CMOS based computing is reaching its limits. To take computation beyond Moores law (the number of transistors and hence processing power on a chip doubles every 18 months to 3 years) requires research explorations in (i) new materials, devices, and processes, (ii) new architectures and algorithms, (iii) new paradigm of logic bit representation. The focus is on fundamental new ways to compute under the umbrella of rebooting computing such as spintronics, quantum computing, adiabatic and reversible computing. Therefore, this thesis highlights explicitly Quantum computing and Adiabatic logic, two new computing paradigms that come under the umbrella of rebooting computing. Quantum computing is investigated for its promising application in high-performance computing. The first contribution of this thesis is the design of two resource-efficient designs for quantum integer division. The first design is based on non-restoring division algorithm and the second one is based on restoring division algorithm. Both the designs are compared and shown to be superior to the existing work in terms of T-count and T-depth. The proliferation of IoT devices which work on low-power also has drawn interests to the rebooting computing. Hence, the second contribution of this thesis is proving that Adiabatic Logic is a promising candidate for implementation in IoT devices. The adiabatic logic family called Symmetric Pass Gate Adiabatic Logic (SPGAL) is implemented in PRESENT-80 lightweight algorithm. Adiabatic Logic is extended to emerging transistor devices.

Internet of Things (IoT) is a network of devices that are connected through the Internet to exchange the data for intelligent applications. Though IoT devices provide several advantages to improve the quality of life, they also present challenges related to security. The security issues related to IoT devices include leakage of information through Differential Power Analysis (DPA) based side channel attacks, authentication, piracy, etc. DPA is a type of side-channel attack where the attacker monitors the power consumption of the device to guess the secret key stored in it. There are several countermeasures to overcome DPA attacks. However, most of the existing countermeasures consume high power which makes them not suitable to implement in power constraint devices. IoT devices are battery operated, hence it is important to investigate the methods to design energy-efficient and secure IoT devices not susceptible to DPA attacks. In this research, we have explored the usefulness of a novel computing platform called adiabatic logic, low-leakage FinFET devices and Magnetic Tunnel Junction (MTJ) Logic-in-Memory (LiM) architecture to design energy-efficient and DPA secure hardware. Further, we have also explored the usefulness of adiabatic logic in the design of energy-efficient and reliable Physically Unclonable Function (PUF) circuits to overcome the authentication and piracy issues in IoT devices. Adiabatic logic is a low-power circuit design technique to design energy-efficient hardware. Adiabatic logic has reduced dynamic switching energy loss due to the recycling of charge to the power clock. As the first contribution of this dissertation, we have proposed a novel DPA-resistant adiabatic logic family called Energy-Efficient Secure Positive Feedback Adiabatic Logic (EE-SPFAL). EE-SPFAL based circuits are energy-efficient compared to the conventional CMOS based design because of recycling the charge after every clock cycle. Further, EE-SPFAL based circuits consume uniform power irrespective of input data transition which makes them resilience against DPA attacks. Scaling of CMOS transistors have served the industry for more than 50 years in providing integrated circuits that are denser, and cheaper along with its high performance, and low power. However, scaling of the transistors leads to increase in leakage current. Increase in leakage current reduces the energy-efficiency of the computing circuits,and increases their vulnerability to DPA attack. Hence, it is important to investigate the crypto circuits in low leakage devices such as FinFET to make them energy-efficient and DPA resistant. In this dissertation, we have proposed a novel FinFET based Secure Adiabatic Logic (FinSAL) family. FinSAL based designs utilize the low-leakage FinFET device along with adiabatic logic principles to improve energy-efficiency along with its resistance against DPA attack. Recently, Magnetic Tunnel Junction (MTJ)/CMOS based Logic-in-Memory (LiM) circuits have been explored to design low-power non-volatile hardware. Some of the advantages of MTJ device include non-volatility, near-zero leakage power, high integration density and easy compatibility with CMOS devices. However, the differences in power consumption between the switching of MTJ devices increase the vulnerability of Differential Power Analysis (DPA) based side-channel attack. Further, the MTJ/CMOS hybrid logic circuits which require frequent switching of MTJs are not very energy-efficient due to the significant energy required to switch the MTJ devices. In the third contribution of this dissertation, we have investigated a novel approach of building cryptographic hardware in MTJ/CMOS circuits using Look-Up Table (LUT) based method where the data stored in MTJs are constant during the entire encryption/decryption operation. Currently, high supply voltage is required in both writing and sensing operations of hybrid MTJ/CMOS based LiM circuits which consumes a considerable amount of energy. In order to meet the power budget in low-power devices, it is important to investigate the novel design techniques to design ultra-low-power MTJ/CMOS circuits. In the fourth contribution of this dissertation, we have proposed a novel energy-efficient Secure MTJ/CMOS Logic (SMCL) family. The proposed SMCL logic family consumes uniform power irrespective of data transition in MTJ and more energy-efficient compared to the state-of-art MTJ/ CMOS designs by using charge sharing technique. The other important contribution of this dissertation is the design of reliable Physical Unclonable Function (PUF). Physically Unclonable Function (PUF) are circuits which are used to generate secret keys to avoid the piracy and device authentication problems. However, existing PUFs consume high power and they suffer from the problem of generating unreliable bits. This dissertation have addressed this issue in PUFs by designing a novel adiabatic logic based PUF. The time ramp voltages in adiabatic PUF is utilized to improve the reliability of the PUF along with its energy-efficiency. Reliability of the adiabatic logic based PUF proposed in this dissertation is tested through simulation based temperature variations and supply voltage variations.

La densité de puissance est devenue la principale préoccupation lorsqu'un circuit numérique est conçu. Comme pour tous les systèmes embarqués, chaque nouvelle génération de système numérique a plus d'applications que la précédente et exige en fin de compte une plus grande densité de puissance. C'est pourquoi de nombreux chercheurs et concepteurs industriels se sont penchés sur de nouvelles méthodes de réduction de la consommation énergétique des circuits numériques. La logique adiabatique est un style de conception prometteur qui peut réduire la dissipation d'énergie dynamique. La logique adiabatique est différente de la logique conventionnelle en deux principaux points : 1) l’alimentation d’une porte logique adiabatique est un signal à 4 phases, et 2) l’énergie stockée dans la porte est récupérée. Afin de respecter ces principes, la logique adiabatique nécessite une alimentation spéciale. Étant donné que l’objectif d’une telle alimentation est d’agir comme une horloge, elle est appelée alimentation-horloge. L'objectif de cette thèse est de concevoir et d'optimiser une alimentation-horloge ainsi que son réseau de distribution. Cette thèse a été financée par l'Agence Nationale pour la Recherche, ANR, avec le projet ADIANEMS2 (numéro de subvention : ANR-15-CE24-0013). / Power density has become the primary concern when a digital core is designed. As in any embedded systems, each new digital core generation has more applications than the previous one and ultimately demands more power density. This is why many researchers and industrial designers have been looking into novel methods for reducing power consumption of digital circuit. Adiabatic logic is a promising design style, which can reduce the dynamic energy dissipation. Adiabatic logic is different than conventional logic in two main points: 1) adiabatic gate are charged with a 4-phase power signal, and 2) the energy, which is stored in the gate, is recovered. In order to fulfill these principles, the adiabatic logic needs a special power supply. As the purpose of such supply is to act as a clock also, it is referred as power-clock supply. The aim of this thesis is to design and optimize a power-clock supply and its delivery network. This thesis has been funded by the French National Research Agency, ANR, with the project ADIANEMS2 (Grant number: ANR-15-CE24-0013).

The growing applications for IoT devices have caused an increase in the study of low power consuming circuit design to meet the requirement of devices to operate for various months without external power supply. Scaling down the conventional CMOS causes various complications to design due to CMOS properties, therefore various non-conventional CMOS design techniques are being proposed that overcome the limitations. This thesis focuses on some of those emerging and novel low power design technique namely Adiabatic logic and low power devices like Magnetic Tunnel Junction (MTJ) and Carbon Nanotube Field Effect transistor (CNFET). Circuits that are used for large computations (multipliers, encryption engines) that amount to maximum part of power consumption in a whole chip are designed using these novel low power techniques.