Voltage compensation of manufacturing errors applied to artificial band gaps in nanoscale transistor Authors: Ashkan Aghajani, Avi Karsenty Published: 2026 ABSTRACT A nanoscale MOS transistor possessing an embedded SiO2-Si-SiO2 quantum well serving as source-drain channel is presented. To examine the influence of significant variables on discrete quantum energy levels, band gap, and optoelectronic characteristics of the electrical potential distribution through the channel, an advanced numerical model has been designed. Since the channel thickness – the key parameter for the light emission wavelength – is not always uniform between devices in the same fabrication batch, we propose using activation voltages to compensate for manufacturing errors in the Si-SiO2 interfaces of the quantum well. The expected benefits include the control of light emission through electrical voltages (VGS, VDS) and the channel thickness (tSi), the mitigation of parasitic currents through the insulation buried oxide layer, and a completely VLSI-compatible design for swift industry integration based on Si and SiO2. Through calculating the eigenvalues’ distribution along the channel length (L), we argue that electrical compensation of Si layer process mismatch can modulate the inter-sub-band transitions (ISBT) when applying adequate activation voltages. This is of particular importance for long (L) and large (W) channel transistors, as it is necessary to ensure a significant quantity of hot electrons for triggering the ISBT efficiently. Introduction Embedding silicon nanophotonics in CMOS technology Currently, nanoelectronics and nanophotonics are at the forefront of the nano revolution. Since microelectronics rely on silicon, progress in silicon photonics and photonic integrated circuits (PICs) will enable the implementation of combined optical signaling and electronic data processing. Advancing silicon past its existing limits will necessitate substituting electronic systems with photonic alternatives1–3, and advancements in the field of silicon PICs4,5 in order to combine optical signals with electronic data processing. Recently published roadmaps outlining silicon photonics’ future6–9 discussed the various innovations and benefits expected to result from further development. Development of these devices intended for the upcoming generation of ultrafast computers is advancing alongside broader initiatives aimed at creating the next generation of optoelectronic communication systems, since the constraints of conventional electronic communication require a shift in approach towards electro-optic (EO) communication systems that combine electronic elements with optical coupling methods. Despite this trend, not every existing platform and foundry can facilitate seamless integration of developed coupled devices and PICs10,11, as numerous parameters must be considered. Overall, the field of photonics platforms is quite advanced; for instance, the Scientific Community can utilize a range of industrial photonics platforms12, and Metal-Oxide-Semiconductor (MOS) technology remains one of the most attractive bases for this kind of integration. Even silicon meta-surfaces were proven perfectly compatible with CMOS technology and Si photonics13, and already are in wide service as optical waveguides14, imaging lenses15, and holography devices16. Since one of the basic active components of microelectronics is the MOS transistor, and since silicon photonics becomes more desirable with integrated light sources17,18, combining both aspects in one device seems to present a win-win situation. The nanophotonic MOSQWELL – Metal-Oxide-Semiconductor Quantum Well – is a nanoscale light emitting transistor, using both electrons and photons for communication, which surmounts silicon’s indirect bandgap through its unique geometry. In an indirect bandgap material like silicon, Band-To-Band (BTB) transitions between conduction and valence bands inhibit radiative recombination. However, hot electrons could trigger photonic inter-sub-band transitions (ISBT) within the conduction band19. Positioned within the transistor channel, the quantum well design compels the creation of hot electrons and photonic ISBT. This quantum well features a recessed ultra-thin silicon layer, created using Gate-Recessed Channel (GRC) geometry, situated between two oxide layers. Process technologies considerations For the fabrication of the MOSQWELL transistor, CMOS technology has been chosen since it provides a smooth path for the integration of the SiO2-Si-SiO2 quantum well (QW) structure (Fig. 1) as the channel path between source and drain, and for a first proof of concept. Moreover, the gate oxide (GOX) upper layer and the buried oxide (BOX) were identified as a straightforward choice for the structure to create the necessary energy band levels of the QW (Fig. 2). Since silicon has dominated the microelectronics industry for eight decades as an irreplaceable large scale wafer starting material – almost 28% of the Earth’s crust20–22 is sand – there was no doubt that we should start our investigations with CMOS. Of course, there are alternatives to CMOS, such as FinFET23, GAAFET24, and FATFET25, and comparisons of nano transistors were performed using these techniques26. Oxide replacement materials such as high-K dielectric23,24,27 can be explored as well in the future for the implementation of MOSQWELL fabrication. The pursuit after more advanced technologies and further device shrinking – in particular beyond the end of Moore law28–31– is usually driven by the need of improved performances (reduced area, increased speed, and higher power efficiency). Recent work has shown superior switching properties using different materials such as gallium nitride (GaN) semiconductor on insulator (SOI) FinFET with high-k wrapped underlaps and TiO2 gate oxide23. The usage of a high-K dielectric layer replacing the classic oxide has been extensively discussed in the literature and found useful in electrical performance with GAA FET using silicon-germanium (SiGe) with a 2 nm gate underlap encapsulating a high-k spacer24. Superior electron mobility and high breakdown voltage, which are extremely desirable for high-frequency applications, have been presented in recent work32. Parasitic capacitance and reduced current leakage were also obtained due to the high-K layering system32. SIMOX (Separation by IMplanted OXygen) SOI wafers33 have been selected as the commercial starting material for the MOSQWELL device. In these wafers, the Buried OXide (BOX) is created through the internal oxidation of oxygen ions implanted deeply into silicon34. The 55 nm upper silicon layer is p-type with a resistivity of 14–22 ∧cm over an 80 nm buried oxide. The conventional isoplanar procedure, along with the growing of 25 nm gate oxide, decreased the silicon layer to 40 nm. A 220 nm polysilicon layer was deposited on the gate oxide, then oxidized and patterned to create the gate electrode. Prior to the phosphorus implant for the source, drain, and gate, a 15 nm thin layer of nitride was applied. The aim of this nitride layer was to stop additional oxidation of the thin silicon layer during the thermal annealing after implantation. A 350 nm thick Chemical Vapor Deposition(CVD) oxide was applied, and this was succeeded by the aluminium metallization process. Standard reference MOSFET transistors in bulk silicon were similarly fabricated using the same technology for evaluating and comparing the electrical and optical characteristics of the devices. Electrical compensation correction A semi-classical model was established in a previous study35 for the MOSQWELL nano-transistor, focusing on the electronic aspects of the device and using MATLAB36. We present here a complementary analysis, this time using the COMSOL Multiphysics® software package37. This paper investigates the concept of electrically compensating for non-uniform thicknesses between different silicon-on-insulator (SOI) devices. This proposed adaptable technology holds immense potential for transformative applications in various fields. The electrical compensation and correction in metal-oxide-semiconductor (MOS) structures usually treats two main sub-domains: the compensation of physical defects or interfaces in MOS devices, and the compensation of parasitic electrical effects in MOS circuits. Charge injection error compensation38 falls under the first category, and interface trap correction and charge compensation under the second. In regard to latter, the main needs remain current mirror39 and temperature compensations, as well as MOS model enhancement40. In fact, the original idea of electrical compensation correction was first presented few decades ago, where a back-gate voltage (VBG) solution was applied to compensate for the threshold voltage in devices41,42. Another popular use case of electrical compensation is in the domain of floating gate transistors for mismatch compensation. Indeed, the simplicity of the CMOS current mirror makes it particularly susceptible to device mismatch due to process variations39. Generally speaking, the idea of electrical compensation correction methods belongs to analog design, such as the design of a new analog switch, tested on 0.18 μm standard CMOS technology for charge injection error compensation38. In our case, activation voltages will compensate for the layer thickness and enable several eigenvalues along the channel length. Numerical model concept The physics of an embedded quantum well structure inside an existing functional transistor is not straightforward; first, we need to treat the two separate components, and then to combine their descriptions into one advanced model. The well itself enables ISBT due to the discretization of the energy levels, and it acts as a photonic component by enabling radiative recombination. On the other hand, the transistor is a nano-electronic device, activated at a threshold voltage (VTH) by the gate voltage (VGS), which is responsible for the quantity of electrons injected into the channel, and by the drain voltage (VDS), which is responsible for the velocity of these hot electrons. Through impact ionization effect, these hot electrons will acquire spare energy and will populate the higher discrete energy levels of the conduction band, enabling ISBT and radiative recombination. To explain the details of the fully integrated model, we will progress in a step-by-step manner, as presented below. Starting from a quantum well structure When two different materials or semiconductors are placed adjacent to one to each other they form a heterojunction. The discontinuity in the conduction band can be represented by a constant potential term. As mentioned above, if the band gaps of the materials are different, a potential barrier is formed. If, say, silicon is sandwiched between two layers of silicon oxide, the band offset – the potential difference between the conduction band edges – forms a finite potential well. In the case of the MOSQWELL, the GOX – Si – BOX combination forms a quantum well of height 3.3 eV (Fig. 1). The discrete levels of the channel conduction band heterostructure can allow for radiative recombination by way of ISBT. In this way, MOSQWELL can operate as an optical transistor43. Fig. 1MOSQWELL heterostructure band diagram. MOSQWELL heterostructure band diagram. The application of an external electric field to a quantum well results in a bending of the potential energy bands which form the well. The resultant perturbation to the system’s Hamiltonian causes changes to the confined energy levels and a shift in the wavefunctions. While confinement still exists, it is different due to the change in potential. These phenomena are known as the Quantum Confined Stark Effect (QCSE) and the Stark shift respectively. For the MOSQWELL transistor, the application of gate bias causes a voltage drop along the Z axis over the SiO2/Si/SiO2 heterojunction, thereby resulting in a Stark shift44. The four regions of the MOSQWELL model For the calculation, the well is divided into 4 sections (Fig. 2): the GOX, the quantum well itself, the BOX and the silicon body. The resulting graphs are taken from simulations done on a vertical 1D cross-section model of the component including the above four layers. Fig. 21D semi-conductor, Schrödinger-Poisson analysis 300 K. (a) MOSQWELL four regions; (b) Zoom-in of the QW. 1D semi-conductor, Schrödinger-Poisson analysis 300 K. (a) MOSQWELL four regions; (b) Zoom-in of the QW. By applying a gate voltage VGS on the GOX part of the device, the capacitive coupling of the different interfaces of the MOSWELL will cause the electric potential energy to be linearly bent according to Eqs. (7), (8) and (9) below. In the GOX the electric potential energy EGOX(z), a function of the coordinate z corresponding to the conduction level, is defined using the following equation for 0 < z <\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\:{t}_{Gox}$$\end{document}:1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{{E}_{GOX}\left(z\right)/q=V}_{GOX}-\left(\frac{{C}_{T}}{{C}_{Si}}\right)\cdot\:{V}_{GS}+\left(\frac{{C}_{T}}{{C}_{GOX}}\right)\cdot\:\left(\frac{{V}_{GS}}{{t}_{GOX}}\right)\cdot\:(z-{t}_{GOX})$$\end{document} VGOX is the potential height of the quantum well at VGS = 0 V and is defined as the difference between the top conduction band of the GOX and the silicon channel (VGOX = 3.3 V relative to the silicon conduction band). This value is given by the difference between the respective electron affinities of silicon oxide and silicon, 0.8 V and 4.1 V respectively. Since zero potential energy is defined here to be at the bottom of the quantum well at VGS = 0 V, the electric potential for the channel part (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{t}_{Gox}$$\end{document}< z <\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\:{t}_{Gox}+{t}_{Si})\:$$\end{document}is given by:2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{E}_{Si}\left(z\right)/q=\left(\frac{{C}_{T}}{{C}_{Si}}\right)\cdot\:\left(\frac{{V}_{GS}}{{t}_{Si}}\right)\cdot\:(z-{t}_{GOX}-{t}_{Si})$$\end{document} The electric potential energy for the BOX part (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{t}_{Gox}+{t}_{Si}$$\end{document}< z <\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{t}_{Gox}+{t}_{Si}{+t}_{Box})\:$$\end{document}is given by:3\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{{E}_{BOX}\left(z\right)/q=V}_{BOX}-\left(\frac{{C}_{T}}{{C}_{Si}}\right)\cdot\:{V}_{GS}+\left(\frac{{C}_{T}}{{C}_{GOX}}\right)\cdot\:{V}_{GS}+\left(\frac{{C}_{T}}{{C}_{BOX}}\right)\cdot\:\left(\frac{{V}_{G}}{{t}_{BOX}}\right)\cdot\:\left(z-{t}_{BOX}\right)$$\end{document} Where: VGOX = 3.3 V is the GOX potential height in 1D. VBOX = 3.3 V is the BOX potential height in 1D. CSi is the silicon channel capacitance. CGOX is the gate oxide capacitance. CBOX is the buried oxide capacitance. CT is the total capacitance (serial association of CSi, CGOX and CBOX). tSi is the silicon channel thickness (2.4 nm). tGOX is gate oxide thickness. tBOX is the bottom oxide thickness. Discrete energy levels After creating a model that works well with VGS, we wanted to check whether a change in the voltage would affect the values of the energy levels. The assumption was that an increase in the voltage would affect the values of the energy levels, and in fact a certain eigenvalue will have a lower eV value as the voltage VGS increases. When there is no electric field in the device, many discrete energy levels exist in the quantum well, depending of the well thickness (in our case 2.4 nm). As the electric field in the device increases, the energy levels (En) in the well decrease according to the Quantum Confined Stark Effect (QCSE)45,46. Table 1 shows the changes in λn as a function of the voltage for each of the first ten eigenvalues found, and a plot of λn vs. VGS is presented in Fig. 3. Table 1Energy levels λn (eV) for n = 1 to 4 Eigenvalues depending on VGS. VDS = 0 V.λVGS0 V1 V2 V3 Vλ10.19753− 0.00764− 0.047− 0.07887λ20.796910.059080.550270.25532λ31.80311.59671.22860.48579λ43.07892.19631.40050.51727 Table 1Energy levels λn (eV) for n = 1 to 4 Eigenvalues depending on VGS. VDS = 0 V.λVGS0 V1 V2 V3 Vλ10.19753− 0.00764− 0.047− 0.07887λ20.796910.059080.550270.25532λ31.80311.59671.22860.48579λ43.07892.19631.40050.51727 Energy levels λn (eV) for n = 1 to 4 Eigenvalues depending on VGS. VDS = 0 V. Fig. 3Plot of λn vs. gate voltage VGS. Plot of λn vs. gate voltage VGS. Quantum well embedded in MOS transistor For the investigations and simulations, one-dimensional (1D), two-dimensional (2D), and three-dimensional 3D QW models of the device were built, with the QW structure embedded in the MOSFET device as shown in Fig. 4. The colors red and yellow indicate different materials within the component. The yellow areas are monocrystalline silicon (Si), while red represents silicon dioxide (SiO2). Additionally, between the red layers, there is a thin channel where the quantum well is formed, which is the focus of the project. This simulation is conducted on a 2-axis model, which was a horizontal cut of the component from the GOX (gate oxide) to the BOX (buried oxide) so the quantum well is positioned in the middle. Fig. 4Cross-sectional image of the QW in MOSQWELL. Cross-sectional image of the QW in MOSQWELL. Non-idealities influence While temperature acts as a scaling factor for both phonon activity and carrier distribution, phonon scattering is the dominant mechanism for non-radiative relaxation in quantum wells. This leads to a linewidth broadening of the photon emission. In particular, in the mid-infrared or THz regime, if the sub-band spacing is greater than the optical phonon energy, electrons in the upper sub-band can emit an LO-phonon and relax almost instantly. This significantly reduces the carrier lifetime in the excited state. High carrier density leads to frequent collisions between electrons, which redistributes energy within the sub-bands and contributes to the thermalization of the carrier distribution. As carriers thermalize and spread across different energy states within a sub-band, the range of possible transition energies increases, which can broaden the spectral peaks of the device. Intrinsic body vs. doped body The applicability of the MOSQWELL transistor model to an intrinsic body – meaning one with a silicon channel that is undoped or very lightly doped – is an important question in modern nano-CMOS devices47,48 and silicon photonics, since dopant fluctuations can be critical49. In this state, the device behavior shifts from being dominated by impurity scattering to being governed by pure quantum confinement and field-induced carrier generation. However, we started our investigations with a standard doped MOSFET structure which was turned into a quantum well by thinning the channel to the nanoscale. The potential well is partially defined by the depletion region of the dopants. In the future, we may consider the feasibility of an intrinsic body-based device since the model is arguably most applicable in this case. Indeed, it allows for the creation of a tunable “artificial bandgap” that is not degraded by the statistical randomness of dopants, making it an ideal platform for monolithic silicon light sources. Short channel vs. long channel Regarding the channel length (L) between source (S) and drain (D), we have designed and addressed a case study of a long channel device with longer free length, in order to enable more injection of electron to higher confined states through ISBT mechanism. Today, new technologies’ trend is to work more with short channels (22 nm, 14 nm, 7 nm), which gives better voltage distribution and part of shrinking effort. Moreover, short-channel effects (SCEs) in MOSFETs cause some non-desirable effects such as drain-induced barrier lowering (DIBL), threshold voltage (VTH) reduction, hot carriers trapping due to strong electric fields near the drain, imparting high energy to electrons and more. Our perspective was to first develop a proof of concept, and later on – in further investigations – to address more challenging case studies (after shrinking) and with optimized manufacturing process. This is also why the fabrication process of first generation devices used accurate dry oxidation in order to thin the original BOX (buried oxide) layer from 50 nm to a few nanometers. The process was well-controlled but presented some fluctuations. The purpose was to supply proof of concept of our novelty toward the second expected generation of MOSQWELL devices. Many alternative techniques exist today, such as atomic layer deposition (ALD), allowing for a smoother Si-SiO2 interface, preventing cases of trapped charges. Methods The need for numerical methods COMSOL Multiphysics® is a cross-analysis platform and software package which provides tools to model structures and solve complex physical problems in the fields of Physics and Engineering, based on the Finite Element Method (FEM)50. It allows conventional physics-based user interfaces and coupled systems of Partial Differential Equations (PDE). The FEM, introduced many decades ago by Courant51, is used in Multiphysics software packages in order to support the design and simulation of physical devices52 and phenomena53. The physical equations are discretized on a mesh using the Galerkin FEM, and to a lesser extent the Finite Volumes Method (FVM). The equations can be implemented in a variety of forms such as directly as a PDE, or as a variational integral, the so-called weak form. Boundary conditions may also be imposed directly or by using variational constraint and reaction forces. Both choices have implications for the convergence and physicality of the solution. The mesh is assembled from polygons (typically triangles or quadrilateral elements in two-dimensions), and hexahedral or prismatic elements in three dimensions, using a variety of algorithms. The solution is achieved using direct or iterative linear solvers and nonlinear solvers. Workflow and parameters As mentioned, COMSOL is an advanced analysis software, and therefore it uses many calculation methods and numerical approximations of the solution. This versatility is very significant, since the software is used as a tool for analysis of in-depth mathematics for different types of physical systems, each of which requires a different analysis. Working on the project, various calculations were used in different levels of physics and electronics, each with its own set of equations. Most of the time, the resulting equations are non-linear, and therefore a numerical solution is required. Beyond choosing the method of the mathematical solution (here FEM), it is possible to define to the adapted software what the nature of the requested solution. For example, for problems dealing with a quantum well, the nature of the expected solution is the eigenvalues of the Schrödinger equation. Also, there are calculations that require solutions depending on time, and some of the solutions are required in a state of equilibrium. In our case, this paper will focus on electrical studies, in a combination of both classical and quantum-based physics. The workflow of this research involved several stages (Fig. 5) and the parameters used are summarized in Table 2. Initially, the simulations were used to demonstrate that the component behaves electrically like a MOSFET transistor (Fig. 6). Subsequently, it was proven that discrete energy levels exist within the device (Fig. 10). At this stage of the work, there is an effort to demonstrate correct electrical behavior when an electric field is applied to the device (Figs. 7, 8 and 9). Fig. 5COMSOL workflow of studies of the MOSQWELL device. COMSOL workflow of studies of the MOSQWELL device. Additional considerations like strain, external fields and temperature can be considered. In the first case, one may need to consider the effects of strain on the band structure and energy levels. If there are external electric or magnetic fields, one can include them in the model to study their effects on the inter-sub-band transitions. Finally, temperature can affect the carrier distribution and the broadening of energy levels. By following these steps and leveraging the capabilities of COMSOL, one can effectively simulate ISBT in quantum wells and gain valuable insights into their properties and applications. Table 2Key parameters for simulations.ParameterDefinitionValue Geometry parameters for unidimensional quantum well (1D QW) simulations \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{t}_{GOX}$$\end{document}Gate oxide thickness25 nm \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{t}_{BOX}$$\end{document}Buried oxide thickness, reducing parasitic current71 nm \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{t}_{Si\:channel}$$\end{document}Silicon channel thickness, p-type2.4 nm \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{t}_{Si\:body}$$\end{document}Silicon substrate thickness (for simulation)500 nm Electrical parameters for simulations \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{V}_{GS}=\:{V}_{G}$$\end{document}Gate-Source applied voltageVarying \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{V}_{DS}=\:{V}_{D}$$\end{document}Drain-Source applied voltageVarying \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{V}_{th}$$\end{document}SOI Thin film threshold voltage [calculated]0.713 [V] \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{V}_{fb}$$\end{document}Flat band potential [calculated]0 [eV] \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\upphi _{f}$$\end{document}Fermi potential = kT ln(ni/Na) at equilibrium 300 K in p-doped silicon channel0.36 [eV] \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:q$$\end{document}Elementary charge1.602 10− 19 [C] \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{N}_{a}$$\end{document}Acceptor doping concentration (p-type) – channel1016 [cm⁻³] \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{N}_{d}$$\end{document}Donor doping – source-drain1019 [cm⁻³] \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{n}_{i}$$\end{document}Intrinsic carrier concentration (T = 300 K)1.08 × 1010 [cm⁻³] \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{C}_{gox}$$\end{document}Oxide capacitance [calculated]1.4 mF. m⁻2 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\upvarepsilon _{{ox}}$$\end{document}Oxide relative permittivity3.9 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\upvarepsilon _{0}$$\end{document} \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\upvarepsilon _{0}$$\end{document}Free space permittivity8.854 × 10− 12 [F/m] Table 2Key parameters for simulations.ParameterDefinitionValue Geometry parameters for unidimensional quantum well (1D QW) simulations \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{t}_{GOX}$$\end{document}Gate oxide thickness25 nm \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{t}_{BOX}$$\end{document}Buried oxide thickness, reducing parasitic current71 nm \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{t}_{Si\:channel}$$\end{document}Silicon channel thickness, p-type2.4 nm \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{t}_{Si\:body}$$\end{document}Silicon substrate thickness (for simulation)500 nm Electrical parameters for simulations \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{V}_{GS}=\:{V}_{G}$$\end{document}Gate-Source applied voltageVarying \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{V}_{DS}=\:{V}_{D}$$\end{document}Drain-Source applied voltageVarying \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{V}_{th}$$\end{document}SOI Thin film threshold voltage [calculated]0.713 [V] \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{V}_{fb}$$\end{document}Flat band potential [calculated]0 [eV] \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\upphi _{f}$$\end{document}Fermi potential = kT ln(ni/Na) at equilibrium 300 K in p-doped silicon channel0.36 [eV] \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:q$$\end{document}Elementary charge1.602 10− 19 [C] \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{N}_{a}$$\end{document}Acceptor doping concentration (p-type) – channel1016 [cm⁻³] \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{N}_{d}$$\end{document}Donor doping – source-drain1019 [cm⁻³] \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{n}_{i}$$\end{document}Intrinsic carrier concentration (T = 300 K)1.08 × 1010 [cm⁻³] \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{C}_{gox}$$\end{document}Oxide capacitance [calculated]1.4 mF. m⁻2 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\upvarepsilon _{{ox}}$$\end{document}Oxide relative permittivity3.9 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\upvarepsilon _{0}$$\end{document} \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\upvarepsilon _{0}$$\end{document}Free space permittivity8.854 × 10− 12 [F/m] Key parameters for simulations. Results Numerical classical results Classical output Since the correct functionality of a MOSFET device is evaluated using two types of characteristics – output and transfer curves – it is necessary to demonstrate the accuracy of the model by obtaining representative graphs. When applying positive voltage to the gate terminal, minority electrons from the substrate are attracted toward it and form an inversion layer, or channel, which allows the flow of IDS. The gate bias can be adjusted to change the resistance seen by IDS and thereby modulate its flow. In fact, while VGS is responsible for the quantity of electrons injected into the channel, VDS is responsible for their velocity from the source to the drain. Figure 6 presents the expected output characteristics. Fig. 6COMSOL simulations of the MOSQWELL electrical characteristics. (a) Output curves for VGS varying from 0 to 4 V at 300 K. COMSOL simulations of the MOSQWELL electrical characteristics. (a) Output curves for VGS varying from 0 to 4 V at 300 K. By adding a buried oxide layer underneath the Si junction, parasitic device capacitances can be lowered, resulting in a lower subthreshold swing and improved performance. Transistors of this kind are known as Silicon-On-insulator (SOI) transistors. Using Gate Recessed Channel (GRC) fabrication techniques, the BOX can reduce the channel width to nanometric dimensions, classifying the device as an Ultra-Thin Body (UTB) or Nanoscale Body (NSB), depending on the thickness value. Electrical field and electrical potential The 2D simulation of the electric potential in the channel is presented in Fig. 7. While VDS remains constant at 0.4 V, VGS varies from 0 V to 1 V. Accordingly the intensity of the field increases inside the channel. A zoom-in series is presented in Fig. 8. Fig. 7Simulation of the electrical potential between the source and the drain, as a function of both voltages. VDS = 0.4 V remains constant while VGS increases. (a) VGS = 0 V; (b) VGS = 0.2 V; (c) VGS = 0.4 V; (d) VGS = 0.6 V; (e) VGS = 0.8 V; (f) VGS = 1 V. Simulation of the electrical potential between the source and the drain, as a function of both voltages. VDS = 0.4 V remains constant while VGS increases. (a) VGS = 0 V; (b) VGS = 0.2 V; (c) VGS = 0.4 V; (d) VGS = 0.6 V; (e) VGS = 0.8 V; (f) VGS = 1 V. Fig. 8Zoom-in of the simulation of the electrical potential between the source and the drain, as a function of both voltages. VDS = 0.4 V remains constant while VGS increases. (a) VGS = 0 V; (b) VGS = 0.2 V; (c) VGS = 0.4 V; (d) VGS = 0.6 V; (e) VGS = 0.8 V; (f) VGS = 1 V. Zoom-in of the simulation of the electrical potential between the source and the drain, as a function of both voltages. VDS = 0.4 V remains constant while VGS increases. (a) VGS = 0 V; (b) VGS = 0.2 V; (c) VGS = 0.4 V; (d) VGS = 0.6 V; (e) VGS = 0.8 V; (f) VGS = 1 V. Numerical and analytical quantum results The challenge of this research remains the correct interactions between the QW structure (i.e. the silicon channel) and the MOSFET device. The step-by-step workflow enabled both an accurate model and a deep understanding of the interactions. Several series of simulations were performed to validate the presented, optimized model. The following are the main milestones. Activation of VGS voltage and band bending Following previous incomplete models, there was a need to model the QW in a manner that all the physical features will be present in one comprehensive model. The main goal of the study consisted in analyzing three separated models – the well bottom changing shape from rectangular to triangular, the bending of the GOX and BOX slopes, and the embedded wavefunctions – and in understanding the correct parts of each model to combine together. It can be observed that the advantages of both models correspond to all the behaviors that we expected to obtain. In fact, it can be said that the current stage was a process of customization until obtaining a correct physical result that matches our expectations. Figure 9(a) depicts the initial state of the device before the application of an external field. It is important to note that in this initial state our quantum well is rectangular, and there is no slope in any part of the device. Additionally, the energy levels within the quantum well are clearly visible. Describing this initial state allows us to analyze the other states where an electric field is applied more effectively. The sequence of graphs presented in Fig. 9b–d illustrates the device’s behavior under the application of different external electric fields while the gate voltage varies between the figures by increments of 1 V, up to a voltage of 3 V. As can be observed, as the gate voltage increases, the phenomena we discussed earlier become more pronounced. The quantum well gradually transforms from rectangular to triangular, the slope of the GOX section becomes increasingly positive (depending on the applied voltage), and the eigenvalues are affected by the changing slope (the BOX slope was presented in Fig. 2). Essentially, the simulation has succeeded in creating a physical model that exhibits the three effects predicted earlier to occur when applying an electric field to the device, causing a secondary effect of the lateral (source-drain) electric field. This is a significant advancement because the model now not only demonstrates the electric characteristics of a MOSFET transistor, as proven in previously published papers43,54,55 but also exhibits quantum behavior alongside the regular electrical behavior. Fig. 9The model is in operation mode with applied gate voltage. (a) VGS = 0 V; (b) VGS = 1 V; (c) VGS = 2 V; (d) VGS = 3 V. The model is in operation mode with applied gate voltage. (a) VGS = 0 V; (b) VGS = 1 V; (c) VGS = 2 V; (d) VGS = 3 V. As observed, all the expected features appear: the well changes from a rectangular well to a triangular one, the slope of the GOX starts to incline and the eigenvalues slightly decrease with the slope of the well. Schrödinger-Poisson simplified model for MOSQWELL When an external voltage is applied, an electric field is generated across the quantum well, modifying its electronic and optical properties56. The impact of an external voltage on the QW, confined in z direction, can be understood through the simplified 1D time independent Schrödinger–Poisson equations. The electron concentration in the n-channel silicon depends on the applied voltage through the electrical potential \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\upvarphi\:}\left(z\right)$$\end{document}, which is directly related by the space-charge density \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\rho\:\left(z\right)$$\end{document} described by Poisson’s Eq. (4)(5)57.4\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\frac{{\partial\:}^{2}{\upvarphi\:}\left(z\right)}{\partial\:{z}^{2}}=-\frac{\rho\:\left(z\right)}{{\epsilon\:}_{Si}}$$\end{document} Where:5\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\rho\:\left(z\right)=q(p\left(z\right)-n\left(z\right)+{N}_{d}\text{}\left(z\right)-{N}_{a}(z\left)\right)$$\end{document} In the undoped (intrinsic) semiconductor, n = p = ni and Nd = Na= 0, so the space-charge density is zero. In a strongly inverted n-channel, the electron concentration n is much larger than Na (so p < < n while Nd = 0), and so the space-charge density is no longer zero and is related to the wave function as shown below (5)57.6\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\rho\:\left(\mathrm{z}\right)=-qn\left(z\right)=-q\sum\:_{i=1}^{n}f({E}_{i},{E}_{F}){\left|{\psi\:}_{i}\left(z\right)\right|}^{2}$$\end{document} Where f is the Fermi-Dirac distribution, EF is the Fermi level in an n level quantum well and q is the elementary charge. Within the biased quantum well, the electron state in a given energy level Ei along the confined direction z is described by the stationary Schrödinger-Poisson Eq. (7)8.7\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:-\frac{{{\hslash\:}}^{2}}{2{m}^{\mathrm{*}}}\left(\frac{{\partial\:}^{2}{\psi\:}_{i}\left(z\right)}{\partial\:{z}^{2}}\right)+{V}_{z}{\psi\:}_{i}\left(\mathrm{z}\right)={E}_{i}{\psi\:}_{i}\left(z\right)$$\end{document} Where:8\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:V_{z} = - q\upvarphi \left( z \right) + \Delta {\mathrm{E}}_{{\mathrm{c}}} ({\mathrm{z}})$$\end{document} Here, Vz represents the sum of the conduction band-edge offset energy of the Si-SiO2 heterostructure, and the electrostatic electron energy \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:-\boldsymbol{q}\boldsymbol{\varphi\:}\left(\boldsymbol{z}\right)$$\end{document} (8). The applied (external) voltage enters the equation through φ(z), directly modifying the potential energy term and, consequently, the energy eigenvalues and wavefunctions58. Although the gate voltage (VGS) remains constant, the total applied voltage varies along the n-channel as a function of position and distance from the voltage sources (VGS and VDS). This variation arises from the nanometric dimensions of the silicon channel43, as simulated in this article. Activation of VGS and VDS voltage simultaneously and band bending This section presents the dependance of the QW eigenvalues in different channel locations: near the source, in the middle of the channel, and near the drain. To simulate the VDS effect on the bending, VGS and VDS are both set to 1[V]. Figure 10 shows the electrical potential distribution all along the channel. Fig. 10Simulation of the electrical potential between the source and the drain, as a function of voltages VGS and VDS both set at the value of 1 V. (a) MOSQWELL electric potential distribution; (b) Si channel zoom in. Simulation of the electrical potential between the source and the drain, as a function of voltages VGS and VDS both set at the value of 1 V. (a) MOSQWELL electric potential distribution; (b) Si channel zoom in. Figure 10 demonstrates how the electrical potential is not uniformly distributed across the silicon channel, gate oxide (GOX), and buried oxide (BOX). The electrical potential varies depending on the location relative to the voltage source (VGS or VDS). This observation implies that the quantum well (QW) eigenvalues vary along different positions within the channel according to the Schrödinger-Poisson equation presented above. To investigate this distinctive characteristic of the MOSQWELL device, three sectional analyses were performed, as shown in Fig. 11(a). The locations of the sectional cuts are defined by their distance from the voltage origin: near the source, mid-channel, and near the drain. Each sectional cut shows the potential distribution across the gate oxide (GOX), the thin silicon channel, and the buried oxide (BOX) at the corresponding location. As shown in Fig. 11(b), the potential distribution across the channel varies with position along the channel, indicating the presence of a continuous multiple-quantum-well structure. Due to the position-dependent electric potential and its variation with distance from the applied voltage sources along the channel, the local electrostatic environment becomes non‑uniform. This spatial variation modifies the potential energy landscape experienced by carriers in the quantum well, leading to shifts in the absolute energies of the bound states, as demonstrated by the simulation results in Fig. 9 and described by Eq. (8). Fig. 11Three cut sections, near source, middle, near drain; (a) Cuts location; (b) Electric potential distribution in section cuts near the source terminal, in the middle of the channel, and near the drain terminal. Three cut sections, near source, middle, near drain; (a) Cuts location; (b) Electric potential distribution in section cuts near the source terminal, in the middle of the channel, and near the drain terminal. The Schrödinger-Poisson COMSOL analysis results are presented in Figs. 12 and 13, in which the eigenvalues’ variation is investigated at three different locations. Fig. 123 Eigen values and wave functions at 3 different locations across the Si channel. (a) Near the source; (b) at the middle point; (c) near the drain. 3 Eigen values and wave functions at 3 different locations across the Si channel. (a) Near the source; (b) at the middle point; (c) near the drain. Fig. 13Comparison chart of four embedded eigenvalues (E1-E4) at different channel locations at VDS = VGS = 1 V in a 2.4 nm thick device. Comparison chart of four embedded eigenvalues (E1-E4) at different channel locations at VDS = VGS = 1 V in a 2.4 nm thick device. Activation voltage influence on eigenvalues and voltage-tunable band gap It appears that the eigenvalues are strongly influenced by the effective voltage experienced within the quantum well, enabling voltage-tunable artificial bandgap transitions (ISBT). We argue that at the 1–2 nm scale, even sub-angstrom variations in channel thickness can lead to significant shifts in the energy eigenvalues and the resulting inter-sub-band transition (ISBT) wavelengths. The practical compensation mechanism relies on the Quantum-Confined Stark Effect (QCSE) and electrostatic field modulation to realign energy levels that have shifted due to fabrication tolerances. The electrical bias induces a spatial separation of the electron wave functions and shifts their energy levels. By increasing the gate bias, the effective “bottom” of the well is lowered, allowing it to “pull” the eigenvalues back to their target positions59. In MOS-type structures, the gate voltage also controls the carrier density. Higher carrier densities increase depolarization shifts, which blue-shift the ISBT resonance60. This allows for a second “knob” to counteract red-shifts caused by a well that was fabricated slightly too wide. The relationship between applied bias and the resulting radiative transitions can be quantified through self-consistent Poisson-Schrödinger simulations. Post-assembly optimization using CMOS-standard voltages (e.g., 5–6 V) has demonstrated the ability to compensate for structural drifts and misalignments in silicon photonic platforms61. Permitted optical transitions Following Fig. 13, and looking forward to the expected radiative ISBT, it is important to remember that there are transition rules which restrict the allowed transitions. The parity should preserve odd numbers when calculating the difference ΔE between two energy levels. In other words, Δn = 1, 3, 5… due to the symmetry of the QW potential. In our case, the corresponding values of the authorized ISBT, a.k.a. “artificial bandgaps”, are presented in Table 3, and a visual representation of the emitted wavelengths is presented in Fig. 14. While for low energy levels, the relative shift is not significant (0% to 1%), it appears that for higher energy levels and transitions (E4-E3), it increases to more than 8%. If in most of the transitions (E2-E1, E3-E2, and E4-E3), the optical emission wavelengths correspond to the short-wavelength infrared (SWIR) range (900–2,500 nm), mainly used for optical communication, one can also identify an optical transition in the visible range (E4-E1). The enhanced relative energy shift observed for higher-order transitions (e.g., E2–E1, E3–E2, E4–E3) arises from the intrinsic properties of quantum well eigenstates under a triangular potential, as described by analytical Schrödinger–Poisson models. In this framework, the sub band energies scale with the local surface electric field as mentioned62 in Eq. 11, where j is sub band index. As a result, higher-order eigenstates exhibit an amplified sensitivity to variations in the electrostatic field. The control of light emission in low-dimensional semiconductor structures relies on the manipulation of quantum confinement and carrier dynamics. In III-V systems, this is often achieved through the Quantum-Confined Stark Effect (QCSE), while in silicon photonics, it involves field-effect modulation within gated structures like MOSQWELL. Electrical biasing allows for real-time tuning of emission intensity and wavelength by modifying the internal electric fields and carrier populations. In III-V multi-quantum well (MQW) structures, applying a bias shifts the energy levels of electrons and holes. This results in a “redshift” or “blueshift” of the peak emission wavelength depending on the polarity and internal piezoelectric fields63. Beyond vertical biasing, lateral electric fields from side gates can manipulate the Stark shift in single quantum dots for electroluminescence (EL) control64. Another study in InGaN LEDs with various QW thicknesses, showed a blue shift of the emission spectra which stemmed from the presence of built-in polarization and differing carrier densities within the quantum wells. Specifically, in thinner quantum wells, the carrier density starts at a lower level but rises alongside the current density; this process leads to the screening of the internal electric field, which ultimately causes a blue shift in the emission spectrum65. Table 3ISBT transitions expressed in ΔE (eV) and emission wavelengths λ (nm).Relative shift (%)Near SourceMiddleNear DrainΔE allowed (eV)Δn0.11%0.5992260.597932390.598887E2–E110.00%1.005691.005631.0057E3–E212.09%2.1069162.196762392.176187E4–E138.33%0.5020.59320.5716E4–E31Relative shift (%)Near SourceMiddleNear Drainλ emitted (nm)Δn0.11%206920742071E2–E110.00%123312331233E3–E212.09%589564570E4–E138.33%247020902169E4–E31 Table 3ISBT transitions expressed in ΔE (eV) and emission wavelengths λ (nm).Relative shift (%)Near SourceMiddleNear DrainΔE allowed (eV)Δn0.11%0.5992260.597932390.598887E2–E110.00%1.005691.005631.0057E3–E212.09%2.1069162.196762392.176187E4–E138.33%0.5020.59320.5716E4–E31Relative shift (%)Near SourceMiddleNear Drainλ emitted (nm)Δn0.11%206920742071E2–E110.00%123312331233E3–E212.09%589564570E4–E138.33%247020902169E4–E31 ISBT transitions expressed in ΔE (eV) and emission wavelengths λ (nm). Fig. 143D visualization of the non-uniformity of the localized light emission wavelengths along the channel length. 3D visualization of the non-uniformity of the localized light emission wavelengths along the channel length. Discussion Electrical analysis The main goal of this research was to create a correct model that would work as expected and include the changes obtained when applying the voltage VGS. In the earlier parts of this study, many simulations were performed to show that the component behaves electrically like a MOSFET transistor. The measurement characteristic graphs of the MOSQWell component correspond to the known characteristic graphs of the MOSFET transistor66,67. If in the past, it was already shown that discrete energy levels of eigenvalues exist in the well60,68,69, today one can observe varying values along the channel. Another accomplishment was to simulate two electric fields at the same time by applying VGS voltage and VDS voltage. The results obtained from this simulation confirmed the correct electrical behavior of the device. In fact, only the electrical potential in the p-type Si substrate is shown as function of both VGS and VDS (Fig. 10). In our model, the change in BOX and GOX slopes over all the structure as a function of the applied external field was accounted for, as Eqs. (7)-(9) account for it analytically. The “final model” based on Eqs. (7)-(9) is presented in Fig. 9. Developing equations which incorporate both factors will create a highly reliable model from the perspective of the physical theory of quantum wells. Controlled tuning of artificial band gap to compensate for differential channel thickness As shown in Figs. 13 and 14, the eigenvalues are closely related to the actual activated voltage experienced by the quantum well (QW). This unique property of the MOSQWELL structure enables the formation of artificial band gaps based optical transitions (ISBT), whose values can be tuned through the applied voltage. Indeed, since the quantum well channel thickness controls the radiative wavelength43,60 and since a thickness of 1–2 nm is difficult to fabricate70, the option of fine tuning of the radiative transition using the voltage is a good compensating opportunity. Figure 15 presents several structural results (HRTEM, EDS, and interferometry) showing non-uniform silicon layer thickness. Fig. 15Structural characterization of the channel silicon thickness distribution. (a) HRTEM cross-section of the layers at the gate location; (b) EDS result showing over oxidation at channel extremity; (c) 3D map of the SOI layer distribution after GRC oxidation process. The initial thickness of the SOI was 46 nm. Wafer diameter in 2”. Structural characterization of the channel silicon thickness distribution. (a) HRTEM cross-section of the layers at the gate location; (b) EDS result showing over oxidation at channel extremity; (c) 3D map of the SOI layer distribution after GRC oxidation process. The initial thickness of the SOI was 46 nm. Wafer diameter in 2”. Using our voltage-biased approach, we argue non-uniformity can be compensated by applying specific voltages. By utilizing the Quantum-Confined Stark Effect (QCSE), we can dynamically engineer the effective bandgap of a heterostructure. For instance, by applying a specific bias voltage to a silicon-based quantum well, we can tune the ISBT to 0.67 eV effectively emulating the optical application of germanium in the NIR domain within a Silicon platform. Possible location and parameters of light emission While emission is primarily localized near the drain – suggesting minimal influence from eigenvalue variations in the source or mid-channel – electrical compensation remains critical for addressing thickness non-uniformity across devices from the same batch. Beyond analyzing thickness dependency, it is essential to characterize how these eigenvalues can be dynamically retrieved or tuned by modulating drain and gate voltages. To achieve the required emission intensity, several critical criteria must be met, primarily: A first condition would be smooth interfaces between the SiO2/Si/SiO2 layers of the quantum well. Technology has progressed, and today smooth interfaces71 can be made by ALD72 with uniform layers73, and are used in gate oxide and all-around gate devices. Moreover, new materials such as High-K dielectric material74 have already replaced the classic oxide layer. Nevertheless, for a given batch, non-uniformity can still be a problem when comparing between devices.The second one would be to use a long (L) and large (W) channel structure, since we need a large number of electrons inside the channel per unit of surface (W × L). In other words, a high concentration is desirable75,76 so the surface should be large, with W being more important than L in this case. The more electrons, the more photons; this is linked to the expression W × L × Cox (VGS - VTH) for the total charge inside the channel.Hot electron generation is essential to reach the higher energy sub-levels in the conduction band to allow spontaneous transitions. For this purpose, while VGS is responsible for the number of injected electrons inside the channel, VDS is responsible for their velocity. Near the drain (pinch-off) the probability of light emission is higher77, since radiative recombination may occur there. Indeed, in MOS transistors, pinch-off luminescence, also known as hot-carrier electroluminescence, may occur when a MOSFET, operating in the saturation region (VDS \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\ge\:$$\end{document} VGS - VTH), emits visible or near-infrared light from its channel. The relaxation or recombination of “hot” electrons, which have gained substantial energy from the intense electric fields in the pinch-off region close to the drain is the main cause of this light emission. In such cases, the usage is for integrity check of the device77, since causing damage78.Light emission may also occur away from the drain, all along the channel, as argued here. In fact, one can expect not only monochromatic light emission, but also several peaks of higher intensity with different wavelengths. Of course, hot carrier injection can also cause leakage/break in oxide, and we may be careful with VDS applied values.In a VLSI perspective, one can expect fluctuations from one device to another one due to non-uniform thicknesses. In particular, the critical dry oxidation step involves non-uniformities, so an interesting point would be to correct from one device to another one in the same wafer, while we will recalibrate the voltages per identified needs. This is why the approach of applied voltage compensation is acceptable as an efficient solution for manufacturing errors. In future investigations and publications, two additional domains will be addressed: Process optimizations and device validation. Regarding fabrication, these investigations will include the definition of the process tolerances, reproducibility, and compatibility with current CMOS fabrication nodes. Regarding the validation, again combined and complementary numerical and analytical analyses will be checked versus experimental characterization. A first condition would be smooth interfaces between the SiO2/Si/SiO2 layers of the quantum well. Technology has progressed, and today smooth interfaces71 can be made by ALD72 with uniform layers73, and are used in gate oxide and all-around gate devices. Moreover, new materials such as High-K dielectric material74 have already replaced the classic oxide layer. Nevertheless, for a given batch, non-uniformity can still be a problem when comparing between devices. A first condition would be smooth interfaces between the SiO2/Si/SiO2 layers of the quantum well. Technology has progressed, and today smooth interfaces71 can be made by ALD72 with uniform layers73, and are used in gate oxide and all-around gate devices. Moreover, new materials such as High-K dielectric material74 have already replaced the classic oxide layer. Nevertheless, for a given batch, non-uniformity can still be a problem when comparing between devices. The second one would be to use a long (L) and large (W) channel structure, since we need a large number of electrons inside the channel per unit of surface (W × L). In other words, a high concentration is desirable75,76 so the surface should be large, with W being more important than L in this case. The more electrons, the more photons; this is linked to the expression W × L × Cox (VGS - VTH) for the total charge inside the channel. The second one would be to use a long (L) and large (W) channel structure, since we need a large number of electrons inside the channel per unit of surface (W × L). In other words, a high concentration is desirable75,76 so the surface should be large, with W being more important than L in this case. The more electrons, the more photons; this is linked to the expression W × L × Cox (VGS - VTH) for the total charge inside the channel. Hot electron generation is essential to reach the higher energy sub-levels in the conduction band to allow spontaneous transitions. For this purpose, while VGS is responsible for the number of injected electrons inside the channel, VDS is responsible for their velocity. Near the drain (pinch-off) the probability of light emission is higher77, since radiative recombination may occur there. Indeed, in MOS transistors, pinch-off luminescence, also known as hot-carrier electroluminescence, may occur when a MOSFET, operating in the saturation region (VDS \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\ge\:$$\end{document} VGS - VTH), emits visible or near-infrared light from its channel. The relaxation or recombination of “hot” electrons, which have gained substantial energy from the intense electric fields in the pinch-off region close to the drain is the main cause of this light emission. In such cases, the usage is for integrity check of the device77, since causing damage78. Hot electron generation is essential to reach the higher energy sub-levels in the conduction band to allow spontaneous transitions. For this purpose, while VGS is responsible for the number of injected electrons inside the channel, VDS is responsible for their velocity. Near the drain (pinch-off) the probability of light emission is higher77, since radiative recombination may occur there. Indeed, in MOS transistors, pinch-off luminescence, also known as hot-carrier electroluminescence, may occur when a MOSFET, operating in the saturation region (VDS \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\ge\:$$\end{document} VGS - VTH), emits visible or near-infrared light from its channel. The relaxation or recombination of “hot” electrons, which have gained substantial energy from the intense electric fields in the pinch-off region close to the drain is the main cause of this light emission. In such cases, the usage is for integrity check of the device77, since causing damage78. Light emission may also occur away from the drain, all along the channel, as argued here. In fact, one can expect not only monochromatic light emission, but also several peaks of higher intensity with different wavelengths. Of course, hot carrier injection can also cause leakage/break in oxide, and we may be careful with VDS applied values. Light emission may also occur away from the drain, all along the channel, as argued here. In fact, one can expect not only monochromatic light emission, but also several peaks of higher intensity with different wavelengths. Of course, hot carrier injection can also cause leakage/break in oxide, and we may be careful with VDS applied values. In a VLSI perspective, one can expect fluctuations from one device to another one due to non-uniform thicknesses. In particular, the critical dry oxidation step involves non-uniformities, so an interesting point would be to correct from one device to another one in the same wafer, while we will recalibrate the voltages per identified needs. This is why the approach of applied voltage compensation is acceptable as an efficient solution for manufacturing errors. In future investigations and publications, two additional domains will be addressed: Process optimizations and device validation. Regarding fabrication, these investigations will include the definition of the process tolerances, reproducibility, and compatibility with current CMOS fabrication nodes. Regarding the validation, again combined and complementary numerical and analytical analyses will be checked versus experimental characterization. In a VLSI perspective, one can expect fluctuations from one device to another one due to non-uniform thicknesses. In particular, the critical dry oxidation step involves non-uniformities, so an interesting point would be to correct from one device to another one in the same wafer, while we will recalibrate the voltages per identified needs. This is why the approach of applied voltage compensation is acceptable as an efficient solution for manufacturing errors. In future investigations and publications, two additional domains will be addressed: Process optimizations and device validation. Regarding fabrication, these investigations will include the definition of the process tolerances, reproducibility, and compatibility with current CMOS fabrication nodes. Regarding the validation, again combined and complementary numerical and analytical analyses will be checked versus experimental characterization. Concept implementation in other technologies In this research, we have considered a SiO2-Si-SiO2 quantum well, with a stack of GOX-Si-BOX, and where the gate oxide layer thickness is 25 nm. As mentioned above, the main reason for this choice was to smoothly integrate into CMOS technology. Looking forward, one can ask if the gate oxide considered thickness of 25 nm will be applicable to more recent technology nodes. In other works, we may think about another kind of quantum well, made of High-K-Si-SiO2, while High-K dielectric layer (HfO2, Al2O3, …) replaces the gate oxide. In such a case, the structure and energy picture of the full stack will be made of an upper metal gate, High-K dielectric layer, thin Si quantum well, SiO2 (BOX) and substrate / back gate. Here also, the Si layer forms the quantum well because carriers are confined vertically by oxide barriers and band offsets at dielectric interfaces. Electrons in the Si layer occupy quantized sub-bands: \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{E}_{1},{E}_{2},{E}_{3},\dots\:$$\end{document} Of course, the design of the full stack (materials and thicknesses) is important; however, one may pay attention first to the work function of the new QW. In a High-K–Si–SiO2 QW, since the dielectric stack is asymmetric, the electric displacement differs across interfaces, the potential drop is uneven, and the confinement potential becomes triangular/trapezoidal instead of nearly square. In such a case, the Schrödinger equation becomes:9\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\left[-\frac{{{\hslash\:}}^{2}}{2{m}^{*}}\frac{{d}^{2}}{d{z}^{2}}+V\left(z\right)\right]{\psi\:}_{n}={E}_{n}{\psi\:}_{n}$$\end{document} Where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:V\left(z\right)$$\end{document} is now strongly asymmetric. Regarding the band offset considerations, the confinement strength depends on the conduction-band offsets. The typical values for interfaces, ΔEc are Si / SiO2 ~3.2 eV, Si / HfO2 ~1.5 eV and Si / Al2O3 ~2.8 eV. Smaller offset means weaker confinement barrier, more possible tunnelling leakage, and broader wavefunction tails. Thus, some high-K materials may reduce the confinement quality. This is why most advanced devices use today a combined solution of hybrid High-K/SiO2 gate stacks (HKMG): Metal, High-K, ultra-thin SiO2 interfacial layer, and Si quantum well. The thin SiO2 layer’s role is to preserve the interface quality and to reduce the trap density. In parallel, the high-K layer boosts capacitance. Such hybrid stacks are now standard in FinFETs79 and GAA FETs79,80, as well as in silicon qubit research CMOS platforms. Non-idealities phenomena and influence In such a kind of MOSFET transistor, where the channel is an embedded SiO2-Si-SiO2 quantum well, the idealized optoelectronic response is usually derived from a scenario assuming perfectly confined carriers, ballistic transport, discrete sharp sub-band energies, negligible scattering, and equilibrium lattice conditions. The dominant non-idealities, such as phonon scattering, carrier-carrier interactions, interface disorder, and temperature effects, may strongly influence and modify both the electronic structure and the optical emission. Robustness of the method – preliminary thoughts In this preliminary research, we propose a voltage compensation concept to mitigate fabrication-induced variations in channel thickness. Of course, further investigations are necessary in the future to evaluate the physical limits of this compensation strategy. The robustness of the proposed method can be challenged by several physical conditions and phenomena such as interface roughness, oxide traps, random dopant fluctuations, and more. It is also possible that beyond some range, the activation voltages (VGS, VDS) tuning will become ineffective. To answer such reasonable concerns, a second generation of fabricated devices will be necessary as well as diverse characterizations: structural, electrical, and optical. We also, believe that improved process – for example, by using atomic layer deposition (ALD) technique instead of gate-recessed channel (GRC) oxidation – can solve many problems by obtaining uniform interface between Si and oxide layers81,82. Until then, we can briefly suggest here some dependencies between the key parameters, and their influence on the model, since such elaborations may require a separated future publication. In an embedded SiO2-Si-SiO2 quantum well MOSFET, fabrication-induced variations in Si channel thickness (tSi) produce threshold-voltage (Vth) and mobility fluctuations because quantum confinement energy depends strongly on well thickness. Voltage compensation (gate bias, body bias, adaptive substrate bias, etc.) can mitigate these effects only within fairly strict limits. As a first shot, one can think about the following main limits: Quantum confinement nonlinearity, carrier mobility degradation (traps), leakage and oxide field constraints, and subthreshold slope degradation. Research on double-gate and ultrathin-body MOSFETs has reported that once the Si quantum well thickness drops below roughly 4–5 nm, variability rises sharply because confinement energy becomes extremely sensitive to thickness fluctuations83. Typical sensitivity values reported in fully depleted SOI / quantum-well-like devices are on the order of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:10\mathrm{\--}50\mathrm{\:mV/nm}\text{}$$\end{document} threshold shift for moderate body thicknesses, and potentially much larger below ~ 5 nm. Experiments on SIMOX SOI devices observed ~ 100 mV threshold variation from ± 15 nm film-thickness variation. The compensation voltage range is fundamentally limited because increasing gate or body bias introduces additional penalties. Again, large literature review can bring legitimate concerns, and of course after this concept preliminary research, further investigations and publications on improved processed devices will finalize the range of the limits in this novel method. Conclusion If the MOSQWELL transistor itself shows significant progress towards the integration of silicon photonics into nanoelectronics circuitry, the voltage-biased compensation of manufacturing errors represents a significant leap forward in the nanoscale industry. The distribution of the eigenvalues along the channel structure implies that different energy levels can be obtained as a function of the location inside the channel and can also be electrically controlled and adapted to the needs of the device specifications. For quantum wells, at VLSI scale, the capability to correct non-uniformities in physical thicknesses is a great advantage, preventing the disqualification of devices excluded from the authorized range and improving process yield. Using only the VGS and VDS activation voltages, this method brings significant cost reduction – there is no need for new processing or batch – and also electrical control of mismatched physical parameters.