly A and Schlapbach L 1996 Self-thickness-limited plasma polymerization of an ultrathin antiadhesive film J. Vac. Sci. Technol. A 14 3043–8  Stoffels W W, Stoffels E and Tachibana K 1998 Polymerization of fluorocarbons in reactive ion etching plasmas J. Vac. Sci. Technol. A 16 87–95  Imamura T and Hasuo S 1992 Fabrication of high quality Nb/AlOx –Al/Nb Josephson junctions: deposition of thin Al layers on Nb films IEEE Trans. Appl. Supercond. 2 84–94  Rangelow I W and L¨oschner H 1995 Reactive ion etching for microelectromechanical system fabrication J. Vac. Sci. Technol. B 13 2394–9  Grynkewich G W, Fedynyshyn T H and Dumas R H 1990 The effect of aluminium etching on the plasma etch rates of polysilicon and silicon nitride J. Vac. Sci. Technol. B 8 5–9