User:Giulio.galderisi/sandbox

NOTE ON EDITING
To all readers and writers: from now on I will use the history "tool" as intended and give a little description of the adopted changes every time a significant change is made. If you want to edit something you might want to check if the edit was already done, and if you want to make some changes you can also write a very small description of what was changed. This will make us work way more quickly! Thank you :)

= RFET = The reconfigurable field effect transistor (RFET), is a field effect transistor whose conduction mechanism can be reversibly reconfigured.

The majority of reconfigurable FETs available today are polarity control reconfigurable FETs (PC-RFETs), that can be switched between n-type and p-type operation modes. This class of devices combines both functionalities, that are usually realized by two different devices, in a single one. Such an RFET is usually controlled by at least two independent gates: one is used to select the kind of charge carrier (electrons or holes) and the second one modulates the channel conductance and so the amount of current. In comparison, a classical MOSFET can be switched on and off but it is fixed to either n-type or p-type operation by the underlying fabrication process. Although RFET technology doesn't aim to replace standard CMOS technology, it comprises interesting features that can lead to the coexistence of both in specific fields of application.

RFETs have been proposed or realized based on various architectures: they all share the possibility to use both valence and conduction band of the semiconducting channel for operations. Injection can happen from a metal into the semiconductor (Schottky barrier) or from one band of the semiconductor to the other (band-to-band tunneling).

The term RFET includes a broad family of devices that can be reconfigured using novel approaches. Since the first attempts to realize such a device were made by modulating the Schottky barrier at the metal/semiconductor interface of Schottky barrier FETs, it is possible to differentiate a first classification of RFETs between polarity-controllable Schottky barrier field effect transistors (PC-SBFETs) and non-Schottky barrier based devices. The former approach includes devices like electrostatically doped FETs, bias-induced doping FETs, filtered ambipolar operation FETs, independent control of carrier injection FETs, and more. In the latter approach are included devices like tunnel field effect transistors (TFETs), electron-hole bilayer tunnel field effect transistors (EHB-TFETs), single electron spin transistors (SESTs) and more.

Background
With the continued scaling of device sizes according to Moore`s Law, the stable and reliable operation of MOSFETs is challenged by the nanoscale device dimensions. For example, doping by incorporation of chemical impurities, and especially the precise control of the dopant fluctuations became increasingly difficult for ultra-scaled technology nodes. Since the first Schottky-barrier-based RFET did not require any channel doping, but relied only on the electrostatic doping induced by the structure with multiple independent gates, it was considered a promising device to overcome the issues related to aggressive scaling. In addition, the added functionality provided by the reconfigurability holds the promise to increase the system functionality, even without scaling the individual device sizes.

Early phase
In 2001, H.C. Lin, K.L. Yeh, R.G. Huang, C.Y. Lin and T.Y. Huang realized a polysilicon-based TFT showing an ambipolar operation mode with the possibility to suppress p-type or n-type conduction accordingly to the combination of the applied gate voltages. Even though with this device such a feature was demonstrated for the first time, due to the very special target application the potential of the technology was still not recognized in the scientific community.

A couple of years later, in 2004, ''Y.-M. Lin, J. Appenzeller and P. Avouris'' from the IBM T.J. Watson Research Center realized a reconfigurable FET based on a carbon nanotube as channel material. Exploiting the exceptional transport properties of CNTs that were arising in those years a very small inverse subthreshold swing of 63 mV/dec and a high on-state current, were achieved.

In 2008, for the first time, silicon was used as channel material by W. Weber et al. to realize a concept of polarity tunable device. Bottom-up grown silicon nanowires were fabricated, and the modulation of the silicidated interfaces between the channel and the nickel source/drain contacts, lead to the observation of both n- and p-type behavior in the same device.

Device outgrowth
The name "Reconfigurable Field Effect Transistor" (RFET) itself was at first introduced by Heinzig et al. in 2012 in a work employing silicon nanowires as channel material. Here, many technology enablers like silicidated source/drain contacts with mid-gap band alignment and a Si-SiO2 core-shell structure minimizing hysteresis effects were employed to realize a record on/off current ratio up to 109. The device was refined just one year later, showing the necessary prerequisite for the use of reconfigurability in CMOS-based circuits, which is the symmetry between n-type and p-type conduction mechanisms. Employing oxidation-induced mechanical stress, equal p- and n-type currents have been achieved in a structure needing only two distinctive voltage potentials. The operation was verified in a reconfigurable complementary inverter circuit. From that moment, a broad range of device concepts have been implemented employing different materials, all featuring low dimensionality, ranging from 1-D, like nanowires and nanotubes, to 2D (2D materials or quasi-2-D industrial platforms like FinFETs or FDSOI).

1-D devices
Many efforts have been made in the past decade to employ silicon as channel material for such reconfigurable devices. In particular, silicon nanowires (SiNWs) have been employed by Heinzig et al. , De Marchi et al., Wessely et al. and Mongillo et al. to realize scalable, low-dimensional polarity control devices. Also germanium nanowires have been employed by Trommer et al. , exploiting the smaller band gap of such material with respect to silicon.

2-D devices
Nakaharai et al. showed in 2012 that even graphene, despite the absence of a band gap, can be used to realize polarity control devices exploiting a double gate structure. To realize such a device employing a material that doesn't posses a band gap, additional process steps are needed. For this reason, the same author later realized a RFET based on semiconducting MoTe2 monolayers. Another RFET realized employing a transition metal dichalcogenide (TMDC) 2D channel material (WSe2) was implemented in 2016 by G. Resta et al.. Lately, also black phosphorus has been employed in the realization of polarity controllable devices by P. Wu et al..

Industry-oriented platforms
With the successful demonstration on nanostructures, research also started to employing more mature devices architectures like FinFETs or FDSOI channels, e.g. by Zhang et al. and Krauss et al. .

Other material systems
To extend the field of application also to mechanically flexible substrates, also organic materials have been used to realize polarity control devices such as split-gate organic FETs from B.Y. Hsu et al. (2010) and ambipolar organic transistors from H. Yoo et al. (2016). In this last work also logic gates were experimentally demonstrated.

RFET concepts and working principles
The reconfigurability of RFETs can be realized by different mechanisms. In most RFET concepts, a modulation of the Schottky barrier at the interface between semiconductor and metal is used but in some embodiments band-to-band tunneling is utilized instead.

Schottky-barrier-based devices
A Schottky barrier is created when a semiconductor is put in contact with a metal, under certain boundary conditions. Depending on both the semiconductor and metal work functions as well as the thickness of the barrier itself, the created junction can have a rectifying behavior or not. If a heterostructure with two Schottky barriers is assembled, and both junctions are steered by a controlling gate, one can obtain a device called Schottky barrier FET (SB-FET). By adding a second gate to the Schottky barrier FET and steering the two resulting gates independently, one can then control the polarity of the device obtaining unipolar transfer characteristics with proper switch-off currents then suppressing the undesired flow of opposite charges in the off-state. Thus, the so called program gate (or polarity gate, PG) controls the kind of carrier that is allowed to generate a current, while the control gate (CG) actually switches on and off the transistor. Different positioning of the second gate and overall steering of the two lead to different RFET concepts and working principles that are going to be explained hereafter.

Electrostatic doping
This was the first approach used to reconfigure a SBFET, and it introduces the presence of a back gate that overlaps both Schottky barriers. This gate is used as program gate to select the polarity of the device by bending the band diagram of the structure. Since the back gate covers the whole structure, positive (negative) voltages allow an injection of electrons (holes) in the entire channel thus electrostatically doping the device itself. The top control gate modulates an additional energy barrier switching on and off the transistor. This is an example of dependent Schottky barriers gating, since both junctions are steered by the same gate.

Filtered ambipolar operation
By reversing the roles of control and program gates from the previous structure, one can obtain a different and alternative concept that is still an example of dependent Schottky barrier gating. In this case both carriers can be injected in the channel during the on-state operations and a potential barrier is formed in the middle of it when a voltage is applied. Thus the buried control gate provides an ambipolar behavior when steered while the top program (polarity) gate blocks the undesired charge injection. This two concepts may appear very similar, if not equivalent: in reality, the electrostatic doping concept does not have an ambipolar behavior since the polarity of the device is programmed directly at the Schottky junctions, while in the filtered ambipolar operation concept both carriers can be injected but one type is blocked in the middle of the channel by the band bending provided with the top program gate.

Schottky barrier biasing
In order to simplify the structure and to provide a better gating control through potentially thinner oxides, the back gate can be replaced by two additional top gates placed above the Schottky junctions. By steering these two gates as polarity gates with the same fixed voltage applied, one can control the type of carrier injected at the junctions while modulating the channel conductance through the control gate, placed like before in the middle of the device. Main benefit over the electrostatic doping concept is, that the gates do not compete with each other regarding the control of a certain channel region. As a result it shows itself to be more performing since lower voltages are required to operate it. Also implementation simplifies the device fabrication as all gates are patterned from the front.

Independent control of carrier injection
In this approach, differently to the previous ones, the Schottky junctions are steered independently: the program gate overlaps the drain-sided barrier and sets the device polarity of the device by blocking the undesired carrier type, while the control gate at the source-sided junction actually controls the carrier flow of the other carrier type through the channel. It is interesting to note that in this concept the channel is mostly ungated, since the two gates cover only the junctions. Moreover, potential accumulation of charge in the channel is totally avoided with this structure, unlike the previous concepts.

Multiple independent gates
The benefits of the two latter concepts can be combined in a single device with multiple independent gates. For example a three independent gates RFET (TIG-RFET), albeit structurally identical to the SBB-RFET concept, is programmed on the drain-sided barrier alone. The other two gate can be used to turn the transistor on or off. In such a structure, a current flow is actually allowed only if all the gates are biased to be 'open'. This enables improved functionalities: for example an AND-wired logic gate built over a single transistor has been demonstrated. In addition, a steeper subthreshold slope is achieved with the middle gate with respect to the source-sided gate. This is because the carriers are already injected through the Schottky barrier when the transistor switches, unlike in the previous concept where the switching was controlled directly at the junction by the only control gate. The effect can be exploited on the circuit level for power saving techniques.

Non-Schottky barrier approaches
All the concepts presented so far are based on the Schottky barrier modulation that can induce the selective blocking of one kind of charge carrier realizing in this way the device reconfigurability. There are though many other approaches to realize a RFET that are going to be presented hereafter.

TFET-based polarity control devices
Reconfigurable devices have been proposed and realized also employing TFETs (tunnel field effect transistors) exploiting band to band tunneling. In the first work a double gated structure is employed in order to steer the injection of carriers due to BTBT suppressing at the same time the ambipolar behavior of the device. Since the conduction mechanism does not relies on thermally generated carriers and there is no Schottky barrier between the carriers reservoirs and the channel, very steep subthreshold slopes are foreseen. In any case, this kind of device suffers from low values of on current, that make it slow during logic switching events. Also, heavily doped contact regions are needed in order to obtain band to band tunneling. This kind of device is generally composed by two highly doped silicon regions separated by the channel region, nearly intrinsic. By steering the two contact regions with sufficient voltages, it is possible to induce enough band bending to allow electrons to tunnel from the valence band of the p+ doped contact region to the conduction band of the intrinsic region and thus to be collected at the other n+ doped contact region. If then another gate is employed to suppress the ambipolar behavior (due to opposite applied voltages), one can obtain a reconfigurable device. Therefore, one of the gates actually switches the device on and off, while the other one blocks the possibility for the carriers to be injected also when the polarity is reversed avoiding in this way the ambipolar behavior. The mechanism is different from a Schottky barrier based RFET even if, also in this case, two gates are employed. In any case, the result is still a transistor that shows unipolar n or p characteristics depending on how the two gates are steered.

IMOS/TFET switch devices
With this kind of device, proposed by Singh et al. in 2015, the concept of reconfigurability is extended from polarity control (switch between n-type and p-type conduction) to charge injection mechanism control. Exploiting a three independent gate (TIG) wrap-all-around (WAA) structure with a silicon nanowire used as channel material, the authors here achieved dynamic reconfigurability of the device between IMOSFET (impact ionization MOSFET) and TFET operation modes. The former injection mechanism provides very high output currents (up to mA) and a ultra steep switching behavior, although it requires high operating voltages. The latter, instead, presents very low off-currents (and then a high power efficiency) and lower operating voltages, but also low drive currents and slower switching. The concept of reconfigurability tries to overcome the limitations of these injection mechanisms, exploiting their strengths in a device that merges both. In addition to the injection mechanism reconfigurability, here the authors showed also the possibility to switch between conduction modes.

MOSFET/TFET switch devices
A dynamically reconfigurable FET, capable of switching charge injection mechanism between MOSFET and TFET operation modes, has been theoretically proposed by S. Bhaskar and J. Singh in 2015. Also in this simulated device, an undoped silicon nanowire is used as channel material, while the gating is operated by three independent gates, surrounding the overall structure. The device exploits the presence of a charge plasma electrode used to induce electrons at the drain side of the structure. Such a device allows to switch between high performance MOSFET mode and low power consumption TFET mode, just changing the polarity of a steering gate. Later on (2019) P. Wu et al. fabricated such a switching device employing black phosphorus as channel material, realizing both injection mode and conduction mechanism reconfigurability.

Technology enablers
Independent of the chosen concept and working principle a set of technology enablers, i.e. design and processing key elements that make the realization of an RFET possible, have been identified:


 * 1) Independent gating: all the proposed structures show the presence of at least two independent gates. This feature deeply characterizes reconfigurable devices since it is always needed an additional gate to program the device to switch between operational modes.
 * 2) Nanoscale channels: to obtain an adequate gate control on the carrier transport, nanoscale structures are required as channel materials;
 * 3) Sharp junctions: for both Schottky barrier related and band-to-band-tunneling concepts the importance of obtaining a controlled sharp interface is of importance;  this can be achieved e.g. through a silicidation process;
 * 4) n-type/p-type symmetry: the application of these devices in CMOS circuits made necessary the realization of equally large currents for electron and holes in both conduction modes.  In the case of polarity-controllable devices, a midgap alignment of the source/drain contacts, like present in the NiSi2/Si material system is crucial. In addition, stress, e.g. induced by self-limiting oxidation, is used as second enabler to fine tune the symmetry between electrons and holes currents.
 * 5) Low bandgap materials: the injection of carriers via tunneling mechanisms at a given electric field increases with use of materials with comparatively low bandgap, such as Ge or InAs. The use of such materials has been proposed as the main enable to increase switching speed and lower threshold voltages;

Applications
A number of gate level features have been demonstrated for RFETs, which provide an added benefit over their CMOS counterparts: dynamic reconfiguration, intrinsic XOR and wired-AND capabilities , control of threshold voltage and suppression of parasitic charge sharing effects in dynamic logic gates. This higher expressive capability of RFETs can be exploited to yield circuit designs with a higher compactness. Pioneer studies have shown that overall chip area can be saved albeit the larger feature size of the individual devices. Reconfigurable transistor concepts have also been proposed for the co-integration of a number of add-on functionalities into classical CMOS, which go beyond general computing purposes. The polymorphic nature of RFET circuits enables new takes on hardware security solutions, such as logic locking, camouflaging, physically unclonable functions (PUFs) or chip authentication. RFET based XOR cells and flip-flops have been shown to be less prone to delay-side-channel attacks as their CMOS counterpart. Beyond that, RFETs have a high potential for new operation schemes, such as analog/mixed-signal, asynchron or neuromorphic computing. For example synaptic cell for emulating binary neural networks or spike-time-dependent-behavior have been proposed.