All of us can see the next industrial revolution on the horizon. There are generational opportunities in hyperscaled computing, artificial intelligence (AI), autonomous driving, 5G/6G communications, integrated photonics, and even quantum computing and quantum sensing. Yet, as computing systems create new markets and move into these domains, the underlying hardware performance required to truly unleash these markets at scale is faced with constraints driven by fundamental physical limits. The opportunities of tomorrow are simply unachievable through anticipated hardware technologies and our old two dimensional (2D-) scaling playbook. A paradigm shift is underway to create new value propositions with semiconductor technologies and fabrication as the key drivers.

In response to this emerging crisis, SRC convened scientists from industry, academia, and government to outline a ten year R&D plan, the 2030 Decadal Plan for Semiconductors, as a scientific call to action.^[1,2] Therein, we have outlined the global drivers and constraints for future information and communication technology (ICT) systems, providing goal-driven yet creative solutions to seek measured progress and real world impact. The document outlines five “seismic shifts” that are shaping the future of chip technologies for smart sensing, memory and storage, communication, security, and energy efficient computing.

At the heart of these advances lies chipmaking which requires materials innovation, new precursors, new chip fabrication equipment, an understanding of material interfaces, the continued development of process knowledge, and skilled workers that embody these essential elements. The tribal knowledge and research activity in these areas must evolve beyond 2D-scaling to contribute to heterogeneous and 3D-monolithic integration that is informed by a rapidly-changing state of the art in packaging and assembly, design, and novel ICT architectures.

In this talk, I will share my perspective on the evolving global R&D landscape for semiconductors, illustrate the five “seismic shift” drivers behind future ICT chip advances, and give specific examples of ALD- and ALE-based research that map to today’s and tomorrow’s efforts. I hope to leave the audience with a clearer picture of the needs and opportunities we see for the years ahead. The “Roaring 20s” will not be boring if you are interested in materials-driven semiconductor advances and are a disciplined imaginator!

[1] 2030 Decadal Plan for Semiconductors, https://www.src.org/about/decadal-plan/

[2] Newly-Announced $3.4 Billion Plan Aims to Stimulate US Semiconductor R&D, https://www.allaboutcircuits.com/news/newly-announced-3point4-billion-plan-aims-stimulate-us-semiconductor-rd/

Ru metal has high work function of 4.7 eV, a low bulk resistivity (∼7 μΩ∙cm). It has good thermal and chemical stability on metal state. Furthermore, Ru forms a conductive oxide phase, RuO₂, with work function of ~5 eV and resistivity as low as ~ 30 μΩ∙cm, which prevents the formation of insulating interfacial layers in contact with oxides. As the size of semiconductor devices decreases, electronic applications require a nanometer-range of conformal thin film deposition in complex 3D structures. Among various thin film deposition technologies, atomic layer deposition (ALD) is considered to be promising in the development of nanometer device manufacturing technology due to its excellent step coverage and thickness control. [1]

However, continuous Ru film of under certain thickness (~6 nm) is difficult to form since they exhibit significant nucleation delay on various surfaces. The nucleation delay leads to increased process time and consumes a lot of precursor. For short nucleation delay time, high nucleation density is required.The high nucleation density forms continuous thin film with a smoother surface than obtained with low nucleation density. Therefore, the early nucleation density is important for metal film formation.

Using electric field/potential assisted atomic layer deposition (EA-ALD) can reduce the grain size and the critical thickness of the continuous film. When voltage is applied during precursor injection, it changes the surface potential of the substrate to promote chemical reactions. The electric field attracts the metal precursor molecules onto the substrate to improve the early nucleation density.As a result, the saturation of Ru layer density was increased on the substrate and a higher density thin film was deposited. It is possible to deposit continuous Ru ultrathin films with superior physical and electrical properties.

This work discusses the role and impact of the plasma on film conformality during plasma ALD, using the material systems SiO₂ and TiO₂ as industry-relevant case studies. Specifically, detailed insight into the impact of radical recombination and ion bombardment is provided. This is essential for predicting and further advancing the film conformality, especially for future applications with extremely high-aspect-ratio (AR) features.

First of all, the loss of reactive plasma radicals through surface recombination is often considered a major factor limiting film conformality during plasma ALD. To pinpoint what level of film conformality can still be achieved, we have developed a method to determine the surface recombination probability r of the radicals.¹ This method is based on the film penetration into high-AR trench structures. For plasma ALD of SiO₂ and TiO₂ very low values of r~10^-4 are determined, where r is observed to further decrease with temperature and pressure down to ~10^-5. Accordingly, film growth up to an AR as high as 200:1 or even 800:1 is achieved within reasonable cycle times, depending on temperature and pressure.² These results demonstrate that extremely challenging applications such as the coating of porous materials are feasible under specific process conditions.

Secondly, we demonstrate that ion bombardment can also have an important impact on film conformality during plasma ALD, even under mild plasma conditions with low-energy (<20 eV) ions. Specifically, it is observed that (low-energy) ions contribute to the film quality of SiO₂,³ can induce crystallization during plasma ALD of TiO₂,⁴ and can alter the growth per cycle by a factor of up to ~2 (for both SiO₂ and TiO₂). This can significantly affect the film conformality obtained on 3D nanostructures. Furthermore, we reveal that the magnitude of the influence of ions can be controlled by the ion energy dose, where a minimal effect is obtained when supplying a dose of <1 eV nm^-2 cycle^-1, or a strong effect when supplying a dose of >100 eV nm^-2 cycle^-1.^3,4

In conclusion, we provide key insights that can further advance plasma ALD, particularly to meet the conformality requirements in demanding future applications.

1. K. Arts et al., J. Phys. Chem. C 123, 27030 (2019).
2. K. Arts et al., Oxygen recombination probability data for plasma-assisted atomic layer deposition of SiO₂ and TiO₂ (to be published)
3. K. Arts et al., Appl. Phys. Lett. 117, 031602 (2020).
4. K. Arts et al., Impact of ions on film conformality and crystallinity during plasma-assisted atomic layer deposition of TiO₂ (to be published).

Electron-enhanced atomic layer deposition (EE-ALD) has been demonstrated as an effective method to rapidly nucleate and grow thin films at low temperatures [1,2]. During EE-ALD, electrons are used as a “reactant” in an ALD process. The role of the electrons is to remove surface ligands via electron stimulated desorption. In this work, EE-ALD was used to grow Ru thin films using dimethylbutadiene (DMBD) ruthenium tricarbonyl, or (DMBD)Ru(CO)₃, at low temperatures.

The sequential surface reactions for Ru EE-ALD are shown in Figure 1. During this reaction sequence, a hollow cathode plasma electron source (HC-PES) provided a high flux of electrons. The HC-PES is chemically robust and can operate at reactor pressures as large as 1 mTorr. Our HC-PES can emit electron currents ≥ 100 mA over surface areas of 50 cm². The electron energy employed for the Ru EE-ALD was 125 eV. Ru growth during EE-ALD was measured with in-situ spectroscopic ellipsometry. The Ru films were grown on silicon substrates with a native oxide.

Ru film growth was demonstrated at both 100°C and 160°C. Film thickness measured by ellipsometry showed that the Ru films nucleate and grow from the very first EE-ALD cycle. Linear film growth was observed over 600 cycles with growth rates of 0.17 Å/cycle and 0.23 Å/cycle at 100°C and 160°C, respectively. Spectroscopic ellipsometry measurements for 600 cycles of Ru EE-ALD at 160°C are shown in Figure 2. Self-limiting growth behavior was observed with respect to both (DMBD)Ru(CO)₃ and electron exposures. XPS was used to determine the purity of the Ru films. Oxygen was observed to be as low as 0.3 at%. Carbon was also low but could not be quantified because of the overlap of the C and Ru XPS peaks.

[1] Z.C. Sobell, A.S. Cavanagh and S.M. George, “Growth of Cobalt Films at Room Temperature Using Sequential Exposures of Cobalt Tricarbonyl Nitrosyl and Low Energy Electrons”, J. Vac. Sci. Technol. A37, 060906 (2019).

Atomic layer deposition (ALD) and atomic layer etching (ALE) are defined by two sequential, self-limiting surface reactions. ALE is based on surface modification by the first reactant. The second reactant then leads to volatile release of the modified surface. There are two types of ALE:plasma and thermal. Plasma ALE methods employ energetic ion or neutral species to release the modified material anisotropically using sputtering. Thermal ALE processes utilize gas species to release the modified material isotropically using thermal reactions [1]. Thermal ALE can be viewed as the “reverse of ALD”.

One important mechanism for thermal ALE is surface modification by fluorination and volatile release by ligand-exchange reactions. Fluorination is thermochemically favorable because metal fluorides are more stable than metal oxides. The metal fluorides can then be volatilized when a ligand from an incoming metal precursor exchanges with F from the metal fluoride. This fluorination and ligand-exchange mechanism is applicable to the thermal ALE of many metal oxides such as Al₂O₃. Conversion reactions are able to extend the range of materials that can be etched using thermal ALE. Conversion reactions occur when an incoming metal precursor reacts with the surface of an initial material and converts the surface layer into another material. This new material may have etching pathways that were not accessible to the initial material.

Another thermal ALE mechanism is based on oxidation and ligand-addition reactions. This mechanism is particularly important for etching elemental metals. Oxidation first changes the oxidation state of the elemental metal. Subsequently, ligands can add to the oxidized metal center and form stable and volatile metal compounds. An example of oxidation and ligand-addition is thermal Ni ALE using SO₂Cl₂ and P(CH₃)₃. Chlorination by SO₂Cl₂ changes the oxidation state and then P(CH₃)₃ adds to the oxidized Ni center and forms volatile NiCl₂(P(CH₃)₃)₂.

Different mechanisms for thermal ALE can lead to selectivity in etching between various materials. Etching results if the sequential, self-limiting surface reactions produce stable and volatile compounds. Selectivity between materials occurs if the etching products for one material are not stable or volatile. A good example is the selectivity between Al₂O₃ and ZrO₂ using HF and Al(CH₃)₃ as the reactants. ZrO₂ is not etched and can be used to define a monolayer etch stop for the HF and Al(CH₃)₃ reactants.

[1] S.M. George, “Mechanisms of Thermal Atomic Layer Etching”, Acc. Chem. Res. 53, 1151 (2020).

Recent years saw a rise of solvent-free methods for synthesis of metal organic frameworks with the aim of accessing new application fields and facile processing. Since vapor deposition excels at these key points we employed a two-step chemical vapor deposition process, that allows for the delivery of high-quality, homogeneous thin films of zeolitic imidazole framework 8 (ZIF-8) from ZnO¹.

First, an ultrathin ZnO seed layer is deposited via plasma-enhanced atomic layer deposition (PE-ALD). Acting on the substrate temperature, ranging from room temperature to 200°C, the preferred crystal orientation can be switched from (100) to (002). ZIF-8 thin films are subsequently grown by subjecting the ZnO-layer to a 2-methyl imidazole vapor. To our knowledge, this is the first time that PE-ALD has been employed for the growth of ZIF-8.

To gain better control over the still novel deposition technique, the impact of crystal orientation and thickness of the ZnO precursor onto the resulting ZIF-8 thin films was investigated. The results show that ZIF-8 was successfully synthesized. Furthermore, X-ray diffraction studies reveal a powder-like structure together with a strong (100) orientation of ZIF-8 crystals. ZIF-8 coverage on the substrate increases for thicker ZnO layers with ZIF-8 particles exhibiting average thicknesses as high as (93 ± 9) nm for ZIF-8 from 3 nm ZnO and (192 ± 11) nm for ZIF-8 from 10 nm ZnO. The thickness increase during conversion rises to about 1600% as a function of ZnO thickness. ZnO orientation weakly influences the thickness increase during conversion and, via differing densities, also the crystallinity of the resulting ZIF-8. Our results provide vital knowledge about the link between deposition parameters of ZnO and properties of the resulting ZIF-8 thin films, namely coverage, thickness, roughness and orientation, thus making it possible to tailor them towards specific applications.

(1) Stassen, I.; Styles, M.; Grenci, G.; Van Gorp, H.; Vanderlinden, W.; De Feyter, S.; Falcaro, P.; De Vos, D.; Vereecken, P.; Ameloot, R. Chemical Vapour Deposition of Zeolitic Imidazolate Framework Thin Films. Nat. Mater.2016, 15 (3), 304–310. https://doi.org/10.1038/nmat4509.

Thermal metal ALE is challenging because the oxidation state of the volatile metal etch products are usually different than the zero oxidation state of the elemental metal.In this work, cobalt ALE was developed by first oxidizing the cobalt via chlorination using SO₂Cl₂. Then the Co was etched by the binding of trimethyl phosphine (PMe₃) ligands that can volatilize the cobalt chloride.Thermal Co ALE was achieved at 150 °C by properly balancing the SO₂Cl₂ exposures that chlorinate cobalt and the PMe₃ exposures that remove the cobalt chloride (Figure 1).

The logic of this approach is based on the Covalent Bond Classification (CBC) method.The CBC method indicates that most organo-metallic complexes obey the 18 or 16 electron rule. We have shown previously that this method works for the thermal ALE of nickel. In the CBC method, a chloride ligand donates one electron to the metal center and is referred to as an X ligand. PMe₃ is a two-electron donor and is designated an L ligand. The nickel ALE process was accomplished using these sequential steps: chlorination to NiCl₂ and ligand addition to form NiX₂L₂, or NiCl₂P(Me₃)₂, resulting in an etch process. For cobalt, the most likely compounds to be created using chloride and trimethyl phosphine ligands are CoX₂L₄, CoX₂L₂, or CoX₃L₃. These compounds have cobalt in the +2 or +3 oxidation states.

Co ALE process was demonstrated with SO₂Cl₂ and PMe₃ exposures at 150 °C using in situ quartz crystal microbalance (QCM) measurements. The QCM experiments indicated that cobalt very readily chlorinates when exposed to SO₂Cl₂. A typical mass gain with one exposure of SO₂Cl₂ on cobalt can be up to 1400 ng/cm².In contrast, an identical exposure of SO₂Cl₂ on nickel results in a mass gain of only 100-150 ng/cm². The PMe₃ step was also different on cobalt when compared with nickel. For nickel ALE, one exposure of PMe₃ removes all available NiCl₂. An identical exposure of PMe₃ on CoCl₂ only removes ~20% of the available CoCl₂.Consequently, multiple PMe₃ exposures are required to remove the CoCl₂.

Co ALE requires a careful balance between the SO₂Cl₂ and PMe₃ exposures.Too much cobalt chlorination with SO₂Cl₂ will lead to difficulties removing all the cobalt chloride with PMe₃.Cobalt chloride that is not removed during one ALE cycle will build up and a cobalt chloride top layer will grow with successive ALE cycles.Optimizing Co ALE was based on determining the number of PMe₃ exposures required to remove the top cobalt chloride layer.Using the optimum exposures of SO₂Cl₂ and PMe₃ results in high etch rates for cobalt ALE of 3-5 Å/cycle at 150 °C (Figure 2).