1 Introduction
The development of nanoelectronic devices requires controlled fabrication of nanostructures on the surface. Two parallel routes have been taken toward this goal: one is the top-down approach, such as nanopatterning, and the other is the bottom-up approach, such as self-assembly. In this article, we will demonstrate a novel effect of Coulomb charging on the coarsening of metal nanoclusters on surfaces, providing a potentially useful method for controlled fabrication of metal nanostructures.
The Coulomb effect is ubiquitous in physics, chemistry, and biology. One well-known manifestation of the Coulomb effect on the stability of a cluster is Coulomb explosion. It is defined classically by the Rayleigh instability limit [
1], above which an excessively charged cluster becomes unstable and explodes into smaller fragments. For a nanocluster, the critical size for Coulomb explosion depends on the nature of chemical bonding [
2].
When a metal cluster is charged on a surface, it may not explode, but instead grow its size by “sinking” atoms from its neighboring neutral clusters to reduce its Coulomb energy. We call this effect the “Coulomb sink”. The first experimental evidence of the Coulomb sink is from Okamoto
et al. [
3], who attributed this phenomenon to the high Maxwell stress of the local electrical field lines which draw the atom towards the high field region on top of the island. However, their theoretical explanation was unsuccessful [
3]. In the following sections, we will elaborate, by both theory and experiment, on the phenomena of Coulomb sink with metal nanomesas grown on insulator and semiconductor surfaces. The basic viewpoint is that charging can reduce the chemical potential of the charged mesa relative to its neighboring neutral mesas, and consequently it grows at the expense of its neighbors via a coarsening process. Because one can selectively charge any chosen mesa with a controllable amount of charge,
Coulomb sink provides a unique and effective method for manipulating the growth of metal mesas with a size control up to millions of atoms.
2 Theoretical models
2.1 Hemisphere-capacitor model
The capacitance of a conductor monotonously increases as a function of its size. Thus, to lower its Coulomb energy, the charged mesa (cluster in general) may increase its size to maximize its capacitance. For coarsening, this implies that the chemical potential of the charged cluster is instantaneously reduced, with respect to its neighboring uncharged clusters, so it grows at the expense of its neighbors. Let us consider a metal cluster in a hemisphere shape with radius R grown on an insulator surface, as shown in Fig. 1. If the cluster is neutral, the energy in reference to the bare surface is
where and are the cluster-substrate interface energy, the substrate surface energy, and the cluster surface energy, respectively. The chemical potential can be defined as
where Ω is the atomic volume, V is the volume of the hemisphere cluster, N is the total number of atoms in the hemisphere. In writing Eq. (2), the relation
and Eq. (1) have been used.
When an isolated conductor is charged with Q, the excessive Coulomb energy is
where C is the capacitance of the isolated conductor. Now, for a hemisphere conductor charged with Q, its Coulomb energy is
where ε is the dielectric constant. Therefore, the total energy of the charged cluster is
Here, the capacitance of the metal hemisphere has been assumed to be C = 2πεR. Then, the chemical potential becomes
Now, consider two metal nano-hemisphere clusters, A and B, with the same initial radius R. Let A be charged by Q, and B is neutral. Apparently, the chemical potential of A will be reduced by , as expressed in Eq. (7). It follows that the atoms of B will diffuse onto A in a coarsening process, i.e., the charged cluster A sinks the atoms from the neutral cluster B. During the process of coarsening, the system total energy is
where RA and RB are the radii of A and B, respectively. Also, the volume conservation requires
From Eqs. (8) and (9), it is shown that the minimum of Etot is at RB = 0, and then the maximum of RA equals to . This indicates that the neutral cluster will eventually disappear.
2.2 Cylinder-capacitor model
The growth of metal film on semiconductor substrates proceeds via Volmer-Weber mode [
4,
5], forming three-dimensional (3D) islands, when the surface energy of the film is much higher than that of the substrate. The islands may adopt an equilibrium mesa shape defined by surface energy anisotropy, such as Au mesas on C(0001) surface [
6] and Pb mesas on Si(111) surface [
7−
9]. For simplicity, however, we assume a mesa has the perfect cylindrical shape with radius
R and height
H, as shown in Fig. 2. The energy of a neutral mesa, in reference to a bare surface, is then
where α1 = γmt + γint − γsub and α2 = γms. Here, γmt and γms are, respectively, the surface energy of the mesa top and the mesa sidewall, and γ int and γsub are, respectively, mesa-substrate interface energy and substrate surface energy.
Now, consider one chosen mesa is charged by Q, using a scanning tunneling microscope (STM), and the charge will subsequently discharge to the substrate within a period of time, as illustrated in Fig. 2. By considering a capacitor formed between the charged mesa and its surroundings, its energy is instantaneously increased by Ec [Eq. (4)].
Based on the mesa geometry, we model the charging effect by a coaxial cylindrical capacitor of height H, as shown in Fig. 2. The inner conductor is the charged mesa having a radius R with charge Q, and the outer conducting shell (dotted lines) represents the distant groundings, having a much larger radius Rg. Its capacitance is then calculated as
where ε is the dielectric constant. Therefore, the excessive Coulomb energy of the charged mesa is
Thus, from Eqs. (10) and (12), the total energy of a charged mesa is
An increase of either H or R results in a decrease of E. To specify the growth of height versus radius, we define a “partial” chemical potential for height growth as
representing the change of Coulomb energy with respect to changing height at a fixed radius, and a “partial” chemical potential for radius growth as
representing the change of Coulomb energy with respect to changing radius at fixed height. Here, Ω is the atomic volume, V is the volume of the charged mesa, and N is the total number of atoms contained in the mesa. Also, in writing Eqs. (14) and (15), we have used Eq. (13) and the relation V = NΩ = πR2H. From Eqs. (14) and (15), a growth index function can be defined as
which tells the energetically favored growth direction. It can be easily verified that if Q is not too small, the surface tension effects can be neglected, and then Eqs. (14), (15), and (16) become
and
respectively. Because , S < 0. It is readily seen from Eq. (19) that the charged mesa always favors growth in height.
Charging affects coarsening by breaking down the chemical-potential balance between the original neutral mesas, because the energy of the charged mesa is decreased by an amount ~
Q2. It makes the charged mesa act effectively as a Coulomb sink to attract atoms from the surrounding neutral mesas. Therefore, it triggers the otherwise “ceased” coarsening process by creating a high supersaturation around the charge mesa to overcome the energy barrier for nucleating new layers [
7,
8,
10] to grow in height. To assess the amount of growth induced by charging, we consider the coarsening process to be controlled by the attachment of adatoms to the mesa [
11], and assume that the rate of volume growth is simply proportional to the chemical-potential difference between the charged mesa and the surrounding neutral mesas. It follows that the rate of height growth can be expressed as
where σ is a coefficient related to the adatom attachment rate and the atomic volume, µ is the chemical potential of the charged mesa, and is the mean chemical potential of the surrounding neutral mesas. The proportionality coefficient σ can be estimated by
where ∆ is the energy barrier of attachment, k is Boltzmann’s constant, and ν is the atom-jumping attempt frequency. If ν is approximately taken as kT/h, where h is Planck’s constant, and T is temperature, then Eq. (21) can be rewritten as
From the fundamental thermodynamic theorem, we have
where S is the entropy of the system, P is the external pressure, Vs is the total volume of the system, is the number of atoms attached onto the charged mesa, and dN is the number of atoms detached from the surrounding neutral mesas. We assume that in the mass flow process between the charged mesa and the neutral mesas, (1) , (2) no volume work is done (PdVs = 0), and (3) the process is adiabatic (dS = 0). Then, Eq. (23) can be written as
or
If R is constant, , and then comparing Eq. (25) with Eq. (14) yields
Substituting Eq. (26) into Eq. (20), we have
Next, substituting Eq. (17) into Eq. (27) we get
where
Similar to the hemisphere model in Section 2.1, it can be proved that if the charge Q carried by the charged cylindrical mesa remains unchanged, the surrounding neutral mesas will completely disappear. However, if the substrate is a semiconductor, there will be a charging-discharging process, and consequently Q is time-dependent. Generally, the metal-semiconductor interface can be considered as a depletion-layer capacitor. If the depletion-layer capacitance is assumed to be Cd, then when a charging voltage of U is applied, the maximum charge amount of the capacitor is
The time-dependent charge carried by the metal mesa can be expressed as
where t1 is the total time of the charging process after which the applied voltage is removed and then the mesa discharges, τ1 and τ2 are, respectively, the time constants corresponding to the charging and discharging processes, and
which is the charge amount at t = t1 when the charging process terminates. The typical charging and discharging curve of a capacitor is shown in Fig. 3. The solid curve and the dashed curve, respectively, represent the charging and the discharging processes described in Eq. (31).
Substituting Eq. (31) into Eq. (28) results in the integration
where H0 is the initial height of the charged mesa. From Eqs. (33) and (32), we obtain
When , Eq. (34) becomes
Here, we have assumed that there are a sufficient number of the neutral mesas surrounding the charged mesa for its height growth.
3 Explanations for experiments
3.1 Selective growth of Pb nanomesas triggered by STM tip
Figure 4 shows the quick growth of a charged Pb mesa on Si(111). The experiment was by the Beijing Group [
12] performed on an OMICRON UHV molecular beam expitaxy-STM system. About 3 monolayers of Pb were deposited on the clean Si(111)−(7 × 7) surface at ~ 150 K, forming flat-top mesas. The charging was applied at room temperature by a pulse of 10 V and a tunneling current 20 pA with feedback on for about several ms. Right after the pulse, normal room temperature STM scanning (1.5 V and 20 pA) was resumed to monitor the morphology evolution.
For an actual charged mesa, some mesa-edge positions have relatively higher charge distributions (the fringing effect) because these edge positions have relatively larger geometric curvatures. This indicates that the height growth first starts along these mesa-edge positions, creating many steps, as shown in Fig. 4. The charged mesa continues to grow even after discharging, since coarsening can now proceed via step flow without the need of nucleating new layers. The growth after discharging is much slower, as it is driven by a smaller chemical-potential difference due to mesa size difference and by the tendency for the growing mesa to eliminate steps to resume the flat-top equilibrium shape. The growth stops when a flat top is formed. Further growth can be triggered only by another charging pulse. Below is a rough estimation for the increased height of the charged Pb mesa during the charging-discharging process. The time constants are defined by
and
where R2 is the interface resistance, and R1 is the sum of R2 and the STM contact resistance Rc. In experiments, R2 ~103 Ω, Rc ~ 108 Ω, and Cd ~ 10−16 F, so τ1 ~ 10−8 s, and τ2 ~10−13 s. Substituting these values of time constants into Eq. (31), the charging and discharging curves, respectively, can be plotted as Figs. (5) (a) and (b). The experimental charging ending time t1 ~ 10−3 is much larger than τ1 and τ2, so Q1 is close to Q0, and Eq. (35) can be simplified as
This indicates that the main contribution to height growth is from the charging-saturation period
in the charging process, and there are almost no contributions from the period
before the charging-saturation as well as the
discharging process. Typically, for the Pb and Si system in the experiment,
Ω ~ 0.03 nm
3,
∆ ~ 130 meV, and
ε ~ 2
ε0, where
ε0 is the permittivity of free space, and according to Fig. 4, we take
R ~ 300 nm,
Rg/
R ~ 10
3,
H0 ~ 10 nm,
U = 10 V, and
T = 300 K. Substituting these parameters into Eq. (38) with the use of Eqs. (29), (22) and (30), we obtain a value of ∆
H =
H − H0 ~ 13 nm, which agrees with the experimental measurement [
12].
3.2 Fabrication of artificial nanowells with tunable size and shape by using STM
The construction of artificial nanostructures is of great significance in both the understanding of physical and chemical properties at nanoscale and potential nanodevice technologies. One effective way to construct artificial nanostructures is by using STM. In the early 1990s, STM was applied to move and position single atoms or molecules [
13−
17]. Larger nanostructures involving a huge quantity of atoms can be fabricated by self-organization growth [
18−
21], but the precise control in local structures is lacking in these studies. Recently, the STM manipulation based on the Coulomb sink effect has been applied to precisely control the thickness of the nanostructure with single atomic layer precision, and the engineered nanostructure can involve a macroscopic number of atoms up to 10
5−10
6 [
8,
12].
In this section, we illustrate a couple of examples of employing the Coulomb sink effect to extend the STM-based nanofabrication of metal structures. The experiments have been performed by the Beijing Group [
22]. In the experiments, by applying a series of appropriate voltage pulses to the STM tip, the control of the lateral shape as well as the exact location of the targeted Pb island grown on Si(111)−(7 × 7) substrate can be achieved through regulating the pulse parameters. The resulting nanostructures have interesting shapes and selected locations with atomic layer precision, implying the potential in the research and application of size- and shape-sensitive properties at nanoscale.
In experiments, the high purity Pb is thermally evaporated from a homemade tungsten cell and deposited onto the Si(111)−(7 × 7) substrate, and the Pb mesas can be grown in an ultrahigh vacuum molecular beam epitaxy chamber. More details for the description of the instrument and experimental conditions have been reported [
22,
23]. Here, we only quote the most interesting experimental results to analyze how the Coulomb sink effect is functioning in the fabrication of nanostructures.
Figure 6 (a) shows a flat-top Pb mesa grown by depositing Pb at 150 K followed by slowly warming up to room temperature. The mesa height has a range of 3.1 nm to 4.9 nm from the upper left to the lower right due to the stepped Si(111) substrate surface. When a voltage pulse of ~10 V is applied upon the mesa edge by the STM tip, as indicated by a white arrow in Fig. 6 (a), for 10 ms during scanning, the flat mesa top was triggered to grow around the edge at the same time, as shown in Fig. 6 (b). From the viewpoint of Coulomb sink effect, this is because when the charge is injected into the Pb mesa, the chemical potential of the mesa, especially around the mesa edge with a much larger curvature and therefore the much higher charge distribution, is drastically dropped, and then the mesa top edge is triggered to quickly grow at the expense of the neighboring neutral mesas. The outside profile of the nanowell remains nearly unchanged relative to the original mesa shown by Fig. 6 (a), and the nanowell bottom has the same height as that of the original flat-top mesa, but the height of the sidewall increases by 3.58 nm, which is approximately equal to 12 Pb(111) monolayers. This indicates that a center-half-hollowed nanowell is fabricated successfully with atomic layer precision. It is found by STM observation that such the center-half-hollowed nanostructure is very stable, even taking a couple of days to resume the flat top of the mesa if no further STM voltage pulse is applied on it.
It is also possible that the Pb mesa transforms into some interesting shapes if a series of STM voltage pulses is applied at different sites on the top of the Pb mesa. When the center-half-hollowed nanowell shown in Fig. 6 (b) is manipulated with another STM voltage pulse at the bottom, the growth will start from the inner edge side where the charge distribution is much higher than the bottom due to the large curvature. The bottom is relatively neutral with the higher chemical potential due to the nearly zero curvature, and then completely broken up, so that the bottom atoms climb up and attach to the inner sidewall edge to reduce the local chemical potential. Thus, the center-half-hollowed nanowell becomes a center-full-hollowed structure finally, as shown in Fig. 6 (c). Unlike the center-half-hollowed nanowell, the inner sidewall of the center-full-hollowed nanowell forms a triangle or hexagon hole. This indicates that except for the Coulomb sink, the diffusion effect will also concurrently occur, which is related to the mass transition of inner walls from a stepped structure in the center-half-hollowed well to a faceted structure in the center-full-hollowed well. The inner-outer diameter ratio can be increased by charging the nanowell from the inner edge side, which makes the inner Pb atoms climb up. The thickness can be tuned by charging from the outer edge side which induces mass transport from the surrounding neutral islands to the charged one. Figure 6 (d) shows the same nanowell but with tuned inner-outer diameter ratio and thickness by a series of STM voltage pulses. More details for how the size and shape of a nanowell depend on STM pulse parameters such as the amplitude or the time constant have been also reported in Ref. [
22].
The above STM manipulation can be extended to fabricate predesigned nanoarray patterns even in the microscale [
22]. Figure 7 shows an example, where five selected islands, indicated by the five white arrows in Fig. 7 (a), are made to grow, while all the others are “erased”, forming the letter “M”, as indicated by the white line in Fig. 7 (a). The mass conservation has been also checked, and consequently it is confirmed that the supply of atoms for the growth of selected islands comes from their neighboring islands.
4 Summary
We have demonstrated a new concept of the Coulomb sink to elucidate the coarsening of metal mesas on a surface. Charging reduces the chemical potential of the charged mesa, making it act as a Coulomb sink to grow at the expense of its neighboring neutral mesas. Thus, it triggers an otherwise ceased coarsening process. For the cylindrical mesas, by introducing a growth index function, we show that the charged mesa grows preferentially; its height and the growth proceeds first rapidly by nucleation of new layers around the island edge creating many steps, and then continues slowly via step flow after discharging to resume a flat top. The theory has successfully explained most salient qualitative features observed in the coarsening of charged Pb mesas on Si(111).
The Coulomb sink, leading to cluster agglomeration, is effectively a reversed process of Coulomb explosion leading to cluster fragmentation. It provides a unique and effective method for manipulating the growth of metal nanoclusters on semiconductor or insulator surface with a size control up to millions of atoms. Various nanostructures by “massive” STM manipulation of millions of atoms via the Coulomb sink effect have been created. In this way, large-scale nanostructures can be fabricated with high precision. The nanostructures so obtained have controllable shapes and sizes. An island array to build predesigned architectures by growing, moving, and erasing entire islands can also be manipulated.
Higher Education Press and Springer-Verlag 2008