Ion Beam Thin Film Processing
In conventional sputtering a glow discharge is established in front of a cathode, and ionisation within the discharge provides a pool of energetic ions to bombard the electrically biased target surface. The plasma sheath provides the acceleration potential for the ion bombardment. Through a process of momentum transfer the bombardment results in material ejection from the target cathode surface. When a substrate is placed in the path of sputtered material it becomes coated with the flux of material.
Imagine taking the glow discharge and confining it within an enclosure, coupled to a vacuum chamber along one side. Placed within the connecting side is an electrode structure that can be used to accelerate ions that impinge upon it. This is the basis for a plasma source and ion accelerator, the pair comprising an ion source, as illustrated in figure 1.
What do we mean by an ion beam? The Oxford English Dictionary provides several definitions of a beam. The one that interests us is stated simply as: a directional flow of particles or radiation. We are interested in the beam of particles rather than radiation. Here the particles are ions, and in particular positive ions, typically Argon ions.
In order to have a directional flow of ions within a vacuum chamber we require a pressure low enough that it may propagate a reasonable distance without interacting with the background gas. This is a function of mean free path. This was covered in our introduction to sputtering, sputtering basics. At a pressure of 0.025 Pa (ca 2 x 10-4 Torr) Argon atoms have a mean free path of about 0.3 m (1 ft).
The ions generated within the plasma source are thermal ions and have no preferred direction of travel. They move, or drift, randomly.
It is possible to impose upon the ions a substantial drift, for example by exploiting the Hall effect (ie an E x B force). This can accelerate the plasma as a whole. This forms a relatively low energy ion beam embedded in the cloud of background gas and cold electrons. A device that produces a streaming plasma of this kind is sometimes referred to as a plasma gun. They are also sometimes called a gird-less ion source, Hall ion source or end Hall ion source. We shall confine ourselves here to what are normally referred to as gridded ion sources. The grids are perforated electrodes that comprise the beam accelerator assembly.
We have established that we can create a glow discharge that contains ions and electrons with no overall preferred momentum. Due to the relative mobility of electrons compared to ions (we shall normally be considering Argon ions) they are lost to the discharge boundary until an equilibrium is established, the discharge takes up a positive potential and sheaves are formed at its boundaries. Ions then will move randomly through the discharge until they encounter a sheath. At the sheath they will be accelerated towards whatever lies beyond the discharge boundary. This might be the wall of the plasma source, or it might be part of the accelerator assembly.
The accelerator assembly is a physical structure that contains a set of electrodes that are appropriately biased. Normally the electrodes comprise parallel plates with an alignment of small apertures (forming grids). When an ion crosses the discharge boundary at the accelerator it might strike the body of the first electrode and would be lost. Alternatively it may be gently accelerated, by the plasma sheath, into an electrode aperture. In this latter case the ion will now encounter the electrostatic field established within, and between the accelerator electrodes. Where the electrodes are appropriately biased this results in acceleration of the ion. The electrodes comprise an array of apertures, so an array of beams, or beamlets are formed downstream of the accelerator. As they propagate, the inherent divergence of the individual beamlets cause them to coalesce into a single broad ion beam.
It is possible to use a single mesh grid accelerator as described by Forrester where the requirements of the beam are not rigorous. It is more usual for a diode, or triode accelerator to be used. These configurations are illustrated in figure 3. In both cases the first electrode that the plasma sees is positively biased (to extract electrons it would be negatively biased). Between the first and second electrodes there is a strong electrostatic field and the ions are accelerated to a potential equal to the arithmetic sum of the two applied voltages. For example where +1000 V is applied to the first grid and -1000 V is applied to the second grid then the ions are accelerated to 2000 V. In the case of the diode structure, on emerging from the accelerator, the beam is now decelerating from 2000 V back to 1000 V. There is a retarding potential difference (-1000 V) between the second grid and the earthed chamber. The location of the earth plane is nebulous in this instance. It is more straight forward in the case of the triode. Ions emerging from the apertures in the second electrode are subjected to a reverse field in their passage between the second and third grids and are decelerated by 1000 V, so the beam is propagated into the chamber at 1000 V.
Note that the plasma source itself is bonded electrically to the first gird and forms a Faraday cage. Although the first grid is biased at +1000 V the only potential between the glow discharge and first grid is the plasma potential and this is dropped across the plasma sheath. So where we said above that the ion beam is propagated into the chamber at 1000 V, this isn’t quite true since the plasma potential is +ve with respect to ground, so the ions will have an energy of: 1000 + Vplasma potential. It is usual to ignore the plasma potential, as it is small (a few tens of volts). When very low voltage beams (less then 100 V) are considered then it becomes significant. These configurations are normally termed as Accel-Decel accelerators.
Nordiko’s application of ion beams concerns broad ion beams for milling (etching) and for deposition. The beam energy range of interest is typically from 200 to 2000 V. There are other applications where far higher beam energies are required. Examples are ion implantation used in semiconductor manufacturing and tokamak fusion fuelling. The beams used for ion implantation normally constitute a high energy beam from a single aperture, often a slit rather than a circular hole. The energies required for implantation are high and typically span the range from a few 10s of keV through to a few MeV. The energy used is specific to the application step in the the device fabrication process. In fusion applications often neutral beams are needed. To generate a high energy neutral beam a high energy ion beam is first formed, that is then neutralised. Such beams need to be large. For the acceleration of beams to these very high energies it is usual to use more than three electrodes (grids). It is not a good idea to place 1 MeV across a single electrode gap and it is good practice to divide this into two or three stages, reducing the stress on the accelerator structure and the likely damage caused from a breakdown.
In figure three are depicted a diode and a triode configuration. In many respects they are very similar, but they fundamentally differ in respect to the space over which the beam is decelerated. For the triode the earth plane is well defined. In the case of the diode the beam energy varies between the point it emerges from the accelerator and the point it encounters an object. The object it strikes may be the opposite wall of the vacuum chamber or a piece of furniture within the chamber. The passage of the charged ion beam passing through the background gas causes ionisation and the generation of a beam plasma. In the case of the diode the beam plasma is in contact with the outer electrode. The result is that unless the second grid is earthed the ions from the plasma will bombard and sputter the electrode. It is normal practice to apply at least a -200 V bias to the second electrode, both to enhance beam formation and more importantly to suppress electrons entering the accelerator from the beam side. Without the negative barrier electrons can readily back-stream through the accelerator causing instability and potentially damage to the plasma source. Whether or not sputtering of the outer grid is a problem may depend upon the the application. In a deposition system it can cause contamination of the growing film if the arrival rate of target material at the accelerator is lower than the re-sputtering rate. In an etch system it can cause issues if the material sputtered from the electrode is not quickly removed from the substrate illuminated by the ion beam. For example the native grid material may have a low sputter yield and cause micro-masking.
There are a variety of techniques that can be used to generate a low pressure gas discharge within a plasma source. The first ion sources used DC arc discharge generation. An early example was reported by Thomson. This employed a Canal Ray (from the German Kanalstrahlen) discharge source. Another high voltage discharge source is the Capillaritron as described by Mahoney et al. These are high voltage devices and better suited to high energy physics experiments than for deployment in thin film process equipment. For number of years the main stream solution was a simple thermionic hairpin filament, or multiples thereof, as illustrated schematically in figure 4. This approach lends itself to use in a large area source, used to project a broad ion beam. This style of embodiment can be enhanced by the application of a magnetic confinement field, as used by Holmes et al. Figure 5 illustrates a source with a peripheral multi-cusp permanent magnet array around the discharge volume. The magnetic field helps stop the electron loss to the walls, making the discharge both more efficient and improving the homogeneity of the discharge. The central volume of the source is field free. As opposed to the magnetron case, the walls are the anode rather than the cathode. Although the discharge density close to the source walls is increased by the presence of the magnetic field, sputtering of the wall by positive ions is not a problem.i
For production application the reliability of the ion source is paramount. In many situations the system is vacuum load locked and needs to remain closed under vacuum for many weeks between maintenance intervals.
The filament, as in an incandescent light bulb has a finite and unpredictable lifetime. This situation is further impacted by the use of Oxygen in processes for the deposition of optical films, for example. Nordiko introduced our first ion sources in 1989. In the development phase we considered various options for the excitation of a filament-less ion source. The technique we selected was RF excitation. We make a range of ion sources with broad beam capabilities from 100 mm up to 450 mm diameter. Each model is inductively RF excited at 13.56 MHz.
Nordiko RF Board Ion Beam Source
There are a number of features that are common throughout our range of sources. They all have a cooled aluminium plasma source. Embedded within the structure of the source is a permanent magnet array. This uses rare earth (NeFeB) permanent magnets in an arrangement to confine the plasma and provide a central field-free volume within the source. RF at 13.56 MHz is inductively coupled in to the source through a dielectric window.
The accelerator is built as a separate assembly that fastens either to the front of the ion source or to the chamber wall. The electrodes (grids) are held in a rigid support structure in which they are held parallel to one another and the apertures are self-aligned.
On the outside of the system there is a safety enclosure where the service connections, cooling water, RF power, high tension DC for the accelerator and gas metering are made. To this is interfaced an automatic RF impedance matching transformer. An installation on the Nordiko 3400 is illustrated in figure 6.
High tension DC power supplies provide the biasing of the first and second electrodes within the accelerator. The first grid, that is the one in contact with the plasma inside the source, is positively biased and the adjacent grid is negatively biased. The third and outer grid is bonded to earth. The Nordiko Ion Source is shown schematically in figure 7.
Broad Ion Beam Process Systems
Other than ion implantation mentioned earlier, two styles of ion beam systems are in common use for thin film processing, these are; Ion Beam Deposition (IBD) or Broad Ion Beam Deposition (BIBD) and Ion Beam Milling (IBM) or Ion Beam Etching (IBE).
Broad Ion Beam Deposition
This is a form of sputter deposition. Rather than sputtering a cathode directly by applying a high voltage (DC or RF), a glow discharge is established and held within an enclosure; subsequently ions from the enclosed plasma source are accelerated to form a broad ion beam by an adjacent accelerator assembly. The directional ion beam is aimed at, and illuminates a target. The ion bombardment gives rise to sputtering in the same manner as in DC or RF cathodic sputtering. In a similar fashion to a conventional sputtering system a substrate to be coated is placed in the path of the evolving material flux.
The principle differences in the ion beam technique is that the ion bombardment energy is now well controlled and can be varied at will. There is also a significant reduction in background electron flux. The process pressure is lower. Not only is the pressure lower, the dynamic range over which it can be used is narrower. To effectively form a particle beam the pressure needs to be below 0.13 Pa (1 x 10-3 Torr). The lower limit of operation tends to be about 4 x 10-3 Pa (3 x 10-5 Torr).
Another big difference between ion beam sputtering and conventional sputtering is the geometry. When coating a large wafer the normal configuration for conventional sputtering is that the wafer is held static and parallel to the sputtering cathode, as with a relatively modest source to substrate separation. A variation often used in the laboratory or for very thin films is confocal sputtering.
For the ion beam deposition case the geometry defines things. Space is needed to illuminate the target with the ion beam and for the resulting sputtered material flux to reach the intended substrate without shadowing. To achieve this the ion beam needs to illuminate the target at an angle.
Unlike a laser beam that can exhibit very low levels of divergence (it is possible to shine a laser across a room and illuminate a spot only a few millimetres in diameter). Physics is less kind to ion beams. An ion beam, especially a low energy (less that 10 kV) has a finite and significant divergence. This in practical terms can be relatively large, with a half angle of 4 to 10 degrees of arc.
Consider an individual beamlet that emerges from the accelerator with a diameter of 3 mm, and propagates a distance of 300 mm before striking the target, as in figure 8. Below is a table illustrating the impact of the inherent divergence on the size of the beam downstream of the accelerator.
It is common for an ion beam deposition system to be fitted with a second auxiliary ion source. This second source is typically aimed at the substrate and is used for sputter-etch cleaning and can also be used to illuminate the substrate during film deposition (surface modification). This general arrangement is shown in figure 9. In practice is not uncommon for ion beam deposition systems to be equipped with a dual axis substrate table. This allows both on axis substrate rotation and the ability to present the substrate to the material flux at a variable angle. By off-setting the axes of the target and substrate it is simpler to find the room to satisfy the requirements. With an offset geometry the static uniformity will be poor, but by exploiting the on-axis substrate rotation it is possible to achieve very good within substrate non-uniformity of the deposited film thickness. The target is often mounted on a rotary platform, arranged either as a carousel or as a prism. This allows the adoption of multiple targets for the sequential deposition of more than one material. Figure 10 illustrates a more representative arrangement of features within a practical ion beam deposition system. Other factors that can affect the final arrangement are the requirements for deposition rate, within substrate non-uniformity and access for substrate loading and exchange where the system is equipped with a vacuum load lock.
Note that the deposition ion beam is smaller than the auxiliary beam. This not uncommon; the deposition beam would normally be accelerated to a higher voltage than the auxiliary beam. Representative values for the two voltages ranges are: from 800 to 2000 V for deposition and from 150 to 500 V for the auxiliary beam. There are two applications for the auxiliary beam; one is for pre-cleaning (pre-etching ) the substrate; the second is to modify the properties of the deposited film. Depending upon the nature of the substrate there are often constraints on how energetic the illuminating beam should be for pre-cleaning. Where the substrate is an electronic device there may be concerns over inflicting device damage at too high a beam voltage. For optical applications the surface morphology can be paramount and even the weakest of beams will have a tendency to roughen the surface. Similar arguments are true for the substrate illumination during film deposition. As a result the auxiliary beam is often used at a maximum of 200 to 250 V.
The auxiliary source can also be used effectively in reactive deposition. Here a metallic target may be sputtered by an Argon beam and Oxygen, or another reactive gas, is metered into the chamber. There is the option of metering the reactive gas simply into the chamber, or through the auxiliary ion source. In the latter instance there are further permutations that can be exploited. The source could be energised; that is the discharge lit within it, but the accelerator left idle. In this instance a flux of low energy ions, dissociated (free radical) and excited molecules are directed towards the substrate. Alternatively the accelerator can be active and there is a choice as to the beam voltage and current employed.
It is worthwhile now considering the beam extraction. It was mentioned in the previous section that low energy beams exhibit poor divergence and implied that the divergence improves with increasing beam voltage. As a generalisation this is indeed true. Before considering this topic in more detail, let us visualise the plasma/accelerator interface and the passage of the beam through the accelerator as a whole. A representation of a beam extracted through a triode accelerator is show in figure 11. The electrodes (grids) are shown with their apertures axially aligned. To the left hand side the plasma within the source is shown. In the region of the first grid aperture a meniscus forms, distorting the plasma sheath. It is across this meniscus that the ions are launched as they enter the accelerator channel formed by the aligned apertures of grids.
The black lines represent the ion trajectories. The complex pattern of coloured lines are electrical equip-potential contours of the electric field within the accelerator. The electric field focusses the beam. As it does this the space-charge is trying to make the beam expand.
The beam current extracted through an aperture is determined by the Child-Langmuir equation. This expression was first described by Child but he only modelled space charge limited current flow between infinite planar electrodes. Langmuir and Blodgett extended this work to cover spherical and cylindrical electrodes which is more pertinent to ion accelerators, particularly plasma ion emitters that form part of a spherical surface. The full equation takes the form of an infinite series of logarithmic terms but it can be readily simplified to the form below which is good to about ±1% accuracy:
The term ε0 is the permittivity of free space (= 8.85 E-12 farads/m), e is the ionic charge (usually 1.602 E-19 coulombs), Mi is the ion mass in kilograms, N is the number of apertures, a1 is the radius of the first aperture, d1 is the first gap distance (see below), U1 is the voltage across the first gap and is also the beam kinetic energy (in eV) at the second electrode and lastly, θ is the angle the edge trajectory makes to the axis at the plasma boundary with converging rays having negative values of θ. This first gap voltage, U1 (= V1-V2) is always bigger than V1 and particularly so if a large negative potential is applied to the second grid.
The beam perveance, P, is defined as I/V13/2 where V1 is the first electrode potential (which is smaller than U1), so the perveance can be increased by using large negative potentials on the middle electrode. The reason for the above definition of perveance is that it is related for historical reasons to what happens to the beam downstream of the accelerator, in particular beam drift in the absence of any plasma. However, in situations where there is a dense beam plasma (which compensates the beam space charge) the beam perveance has little meaning and the first gap definition used in the equation above is more meaningful.
Despite the apparent complication of the above expression, it is surprisingly simple. Firstly, the first pair of terms on the right hand side are all universal constants and can be given their usual values including the ion mass which is 6.68 E-26 kilograms for singly charged Argon ions. Most of the rest of the terms are set by the accelerator geometry apart from the extraction potential, U1, which is the algebraic sum of the voltage on the plasma electrode and that on the negative second electrode. This yields an even simpler expression:
U1, the voltage across the first gap is often referred to as the extraction potential, or extraction voltage. It should be noted that the ions are not actually extracted from the plasma, they are accelerated having randomly drifted into an accelerator aperture.
However, the value of θ then also changes and becomes more negative when this is done and hence opposes the increase in perveance. This is discussed further in the next section. However the value of θ is also a function of the second grid potential and becomes more negative for high negative potentials on this grid when this is done and hence opposes the increase in perveance.
The core beam can be bought into focus by either adjusting the plasma current density which leaves the plasma generator (via the RF power) or by altering the extraction potentials. Both of these processes modify the radius of curvature, R1, of the plasma ion emitter meniscus which in turn alters the angle θ at the plasma edge. For linear (small angle) optics this angle is given by:
Where R1 is defined as negative if the surface is concave to the beam.
The accelerator can be described by a series of optical lenses, the first of which is the meniscus itself which can be seen in figure 11 as the second leftmost equi-potential surface. Near the edge it becomes non-spherical and the beam aberrates, see figure 12. Also in figure 11, note the way the ion trajectories graze the inner circumference of the first aperture before becoming properly focussed. There are two other optical lenses, one located at the middle plane of the second negative electrode and a final one at the middle of the ground electrode.
The focal length, F, of these two lenses is governed by the Davisson-Calbick equation which is:
Where Edown is the field downstream of the lens position on axis, and Eup is the upstream field on axis, while U is the beam kinetic energy (in electron volts) at the lens itself. The lens convention is that a positive focal length corresponds to a focussing lens. Using the lens at the ground electrode as an example, Edown is zero, Eup is a retarding field and hence is defined as a negative value. Thus F for this lens is positive and is thus a focussing lens. The lens at the middle electrode is invariably a diverging lens as the upstream field is always accelerating and hence positive, and the downstream field is a retarding field and hence negative.
Solving this system as a set of lenses with an initial angle, θ, at a radius, a1 can be done via a matrix approach and this is described by Holmes and Thompson. This gives a final exit angle, Ω, as a function of θ, a1, and the accelerator potentials and gaps. This is not described again here but for a low potential on the middle electrode, corresponding to a weak field in the final gap, the value of θ for a collimated beam output is -0.25a1/d1. For a high potential on the middle electrode, the value of θ becomes more negative and the beam becomes more compressed as it passes through the middle electrode and then expands again in the second gap and the final lens at the third electrode collimates the beam.
Based on this linear optics model, it would be expected that a zero divergence beam would be possible. However the beam has finite disorder or entropy and this is described by the beam emittance, ε, and this controls the minimum divergence angle. As the emittance arises from the concept of a beam “temperature”, it is related to the beam halo shown in figure 12. In that figure the two ellipses drawn round the beam correspond to the inner and outer emittances. The emittance is the area of the ellipse divided by pi (= 3.142). The inner ellipse, containing about 39% of the beam current and called the beam core, is free of much of the beam aberrations and has a temperature of 1eV or 11600 Kelvin. The outer ellipse contains the entire beam and has a much higher temperature of around 50eV. The minimum divergence angle that the beam can have is set by these temperatures. The core beam has a minimum divergence of:
Where r3 is the beam radius at the ground electrode and T is the beam temperature. Thus the lower the beam energy becomes, the harder it is to form a highly collimated beam. Beam contraction inside the accelerator also causes divergence as it heats the beam by adiabatic compression.
Ion Beam Transport
As the beam has passed through the accelerator and is in flight within the vacuum chamber there can be many influences on the beam propagation and trajectory. The topic of beam transport is pertinent to a wide variety of beam applications including ion implantation, particle accelerator design, heating of fusion plasmas by neutral beam injection, electric propulsion and thin film processing. For each of these areas the motion of ions can be considered under the influence of electric and magnetic fields. From the laws of motion and knowledge of the fields encountered, in principle, the full description of the beam behaviour can be obtained during the acceleration and in subsequent flight. Taking this general approach encounters severe mathematical problems. An alternative approach remains very mathematical and the interested reader is referred to a detailed treatment by Holmes (The Physics and Technology of Ion Sources, Edited by Ina G Brown, John Wiley and Sons 1989, first Edition).
There are two types of neutralisation we shall consider; these are space-charge neutralisation and current neutralisation. Space-charge, arises when a collection of particles with a net electric charge occupy a region, either in free-space or in a device. In applications used for thin film processing we are concerned with positive ion beams. Electrons, therefore, play in important role in preserving overall neutrality. When the space-charge becomes too high the beam will stall. Stalling of the beam in practice tends to be avoided due to neutralisation of the positive ion space-charge by electrons trapped within the beam. An ion beam will create electrons through secondary emission at all surfaces that the beam illuminates. For sufficiently energetic beams electrons will also be generated by ionisation of the background gas simply as the beam passes through it. The electrons generated become trapped within a potential well created by the positive ion space-charge. The electrons, the ions and the background gas interact with one another to create the glow that we refer to as the beam plasma. Not only does this glow contain the fast ions associated with the beam, it also contains thermalised electrons and slow ions generated through charge transfer between the beam and the background gas. The electron density must approximately equal the sum of the ion densities to preserve neutrality.
For beams used in thin film processing space-charge tends to take care of itself and in a system used for the deposition of conducting films additional neutralisation is not generally necessary. For a system where insulators are deposited, or when the substrate to be coated is either insulating, or electrically isolated, surfaces illuminated by the beam will become charged. The charging of an insulating surface tends to be exacerbated by the fact that dielectric materials have a relatively high secondary electron emission yield.
To compensate for the surface charging, electrons may be injected into the beam plasma. A device originally developed for net neutralisation of electric ion thrusters is commonly used for this purpose. Ion Beam engines used as a space thruster cannot eject a positive ion beam without emitting an equal current of electrons. Without the balancing flow of electrons the vehicle would become so negatively charged that the ions could no longer escape. There would also result a net attractive force between the charged vehicle and the ejected positive beam. A very simple way to provide these electrons is from a filament immersed within the beam. This can provide low energy electrons that couple very well to the beam. The immersed filament, however, ages quickly. A high reliability solution takes the form of a plasma bridge neutraliser.
Nordiko ion beam systems employ a filament-less plasma bridge neutraliser for current neutralisation in both our Broad Ion Beam Deposition and Broad Ion Beam Milling systems. Figure 13 shows an ion source with an accompanying Plasma Bridge Neutraliser (PBN).
Ion Beam Milling (Ion Beam Etching)
In thin film processing ion beams are used for milling (etching) as well as deposition. In Semiconductor processing many films are etched reactively using plasma etching, or reactive ion etching. The benefit of invoking a chemically reactive process is to achieve both higher etching rates than would otherwise be possible by a physical process alone, and to achieve a good selectivity over other materials. There remain, however, materials where there are no effective chemical processes that can be exploited.
It is possible to physically etch materials in a reactive ion etcher using a noble gas discharge. This technique is termed conventionally as sputter etching. The problem with this approach is that at the lowest pressure to sustain a discharge the mean free path is of a level where the sputtered material is readily re-deposited. The preferred method for physical sputtering, or milling as it is often referred to, is Ion Beam Milling (Ion Beam Etching).
Where films of noble metals, such as gold and platinum need to be patterned, ion milling is commonly used. The emergence of non-volatile magnetic memory elements such and MRAM (Magnetic Random Access Memory) that incorporate very thin magnetic, as well as noble metal, films has sparked a resurgence of interest in ion beam milling.
It is possible to add a reactive component to ion beam milling systems. There are essentially two ways such a system may be configured. In one approach the reactive precursor gas is injected directly into the ion source, generally together with Argon. In this instance the reactive gas is ionised within the plasma source and accelerated as a component of the ion beam. This is called Reactive Ion Beam Etching (RIBE). An alternative approach is to meter the reactive gas into the system downstream of the ion source. Here a common approach is to introduce the gas around the substrate to be etched, or through a nozzle or gas ring that floods the substate area with the entrained gas. This technique is called Chemically Assisted Reactive Ion Beam Etching (CARIBE). The ion beam in this latter instance is maintained as an Argon beam.