Search

Despite tremendous interests in ferroelectric field-effect transistors (FEFETs) for embedded, data-centric applications, the fundamental trade-offs between memory window (MW) and write voltage to optimize performance remains poorly understood. To that end, we fabricated ferroelectric (FE) $ZrO_{2}$ based, p-type FEFETs and studied the impacts of FE and the interfacial oxide layer (IL) thicknesses ( $t_{FE}$ and $t_{IL}$ , respectively) on device performance. We observe that a decrease of $t_{FE}$ and $t_{IL}$ reduces not only write voltages for erasing and programming, but also the memory window. A quantitative analysis of these results offers the following insights and guidelines for FEFET design: to decrease write voltages, all of $t_{FE}$ , $t_{IL}$ and coercive field of FE needs to decrease, and to compensate for the subsequent decrease in MW, the polarization of the FE needs to be increased - notwithstanding the fact that the reliability implications of the magnitude of FE polarization still need to be understood.

We demonstrate a novel process for building a Resistive RAM (ReRAM) stack which reduces the forming voltage ( $\text{V}_{\textit {form}}$ ) and increases the switching resistance, both characteristics that are important ingredients for the use of ReRAM in scalable analog compute for AI. Utilizing this process, we explore analog switching characteristics above 100k $\Omega $ and demonstrate 4-bit programming at Rmax $=1\text{M}\Omega $ . Utilizing the same writing characteristics, CIFAR-10 inference simulation shows 90% accuracy, comparable to the full precision model accuracy.

There is a pressing need to improve the efficiency of drug development, and nowhere is that need more clear than in the case of neglected diseases like malaria. The peculiarities of pyrimidine metabolism in Plasmodium species make inhibition of dihydroorotate dehydrogenase (DHODH) an attractive target for antimalarial drug design. By applying a pair of complementary quantitative structure–activity relationships derived for inhibition of a truncated, soluble form of the enzyme from Plasmodium falciparum (s-PfDHODH) to data from a large-scale phenotypic screen against cultured parasites, we were able to identify a class of antimalarial leads that inhibit the enzyme and abolish parasite growth in blood culture. Novel analogs extending that class were designed and synthesized with a goal of improving potency as well as the general pharmacokinetic and toxicological profiles. Their synthesis also represented an opportunity to prospectively validate our in silico property predictions. The seven analogs synthesized exhibited physicochemical properties in good agreement with prediction, and five of them were more active against P. falciparum growing in blood culture than any of the compounds in the published lead series. The particular analogs prepared did not inhibit s-PfDHODH in vitro, but advanced biological assays indicated that other examples from the class did inhibit intact PfDHODH bound to the mitochondrial membrane. The new analogs, however, killed the parasites by acting through some other, unidentified mechanism 24–48 h before PfDHODH inhibition would be expected to do so.

Recently, multilevel Resistive RAM (ReRAM) stacks have been demonstrated to have promising applications related to inference and learning in artificial intelligence. However, extensive studies on energy efficiency, repeatability, and retention during multilevel operations have not been undertaken previously. It is, therefore, required to investigate the device characteristics such as compliance current (CC) dependence, forming voltage, memory window, symmetry, and switching energy. Furthermore, understanding the impact of dielectric composition and/or electrode materials is necessary to optimize the device performance, such as switching characteristics, endurance, and energy efficiency. In this work, we have investigated a two-terminal ReRAM device, 10nm Ti/50nm TiN/7nm HfO2/1nm Al2O3/PVD Ti/TiN (Fig. 1), with bilayer switching dielectrics of HfO2 and Al2O3. In this case, the embedded Al2O3 layer was placed between the HfO2 layer and the top electrode (TE). We have evaluated the gradual resistance change capability with varying CC. The I-V characteristics of the ReRAM device, measured by a dc double sweep, is shown in Fig. 2. The device switched at a compliance current of 50nA with a forming voltage of 1.85 volts to set to low resistance state (LRS) from the high resistance state (HRS). The subsequent reset from LRS to HRS was obtained during the negative voltage ramp showing a bipolar switching operation (Fig. 2). With the variation of compliance current (50 nA to 70 nA), the LRS was gradually changed indicating multilevel LRS. It is known that by increasing the compliance current with a gradual set process, multiple conducting filaments (CFs) or thicker CFs can form setting the different LRS values and the annihilation of the CFs during the reset process can also moderately vary the HRS values. The resistance ratio between the HRS and LRS was observed to be more than 1000. The devices were subjected to 100 cycles of SET and RESET operations at different CCs without any failures. When the embedded Al2O3 layer was placed between the bottom electrode (BE) and the HfO2 layer, the switching compliance current was increased to 10 mA with a reduced forming voltage of 1.6V. Multilevel LRS states were observed for this device when the CC was varied from 10 mA to 12 mA for 100 cycles at each CCs. To further understand the behavior, TEM depth profile measurements were carried out, and it was observed that Al2O3 diffused from the peak at the as-deposited location into the HfO2 layer in both cases. While this improved the uniformity in switching behavior, the low switching energy consumption in the case of an embedded Al2O3 layer in between the HfO2 layer and the TE layer remains to be investigated. Replacing the PVD Ti layer (Fig. 1) with ALD Ti in the TE when an Al2O3 layer was between the bottom electrode (BE) and the HfO2 layer reduced the switching compliance current from 10 mA to 100 nA. This suggests that the metal quality of the TE can impact the switching behavior of the multilevel ReRAMs. Figure 1

‘Deep learning’ is causing rapid technological changes in many fields of science, and conjectures about its potential for transforming everyone's work and lives is a matter of great debate. Unfortunately, it is all too easy to apply it as a ‘black box’ tool with little consideration of its potential limitations, especially when the data it is being applied to is less than perfect. In this Perspective, I try to put deep learning into a broader mechanistic and historical context by showing how it relates to older forms of artificial intelligence; by providing a general explanation of how it operates; and by exploring some of the challenges involved in its implementation. Examples wherein it has been applied to pest management problems are provided to illustrate how the technology works and the challenges deep learning faces. At least in the near term, its biggest impact on agrochemical development seems likely to come in automating the tedious work involved in assessing agrochemical efficacy, but getting there will require major investments in building large, well‐curated data sets to work from and in providing the expertise required to assess the resulting model predictions in real‐world scenarios. Deep learning may also come to complement the machine learning methodologies already available for use in pesticide discovery and development, but it seems unlikely to supplant them. © 2020 The Authors. Pest Management Science published by John Wiley & Sons Ltd on behalf of Society of Chemical Industry., To put deep learning into a broader context, this Perspective relates it to older machine learning tools; explains the basics of how it works; and explores some of its limitations.

The recent discovery of ferroelectric behavior in doped hafnia-based dielectrics, attributed to a non-centrosymmetric orthorhombic phase, has potential for use in attractive applications such as negative differential capacitance field-effect-transistors (NCFET) and ferroelectric random access memory devices (FeRAM). Alloying with similar oxides like ZrO2, doping with specific elements such as Si, novel processing methods, encapsulation and annealing schemes are also some of the techniques that are being explored to target structural modifications and stabilization of the non-centrosymmetric phase. In this study, we utilized synchrotron-based x-ray diffraction in the grazing incidence in plane geometry (GIIXRD) to determine the crystalline phases in hafnia-zirconia (HZO) compositional alloys deposited by atomic layer deposition (ALD). Here we compare and contrast the structural phases and ferroelectric properties of mechanically confined HZO films in metal-insulator-metal (MIM) and metal-insulator-semiconductor (MIS) structures. Both MIM and MIS structures reveals a host of reflections due to non-monoclinic phases in the d-spacing region between 1.75A to 4A. The non-monoclinic phases are believed to consist of tetragonal and orthorhombic phases. Compared to the MIS structures a suppression of the monoclinic phase in MIM structures with 50% zirconia or less was observed. The correlation of the electrical properties with the structural analysis obtained by GIIXRD highlights the importance of understanding the effects of the underlying substrate (metal vs. Si) for different target applications.

Ferroelectric and antiferroelectric Hf/Zr-based oxide films have recently gained interest for memory and AI applications due to their promise of low power and CMOS compatibility. As Hf/Zr-based oxides are not ‘new’ materials, this paper will start with an overview of past learning on this material system. Recent results on fundamental understanding of the mechanism of ferroelectric and antiferroelectric switching will be presented. Challenges in implementation of this material system in high volume manufacturing will be discussed.

Metal-insulator-metal (MIM) stacks, though simple in design, are the backbone device for resistive random access memories (ReRAM) and, as such, play a vital role in emerging memory technologies. In this work, we characterize the impact of Ar, Ar/N2, and BCl3/Cl2 plasma processes on the physical properties of the TiN bottom electrode (BE) surface. The BCl3/Cl2 process increases roughness by 40%. Ar and Ar/N2 processes both decrease PVD TiN roughness by 50%. This reduction propagates through the entire MIM stack. X-ray photoelectron spectroscopy (XPS) indicates that both plasma processes alter the Ti-ON concentration on the TiN surface. The impact of BE surface roughness and composition on MIM electrical properties is currently under investigation.

Direct, atomic-scale visualization of polarization switching in a functional, polycrystalline, binary oxide via insitu high-resolution transmission electron microscopy (HRTEM) biasing is reported for the first time. Antiferroelectric (AFE) ZrO 2 was used as the model system, which is important for commercial DRAMs and as emerging NVMs (through work-function engineering). We observed (1) clear shifting and coalescing of domains within a single grain, and (2) dramatic changes of the atomic arrangements and crystalline phases-both at voltages above the critical voltage measured for AFE switching. Similar synergistic in-situ structural-electrical characterization can pave the way to understand and engineer microscopic mechanisms for retention, fatigue, variability, sub-coercive switching and analog states in ferroelectric and AFE-based memory devices.

In this work, we presented the Hf-Zr system as a case study of materials that were evaluated in the past and are now receiving renewed interest due to its potential use in artificial intelligence (AI). A historical overview of Hf-Zr research and its promising new applications for emerging memories will be shared.

Cannabis is a plant with a long history of human pharmacological use, both for recreational purposes and as a medicinal remedy. Many potential modern medical applications for cannabis have been proposed and are currently under investigation. However, its rich chemical content implies many possible physiological actions. As the use of medicinal cannabis has gained significant attention over the past few years, it is very important to understand phytocannabinoid dispositions within the human body, and especially their metabolic pathways. Even though the complex metabolism of phytocannabinoids poses many challenges, a more thorough understanding generates many opportunities, especially regarding possible drug-drug interactions (DDIs). Within this context, computer simulations are most commonly used for predicting substrates and inhibitors of metabolic enzymes. These predictions can assist to identify metabolic pathways by understanding individual CYP isoform specificities to a given molecule, which can help to predict potential enzyme inhibitions and DDIs. The reported in vivo Phase I and Phase II metabolisms of various phytocannabinoids are herein reviewed, accompanied by a parallel in silico analysis of their predicted metabolism, highlighting the clinical importance of such understanding in terms of DDIs and clinical outcomes.

Pesticides must be effective to be commercially viable but they must also be reasonably safe for those who manufacture them, apply them, or consume the food they are used to produce. Animal testing is key to ensuring safety, but it comes late in the agrochemical development process, is expensive, and requires relatively large amounts of material. Surrogate assays used as in vitro models require less material and shift identification of potential mammalian toxicity back to earlier stages in development. Modern in silico methods are cost-effective complements to such in vitro models that make it possible to predict mammalian metabolism, toxicity and exposure for a pesticide, crop residue or other metabolite before it has been synthesized. Their broader use could substantially reduce the amount of time and effort wasted in pesticide development. This contribution reviews the kind of in silico models that are currently available for vetting ideas about what to synthesize and how to focus development efforts; the limitations of those models; and the practical considerations that have slowed development in the area. Detailed discussions are provided of how bacterial mutagenicity, human cytochrome P450 (CYP) metabolism, and bioavailability in humans and rats can be predicted. © 2018 The Authors. Pest Management Science published by John Wiley & Sons Ltd on behalf of Society of Chemical Industry.

We fabricate, characterize, and establish the critical design criteria of Hf0.5Zr0.5O2 (HZO)-based ferroelectric field effect transistor (FeFET) for nonvolatile memory application. We quantify ${V}_{\textsf {TH}}$ shift from electron (hole) trapping in the vicinity of ferroelectric (FE)/interlayer (IL) interface, induced by erase (program) pulse, and ${V}_{\textsf {TH}}$ shift from polarization switching to determine true memory window (MW). The devices exhibit extrapolated retention up to 10 years at 85 °C and endurance up to $5\times 10^{6}$ cycles initiated by the IL breakdown. Endurance up to 1012 cycles of partial polarization switching is shown in metal–FE–metal capacitor, in the absence of IL. A comprehensive metal–FE–insulator–semiconductor FeFET model is developed to quantify the electric field distribution in the gate-stack, and an IL design guideline is established to markedly enhance MW, retention characteristics, and cycling endurance.

This paper investigates the p-Ge/Al2O3/ZrO2/TiN gate stacks that were subjected to different slot plane antenna oxidation (SPAO) conditions: 1) prior to any high-k atomic layer deposition (ALD); 2) in between Al2O3 and ZrO2ALD layers; and 3) after the ALD of high- ${k}$ layers. The carrier transport mechanisms as a function of temperature on these samples were observed to be different. The SPAO treatment effectively removes the trap centers in ZrO2 and Al2O3 layers depending on the SPAO treatments. Fowler-Nordheim tunneling seems to be the dominant transport mechanism at high field range when SPAO was performed after the high- ${k}$ layers were deposited. On the other hand, Poole-Frenkel emission and hopping conduction mechanism are dominant for other two samples in both gate electron injection and substrate electron injection (SEI) modes. The trap center ( $ {\phi }_{\text{t1}} = {0.13}$ eV) remains in the ZrO2 when ZrO2 is not subjected to the SPAO. Trap energy level $ {\phi }_{\text{t1}}$ (0.13 eV) was removed and $ {\phi }_{\text{t2}}$ (0.27 eV) still exists when only Al2O3 oxide layer was exposed to SPAO. Hopping distance, ${a}$ (about 0.3 nm) is extracted by hopping conduction model. After time dependent dielectric breakdown measurement, it was found that if GeO2 exists at the interface then it can be degraded easily by SEI stress. On the other hand, if SPAO was processed in between the high- ${k}$ layers the formation of a GeOx layer enhances the interface quality and provides better immunity to degradation under stress.

Deep Learning (DL) has gained wide adoption and different DL models have been deployed for an expanding number of applications. It is being used both for inference at the edge and for training in datacenters. Applications include image recognition, video analytics, pattern recognition in networking traffic, and many others. Different applications rely on different neural network models, and it has proven difficult to predict resource requirements for different models and applications. This leads to the nonsystematic and suboptimal selection of computational resources for DL applications resulting in overpaying for underutilized infrastructure or, even worse, the deployment of models on underpowered hardware and missed service level objectives. In this paper we present the DL Cookbook, a toolset that a) helps with benchmarking models on different hardware, b) guides the use of DL and c) provides reference designs. Automated benchmarking collects performance data for different DL workloads (training and inference with different models) on various hardware and software configurations. A web-based tool guides a choice of optimal hardware and software configuration via analysis of collected performance data and applying performance models. And finally, it offers reference hardware/software stacks for particular classes of deep learning workloads. This way the DL Cookbook helps both customers and hardware vendors match optimal DL models to the available hardware and vice versa, in case of acquisition, specify required hardware to models in question. Finally, DL Cookbook helps with reproducibility of results.

Knowing the physicochemical and general biochemical properties of a compound is critical to understanding how it behaves in different biological environments and to anticipating what is likely to happen in situations where that behavior cannot be measured directly. Quantitative structure-property relationship (QSPR) models provide a way to predict those properties even before a compound has been synthesized simply by knowing what its structure would be. This chapter describes a general workflow for compiling the data upon which a useful QSPR model is built, curating it, evaluating that model's performance, and then analyzing the predictive errors with an eye toward identifying systematic errors in the input data. The focus here is on models for the absorption, distribution, metabolism, and excretion (ADME) properties of drugs and toxins, but the considerations explored are general and applicable to any QSPR.

Knowing the physicochemical and general biochemical properties of a compound is critical to understanding how it behaves in different biological environments and to anticipating what is likely to happen in situations where that behavior cannot be measured directly. Quantitative structure-property relationship (QSPR) models provide a way to predict those properties even before a compound has been synthesized simply by knowing what its structure would be. This chapter describes a general workflow for compiling the data upon which a useful QSPR model is built, curating it, evaluating that model's performance, and then analyzing the predictive errors with an eye toward identifying systematic errors in the input data. The focus here is on models for the absorption, distribution, metabolism, and excretion (ADME) properties of drugs and toxins, but the considerations explored are general and applicable to any QSPR.

As device feature size continues to scale, new device architecture and dielectric materials with engineered properties are becoming essential in overcoming the issues in scaled CMOS devices such as short channel effects. Some of the recent innovations device makers are exploring for better electrostatic control and improved performance are the introduction of new device architecture such as gate all around devices (nanowires/nanosheets), and channel engineering such as high mobility channel materials (Ge and III-V). Ge is particularly attractive for use as a channel material for P-type MOSFETs due to its high hole mobility and Si VLSI high volume manufacturing (HVM) compatibility. However, there are couple significant remaining challenges to integrating Ge into 300mm Si VLSI HVM which are a stable gate-stack formation and integration of Ge on Si (1-5). In this talk, advances in Ge gate stack engineering will be discussed. Using Si compatible technologies we have demonstrated scaled EOT by engineering both the interface and high-k formation. Reference Swaminathan, Y. Oshima, M. Kelly, P.C. McIntyre, Appl. Phys. Lett., 95, 032907 (2009). Zhang, M. Gunji, S. Thombare, P.C. McIntyre, IEEE Electr. Device. L., 34, 736 (2013). H Lee, T. Nishimura, C. Lu, S. Kabuyanagi, A. Toriumi, IEDM, 32.5.1 (2014). S. Goley, M. K. Hudait, Materials, 2301 (2014). Xie, S. Deng, M. Shaekers, D. Lin, M. Caymax, A. Delabie, X-P Qu, Y-L. Jiang, D. Dedytsche, C. Detavernier, Semicond. Sci. Technol. 074012 (2012)

Neuromorphic computing represents a potential paradigm shift from conventional von Neumann computing architecture and shows promise for achieving massive parallelism and power efficiency for such data-centric tasks as image recognition and language processing. Based on the concept of synaptic plasticity, human-like machine learning can be potentially realized by use of arrays of electronic synapses that function in an analogous manner to biological neurons. One of the key features of this type of computing is the ability to control synaptic weights in an analog-like fashion for use in both inference and training applications. A number of existing device technologies in non-volatile memory systems exhibit attractive characteristics for such synaptic devices.[1,2] In particular, resistive switching devices (resistive random-access memory or ReRAM) can change and store their conductance value (G) in response to electrical stimuli making them potentially enabling for deep learning applications involving synaptic weights. For ReRAM devices, HfO2-based thin films can be utilized for filamentary oxide ReRAM and are an attractive option due to their fab-friendly processing and current implementation in high-volume manufacturing. In this study, we evaluated atomic layer deposition (ALD) for the growth of HfO2 for integration in both front-end-of-line (FEOL) and back-end-of-line (BEOL) compatible test structures on 300 mm wafers in order to optimize electrical performance for use as synaptic device elements in neuromorphic architectures. The effect of oxidant in the ALD process was evaluated and it was shown that H2O outperformed O3 in terms of better uniformity and lower forming voltage. By utilizing a hydrogen-based plasma either after the deposition or inserted as an intermediate step during deposition we were able to further decrease forming voltage for a fixed dielectric thickness. Reducing deposition temperature to 200°C in conjunction with the hydrogen-based plasma treatment offered an additional tuning knob to further reduce forming voltage. Stable high-resistance switching (> 100 kΩ) with analog behavior in scaled BEOL devices was also obtained using this optimized HfO2-based ReRAM. Additionally, a tight distribution of forming voltage was obtained ensuring that 99.9999% devices in a 14 nm ReRAM module can be formed below the targeted voltage. References Kuzum et al., Nanotechnology, 24, 382001 (2013) W. Burr et al., Advanced in Physics:X, 2, 89 (2016)

In this Letter, we report the percentage of the ferroelectric phase in a 10-nm-thick Hf0.5Zr0.5O2 (HZO) film deposited in a metal-insulator-metal stack by atomic layer deposition. The ferroelectric behavior was confirmed by polarization measurements and piezoresponse force microscopy. Ferroelectric behavior in this material has been attributed most likely to the formation of the polar non-centrosymmetric orthorhombic phase [Muller et al., Appl. Phys. Lett. 99, 102903 (2011)], which is difficult to distinguish from the tetragonal phase in x-ray diffraction due to peak overlap. Using a model for each of the crystal phases of hafnia-zirconia, the phase percentages were estimated using a Rietveld refinement method applied to grazing incidence x-ray diffraction data and a linear combination fit analysis procedure [McBriarty et al., Phys. Status Solidi 257, 1900285 (2020)] applied to grazing incidence extended x-ray absorption fine structure data. Using these methods, it was found that the tetragonal (P42/nmc) phase is the most prevalent at 48–60% followed by the polar non-centrosymmetric orthorhombic (Pca21) phase at 35%–40% with the remainder consisting of the monoclinic (P21/c) phase. Understanding the details of the effect of the phase structure on the electrical properties of these materials is extremely important for device engineering of HZO for logic and emerging nonvolatile memory applications.

One of the key challenges for Ge gate stack is to achieve an improved interface quality between Ge and the high-K dielectric. In order to overcome this limitation, various interface treatments like ozone pre-gate treatment and ozone ambient annealing (1), nitridation of Ge surface (2),formation of a GeO2 layer between Ge and high-K dielectric using ECR plasma technique (3) and formation of an ultra-thin SiO2/GeO2bilayer have been tried. It has recently been reported that slot plane antenna(SPA) microwave plasma treatment helps to improve the EOT and enhances the oxide quality by reducing impurities and leakage current when the dielectric is exposed during or after deposition (4). Moreover SPA oxidation possesses the unique property of having low electron temperature, which makes it a very low damage process compared to other techniques,either inductively coupled plasma(ICP) and electron cyclotron resonance(ECR) plasma (5). In this study we investigated the impact of SPA plasma oxidation (SPAO) treatment in improving the interface quality when MOS capacitors were fabricated on 300mm p-type Ge substrates. SPAO plasma treatments were performed at different steps in the high-k deposition process. The substrate was initially subjected to a dry chemical oxide removal (COR) process. Subsequently, three different sets of samples were prepared with the exposure of SPAO plasma at following locations (Fig.1): CASE-1: SPAO plasma treatment after Al2O3(1nm)/ZrO2(3.5nm) deposition; CASE-2: SPAO plasma treatment was carried out in between Al2O3(1nm) and ZrO2(3.5nm) deposition; CASE-3: the Ge substrate was exposed to SPAO plasma treatment after COR and prior to Al2O3(1nm)/ ZrO2(3.5nm) layer deposition. Al2O3(1nm) and ZrO2(3.5nm) layers were deposited by atomic layer deposition. Capacitance-voltage at multi-frequencies,current-voltage and conductance measurements were performed at room temperature. Parameters like flatband voltage (Vfb), equivalent oxide thickness(EOT), interface state density(Dit),and leakage current were estimated accurately after quantum mechanical corrections and eliminating the series resistance effect. As shown in Fig. 2 lowest fast interface state density is achieved when SPAO plasma treatment was performed after Al2O3 and prior to ZrO2 deposition (CASE-2). A comparable Dit is also observed when plasma exposure was done after the entire gate stack deposition (CASE-1) (Fig.2). From different experimental results significant degradation to both the dielectric quality and interface was observed when SPAO is performed just after COR (CASE-3). References: 1) Zhao Mei et al, Journal of Semiconductors, Volume 34, No. 3, 2013. 2) A. Dimoulas et al, Appl. Phys. Lett., volume 86, no. 3, 2005. 3) Yukio Fukuda et al, IEEE transaction on electron devices, volume 57, No. 1, 2010. 4) M.N. Bhuyian et al, ECS Journal of Solid State Sci and Techn, vol. 3(5), N83, 2014. 5) C. Tian et al, J. Vac. Sci. Technol. A 24, 1421, 2006. Figure 1

For exploring the prospect of higher-k dielectric phase engineering on a high mobility substrate, films of Hf1-xZrxO2 with varying x-values (0 ≤ x ≤ 1) were deposited on Al2O3 passivated Ge substrates using atomic layer deposition (ALD) with a cyclic deposit-anneal-deposit-anneal (DADA) scheme. The evolution of monoclinic to higher-k tetragonal structure with increasing ZrO2 concentration was probed by grazing incident x-ray diffraction and partial reciprocal space maps using the highly brilliant synchrotron x-ray source at the Cornell High Energy Synchrotron Source (CHESS). A primarily amorphous/nano-crystalline matrix of the as-deposited films changed to randomly aligned grains of nanocry stalline MO2 (M=Hf, Zr) after post deposition annealing at 800 °C for 200 seconds. In contrast, the DADA films annealed for same thermal budget showed high degree of preferred orientation along certain crystallographic directions. With increasing ZrO2 content, the structure of the films changed from a monoclinic to a tetragonal phase. A lower amount of ZrO2 (x = 0.33) was required for stabilizing the tetragonal phase in films grown on Al2O3 passivated Ge substrate as compared to similar films grown on a Si substrate via the same DADA process (x ≥ 0.50).

Miniaturization in Integrated Circuit fabrication has scaled to atomic dimensions in recent times. With critical device dimensions moving into the range of ~10nm, atomic layer processing control is becoming essential. Atomic Layer Deposition (ALD) uses separated film components at (or near) self-limiting conditions for atomic level uniformity and thickness control. The unique nature of ALD processing has opened the opportunity for a deposition technique that presents deposition components in a spatially separated method compared to sequential component exposure of traditional systems. Spatial ALD has differentiating characteristics that can benefit IC Fabrication at current fabrication dimensions. Historically, deposition systems for IC fabrication use batch furnace or single wafer configurations. The economic benefit of batch furnace processing has been challenged by the enhanced capabilities of single wafer processing as IC scaling emphasized critical film properties and uniformity. ALD processing has rejuvenated batch furnace benefits since separate exposures of deposition components provides excellent uniformity from self-limiting deposition layers, and lower particle defectivity since the components don’t mix. However, the sequential processing of individual components has strained the efficiencies of batch furnace and single wafer ALD processing. Spatial ALD moves substrates (wafers) through each deposition component separately in space, compared to the time sequential processing of batch furnaces and single wafer processing. There are several advantages to Spatial ALD processing, and several unique capabilities that can apply to IC fabrication. By depositing through movement, the deposition components can be maintained through a steady state, which is advantageous for plasma processing and for limiting peripheral equipment film accumulation. By its nature, the deposition rate of Spatial ALD is controlled by the speed of substrates passing through the deposition zones, which can provide unique controls for deposition. Deposition chemistry and process conditions can be designed to take advantage of Spatial ALD properties for unique benefits in IC fabrication. The presentation will explore some fundamental aspects of Spatial ALD processing, the dependency and opportunities of relevant chemistries, and potential applications for the unique capabilities of Spatial ALD, including unsaturated process conditions.

In this paper we present an overview and examples of thin films processing technologies for future generations of leading edge semiconductor devices. We introduce the main driving forces affecting future thin film deposition and etch technologies including the push for 3D (vertical) power, performance and area scaling. We discuss new thin films processes on the near term horizon that enable future devices, improved contacts, scaled 3D interconnects, and advanced patterning technologies followed by a future outlook for scaling.

When medicinal chemists need to improve bioavailability (%F) within a chemical series during lead optimization, they synthesize new series members with systematically modified properties mainly by following experience and general rules of thumb. More quantitative models that predict %F of proposed compounds from chemical structure alone have proven elusive. Global empirical %F quantitative structure-property (QSPR) models perform poorly, and projects have too little data to train local %F QSPR models. Mechanistic oral absorption and physiologically based pharmacokinetic (PBPK) models simulate the dissolution, absorption, systemic distribution, and clearance of a drug in preclinical species and humans. Attempts to build global PBPK models based purely on calculated inputs have not achieved the2-fold average error needed to guide lead optimization. In this work, local GastroPlus PBPK models are instead customized for individual medchem series. The key innovation was building a local QSPR for a numerically fitted effective intrinsic clearance (CL

When medicinal chemists need to improve oral bioavailability (%F) during lead optimization, they systematically modify compound properties mainly based on their own experience and general rules of thumb. However, at least a dozen properties can influence %F, and the difficulty of multiparameter optimization for such complex nonlinear processes grows combinatorially with the number of variables. Furthermore, strategies can be in conflict. For example, adding a polar or charged group will generally increase solubility but decrease permeability. Identifying the 2 or 3 properties that most influence %F for a given compound series would make %F optimization much more efficient. We previously reported an adaptation of physiologically based pharmacokinetic (PBPK) simulations to predict %F for lead series from purely computational inputs within a 2-fold average error. Here, we run thousands of such simulations to generate a comprehensive "bioavailability landscape" for each series. A key innovation was recognition that the large and variable number of p K

Even through Ge/high-k interface has been extensively studied the high leakage current associated with these gate stacks continues to introduce frequency dispersion and hysteresis in capacitance-voltage (CV) and conductance-voltage (GV) characteristics. These dispersions severely limit the understanding the interface and accurate estimation of interface state density, Dit and equivalent oxide thickness (EOT). Since several parameters like EOT, flatband voltage, bulk doping, surface potential as a function of gate voltage are important and bulk defect and interface defect concentration affect the CV and GV characteristics several corrections to the characteristics are required to obtain accurate values. We have measured the CV and GV characteristics of Ge/ALD 1nm-Al2O3/ALD 3.5nm-ZrO2/ALD TiN MOS capacitors with three different interface treatments by HP4284 LCR meter using different frequencies. The interface treatments are (i) simple chemical oxidation (Chemox); (ii) chemical oxide removal (COR) followed by 1 nm oxide by slot-plane-antenna (SPA) plasma (COR&SPAOx); and (iii) COR followed by vapor O3 treatment (COR&O3). The plots were corrected by three different methods. Firstly, we removed the effect of sheet resistance, RS . The second approach uses data (capacitance and conductance) measured at two different frequencies to solve two unknown variables (capacitance and conductance) (1). The solved CV plot is closer to the corrected CV plot at 1 MHz rather than that at 100 KHz. This allows us to use 1MHz corrected data to obtain the EOT information. Additionally, this approach helps to verify that 1MHz data is robust assuming that defects do not respond to the other frequencies used here. The third approach is to modify the measurement circuits (2) to enhance CV and GV characteristics. Fig. 1 shows the CV characteristics after the corrections by three different approaches for three different samples. COR&SPAOx samples show excellent CV characteristics. The details of all the correction methods and the advantages of these methods to correct CV and GV will be discussed. We will demonstrate that the correction methods are suitable for accurately measure the interface characteristics. The Dit calculated by two method (conductance and capacitance spectroscopy) (Fig. 2) are compared and discrepancy of the results between different samples and different methods will be explained. References Y. Ding and D. Misra, J. Vac. Sci. Technol. B, 33, 021203, 2015. K.J. Yang and C. Hu, IEEE Trans on Electron Dev, 46(7), 1500, 1999. K.S.K Kwa, S. Chattopadhyay, N.D. Jankovic, S.H. Olsen, L.S. Driscoll, and A.G. O'Neill, Semicond. Sci. Technol., 18(2), 82, 2003. Figure 1

Curating the data underlying quantitative structure-activity relationship models is a never-ending struggle. Some curation can now be automated but much cannot, especially where data as complex as those pertaining to molecular absorption, distribution, metabolism, excretion, and toxicity are concerned (vide infra). The authors discuss some particularly challenging problem areas in terms of specific examples involving experimental context, incompleteness of data, confusion of units, problematic nomenclature, tautomerism, and misapplication of automated structure recognition tools.

In a unique collaboration between a software company and a pharmaceutical company, we were able to develop a new in silico pKa prediction tool with outstanding prediction quality. An existing pKa prediction method from Simulations Plus based on artificial neural network ensembles (ANNE), microstates analysis, and literature data was retrained with a large homogeneous data set of drug-like molecules from Bayer. The new model was thus built with curated sets of ∼14,000 literature pKa values (∼11,000 compounds, representing literature chemical space) and ∼19,500 pKa values experimentally determined at Bayer Pharma (∼16,000 compounds, representing industry chemical space). Model validation was performed with several test sets consisting of a total of ∼31,000 new pKa values measured at Bayer. For the largest and most difficult test set with16,000 pKa values that were not used for training, the original model achieved a mean absolute error (MAE) of 0.72, root-mean-square error (RMSE) of 0.94, and squared correlation coefficient (R(2)) of 0.87. The new model achieves significantly improved prediction statistics, with MAE = 0.50, RMSE = 0.67, and R(2) = 0.93. It is commercially available as part of the Simulations Plus ADMET Predictor release 7.0. Good predictions are only of value when delivered effectively to those who can use them. The new pKa prediction model has been integrated into Pipeline Pilot and the PharmacophorInformatics (PIx) platform used by scientists at Bayer Pharma. Different output formats allow customized application by medicinal chemists, physical chemists, and computational chemists.

We report, for the first time, a gate last process, used to fabricate Negative Capacitance field effect transistors (NCFETs) with Hf 0.5 Zr 0.5 O 2 (HZO) as ferroelectric (FE) dielectric in a metal/ferroelectric/insulator/semiconductor (MFIS) configuration. Long channel NCFET's with HZO thickness down to 5 nm exhibit consistent switching behavior with switching slope (SS rev ) below kT/q over four decades of drain current. Temperature dependent transport study shows that, the effective mobility of HZO NCFETs is 15 % higher than that of HfO 2 based control MOSFETs due to suppression of Hf diffusion into the interfacial SiO 2 layer (IL). Using the Preisach hysteresis model, which models dynamics of FE switching through a cluster of independent switching dipoles at arbitrary electric field, we (a) explain the asymmetric SS behavior of NCFETs in MFIS configuration, and (b) establish design guidelines for achieving sub-kT/q SS in both forward and reverse sweep direction.

We report on Atomic Layer Deposition Titanium (ALD Ti) for FinFET source/drain contact applications. On planar test structures, we accurately benchmark contact resistivity (ρc) of ALD Ti, ∼1.4×10–9 Ω·cm2 on Si:P and ∼2.0×10–9 Ω·cm2 on SiGe:B, among to lowest reported values in literature. Ultralow ρc is resulting from enhanced Ti/Si(Ge) reactivity originating in the ALD process. We also demonstrate capability of this process to significantly lower Rc on FinFETs by allowing a lateral contact into the S/D area effectively maximizing the contacting area.

The emergence of in silico tools, such as physiologically based pharmacokinetic modeling, to predict human pharmacokinetics for new compounds now provides the ability to anticipate the advantages and liabilities associated with compounds in discovery. Physiologically based pharmacokinetic modeling is being used in diverse applications including simulation of preclinical, Phase 1 healthy volunteer, and special population pharmacokinetics; formulation optimization and food effects; and predicting the magnitude of drug-drug interactions. These applications are having a significant impact throughout drug discovery and development to increase efficiency, reduce the need for animal studies, and increase pharmacokinetic understanding in order to optimize development programs.

As Si metal oxide semiconducting field effect transistors have continued to scale, SiO2-based gate dielectrics and polySi gates have beensuccessfullyreplacedbyHfO2-basedhigh-kdielectricsandmetal gates at the 45nm technology node and beyond. In order to continue scaling, improvements of key film properties of the high-k dielectric, such as k-value, equivalent oxide thickness (EOT) and leakage current, are needed. A reduction in EOT can be achieved by reducing the thickness of the gate dielectric with the associated drawback of increased gate leakage, or by increasing its effective dielectric constant (k-value). In this regard, the electrical properties of HfO2-based dielectrics may be enhanced by structural modifications, since the kvalue has been shown to be dependent on the crystalline phase (monoclinick =16,tetragonalk =70andcubic =29). 1 HfO2 filmstypically crystallize into the monoclinic phase, which is its most thermodynamically stable phase. 2,3 One method of stabilizing the metastable cubic

As Si metal oxide semiconducting field effect transistors have continued to scale, SiO2-based gate dielectric and polySi gates have been successfully replaced by high-k dielectrics and metal gate at the 45nm technology node and beyond. In order to continue scaling, improvements of key film properties of the high-k dielectric, such as k-value and leakage current, are needed. The electrical properties of HfO2-based dielectrics may be enhanced by structural modifications, since the k-value is dependent on the crystalline phase (monoclinic k=16, tetragonal k=70 and cubic =29) [1]. HfO2 films typically crystallize into the monoclinic phase, which is its most thermodynamically stable phase [2,3]. Alloying HfO2 with other oxides such as ZrO2, TiO2, Al2O3offer a viable option towards engineering the crystalline phase. In this regards, we have deposited 40 cycles of Hf1-xAlxOy films by atomic layer deposition with Al/(Hf+Al)% ranging from 0 to 25%. A set of samples were annealed under N2 at temperatures ranging from 680 oC to 800oC and compared to the as deposited samples. To investigate the crystallinity of the films, we performed grazing incidence in-plane X-ray diffraction (GIIXRD) on the annealed films and in-situ XRD during an anneal process using synchrotron radiation was used to determine the crystallization temperature (Tc) for the unannealed films. The GIIXRD measurements revealed the presence of non-monoclinic peaks which was also confirmed by the in-situ anneal XRD measurements. The crystallization temperature of the non-monoclinic phase obtained from the in-situ anneal XRD is plotted in Fig.1, where it is shown that higher Al/(Hf+Al)% results in higher Tc. In order to resolve the non-monoclinic phase, extended X-ray absorption fine structure (EXAFS) at was utilized. The EXAFS measurement combined with GIIXRD showed the ALD Hf1-xAlxOy films crystallize into a mixture of the tetragonal and monoclinic phases. The electrical properties were investigated by MOS capacitors with ALD TiN as the metal gate. The electrical measurements show a reduction in EOT after post deposition anneal (PDA). In addition, the leakage current was also reduced by a factor of 10 while maintaining a flat-band voltage that is comparable to PDA HfO2 films under identical conditions. The C-V plots of the as deposited samples and annealed samples are shown in Fig. 2 and Fig.3 respectively. We also investigated the effect of different thermal budgets on the Hf1-xAlxOy films. A ~2Å EOT reduction with lower gate leakage was obtained for devices with post deposition anneal near the crystallization temperature of the Hf1-xAlxOyfilms. However, the electrical performance is degraded for devices with higher or lower post deposition anneal temperature than the films’ crystallization. In summary, we have successfully deposited ALD Hf1-xAlxOy. The crystal phase was confirmed to be a mixed phase of tetragonal with monoclinic. The data shows an enhancement in electrical properties near the crystallization temperature of the Hf1-xAlxOyfilms. Acknowledgements Diffraction measurements were carried out at NSLS (X20A and X20C), BNL, which is supported by the DOE, Division of Materials Sciences and Division of Chemical Sciences, under Contract No. DE-AC02-98CH10886. References [1] X. Zhao et al., Phys. Rev. B, 65, 233106 (2002). [2] E.P Gusev, V. Narayanan, M. M Frank, Phys. Status Solidi A, 201, 1443 (2004). [3] D. A Neumayer, E. Cartier, J. Appl. Phys., 90, 1801 (2001).

CMOS device scaling below the 22 nm technology node requires gate dielectric materials with superior properties to those of conventional high-k materials. It was found that, Al incorporation into HfO2 results in an increase in the transition temperature from amorphous to polycrystalline state (1-2). Improved thermal stability (2) and enhanced permittivity (3) by Al doping into ALD HfO2 was also reported. As a result of the improvement in film properties, improved equivalent oxide thickness (EOT) values, reduced gate leakage current, lower hysteresis and improved interface quality were also observed for Al doped HfO2 (4-6). However, intermixing of HfO2 and Al2O3 depends significantly on the stack structure, which results in variations in the stoichiometry in the film (7). The relative position of HfO2 and Al2O3 in HfO2/Al2O3bi-layer has a significant influence on positive charge formation in the dielectrics and also on the quality of the interfacial layer (IL) (8-10). In this work, we deposited and studied the reliability of 40 cycles ALD HfO2 doped with Al in the deposition process. The ALD thin film composition was measured by X-ray photoelectron spectroscopy (XPS). Thin films with Al/(Al+Hf)% composition ranging 0% to 8% were deposited and annealed at 800oC in N2 ambient. MOS capacitors were then fabricated with ALD TiN as the metal gate. The electrical measurements show that, devices with ~2% Al/(Hf+Al)% have an optimized performance with 17% lower EOT than undoped HfO2. The amount of positive charge and gate leakage current density were also evaluated for all devices. The reliability of these devices was monitored by subjecting them to a constant field stress at E= 27.5 MV/cm in the gate injection mode, where the applied voltage to the gate was modified according to the EOT to have equal field across all the dielectrics. Stress-induced trap generation in the dielectrics was observed due to bulk and interface defects. Trap formation followed a power law function with stress time. Observed stress induced flat-band voltage shift (Fig. 1) shows that devices with Al/(Hf+Al)% =2% have 36% reduction in the rate of trap generation as compared to HfO2, whereas degraded flat-band voltage shift was observed for devices with Al/(Hf+Al)% =7%. It can be inferred that an optimized Al concentration in HfO2can lead to an improvement of the device performance. In order to further understand the reliability characteristics we plan to evaluate the change in interface state density with stress and their time dependent dielectric breakdown (TDDB) characteristics. References M.-Y. Ho et al., Appl. Phys. Lett., 81, 4218 (2002). H. Y. Yu et al., Appl. Phys. Lett., 81, 3618(2002). P. K. Park et al., Appl. Phys. Lett., 89, 192905(2006). W. J. Zhu et al., IEEE Electron. Dev. Lett., 23(11), 649(2002). V. Mikhelasvili et al., Appl. Phys. Lett., 85, 5950(2004). Y.-K. Chiou et al., J. Electrochem. Soc., 154 (4) G99(2007). M. H. Cho et al., Appl. Phys. Lett., 81, 1071(2002). B. H. Lee et al., Appl. Phys. Lett., 76, 1926 (2000). L. G. Gosset et al., J. Non-Cryst. Solids, 303, 17(2002). W. Wang et al., J. Appl. Phys., 105, 064108 (2009).

Use of Hf based high-k dielectric materials enabled CMOS device scaling beyond the 22 nm technology node. In order to scale the equivalent oxide thickness (EOT) to 0.7 nm and beyond, while achieving good reliability performance, improvement of the film properties and the deposition methods are highly desirable. It was observed that, the dielectric constant of ALD HfO2 can be improved by the addition of ZrO2 as it offers higher-k tetragonal stabilization when Zr/(Hf+Zr) percentage is more than 50% (1-2). It has recently been reported that the poor dielectric characteristics of ALD grown oxide films, such as the leakage current, can be improved by exposing the dielectric films to a slot-plane-antenna (SPA) plasma (3). Because of low electron temperature the SPA process causes very little damage as compared to conventional inductively coupled plasma (ICP) or electron cyclotron resonance (ECR) plasma (4). Also cyclic treatment during the ALD deposition process by using room temperature ultraviolet ozone, D2O radical, remote microwave N2O plasma, and/or thermal annealing have shown promising results toward improving film properties (5-7). In this study, for the first time, we examined the performance benefits due to the use of intermediate SPA Ar plasma treatment (DSDS) in the ALD deposition process of Hf1-xZrxO2 with 0%, 31% and 80% Zr/(Hf+Zr) percentages and compared them with the standard deposition process (As_Dep). Zr percentage was varied by precisely controlling the Hf-precursor to Zr-precursor pulse ratio in the ALD cycle. MOSCAPs were formed with TiN as gate material, where ALD Hf1-xZrxO2 was deposited on a SiON interface formed by remote plasma radical flow nitridation of chemically grown oxide on a p-Si substrate. It is observed that, DSDS Hf1-xZrxO2with 80% Zr has the lowest dielectric thickness and interfacial layer thickness as compared to others (Fig. 1(a)). Equivalent oxide thickness (EOT) for different Zr percentages for DSDS and As-Dep processing conditions were also compared. Effect of SPA Ar plasma and Zr addition on the reliability of ALD Hf1-xZrxO2 were monitored from the flat-band voltage shift due to a constant field stress at E= 27.5 MV/cm in the gate injection mode (Fig. 1(b)). Also, the gate leakage current density, before and after stress for different devices were monitored. When subjected to a constant field stress traps are generated in the gate dielectric and at the Si-IL interface. DSDS Hf1-xZrxO2with 80% Zr showed the lowest initial flat-band voltage shift after the first 20-second stress and the subsequent increase of positive charge formation was observed as the stress continued. Devices with intermediate Ar plasma have reduced gate leakage current density before stress and after 1000s stress, which implies suppression of oxide trap formation due to plasma exposure. Intermediate Ar plasma exposure to ALD Hf1-xZrxO2with 80% Zr seems to deposit superior dielectric with better EOT downscaling ability and good reliability performance. While Zr addition helps to produce a fine grain microstructure with less oxygen vacancies (2), SPA plasma further helps by reducing contaminants, increasing mass density, and superior bond structure of the film (3). References 1. K. Tapily et al,ECS Trans., 45 (3) 411-420 (2012). 2. R.I. Hegde, et al, J. Appl. Phys. 101, 074113 (2007) 3. T. Tanimura et al, J. Appl., Phys., 113, 064102 (2013) 4. C. Tian et al, J. Vac. Sci. Technol. A 24(4), 1421(2006). 5. S. Consiglio et al, ECS Trans., 41, 89 (2011). 6. R.D. Clark et al, ECS Trans., 35(4), 815 (2011). 7. A. Delabie et al, J. Electrochem. Soc.,153, F180 (2006). Figure 1(a): Dielectric thickness (filled symbols in the left scale) for DSDS and As-Dep gate oxides and interfacial layer (IL) thickness (open symbols in the right scale) for MOSCAPs with DSDS and As-Deposited HfO2 (0% Zr), Hf1-xZrxO2 (31% Zr), and Hf1-xZrxO2 (80% Zr). (All processes used 44 ALD cycles of HfO2/Hf1-xZrxO2 deposition at 2500 C) Figure 1(b): Flat-band voltage values obtained from Capacitance-Voltage (CV) characteristics for unstressed devices (filled symbols) and after 1000s stress (open symbols) with a constant field stress at E= 27.5 MV/cm in the gate injection mode. (MOSCAPS having HfO2 (0% Zr) are denoted by squares, Hf1-xZrxO2 (31% Zr) are denoted by circles, and Hf1-xZrxO2(80% Zr) are denoted by triangles.)