NEWS
PROCESS NEWS A Newsletter from Oxford Instruments Plasma Technology Spring 2019
PLASMA.OXINST.COM
Welcome
which in turn releases new opportunities. We have etch & deposition processes to enable optimum device performances at excellent Cost of Ownership. GaN RF Devices For a some time now GaN has been proposed for RF devices up to 650V and whilst it has always had a place within military applications, it’s nowapproaching the stage when it can be offered to the general market. Companies are already offering the advantages of GaN: fast switching, reduced packaging & higher efficiency. The innovation around the manufacture of normally OFF (E mode) devices are bringing new ideas such as atomic layer etching & production atomic layer deposition into the mainstream. OIPT can offer these technologies on a single cluster system with industry standard cassette handling for high throughput. InP Lasers Data, data, data - if you thinkwe have high data traffic now just wait and see! Internet of Things, 5G, video calls all these and more require communications at a scale we have never seen before. InP lasers are at the heart of the communications systems - without them, you cannot get the speeds needed to satisfy the volume. Over the years we’ve developed facet etches, dielectric passivation and many other process solutions for our customers. 2018 was a great year, with compound semiconductors moving into mainstream manufacturing in many exciting ways. OIPT is THE compound semiconductor plasma process solutions company. 2019 will be even more exciting still!
Here at Oxford Instruments Plasma Technology (OIPT) we have a rich history of working with the best researchers in the world and helping them take their work from the Lab into the Fab. 2018 has been a fantastic example with many innovations reaching a stage where they are being offered to the market. GaAs/AlGaAs VCSELs With the inclusion of facial recognition in the iPhone X late 2017 VCSEL arrays had entered the consumer market. While the forecast may have softened slightly this technology is here to stay and is planned to be introduced by many other handset manufacturers. This is scratching the surface of the impact VCSELs will have. We’ve engaged with most of the VCSEL manufacturers finding out what they need from us and then developing targeted process solutions. Through this approach we have managed to deliver process and device improvements that have really helped our customers move forward. SiC Power Devices Having attended both ISPSD and ECSRM this year it struck me the amount of excitement around both SiC & GaN materials. The atmosphere at ECSRMwas reminiscent of the pre-HBLED boom - just before the market exploded, and this time the market forecasts are based on a solid need and performance fit. Substrate suppliers say their biggest headache is that they simply cannot produce enough wafers to satisfy the market. With the growthofElectricandHybridVehicles that will happen, SiC power devices are only heading in one direction. This will drive innovation & reduce manufacturing costs
IN THIS ISSUE:
Page Article
Welcome note
2.
Thinking strategically: Ensuring we deliver what our customers need Ion Beam Nano Patterning: Large-area VR & AR devices become a reality Atomic Layer Etching
3.
6.
8.
13. Oxford
Instruments Supplies First Mass VCSEL
Production Fabrication Facility in China 14. Dry Etching of AlGaN 16. Control is everything: Control Sytems and PIDs 18. Oxford 19. White paper: Device fabrication for life sciences & biomedical applications 20. Technologies to look out for in 2019 and learnings from 2018 22. White papers 23. Webinars Instruments at European partners meeting for photonic chips
Dr Mark Dineen, Technical Marketing Manager
If you ha ve any articles that you would like to submit to the next edi tion of PROCESS NEWS please send them to: plasma-experts@oxinst.com
2 Winter 2019
Opinion
Thinking strategically: Delivering what our customers need
Author: Klaas Wisniewski, Strategic Business Development Director
Plasma Technology enables performance material solutions to semiconductor needs. We are at the heart of global technology trends, so strategic planning has a permanent place in my diary for the purpose of obtaining new market intelligence and making assessments to improve our production solutions.
Datacom / Telecom Lasers that are produced using our process solutions are one of the most important components for optical transmission. The volume of data transmitted via fibre optic cables is currently growing exponentially, driven by the increased use of internet services, supporting the build out of the 5G netwrok eco system. The increase in worldwide data traffic due to mobile telecommunications and data transfer via fibre optics raises demand for lasers as optical signal transmitters, photodiodes as receivers as well as optical amplifiers and switches. Infrastructure investments (high- speed data communications, applications, data centers, etc.) are pioneering the use of consumer electronics, such as the rising electrification of automobiles, Industry 4.0 and IOT. These are being pioneered by the big corporations such as Apple, Samsung, Google, Facebook, Microsoft, Sony and Huawei etc. Driven in the US, has
We are in the ‘Decade of Materials’ - performance materials will enable applications silicon cannot do. Chip performance needs will drive new deviceconcepts from2020onwards, as the physical limits will soon be reached and new performance materials are being considered. The trend is from “Moore’s Law” to “More Than Moore” Our “Lab to Fab” solutions for the fast-growing Compound Semiconductor market can enable faster commercialisation of technology than any other solution- provider. Global Megatrends There are two global megatrends affecting demand for our products: Mobility and Energy Efficiency. In mobility, applications are mainly for electric cars, virtual realities, Internet of Things and smartphones. These applications have an increasing proportion of performance materials. In energy efficiency - such as automotive or consumer electronics, the trend is
ever higher performance and an increasing energy demand. This requires high volumemanufacturing of new materials that deliver the demanded higher performance and lower energy consumption which are enabled by our compound semiconductor solutions. Optoelectronics: Laser based 3D sensor market Edge & surface emitting lasers (VCSELs) are increasingly used for 3D sensor applications in consumer electronics, industry, and the automotive sector. This is the main reason for increased laser demand. In 2017, Apple began using laser- based 3D sensor functions in their products. According to a report by a Yole in June 2017, the introduction of 3D sensors in smartphones, automotives, gaming and robotics etc. will result in total estimated market value of USD 1 billion in 2017 and is expected to exceed USD 10 billion in 2020.
Winter 2019 3
are working to develop and qualify application chips with the goal of bringing them to market in 2020 and we have the solutions to get them there. Nanotechnology We further enable the development and production of advanced carbon nanostructures such as CNTs, nanowires and graphene in research and development. Application possibilities for such materials include silicons photonics, energy storage, display technology and semiconductors. The number of systems installed in close co- operation with customers allows us to build on our Lab to Fab strategy as the industry migrates from academic research to application specific equipment investments. Regional Trends Asia is very active overall, with strong regional differences, and in many regions there is an enthusiasm for technical innovation. New technology trends are being adopted very quickly by consumers and are experiencing rapid market penetration, which also has a positive impact on investment. In addition, Asian countries are strong power semiconductors in particular are big in Japan, where it still remains globally competitive. Currently, that market is equally split among the U.S., Europe and Japan. Mitsubishi, Hitachi, Sumitomo and Fuji Electric amongst many are particularly strong in power devices and Japan has the largest program in GaN and SiC processing development activities. The environment inChinawill remain disruptive. Therefore, we have to growth markets. Optoelectronics &
keep our business model adaptable to changing market conditions. We are alreadymaking increasing use of market analyses, which will certainly gain in importance in the future. In the US, R&D investment and acquisition priorities are aimed at ensuring institutes remain global leaders in the semiconductor industry. As economic trends and technologies such as artificial intelligence evolve, the US strategy for manufacturing advanced semiconductors and driving R&D in technology innovation has become an economic priority. European countries are at the forefront in investing into technologies for a modern, clean and fair economy with potential to feed future growth. With our portfolio of materials from the Opto, Discrete, Nano and Lab to Fab markets, we should be guided by the idea: “Be prepared for opportunities you haven’t expected”.
led to increased investment in technology and capacity can be observed. Market research companies such as Ovum, IDC or Yole expect investments in laser communication to increase to support the growth of data traffic. According to a study by IDC, global data traffic will grow from 8 zettabytes in 2015 to 160 zettabytes in 2025. LED In the LED sector, the return to profitability of all participants was the highest goal. However, this only applies to the established LED applications (backlighting (BLU), mobile phone / smartphone applications and general lighting). “MicroLEDs” & emerging displays have potential and could become the upcoming applications. Both trends have a decisive influence on the manufacturing process and requirements and also to us as an equipment manufacturer for critical processing solutions. Power Electronics: SiC to GaN the Nobel Prize winning material Our customers are working to develop new devices with the goal of bringing them to market in 2020, we have key process solutions to get them there. Plasma Technology are proud to be part of the Tesla (Model 3) supply chain by providing critical process solutions to the SiC substrate industry. Manufacturers have expressed the desire for more capacity (increasing the number of systems) or even for larger substrates (from 4 to 8 inches). Gallium Nitride (GaN) development, enabling applications such as Wireless Charging or 5G networks standards. Our customers
4 Winter 2019
Hybrid organic-inorganic perovskite solar cells are heavily researched due to their potential to offer both high conversion efficiency and low cost. However, so far, environmental device stability is a major issue. Many avenues to improve the stability of these cells are being investigated with the added constraint of retaining or reaching a high efficiency. One avenue that is seen as very promising is the use of inorganic thin films in the design of the device. ALD with its excellent control of film growth and high-quality films is seen as a key technology to this end. ALD : Towards Stable & Efficient Perovskite Solar Cells
Mostly for research but also for pilot-production, there is a desire for flexible tools with wide processing range and both plasma and thermal capabilities. Oxford Instruments FlexAL ® tools are ideally suited in this respect. The goal of this white paper is to give a practical overview on what ALD is and how it can be used to benefit perovskite solar cells. To illustrate these benefits we will refer to some recent publications where Oxford Instruments equipment is utilised. In atomic layer deposition (ALD), thin films are built up in cycles in which the surface is exposed to various vapor or gas-phase species in alternating, separated doses. In each cycle, a submonolayer of material is deposited. As illustrated in Fig. 1, a typical cycle consists of four steps: (i) a precursor dosing step, where a precursor is typically an inorganic metal- organic or metal-halide (e.g. TMA); (ii) a purge and/or
pump step; (iii) a co-reactant step, typically involving a small molecule (e.g. H 2 O or O 2 plasma); and (iv) a purge and/or pump step. For the precursor, the element to be deposited is in many cases the metal center (e.g. Al), while for the reactant, it is typically a non-metal such as O. Together these then form the resulting film (e.g. Al 2 O 3 ) For ALD, it is vital that the precursor and co-reactants react with the surface in a self-limiting way. The precursor molecules and co- reactants react neither with themselves nor with the surface groups that they create. In the purge and/or pump steps, the gaseous reaction products that may be generated during the surface reactions, as well as any excess precursor or co-reactant molecules, are removed from the ALD reactor. This is necessary to avoid reactions between precursor and co-reactant molecules directly in the gas phase or on the surface, as this could lead to an undesired chemical vapour deposition (CVD) component.
Half-cycle 1 st
2 nd
Purge
Purge
Half-cycle
reaction products
reaction products
co-reactant
precursor
CVD
CVD
saturation
saturation
saturation
saturation
under saturation
under saturation
Time
Exposure
Exposure
Time
Fig. 1. A schematic representation of the various steps in an ALD cycle consisting of two half-reactions. The exposures in the first half-cycle (precursor) and second half-cycle (co-reactant) are self-limiting such that the process stops when all available surface sites are occupied. The two half-cycles are separated by purge steps. The lower panels show the resulting coverage, or growth per cycle, as a function of exposure or time for that particular step. For sufficient exposure, saturated growth is obtained, while insufficient exposure results in incomplete saturation. For insufficient purging, a CVD component frommixing of the precursor and co-reactant is obtained. 1
Download the full white paper from: Plasma.oxinst.com/ALDPerovskites
Winter 2019 5
ROCESS
Ion Beam Nano Patterning: Large-area VR and AR devices become a reality
Author: Dr David Pearson, Senior Technologist
The fabrication of devices for applications in VR (virtual reality) and AR (augmented reality), as well as in other areas of optoelectronics (e.g. the formation of laser diode facets at the wafer level), are increasingly making use of ion beam etching techniques. In particular, nano-patterning of arrays, gratings or laser diode facet structures with features that are angled relative to the normal to the device surface, notably where there is a re-entrant or acute angle (see Figs. 2a and 2b), cannot typically be etched using conventional plasma etch techniques (ICP or RIE) as the angle of the feature is not easily controlled. This is because the bombardment direction of the ions extracted from the plasma by biasing the substrate is always normal to the overall device surface, whereas in ion beam etching the ion bombardment direction and the orientation of the substrate are totally independent thus enabling nano- features angled relative to the substrate normal to be formed. As mentioned in Process News, Summer 2018 article on ion beam nanopatterning, materials such as SiO 2 and InP or GaAs can be nano-patterned (with depths of hundreds of nanometres to microns) by ion beam etching with high selectivity to the patterning mask if a suitable mask material and etch chemistry are chosen (RIBE or CAIBE). Vertical, or close to vertical, features can be etched with good control of sidewalls and absence or elimination of redeposited material, and with a high uniformity over large substrate areas (typically up to 200mm round or 150mm square, for example). Two examples of these types of etched periodic arrays are shown in figures 1a and 1b. The sidewall shape and
overall uniformity are controlled by the substrate angle to the ion beam direction (and, to a lesser extent, by the ion beam energy and flux which may be independently controlled), where the substrate undergoes on-axis rotation to provide axisymmetry of the feature shape and depth (etch rate) across the substrate. However, for angled etching by ion beam, the angle of the features is determined by the tilt angle of the substrate normal relative to the ion beam direction, where the substrate is now static at a fixed in-plane orientation. The problem that is then posed is the decrease in local ion current density with distance from the beam formation grids, due to a finite divergence of the individual ion beamlets, resulting in a decreasing local etch rate with distance from the grids. Since the substrate is tilted, one side of the fixed substrate will be further from the grids than the other side with a consequent imbalance in the local etch rates across the substrate diameter in the direction of the tilt, as can be seen schematically in fig.3. The final outcome will then be an etch depth uniformity along this direction that is greatly degraded depending on the required tilt angle for the nano-patterned features (typically 30 to 45 degrees). The Ionfab300 ion beam tool for angled grating etch has novel features which can provide a solution to this etch rate variation (tilted etch compensation technology or ‘tilt technology’) resulting in a uniform static tilted etch rate profile as shown in fig.4. This allows angled gratings to be produced for augmented/virtual reality and angled structures for other optoelectronic applications over large area substrates with an excellent uniformity.
6 Winter 2019
Fig.1a Cr/PR etched array as UV polarizer former
Fig.1b Optical grating etched in silica
Fig.2a Angled etched silica grating to be used in augmented reality (AR) applications
Fig.2b GaAs angled etched laser diode facets
Fig.3 Schematic diagram of angled etch showing the ion beam current density fall-off due to beam divergence resulting in a local etch rate decrease across the substrate
Fig.4 Comparison of the uniformity of etch rate with and without the ‘tilt technology’ applied.
1 United Kingdom Patent Application No. 1819600.6
Winter 2019 7
Atomic Layer Etching
Author: Dr Mike Cooke, Chief Technology Officer
Plasma etching has been an unsung hero of the microelectronics revolution of the last four decades. The shrink in minimum feature size from 5μm to below 10nm developed by lithography specialists would be worthless, without a method to transfer such patterns vertically into thin films by etching.
There are two drivers for change in plasma etching: productivity and precision. The first demands more wafers per hour, and hence higher etch rate, together with improved uniformity, repeatability and tool reliability. But while features sizes shrink, the layer thicknesses also shrink until in some cases a layer is only a few nanometres thick. That will become even a single atomic layer thickness if 2D materials such as graphene and molybdenum disulphide enter mainstream fabrication. Precision etching cannot proceed at 10 µm per minute. Here the keys are selectivity to underlying layers, and the damage introduced by etching. Atomic layer etching (ALE) is being developed as a precision etching technique, capable of etching very thin layers without damaging the material beneath. Where plasma etching converts solid material to gas by means of fluxes of chemical radicals and energetic positive ions, plasma ALE separates the two fluxes in time: first a chemical gas dose is delivered, and later ion bombardment acts on adsorbed material. When the ion energy is carefully tuned, bombardment removes both the adsorbed species and some components of the solid surface, but does not remove
Plasma etching is the underpinning technology, providing two inherent benefits: • High temperature chemistry at low substrate temperatures • Positive ions bombarding the surface vertically The first property drives efficient etching. Collisions in a plasma fragment the input gases, turning inert, safer gases such as SF 6 into highly reactive agents. Free fluorine atoms are then ready to combine with the solid surface to create new volatile chemicals such as SiF 4 , turning solid to gas. The gas desorbs and is pumped away, taking the substrate layer with it. The second is an intrinsic property of the plasma: a boundary layer called the plasma sheath exists next to all walls, including next to the substrate to be etched. Positive ions are accelerated towards the substrate by the sheath voltage drop, acquiring far more energy than their initial thermal energy. Like a rock thrown off a cliff, they hit the surface close to vertical regardless of their initial velocity. This directed energy can drive vertical, anisotropic etching when etching reactions are stimulated by this energy, rather than by thermally driven chemical processes, which tend to be isotropic.
8 Winter 2019
Oxidised surface AlGaN
Dose
BCl
BCl
BCl
O O
AlGaN Oxidised surface
AlGaN Oxidised surface
Pump byproducts
Purge
Etch
AlGaN
Figure 1. An ideal atomic layer etch cycle. A chemical dose is adsorbed (left); excess gas is removed (top); ion bombardment removes just the chemically bound species (right); reaction products are pumped away (left bottom). The etching of silicon with chlorine under argon ion bombardment is depicted. because the source gas does not itself bind to the surface (for example, CHF 3 used in ALE of silicon dioxide). Ion bombardment energy should be held below the threshold for sputter etching in the chemical dose step, so a plasma source such as an induction coupled plasma (ICP) which does not couple RF power directly to the substrate is ideal. Fine control of ion energy is required, in the range where most etching thresholds lie: 10 – 50 eV. Ideal ALE requires that the adsorbed group have an etch threshold energy distinctly lower than that for etching the substrate – in practice this is hard to achieve, because there may be only a few electron-volts between thresholds, and the ions from a plasma are not monoenergetic. The gas dose step is critical. Using a standard plasma etcher for ALE will require extended purge times, because the equipment is normally optimized for repeatable, steady state operation. Figure 2 shows the optical emission signal from carbon at 516.5 nm in a steady argon plasma, with a steadily increasing dose time of CHF 3 added, from 11 msec to 7 seconds. For dose times up to about 200 msec, there is a sharp pulse with a rapid decay. For dose times up to about 2 seconds, the decay takes several seconds. Longer dose times do not decay back to floor level, even with 40 seconds before the next gas pulse. The requirement to switch between chemical dose and non- chemical energetic bombardment puts limits on the cycle time, not because of the ability to create short gas pulses, but because of the time taken to
additional, unreacted layers. Time is normally required to clear gas phase chemical species before the etch phase, and possibly to clear reaction products before the next dose. Ideally both the dose and etch phases are self-limiting: the adsorbed layer forms a defined reservoir of reactant, and the etch step should be selective between bound species and the underlying solid matrix. Plasma ALE of silicon was shown by Athavale and Economou in 1996 [1]. Other materials etched by ALE include GaAs [2] , SiO 2 [3] graphene [4] and copper [5] . A purely chemical approach in the gas phase came later, notably through the work of Steven George’s group [6] differing from ALD, where thermal processes preceded plasma ALD. Wet ALE processes exist [7] , and studies have also used ion or neutral beams for the etch step, rather than plasma [8] EQUIPMENT ALE requires a plasma etcher with separate control of plasma generation and ion bombardment energy. Some of the chemical dose steps require radical creation either for a faster reaction (atomic chlorine saturating the surface faster than chlorine gas), or
Winter 2019 9
SiCl would leave surplus silicon chlorine to be recycled. Example data of chlorine/argon ALE of PECVD amorphous silicon from our work [12] is shown below, using a total cycle time of 7 seconds. PECVD silicon was deposited on SiO 2 , so that the etch rate could be measured accurately using ellipsometry, after a partial
Figure 2. 516 nm emission from pulses of CHF 3 added to an argon plasma, with gas pulse times steadily increasing from 6 milliseconds to 7 seconds, spaced 40 seconds apart. clear the chamber. Optical emission spectroscopy is a very useful tool, in revealing any build-up of species in the chamber. An ALE etch tool is also likely to be required to work as a conventional etcher, if the requirement is to etch down through a large amount of material, and land As with ALD, ALE etch rates are expressed in etch per cycle (EPC), sometimes without complete transparency on the total cycle length. To qualify as ALE, rather than quasi-ALE (using the cycle, but without any self- limiting characteristics), certain tests must be satisfied. • Synergy: Both the dose and etch steps should be necessary for etching. If significant etching occurs with only one of these present in the cycle, conditions for ALE are not satisfied • Saturation: There should be evidence of a plateau in EPC against dose time, etch time, purge time, and etch energy, indicating self-limitation in dose and etching, and adequate purging. Silicon ALE of silicon using a chlorine dose and argon ion bombardment in the etch cycle has been studied most extensively. Modeling the cycle [9] now includes non-ideal factors, such as the persistence of background chlorine gas during the etch step and ion bombardment during the dose. Experimentally, widely differing data have been reported on EPC in a plateau regime. [10, 11] . This may be attributable to differing depths of surface amorphisation under ion bombardment, or could be caused by differing amounts of chlorine bonded to surface silicon (SiCl, SiCl 2 , SiCl 3 ), coupled with different distributions of removed species. For example, a SiCl 3 surface which ejected mainly softly at a critical interface [8] . Etching Performance Benchmarks
Figure 3. Etch per cycle of amorphous silicon as a function of DC self-bias in the etch phase, with (solid line) and without (dashed line) a 40 msec chlorine gas dose.
layer etch. The solid line is the EPC with a 40 msec chlorine dose, and the dashed line is the EPC with the dose omitted; DC self-bias during the etch step is varied. The graph shows that synergy is satisfied, and there is a partial plateau from 50 volts bias. In this range, argon sputtering is already above threshold, so the plateau is not ideal. ALE could gain ground in mainstream silicon etching, if aspect ratio dependent etching (ARDE) effects are important. The etch depths of high aspect ratio trenches (>10:1) are often significantly less than more open features, limited by loss of etch species to the sidewalls before they can act at the floor, and by transport of reaction products out of the trench. ALD became compelling for its conformality, derived from surface- limited adsorption. ALE offers a similar benefit, where a surface saturation occurs in the dose step, and over-etching is sufficiently selective after adsorbed species are cleared. Silicon dioxide The ALE cycle for SiO 2 is not strictly self-limiting. In the dose step, a thin (~2 nm) fluorocarbon layer is deposited, which acts as a fixed reservoir of reactants in the etch phase, under noble gas bombardment. This process mimics continuous plasma etching of SiO 2 , where the process uses this layer to achieve selectivity to silicon.
10 Winter 2019
The cycle places strong demands on the repeatability of the tool, both in chamber condition and in process cycle parameters. Too little polymer will reduce the etch rate for lack of reactants, and too much will also reduce it, by blocking positive ions. If the polymer layer builds up over time, etching will stop.
/Ar ALE
Figure 4 shows the EPC from an example CHF 3
Figure 5. Raman spectra of MoS 2
exposed to argon/chlorine
ALE cycles.
in MoS 2
. It seems that ALE is able to proceed in this case
with very little disruption to the underlying lattice.
When heterostructure devices are finally demonstrated using a stack of 2D materials (e.g. hexagonal boron nitride insulator, graphene conductor and TMD semiconductor) – and not by mechanically assembling flakes – it is likely that a patterning process will be needed which can remove one very thin layer, without disrupting the next atomic layer. GaN and AlGaN In order for a normally off HEMT device to be realised with both low leakage current and minimal on resistance an AlGaN layer needs to be etched down to the underlying GaN. This AlGaN layer is thin, typically 20-30nm and there is no inherent selectivity to GaN for a classic etch process. It is often capped with a thin GaN layer to prevent oxidation of the AlGaN in air. Therefore an extremely uniform, low etch rate is essential to allow for the depth accuracy to be achieved across a complete wafer. Etching
Figure 4 EPC of SiO 2
and argon etching, varying the DC self-
bias, with (solid line) and without (dashed line) a CHF 3
gas dose.
cycle, of total duration 40 seconds. This data was taken without a fast chemical dose step hardware kit, so the fluorocarbon polymer deposited was relatively thick at 0.6 nm per cycle, seen in the data at zero bias. There is a reasonably broad 20 eV windowwhere synergy is perfect, and argon ion sputtering is almost zero. There is only a hint of a plateau, and this process is classified quasi-ALE. This style of process may find application for clearing native oxides at critical interfaces, such as ohmic electrical contacts. However, the use of polymer may prove problematic, because using oxygen plasma to clear it risks re-growing the oxide. Molybdenum disulphide There is considerable interest in the broad class of two dimensional (2D) materials, especially transition metal dichalcogenide (TMD) semiconductors such as MoS 2 , because electron mobility may be preserved even as the layer thickness is reduced to single digit monolayers. The material was etched in the same chlorine/argon cycle as silicon, and measured by Raman spectroscopy, for different numbers of ALE cycles [13] . The change in separation of the two principal peaks around 400 cm-1 indicates that etching is taking place; 40 cycles removed the entire layer. The absence of a peak at 227 cm-1 is particularly interesting, because it is attributed to damage
Figure 6 AlGaN etch per cycle in Ar/Cl 2
with and without the
chlorine dose.
at the lowest possible ion energy is sought, to reduce damage to the remaining material. Two ALE methods have been tested for these materials. One is a double chemical process, with an initial oxidising plasma followed by chemical ion bombardment. This gives excellent self-limiting etching, but in a relatively long cycle time, above 30 seconds. The other displays ALE synergy but only limited saturation, using chlorine and argon ions in a six second cycle. CONCLUSIONS Atomic layer etching is applicable to a range of materials, rather than being limited to a few special cases, thereby earning its name as a distinct technique. It is transitioning from a curiosity to a usable technique, with the appearance of specialized ALE equipment from several vendors. ALE is beginning to find application areas where its special properties (reduced aspect ratio dependence, and low ion bombardment energy) are useful advantages. If it follows the path of ALD, more applications will follow. ACKNOWLEDGEMENTS Some of the work here was supported by the EC FP7 grant 318804, ‘Single Nanometer Manufacturing’. The work of the team at Oxford Instruments, especially Andy Goodyear is acknowledged. This article was originaly published in the Fall/Winter 2018 bulletin of the Society Vacuum Coaters.
REFERENCES 1. S.D. Athavale and D.J. Economou, “Realiza- tion of atomic layer etching of silicon” J. Vac. Sci. Technol., B 14, 3702, 1996. http://dx.doi. org/10.1116/1.588651 2. T. Meguro, M. Hamagaki, S. Modaressi, T. Hara, Y. Aoyagi, M. Ishii, and Y. Yamamoto, “Digital etching of GaAs: New approach of dry etching to atomic ordered processing” Appl. Phys. Lett. 56, 1552, 1990. http://dx.doi. org/10.1063/1.103171 3. D. Metzler, R.L. Bruce, S. Engelmann, E. A. Jo- seph, and G.S. Oehrlein, “Fluorocarbon assisted atomic layer etching of SiO 2 using cyclic Ar/ C4F8 plasma” J. Vac. Sci. Technol. A 32, 020603, 2014. http://dx.doi.org/10.1116/1.4843575 4. W.S. Lim, Y.Y. Kim, H. Kim, S. Jang, N. Kwon, B. J. Park, J-H Ahn, I. Chung, B.H. Hong, and G.Y. Yeom, “Atomic layer etching of graphene for full graphene device fabrication” Carbon, 50, 429, 2012 http://dx.doi.org/10.1016/j.car- bon.2011.08.058 5. F. Wu, G. Levitin, and D.W. Hess, “Mechanistic considerations of low temperature hydro- gen-based plasma etching of Cu” J. Vac. Sci. Technol. B 29, 011013, 2011.http://dx.doi. org/10.1116/1.3520461 6. S.M. George and Y. Lee, “Prospects for Ther- mal Atomic Layer Etching Using Sequential, Self-Limiting Fluorination and Ligand-Exchange Reaction”, ACS Nano 10, 4889-4894, 2016. http://dx.doi.org/10.1021/acsnano.6b02991 7. G.C. DeSalvo, C.A. Bozada, J.L. Ebel, D.C. Look, C. Barrette, L.A. Cerny, R.W. Dettmer, J.K. Gil- lespie, C.K. Havasy, T.J. Jenkins, K. Nakano, C.I. Pettiford, T.K. Quach, J.P.S. James Sewell, and G.D. Via, “Wet Chemical Digital Etching of GaAs at Room Temperature” J. Electrochem. Soc., 143, 3652, 1996. http://dx.doi.org/10.1149/1.1837266 8. A. Goodyear and M. Cooke, European patent “Cyclical Plasma etching” EP16187143, Priority date 3 September 2015. 9. C.M. Huard,Y. Zhang, S. Sriraman, A. Pater- son, K. J. Kanarik and M. J. Kushner “Atom- ic layer etching of 3D structures in silicon: Self-limiting and nonideal reactions” J. Vac. Sci. Technol. A 35, 031306, 2017 http://dx.doi. org/10.1116/1.4979661 10. S.-D. Park, K.-S. Min, B.-Y. Yoon, D.-H. Lee, and G.-Y. Yeom, “Precise Depth Control of Silicon Etching Using Chlorine Atomic Layer Etching” Jpn. J. Appl. Phys., Part 1, 44, 389, 2005. http:// dx.doi.org/10.1143/JJAP.44.389 11. S. Tan, W. Yang, K.J. Kanarik, T. Lill, V. Vahedi, J. Marks, and R.A. Gottscho, ECS J. Solid State Sci. Technol. 4, N5010, 2015. http://dx.doi. org/10.1149/2.0031506jss 12. A. Goodyear and M. Cooke “Atomic layer etch- ing in close-to-conventional plasma etch tools” J. Vac. Sci. Technol. A 35, 01A105, 2017 . http:// dx.doi.org/10.1116/1.4972393 13. E. Mercado, A. Goodyear, J. Moffat, M. Cooke, and R.S. Sundaram “A Raman metrology approach to quality control of 2D MoS 2 film fabrication” J. Phys. D 50, 184005, 2017. http:// dx.doi.org/10.1088/1361-6463/aa6786
OIPT Cobra ® cluster system for ALE
12 Winter 2019
NEWS
Oxford Instruments Supplies First Mass VCSEL Production Fabrication Facility in China
Oxford Instruments Plasma Technology (OIPT), a leading supplier to today’s optoelectronics manufacturers announced that Sino-semic have selected OIPT’s Cobra® plasma etch systems for their manufacturing facilities in Taizhou City. Sino-semic, the first all-Chinese developer and manufacturer of VCSELs for face recognition, noted the process capability and local support were key factors in their decision to adopt the high-performance inductively coupled plasma (ICP) etch Cobra systems. Li Jun, Vice General Manager at Sino-semic comments: “We chose Oxford Instruments to supply our ICP etch equipment because they offer cutting edge plasma processing systems and unrivalled process support, which will be invaluable to us during our production scheme”. The Cobra process solutions are designed to support leading edge device applications such as Lasers, RF, Power and advanced LEDs. “VCSEL based devices are entering another exciting phase of growth”, stated the Managing Director at OIPT, “We are thrilled to be providing our VCSEL processing solutions to a pioneering production manufacturer such as Sino-semic”.
Winter 2019 13
ROCESS
Dry Etching of AlGaN Anisha Kalra, Shashwat Rathkanthiwar, Anamika S. Pratiyush Indian Institute of Science, Centre for Nano Science and Engineering (CeNSE) Bangalore
Aluminium Gallium Nitride (AlGaN), a member of the III-nitride semiconductor family finds extensive applications in high power, high speed electronics as well as deep-UV optoelectronics owing to its superior material and electrical properties likewide direct energy band gap which can be tailored by varying the Al mole fraction in the solid alloy, radiation hardness, chemical stability and availability of GaN/AlGaN heterostructures. Anisotropic etching of these III-nitride epilayers with a controlled etch rate and requirement of damage free smooth surface morphology post etching is a crucial fabrication step in realising any vertical device structure. Inert chemical nature and high bonding energies make it difficult to utilise conventional wet etching techniques to achieve high etch rates and anisotropic conformal etch patterns, thus making chemical etch based approach undesirable for commercial applications.
Chlorides of Aluminium and Gallium are volatile, making dry etching based on BCl 3 /Cl 2 chemistry attractive for achieving controlled etch for mesa isolation as well as recess etch, two of the most critical fabrication steps in III-nitride based devices. Fabrication of optoelectronic devices like photodetectors and light emitting diodes (LEDs) working in the deep-ultraviolet regime of the electromagnetic spectrum require the use of high Al- content (>40%) AlGaN layers. Though etching of AlGaN layers for processing of AlGaN/GaN HEMTs has been studied extensively, lower volatility of Aluminium Chlorides and higher Al-N bond energy compared to their Ga counterparts make high Al-content AlGaN etch challenging. In this work, we aimed at exploring the etch rates, selectivity, surface morphology and process repeatability during Inductively Coupled Plasma (ICP)- Reactive Ion Etching (RIE) of c-plane oriented Al 0.5 Ga 0.5 N using BCl 3 /Cl 2 /Ar based chemistry.
14 Winter 2019
NEWS The ICP-RIE experiments were performed on 0.5 - 1 μm thick Al 0.5 Ga 0.5 N layers epitaxially grown on c-plane sapphire substrates by Aixtron low-pressure Metal Organic Chemical Vapor Deposition (MOCVD). Samples were patterned using standard i-line optical lithography, utilising hard baked AZ4562, a 6 μm thick photoresist (PR) as the etch mask. The etching was performed in an Oxford Instruments Plasmalab System 100, in which the ion plasma and energy can be independently controlled by a coil and RF biasing to the substrate, respectively. Stable DC bias, repeatable etch rates, good process selectivity and low etch surface damage was obtained with low RF ICP powers and hence, these along with temperature of the Si carrier wafer (25⁰C) were maintained constant for all the subsequent AlGaN etch experiments. Etch rates were found to drop considerably on lowering the RF power whereas increasing the RF power had no impact on the etch rates but led to significant roughening of the etched surface. Different gas mixtures utilising BCl 3 /Cl 2 and BCl 3 /Cl 2 / Ar based chemistries were explored for etching of the Al0.5Ga0.5N samples. Al0.5Ga0.5N : PR etch selectivity varied from ~0.6 to 1 on the addition of Ar to the plasma. Addition of Ar to the etch chemistry in the mentioned proportion however, led to an increase in the etch rate from 1.4 μm/hr to 2.5 μm/hr as studied using Atomic Force Microscopy (AFM) images. An increased etch rate implies less exposure to plasma for the same etch depth requirement and hence leads to an improvement in the surface morphology as well as sidewall profile due to less plasma damage and lower probability of PR burning and etch product redeposition. Morphology of the etched surface can be further improved by exposing the surface to a low power O 2 /BCl 3 cyclic plasma post- etch.
AFM Step height measurement
Surface morphology
characterisation of the un-etched
Average Roughness: 1.4 nm
and etched regions post etching using AFM
Un-etched
Etched
Average Roughness: 1.7 nm
Scanning Electron Microscope (SEM) images of the un- etched and etched regions, showing smooth morphology post- etch with very low density of photoresist residue
Winter 2019 15
Control is everything: Control Systems & PIDs Author: Rhys Miles, Platform & Control Systems Manager OFTWARE
operations working in parallel on the numerical difference between the current state and desired future state of a system; the first operator is proportional to this numerical difference (the P term), the second an integral of the difference (the I term), and the last being the differential (the D term) of the difference. The PID controller is not a new concept even though it is extensively used in modern computer and software controlled systems. Reference to PID control appeared in theoretical papers as far back as 1922, and many mechanical devices existed for over 100 years before that. Being a well-researched subject, a frequently used model of the PID controller is shown in this diagram:
Under the covers of every Oxford Instruments Plasma Technology product, is a Control System (CS). The CS describes the looms of cables interconnecting sensors and actuators dotted around the equipment to the central processing units that control what happens, and more importantly, in what order things happen. Comparisons can be made to the human body as a way to explain how individual components of the CS operate as a whole. The sensors distributed around the equipment correspond to our human senses (touch, sight, hearing, smell, and taste) while actuators (valves, pumps, flow controllers) are very similar to our muscles moving things and changing the world round us. The web of wires provides the same function as our blood vessels and nervous systems combined, providing power and distributing messages. And the central processing unit is both the heart and brain of the operation. The CS is the collective description of all the internal electronics, mechanical devices and software modules needed to operate our products. Control Loops Once you have a CS, you can start building Control Loops (CL). Our homes have a perfect example of a CL. Our heating or Air Conditioning (AC) unit in combination with a thermostat or temperature sensor operate in a CL. If you want the room temperature to change, then you alter the thermostat. This triggers the boiler or AC unit to heat or cool; the room temperature changes which in turn causes the sensor to see the change. Our products have many hundreds of CLs working in parallel. Some are pressure related, others control temperature, and gas flow too; not forgetting the Radio Frequency (RF) power CLs required to create a plasma. PID Control Loops The king of all CLs is the PIDController. The ever-present and ubiquitous PID Controller. The PID controller is so named because of the three central mathematical
Using our home air conditioning example, we can programme the AC controller with our desired room temperature, which in PID controller terminology is the setpoint or SP. A temperature sensor is used to monitor the room temperature. This is the process variable or PV. The setpoint (SP) represents where we want to get to, while the process variable (PV) represents where we are now. If we calculate the numerical difference between where we want to get to and where we are now, then we have an error value, which for any given point in time t, is e(t)= SP - PV. The AC controller needs towork out howmuch power output is needed to get to the desired temperature. Typically it will operate at 100% power when the error values are large, and much lower power levels <10% when close to the right temperature. In PID terminology this is the controlled variable, u. For any given time is represented as a function of t, u(t).
16 Winter 2019
So a room in a building in a cold climate at 10°C that we want to heat up to 21°C has an error value of 11°C at the start. This becomes an error of 10°C when the room heats up by one degree, and so on until the room reaches 21°C and the error value is 0°C In the simplest of control strategies we can take this error term and multiply it by a fixed factor to determine how much power is needed from the AC unit, which is one of the math operations performed as part of the PID controller. For PID controllers this fixed factor or constant is called K_p. So the power output of the AC unit, u, becomes u(t)=K_p*e(t) If we plug in some numbers from our example, SP = 21°C , PV = 10°C, and pick a high constant value of K_p=10
on the breaks limiting the overshoot. The derivative control from the PID can be used for this purpose. The derivative of the error value is the slope of change. When used in this mode, it can be thought of as detecting high velocity of change which needs to limit the output, and hence usually works in the opposite direction to the other two components of the PID controller. While K_p and K_i are positive factors, K_d would be negative in this mode of operation. In the examples so far the air conditioner heater output is only operating when there is an error in where we want to be at (SP) and where we are at (PV), meaning the AC heater is pulsing between zero output (0%) and some non-zero value as it crosses through the desired temperature. Additionally we know if we were to set the heater output to a fixed value, e.g. 14% heat, then the room temperature would hit equilibrium where the heat added to the room equals the heat losses, leaving the room perfectly stable in temperature, with no oscillations. But our proportional only controller and our combined proportional and derivative controller will eventually deliver zero power output whenever we meet the target temperature, not some constant nonzero value. This is where the integral element, the final element in the PID controller takes its rightful place at the table. It uses the history of the error signal, the integral of error over time, to influence the output. Critically, the integral of the error is not zero when SP = PV, so gets to play its part in “biasing” the output to a non-zero value so we can have constant output to for the control applications where equilibriumplays its part. So putting it all together, the PID controller output, u, can be calculated as follows
to give rapid temperature change then: When t=0, e(t)=SP-PV= 21°C-10°C=11°C and u(t)= K_p*e(t) so u(t)= 10*11=110%
The calculated power output (u) of the air conditioning heater at the beginning is 110%, or limited to 100% heat as this is the maximum the AC unit can output. We can tabulate all the remaining calculations as the temperature of the room increases towards the desired 21°C.
For many systems this control strategy can be quite effective - it becomes a P controller, aswe have not used the integral and derivative components. Unfortunately for many situations this simple approach can lead to oscillations which means the actual temperature of the room overshoots the 21°C making it feel too hot, then cycles to below21° makingyou feel colder for a bit before getting closer and closer through more overshoots and undershoots that are smaller in size eventually resting around 21°C. These oscillations are caused because there is usually some form of delay in the system. The room has both a thermal mass and a distance between the heater and the temperature sensor. This leads to a lag in time between doing something and noticing that something has changed. One way to overcome this problem is to detect the point when the rate of change is too fast and then put
All that is then to do is to select the appropriate〖 K〗 _p, K_i and K_dconstants for the given control application. Most controllers have an auto tune function that will observe the behaviour of the system to a step change in setpoint (SP), and back calculating what the appropriate constant factor values should be eliminating the need to experiment manually. PID controllers are used to ensure that the system set points are achieved rapidly and accurately ensuring repeatable process results day in day out.
Winter 2019 17
Oxford Instruments Plasma Technology joins European project focussed on building a pilot production line for photonic chips
Photonics is an emerging technology with a potential multitrillion euro market. Innovative small and medium sized enterprises (SMEs) are at the forefront of this development, but the R&D costs are prohibitive for them. That’s why 12 partners from northwestern Europe are creating an open access pilot line that will drastically reduce costs and time for the pilot production of new products. This new facility is projected to be the incubator of a thousand new companies and thousands of jobs. The 14 million euro project (OIP4NWE) is supported by the European Regional Development Fund and kicks off this week in Eindhoven. Photonics is much like electronics, but instead of electrons it uses light (photons) as its workhorse. It uses much less energy, it is faster, and it opens up a wealth of new opportunities. One of the key problems photonics will help tackle is the exploding energy consumption of data centers, as photonic microchips consume much less energy than their electronic predecessors. Another example is a high-precision monitoring system for aircraft wings, bridges or tall buildings. After two decades of basic photonics research, the first companies producing photonic integrated circuits (PICs) are now taking off – sparsely. One of the main hurdles is the high cost involved in R&D. Not only does the PIC production require expensive high-tech equipment installed in cleanrooms, but currently the production processes still have a high defect rate and are too slow. This was workable for basic research but not for commercial R&D. The technology readiness level, which ranges from 1 to 9, needs to be jacked up from the current 4 to 7. The new project, led by photonics stronghold Eindhoven University of Technology (in collaboration with its Photonic Integration Technology Center), consists of the realisation of an efficient pilot production line for shared use by European SMEs. It should take the defect rate in pilot production down and the throughput
time will be shorter. All in all, this should lead to a cost reduction which significantly lowers the threshold for developing new photonic products. This should help establish a thousand integrated photonics firms within ten years after the project. The front-end process (production of PICs on indium phosphide wafers) will be realised in the existing NanoLab@TU/e cleanroom facility at Eindhoven University. The PICs of different companies will be combined on one wafer to keep costs low. The back- end process is done at the Vrije Universiteit Brussel (optics for beam shaping and light coupling) and at Tyndall National Institute in Cork, Ireland (assembly of fiber-optic connections and electronics in the package). All steps require nanoscale precision to avoid product defects. The first stage of the project is equipment installation. The second stage focusses on automation of the equipment while a third stage will involve intensive industrial research together with equipment manufacturers to optimize and develop new processes. The line should be fully in operation in 2022. To incentivize the initial uptake by SMEs, a voucher scheme for external SMEs will be set up. The other parties involved are the companies AIXTRON SE (Germany), Oxford Instruments Nanotechnology Tools (United Kingdom), SMART Photonics, VTEC Lasers & Sensors, Technobis Fibre Technologies (all Netherlands) and mBryonics Limited (Ireland) along with research centers Photonics Bretagne (France), Cluster NanoMikroWerkstoffePhotonik.NRW (Germany) and Photon Delta Cooperatie (Netherlands). The project has a total budget of 13.9 million euros. Of this, the EU is funding 8.3 million, with the remainder coming from the participating parties.
18 Winter 2019
Page 1 Page 2 Page 3 Page 4 Page 5 Page 6 Page 7 Page 8 Page 9 Page 10 Page 11 Page 12 Page 13 Page 14 Page 15 Page 16 Page 17 Page 18 Page 19 Page 20 Page 21 Page 22 Page 23 Page 24Made with FlippingBook Annual report