ALD2019 Session AM-MoP: ALD for Manufacturing Poster Session

Cobalt precursors are being used as lining and capping materials since a while. The use is beneficial to limit diffusion of copper into the dielectric layer. Additionally deposited cobalt (and longer term ruthenium) metal will displace copper and tungsten because of lower resistivity and these materials require no liner or barrier when used as an interconnect.

Unfortunately, there are significant uncertainties in the supply chain.

1. Ethical sourcing of Cobalt: While Cobalt is not yet listed as a conflict material like gold, tantalum, tungsten and others, articles published by Amnesty International in Jan 2016 (https://www.amnesty.org/en/latest/news/2016/01/child-labour-behind-smart-phone-and-electric-car-batteries/) and Nov. 2017 (https://www.amnesty.org/en/latest/news/2017/11/industry-giants-fail-to-tackle-child-labour-allegations-in-cobalt-battery-supply-chains/) alerted end users and Semiconductor companies alike about child labor and artisanal mining being used to produce cobalt minerals. Umicore is worldwide one of the largest users of cobalt and has taken significant measures to ensure that cobalt sourced and downstream products supplied by Umicore don’t come from unethical sources. This has been audited and certified by an independent firm. Our poster contains more details on these issues.

2. The manufacture of cobalt precursors requires access to an intermediate called dicobaltoctacarbonyl. This material is currently only manufactured by two minor companies with limited production resources, one of these company is even located in a geographical area, which may add to the supply complexity. The poster details how Umicore plans to overcome this bottleneck and ensure sufficient supply for future growth of the industry.

3. Cobalt is also a raw material for Lithium batteries. This application is forecasted to grow significantly in future because of the electrification of the automobile. This raises questions about future price development and availability of cobalt raw materials. Our poster reviews the impact on the electronics industry, the impact on the electronics industry and possible solutions.

The deposition of homogenous conformal films with a precise thickness control on structured surfaces is essential for numerous applications for semiconductors, sensors and optics. Atomic Layer Deposition (ALD) is the key technology to fulfill the high requirements of such devices e.g. DRAM, MEMS, gratings. Plasma Enhanced Atomic Layer Deposition (PEALD) is an extension of ALD that opens up new possibilities such as an increased choice in materials and precursors, extended process region, and processing at reduced substrate temperatures. The new SILAYO ALD system of SENTECH Instruments GmbH with Planar Triple Spiral Antenna (PTSA) as large planar inductively coupled (IC) plasma source promises highly uniform and conformal ICPEALD deposition. Furthermore, the SILAYO system is equipped with substrate bias to control film stress during the deposition process.

ICPEALD Al₂O₃, SiO₂ and TiO₂ films were deposited on 8-inch silicon (Si) substrates. Trimethylaluminium (TMA), bis(diethylamino) silane (SAM.24) and tetrakis(dimethylamino) titan (TDMAT) were used as metal organic precursors and oxygen plasma was applied as oxygen source for the ICPEALD processes. Due to the low vapor pressure of the Si- and Ti-precursor, the material containers were heated to 70 °C. During the deposition process the substrate temperature of 250 °C was kept constant. Nitrogen (N₂) was used as carrier gas for the precursors and a constant flow of oxygen was applied through the plasma source.

The Al₂O₃ process exhibits a growth rate of 1.4 Å/cycle and an excellent non-uniformity of ± 0.3 % over the 8-inch Si wafer. The 66 nm Al₂O₃ film shows a refractive index of 1.64 with also an excellent non-uniformity of ± 0.08 %. The SiO₂ and TiO₂ processes show a growth rate of 1.1 Å/cycle and 0.6 Å/cycle, respectively. The two processes reveal an excellent non-uniformity of ± 0.7 % or ± 0.8 % over the 8-inch Si wafer. Furthermore, ICPEALD Al₂O₃ films were deposited with substrate bias. Standard Al₂O₃ film shows 97 MPa tensile stress. Applying a bias voltage during the deposition process, the Al₂O₃ film shows compressive film stress between -109 MPa and -144 MPa, depending on the set bias voltage.

Among various Metal Oxide Semiconductor (MOS) gas sensor, gas sensors based on ZnO nanostructures have been widely used due to its large exciton binding energy, wide band gape of 3.37 eV, good electrical conductivity, low cost, and high mechanical stability. In this work, ZnO nanotube nanostructure was introduced into the gas sensor application to ethanol vapor concentration due to its high electrochemical stability, nontoxicity, and, especially, high surface-to-volume ratio. Several coaxial ZnO nanotubes were synthesized on porous substrate to further enhance the sensing response to ethanol vapor with increased reaction surfaces.

In this study, ZnO nanotubes were synthesized by Atomic Layer Deposition (ALD) on Anodic Aluminum Oxide (AAO) with sacrificial Al₂O₃ layers. The Al₂O₃ sacrificial layer was synthesized on AAO by ALD with TMA (Al2(CH3)6) and DI water as precursors. Then ZnO thin films were deposited on the surface of the sacrificial layer with precisely controlled thickness by ALD. The whole procedure is considered as one super cycle. More super cycles were applied to synthesize additional coaxial ZnO nanotubes. To eliminate the sacrificial layer, Precision Ion Polishing System (PIPS) was used to remove the top cover as shown in Figure 1. Therefore, the Al₂O₃ sacrificial layers were exposed for Sodium Hydroxide (NaOH) etching. After the sacrificial layers were removed by NaOH, ZnO nanotube gas sensors were synthesized as shown in Figure 1 (b).

To investigate the sensing performance of ZnO nanotube gas sensors to ethanol vapor, a gas sensor testing system was developed with a sealed reaction chamber and control system with stable temperature control and accurate concentration control. The sensing performance of one ZnO nanotube, two coaxial ZnO nanotubes, and three coaxial ZnO nanotubes to ethanol vapor were measured and analyzed under various ethanol vapor concentrations at different temperatures.

Liquid precursors for ALD processes are commonly delivered in an inert carrier gas from a bubbler (an ampoule with a dip tube) or a vapor draw ampoule (an ampoule with no dip tube: the gas in and gas out ports open directly into the ampoule headspace). Compared to some other delivery methods, the use of such ampoules for precursor delivery is relatively straight-forward. However, a potential drawback of using such ampoules is that evaporative cooling of the precursor can lead to a reduction in the amount of precursor delivered over time, i.e., cooling reduces the water vapor pressure and hence the amount of material entrained in the carrier gas. For example, the vapor pressure of water changes 5% per Kelvin at 20°C, while that of trimethylaluminum changes 6% per Kelvin at 20°C. For ALD processes that involve a chemical vapor deposition component, a decrease in the amount of water delivered can be detrimental to deposited film properties. To overcome this drawback, a device has been developed that controls the amount of precursor delivered based upon the precursor temperature. The concept is as follows. A temperature sensor is inserted into a thermowell on the ampoule to below the liquid level. Then, a target temperature is selected that would result in the desired amount of precursor being delivered based on the vapor pressure at this temperature. When flow is initiated through the ampoule, evaporative cooling of the precursor results in a decrease in precursor temperature which is measured with the temperature sensor. To compensate for this decrease, the amount of precursor delivered is increased in one of two ways: (1) the carrier gas flow rate is increased in a flow control mode or (2) the system pressure is decreased in a pressure control mode. In both modes, the amount of precursor is increased to compensate for the decrease in vapor pressure. To demonstrate the performance of this control device, the precursor partial pressure was measured with and without active control using custom-built optical gas sensors. The time-dependent partial pressure was converted to precursor flow rate assuming that the ratio of the precursor to carrier gas partial pressures equals the ratio of the respective flow rates. A five-element temperature sensor array was used to measure temperature. Flow control was demonstrated by controlling the set point of the carrier gas mass flow controller. Pressure control was demonstrated by controlling the set point of a downstream pressure controller. Water and hexane were utilized to investigate the impact of precursor vapor pressure on controller performance. In addition, both bubblers and vapor draw ampoules were investigated.

Plasma-enhanced ALD (PEALD) is a type of ALD, which utilizes ions and radicals in plasma as oxidants which have higher reactivity than conventional oxidants. Plasma can not only effectively eliminate ligand of precursors but also possibly reduce deposition temperature. While the system of commercial PEALD usually adopt capacitively coupled plasma (CCP) or inductively coupled plasma (ICP), it is usually complicated regardless of plasma generating method; such complexity of system may increase the system cost and hamper the process flexibility.

In this presentation, we show the design and manufacturing of fully functioning compact size PEALD station using ICP plasma chamber. Systematic mechanical design process was applied including quality function deployment (QFD), fluid/thermal analysis, and 3D modeling. In the system manufacturing process, gate valve and stop valve have been modified to increase the versatility of the process conditions as well as the possible deposition of many more materials. In addition, the electrical biasing table was added to the system to control the characteristics of the plasma such as the density and flux of ions, and therefore the crystallinity of the film. Furthermore, the biasing table can be further applied for selectively functionalizing the 2-dimensional materials, e.g., graphene, on metal-patterned substrate, and accordingly selectively depositing ALD films on them. As a result, PEALD station that we have designed and manufactured can provide economic options for researchers who want to explore PEALD processes for novel applications.

ALD deposition has enabled the semiconductor industry to deposit thin, highly conformal films. Foundational to delivering the consistent film quality is managing the quality of the parts within the chamber, especially the recycled parts sent offsite for cleaning. It is an under-recognized challenge to reliably produce “certifiably” clean chamber parts. The highly conformal, high step coverage nature of ALD process results in surface deposits on the parts that are tightly adhered. Additional considerations include: parts fragility, part size, high number of ALD film chemistries, and maintaining tight dimensional tolerances.

To address these challenges, recycled part cleaning processes and metrology validation methods have been developed. The processes are non-destructive and provide for an Atomically Clean Surface™ (ACS™) that meets the micro contamination requirements of the most stringent processes. The validation scope spans impurities identification (e.g. trace metals, particles, and trace impurities) and dimensional measurements (e.g. surface roughness, defects, flatness) via use of equipment such as ICP-MS and CMM. Handling systems to provide for employee ergonomics and to protect the parts will also be discussed with regard to the multi-wafer equipment.

In addition, there is smaller demand to refurbish of valves, fittings, and other parts external to the process chamber that may become coated or damaged. Processes have been adapted to address these components as well as ensuring critical dimensions and surface roughness of mating surfaces is maintained.

This paper will review the technologies developed to clean, rinse, dry, and validate that the recycled parts meet performance requirements for various ALD processes.

Continual device shrinkage and incorporation of new alloy materials has led to difficult thermal constraints for ALD in microelectronic applications. Process temperatures are now required to be to less than 350° C where sub-300° C is desired. PEALD has successfully been used in some applications, but has difficulty for 3D structures and aspect ratios greater than 50.

Therefore more reactive oxidant precursors for thermal ALD have been sought where water and ammonia no longer have sufficient reactivity at reduced temperatures. Examples include anhydrous Hydrogen Peroxide and Hydrazine, which exist as low vapor pressure liquids at room temperature. Delivery of a controlled and stable vapor from a liquid source is more challenging than delivering compressed gases with a mass flow controller. Due to the complex nature of chemical liquid-vapor equilibrium, it is difficult to accurately predict the vapor mass output of a liquid precursor for the wide range of ALD operating conditions when solely relying on theoretical models. The liquid phase is in constant equilibrium with its vapor phase counterpart. Slight changes in one, strongly affect the other.

Transformation of a liquid to a vapor depends on:

• Ampoule headspace pressure
• Carrier gas flow rate
• Liquid precursor temperature
• ALD precursor pulsing recipe
• Saturation efficiency of the carrier gas with the precursor vapor
• Intermolecular forces and binary interactions for multicomponent liquids
• System heat transfer
• Ampoule design

A liquid precursor ALD simulation manifold was built to re-create typical ALD conditions. The system is composed of mass flow and pressure flow controllers, pressure sensors, metering valves, vapor mass flow sensors, and several pneumatic valves. Automated test programs control the valves and sensors to simulate custom ALD process recipes. Data collected from the test fixture can quantify liquid precursor/ampoule vapor mass flow rates at different process pressures, carrier gas flow rates, precursor liquid temperatures, and different pulsing recipes. Mass delivery parameters can thus be optimized for process control and material utilization.

Anhydrous hydrogen peroxide gas can be precisely delivered in the range of 20-300 mg/min at moderate temperatures and pressures. Optimized delivery for several recipes will be presented as a function of pressure and liquid temperature. Conclusions pertinent to several low vapor pressure materials will be discussed.

Silicon dioxide has become an integral part of the microelectronics industry, due to the abundance of silicon, and the convenient electrical properties of the thermally formed native oxide. As a thin film, SiO₂ can also be used to tailor the mechanical properties of film stacks, act as an etch stop layer or prevent gas diffusion. In industrial manufacturing, the rate of thin film deposition is crucial. Additionally, a low temperature SiO₂ process is desirable to minimize any effect the deposition process will have on other device materials or the substrate itself.

In ALD, SiO₂ has many available chemistries, each with its own pros and cons. The chlorides and chlorosilanes require a deposition temperature of >300°C and can cause particles and chloride impurities in the films. Since these are unacceptable in industrial applications, many amine chemistries have become available. They can fulfil the growth rate, if not always the temperature requirements, required for industrial production.

In general, lower temperature processes is needed as interlayer diffusion must be avoided and the substrates are becoming more sensitive – e.g. organic or polymer substrates might require <200°C thermal budget. The simplest solution, PEALD [1], can however result in sacrificing the higher throughput possibilities that thermal batch ALD can provide. Plasma processes also exhibit limited applicability on higher aspect ratio samples, due to the short lifetime of radical species.

Achieving industrially feasible growth rates at low temperatures with thermal processes is challenging. For example, with a commonly used process such as bis(diethylamino) silane and ozone, the growth rate at 100°C is only < 0.1Å per cycle [2]. Most results in literature are also only on single wafers.

We present a reliable low temperature SiO₂ process with growth rates in excess of 1Å per cycle from 60 to 350°C, on single wafers (Figure 1) and in batches, using a precursor with 3 Si atoms and an amine group. While the film thickness increases linearly with the number of cycles, other properties will need to be a compromise between film quality and deposition temperature – a feature common to many ALD processes. In our example, the impurity content was lower at higher deposition temperatures.

The experiments were made with PICOSUN™ R-200 Adv, P-300B and P-1000 hot-wall ALD systems. Si wafers with native oxide layers were used as substrates. The ALD process, its scalability to larger batches, and various properties of the resulting thin films were evaluated in this study.

[1] Won et al. (2010), IEEE Electron Device Lett. 31 (2010)

[2] Hirvikorpi et al., Appl. Surf. Sci. 257 (2010)

In recent years, ALD and short pulse CVD have been integrated into more and more commercial microelectronic processes due to the superior thin film qualities these techniques can deliver. In addition to the slower processing time, many of these ALD/CVD processes use expensive liquid precursors, particularly for metal oxides, adding to the expense of moving from less precise PVD to higher performance ALD/CVD. When using liquid precursors in ALD, CVD and MOCVD, standard practice is to use an ampule or a direct liquid injection vaporization technique with a diverter valve. With an ampule or bubbler, an ALD valve is often used to provide a quick pulses of vapor concentration. However there are some limitations to this technique. The source liquid must be pre-heated, not ideal for thermally sensitive liquids; the delivered vapor concentration can be variable and difficult to control; and mass delivery rates are limited. With direct liquid injection vaporization techniques, a diverter valve is typically used to dump the liquid vapor while the vaporized concentration is stabilized; leading to wasted liquid precursor which can be upwards of $10,000 per liter. This paper presents an alternative to these two conventional techniques to provide a vapor from a liquid source for ALD or short pulse CVD. Using the new Performance Enhanced Turbo-Vaporizer, on/off pulsed data will be presented as an option to improve throughput, reduce processing costs and to widen process options.

The PE Turbo-Vaporizer was designed to minimize dead volume; minimize heat required for vaporization; and provide a wider process window, including maximized vapor concentrations. This direct liquid injection technique features a fast response time and stable concentration output while reducing the need for preventive maintenance. This new vaporizer design enables the possibility of turning on and off the vapor delivery fast enough for ALD processes, but without the need to use a diverter valve – reducing processing time, and reducing the amount of liquid used. Pulsed flow experiments were performed using several different liquid control methods and liquid precursors, including data taken with and without the use of a liquid flow controller. 25ms, 50ms, 150ms and 500ms data will be presented. Repeatability, complexity of the liquid control scheme, key control criteria, and advantages and disadvantages will be discussed.