Modulation Doping of Silicon Nanowires to Tune the Contact Properties of Nano‐Scale Schottky Barriers

Doping silicon on the nanoscale by the intentional introduction of impurities into the intrinsic semiconductor suffers from effects such as dopant deactivation, random dopant fluctuations, out‐diffusion, and mobility degradation. This paper presents the first experimental proof that doping of silicon nanowires can also be achieved via the purposeful addition of aluminium‐induced acceptor states to the SiO2 shell around a silicon nanowire channel. It is shown that modulation doping lowers the overall resistance of silicon nanowires with nickel silicide Schottky contacts by up to six orders of magnitude. The effect is consistently observed for various channel geometries and systematically studied as a function of Al2O3 content during fabrication. The transfer length method is used to separate the effects on the channel conductivity from that on the barriers. A silicon resistivity is achieved as low as 0.04–0.06 Ω ·cm in the nominal undoped material. In addition, the specific contact resistivity is also strongly influenced by the modulation doping and reduced down to 3.5E‐7 Ω · cm2, which relates to lowering the effective Schottky barrier to 0.09 eV. This alternative doping method has the potential to overcome the issues associated with doping and contact formation on the nanoscale.


Introduction
Silicon transistors pose the dominant device in the era of microand nanoelectronics and will remain so in the foreseeable future.This dominance is well-founded by continuous miniaturization, and by keeping low the fabrication cost per electronic device, increasing integration density, performance, and energy efficiency of integrated chips and circuit boards to this effect. [1]In order to hold up with this trend, there is a continuous need for DOI: 10.1002/admi.202300600new material solutions, such as high-k metal gate integration and strained silicon, as well as new device architectures.[5][6] Recently, major semiconductor industries have announced to start high-volume manufacturing of nanosheet devices after entering the 3 nm node. [7]However, with shrinking channel dimensions, especially in the case of a nanoscale transistor, the influence of the contact resistance on the overall device performance gains an increasing importance. [8,9]Forming a contact with low junction resistance is essential to allow charge transport with a minimal voltage drop.Thus, Ohmic contacts providing a negligible resistance are desired for most silicon technologies. [10,11]However, achieving a near-perfect Ohmic contact with a nanoscale semiconductor is generally difficult.If brought in contact with moderately doped silicon most metals form a Schottky contact rather than an Ohmic contact. [12]The rectifying barrier can be reduced by an increased doping level near the interface by either implantation [13] or dopant segregation. [14,15]It has been discussed, that a barrier height below approximately 0.1eV has to be achieved to meet performance requirements. [16,17]However, excessive impurity doping leads to the degradation of carrier mobilities, due to the scattering of mobile charge carriers with a large number of defects in the semiconductor. [18]21][22] This calls for a disruptively new method of providing doping in silicon and lowering the contact barriers of Schottky junctions.
One such method is the Modulation doping approach coined by Störmer and Dingle [21] for group III-V semiconductors already in 1978.Back then a GaAs/AlGaAs superlattice was studied and proven to show two orders of magnitude higher mobility by isolating the dopant atoms spatial from the channel area, by placing them into a high gap capping material (AlGaAs).Ever since then a lot of research has been done on quantitatively describing the effects associated with this type of quantum well and superlattice structures. [23,24]However, it was not before 2017 when Dirk König et al. proposed an adaptation of the modulation doping method to silicon by introducing aluminium-induced acceptor states into the surrounding SiO 2 dielectric.Those defects align at an energy level approximately 0.5eV below the silicon valence band edge as evaluated by deep-level-transientspectroscopy (DLTS). [25]Density-functional-theory (DFT) calculation also indicate the potential significance of Gallium or Scandium as a modulation dopant. [26]Such, surface modifications play an ultra-important role in nanomaterials and the surface defects have a dramatic influence on the transport properties compared to bulk materials. [27]If carefully engineered, the defects near the Si/SiO 2 interface can trap electrons tunneling from the adjacent silicon.The effect creates mobile carriers within the semiconductor, as well as a negative fixed charge within the dielectric.
It has been proven, that this method can be exploited to yield an effective silicon surface passivation. [25,28]The related density of fixed charges (Q fix ) is a key parameter for achieving controlled field-effect passivation in modern solar cells. [29]n this work, the approach to generate spatially separate acceptor states within the dielectric is used to realize Modulation doping of silicon nanowires.Thereto, we characterize the silicon nanowire Schottky barrier contacts as a function of the number of Al 2 O 3 ALD cycles added to the dielectric shell.The individual contributions of the modulation doping on the contact and channel resistances are separated by the transfer length method (TLM).The significance of Q fix in controlling the contact resistance of a metal-semiconductor(M-S) contact is also highlighted.The approach enables a radically new method to establish emerging computing platforms on a nanoscale.

Results and Discussion
Modulation doping of silicon is realized by replacing Si atoms in the tetravalent SiO 2 network with trivalent Al atoms, creating an unoccupied defect state, just slightly below the valance band edge (VBE) of Si.A schematic picture of the modulation doping process by Al-induced acceptor states in SiO 2 is shown in Figure 1a.The density of acceptor states created this way can be quantified by the density of fixed charges (Q fix ) inside the dielectric.If these unoccupied states are located within the tunneling distance, electrons can be captured from the silicon, leaving holes behind as majority charge carriers, giving rise to a p-type conductivity inside the channel.This process is enabled by a few fortunate properties of the SiO 2 : it has a substantially higher bandgap than Si, it forms a nearly perfect interface with a low number of defects and it is thermally stable, i.e., neither the oxygen nor the induced defect states diffuse into silicon. [30]Here, the effect is transferred to silicon nano-structures aiming at improved transport.Note, that a hole transport without actual channel doping can also be enabled by surface charge transfer, which has been demonstrated as a potential approach for shifting and tuning the semiconductivity. [27]odulation doping goes beyond that by inducing a direct interaction with not just the dangling bonds at the interface but also by capturing electrons origin deeper from the silicon crystal.
Our investigation was carried out on a batch of Si nanochannel test structures with various widths and lengths patterned on a silicon-on-insulator (SOI) substrate with a Si thickness of 20 nm.The SOI layer was nominally un-doped with a uniform background boron doping level of 1E15cm −3 , leading to a sheet resistivity in the range of 9-14 Ω •cm as specified by the wafer manufacturer. [31]A 2.2 nm thick SiO 2 was thermally grown on the nanowires, followed by a few cycles of atomic layer deposited (ALD) Al 2 O 3 , and an activation anneal to induce acceptor states into the dielectric.The added defects in the dielectric shell features a homogeneous deposition all around the structure even in high aspect-ratios.See Supporting Information for homogeneity and repeatabality of the doping method.The nanowire channels were deliberately contacted with Ni pads, and subjected to another short anneal to form diffused NiSi x contacts reaching into the channel having a near midgap work function.At these metallic NiSi x /Si junctions, a Schottky barrier is formed.
The transport over these junctions is generally split into two contributions: a thermionic current of carriers passing over the Schottky barrier and a field emission current of carriers tunneling through the barrier.Depending on the work function alignment of the contact metal and on the applied bias conditions, one of the two mechanisms dominates, whereby thermionic field emission [32][33][34][35] dominates carrier transport in the intermediate regime(further information can be found in the Supporting Information).Throughout this work, we will show that the dominant conduction mechanism can be influenced by modulation doping applied to the dielectric shell.Note, that the distribution of background impurity dopants can be also influenced by segregation effects and redistribution of dopants near the contact edges during the contact formation. [15,22]We assume this effect to have no influence on our study as it would affect both reference and modulation doped samples equally.
A systematic variation in the number of Al 2 O 3 ALD cycles was done for otherwise completely identical processed structures.A top-view SEM image, as well as a schematic cross-section of the fabricated structures are shown in Figure 1b.The electrical result of this careful surface engineering is shown in Figure 1c by the current-voltage (I-V) characteristics of an exemplary device having a drawn length of 3.5 μm and a width of 500 nm.For the reference sample, a rectifying behavior with a maximum current below 100 pA is observed.It is evident, that with an increasing number of Al 2 O 3 ALD cycles added during fabrication, the current level increases rapidly.Already with only 2 Al 2 O 3 ALD cycles, acceptor states are created in the dielectric shell, leading to an increase of the current by nearly five orders of magnitude.The current increases by another order of magnitude if the number of Al 2 O 3 ALD cycles is increased from 2 to 6.At 15 cycles, a saturation is observed, reaching a minimal resistance value as low as 60 kΩ.Note, that this value cannot be directly related to the sheet resistance of the silicon, due to the strong influence of contact resistance in Schottky barrier devices.The effect is remarkably consistent over a high number of nanowires with different widths and lengths (Figure 1d).
In addition, the increased overall current level after modulation doping is accompanied by a clear linearization of the I-V characteristics as better visualized in the linear scale plot in Figure 2d for the sample achieving the highest modulation doping.
We explain this effect with a different band alignment at the junction due to the added modulation doping.The work function of the silicon thereby is of extreme significance because of its sensitivity to the doping level. [36]For the reference sample, the Fermi level is placed near the middle of the Si bandgap, leading to a higher work function (>5.1eV [37] ) compared to the metallic NiSi x contacts (4.5-4.8 eV [38] ).The work function difference results in band bending towards the interface leading to the formation of a Schottky barrier as shown in Figure 2a.The tunneling transport across the barrier is dominated by field emission and the corresponding I-V curve is shown in Figure 2b.If p-type doping is added, the Fermi level of silicon shifts toward the valence band, also altering the band alignment at the contact.For a sufficiently low barrier height, the thermal energy of the carriers is enough to directly pass the remaining barrier even at low bias.The work function alignment can be highly influenced by chargeable states in the semiconductor right at the interface, which can naturally give rise to Fermi level pinning.This effect has been reported to be very prominent for metal silicides. [39]Modulation doping creates similar surface states in silicon due to the electron capture process and thus influences the interface between the silicide and silicon.As a result, the NiSi x work function is strongly pinned to the Fermi level near the valence band resulting in an Ohmic-type contact as observed for the high modulation doping of silicon.In our study, Fermi level pinning is proportional to an exponential increase in modulation doping and is attributed to the increasing number of electrons tunneling from Si to unoccupied states in SiO 2 , consequently lowering the potential energy barrier to the flow of charge carriers in the M-S system.This lowering results in a current transport across the NiSi x -Si junction that can ultimately lead to a direct thermionic emission, as shown in Figure 2c.This novel doping method thereby induces a transition from a field-emission-dominant transport to a thermionic-emission-dominant transport at the contacts.This hypothesis is also supported by the transition from a supra-linear to linear shape visible in the I-V characteristics, which originates from a transition of field-dependent contact resistance to a fieldindependent resistance.The transition voltage between the two regimes is also reduced effectively with increasing modulation doping.This transition voltage is derived from the extrapolation of the linear region intersecting the x-axis.We call this intersection as V s in this paper, see Supporting Information for further information.
In order to quantify the tuning of the contact properties the specific contact resistivity ( c ) was quantified as a function of the modulation doping using the transfer length method (TLM).In this method an array of contact pads with varying distances is deposited onto a single channel with equal width as, depicted in Figure 3.The total resistance R tot measured between two individual contacts with a certain channel length is composed of several contributions, as also illustrated in Figure 4a: where R m is the resistance of the metal pads, R ch is the channel resistance, and R c is the contact resistance.R c is assumed to be constant for all channel segments, while the value of R ch depends on the length of the channel segment: where R sh is the sheet resistance of silicon, L ch, i is the length of the respective segment of the TLM structure and w is the channel width.The resistance of the metal R m is usually negligible so that both R c and R sh can be extracted from a plot of R tot over various lengths, as shown in Figure 4b.R c can be related to the specific contact resistivity if the contact area is known.In our work the diffused NiSi x contacts lead to the specific case of a Side-wall transfer length method (STLM), where the contact area is perpendicular to the channel direction.[42][43] The STLM also leads to a better approximation of the contact compared to a conventional TLM where the non-uniform current spreading and crowding limits the resolution of the determined contact resistivity. [43]Consequently, the specific contact resistivity is calculated assuming a 0D model, where  c is seen as a macroscopic quantity.When the current density entering the contact window is uniform, the  c is calculated from the thickness t of the silicon, width of nanowire channel w and the R c : Finally, for a Schottky contact,  c is directly influenced by the barrier height and has a strong dependence on temperature.At room temperature, the work function of the metal silicide and the doped Fermi level of the semiconductor determines  c .Considering a purely thermionic emission model consistent with our hypothesis for the transport across the modulation doped junctions, the specific contact resistivity can be given by: [44,53] where k is Boltzmann's constant, A* is Richardson's constant, q is the carrier charge, qϕ B is the effective barrier height under the applied bias conditions, and T is the sample temperature.
A resulting TLM plot is exemplary, shown in Figure 4 bfor channels having a width of 500 nm and different modulation doping treatments.As expected, the influence of parasitic effects originating from current crowding [40,45,46] is not significant due to the horizontal flow of current at the silicide-silicon interface brought about by the fully silicided contact segments with a sharp interface to silicon (see SEM images in Figure 3).There is no diffusion layer assumed below the silicide area that might give rise to a combination of horizontal and vertical current flow. [36,41,46]Note, that only the linear regime of the I-V characteristics above V S was analyzed using linear regression.When a linear plot with the resulting resistance at different lengths is constructed, the extrapolation to the Y-intercept (L ch = 0 nm) gives twice the effective contact resistance 2R C .The sheet resistance can be inferred from the slope.Good linear fits were achieved for all datasets and lengths below 4 μm.For a higher number of Al 2 O 3 ALD cycles, the linearity improves with the best fit across four lengths, yielding a maximum correlation coefficient R 2 of 99.9%.It is evident that the contact resistance is strongly reduced with an increasing number of Al 2 O 3 ALD cycles.The gradual decrease in the sheet resistance of silicon is clearly visible from the slope.The extracted results are summarized in Table 1 for an STLM structure with a width of 500 nm.While the undoped reference approximates a Si resistivity of 14 Ω •cm, the modulation doped silicon achieves a minimum resistivity as low as 0.04-0.05Ω •cm.Overall, a set of 9 TLM structures with different widths was investigated (see Figure 4c), where all the structures show a consistent decrease in the resistivity of silicon with an increasing number of Al 2 O 3 ALD cycles.This is a clear indication of the presence of holes as majority carriers due to the generation of acceptor states in the dielectric shell.In bulk-Si, such room-temperature resistivity can be achieved by equivalent direct boron doping concentrations of 1E18 cm −3 . [31]This finding demonstrates that the disruptive modulation doping concept presents a unique tool to define and control majority carriers in a nanoelectronic device, featuring a scattering-free carrier transport.
A second interesting trend is the lowering of contact resistance with an increasing number of Al 2 O 3 ALD cycles.It is observed that the R c and the  c drop continuously by one order of magnitude with an increasing number of Al 2 O 3 ALD cycles, but the resistivity of Si starts to saturate after 6 Al 2 O 3 ALD cycles, presumably due to a saturation of effectiveness of the modulation doping.The first decrease in R c and the large drop in resistivity of Si from 2 to 6 Al 2 O 3 ALD cycles can be explained due to the increased density of acceptor states leading to an increased hole concentration in the channel.As already discussed in Figure 2, we attribute this to an increase in the Fermi level pinning with modulation doping.In addition, by further lowering the contact resistance when moving to 15 Al 2 O 3 ALD cycles, the fixed charges that accumulate in the oxide shell by modulation doping have an impact on the Schottky barrier. [47]Such charges near the Schottky barrier interface create an artificial gating of the barrier, similar to what can be observed in a charge trapping device. [48]This will lead to an additional band bending, reducing the remaining barrier and making it quasi-opaque even without an applied external voltage.Planar reference samples with additional HfO 2 capping have shown to reach a fixed charge density of −4.6•10 12 cm −2 after 6 Al 2 O 3 ALD cycles [47] (see also Supporting Information).Simulations of  c as a function of the bulk fixed charge density have been discussed, [49] where a Q fix of nearly 3.10 20 cm −3 is estimated for a  c as low as 10 −3 to 10 −4 Ω • cm 2 .This finding suggests, that the observed lowering of the contact resistance cannot be explained by the Q fix alone, but is a combined result of a high majority carrier density originating from acceptor modulation doping and an associated Fermi level shift.If compared with classical Boron implanted SOI structured STLM with NiPt silicided contacts, which showed a contact resistivity of 0.5•10 −8 Ω • cm 2 , [43] our modulation doping method shows comparable R c results.Using Equation ( 4), these resistivity values can be extrapolated to an effective barrier height.We note that by the applied model, these values do not comprise the actual Schottky barrier height, but the equivalence of a purely thermionic barrier leading to the same transport properties.Our findings by STLM study were found to be in good agreement with the effective Schottky barrier heights obtained from electrical characterization of modulation doped single channels at different temperature. [50]For high modulation doping, a significantly low barrier height for hole tunneling is demonstrated which validates also our proposed model of transport physics.The values reported earlier in literature for classical doping accumulation is also in line with the resulting lowering of the barrier for holes from 0.45 [51] to 0.09 eV.For example, Mantavya et al. [52] reported a barrier as low as 0.12 eV for p-type Si treated with Al-ion implantation using an Al dose of 2E14 cm −2 .
Therefore, in comparison to conventional doping, a similar carrier concentration of approximately ⩾1E18 cm −3 can be induced in silicon by modulation doping while keeping the channel crystal intact and, at the same time, having a significantly lowered contact barrier.Furthermore, with our method the device dimensions can be scaled without any limitation on performance in contrast to conventionally doped channel that suffer from a reduced carrier density with decreasing nanowire radius.

Conclusion and Future Work
Modulation doping of silicon nanowires was introduced as a method to potentially overcome the limitations of conventional impurity doping for highly scaled devices.By adding Aluminium induced acceptor states to a shell comprised of SiO 2 , the overall resistance was lowered by nearly six orders of magnitudes.The individual contributions of the silicon resistivity and the contact resistance have been separated by the side-wall transfer length method.This way, we have demonstrated that a channel resistivity as low as 0.04-0.06Ω •cm could be achieved, which relates to hole carrier concentration equal to that of classical doping in the range of 1E18 cm −3 .In addition, the contact properties of the nanowires could be tuned, effectively lowering  c to 3.5E-7 Ω • cm 2 utilizing the combined influence of the fixed charges in the oxide in addition to the altered Fermi level alignment at the silicide-silicon junction as induced by the modulation doping.Using a purely thermionic emission model, a Schottky barrier height as low as 0.09 eV could be extracted.Future work will be devoted to the development of a suitable capping layer to ensure long-term stability in order to demonstrate a modulation doped p-type FET.Such an FET with essentially impurity free channel as provided by the disruptive doping method would be a breakthrough in low-temperature cryogenic applications and quantum computing.Due to the absence of classical thermal ionization of dopants other detrimental effects of impurity doping, such as mobility degradation, fluctuations, and low-frequency noise would be prevented.

Experimental Section
Fabrication: Structures analyzed in this work had been fabricated using two layers of lithography.Silicon-on-insulator substrates with 20nm thick silicon with a background boron doping concentration of 1E15 cm −3 and 100nm thick buried oxide (BOX) had been used.As a first step, the silicon nanowire channels had been patterned using E-beam lithography.Samples were spin-coated with a negative tone resist, HSQ 2% by weight and exposed to electrons with a beam energy of 20kV.Nanowires with different widths were designed using KLayout.Four identical write fields were patterned with the same beam parameters for processing with different Al 2 O 3 ALD cycles in the later stage.The resist was developed in TMAH 25% and the nanowire was defined using an SF6-based reactive ion dry etch with a Si etch rate of approximately 20 nm min -1 , and RF power of 60 W provided by an inductively coupled plasma (ICP).After the nano-channel fabrication, the resist was removed using HF 1% by volume in water, followed by a RCA cleaning.The two-layer modulation doping stack was added, which consists of 2.2 nm SiO 2 thermally grown by rapid thermal oxidation (RTO) and a subsequent ALD of Al 2 O 3 using trimethylaluminum and water at 200°C.The sample was cleaved into four identical pieces, each with a single write field of identical nanowires.All four samples were processed with the same thickness of SiO 2 and a varying number of Al 2 O 3 ALD cycles.The first sample served as a reference without any Al 2 O 3 ALD cycles and was only exposed to water during the ALD process, the second sample was processed with 2 Al 2 O 3 ALD cycles, the third sample was processed with 6 Al 2 O 3 ALD cycles and the fourth sample was processed with 15 Al 2 O 3 ALD cycles.This deposition was followed by an activation annealing in an argon atmosphere at 850 °C for 30s.
The metal contacts were established by a second lithography process using a double-layer positive tone resist PMMA-MA/PMMA.The double layer provides an undercut, which helps in the lift-off process.The resist was developed in the MIBK solvent AR600-55.The interface oxide in the contact area was etched using a buffered HF solution prepared from a mixture of 100 parts of 40% NH 4 F to 1 part of 38-40% HF in water and immediately transferred into the vacuum chamber of the sputtering unit.Nickel was used as a contact metal and was sputtered to a thickness of 50 nm.The lift-off was done in DMSO and acetone with ultrasonication.Subsequently, thermal silicidation in N 2 at 400 °C was done for 6s.The samples were kept in high vacuum until electrical characterization.
Electrical Characterization: The IV characteristics were measured using Keithley 4200-SCS with two SMUs (Source measurement units).The measurement was performed in vacuum in order to avoid environmental influences such as adsorbates, water vapor, or other airborne contaminations.The side-wall transfer length method (STLM) was applied at room temperature in various geometries.Channel lengths were varied in four steps from 500 nm to 7.6 μm.Channel widths were varied from 50 to 500 nm.A total of six contact combinations were measured by biasing one contact while keeping the other contact unbiased, and measuring the current running through the contacts, as shown in Figure 3a.The voltage was swept between −2V to +2V in steps of 0.05V.
Statistical Analysis: The four different samples investigated had the same sample size of 1.5 cm x 1.5 cm and were identical in design layout.The total resistance, R tot measured for each data point was an average of resistances measured from the linear region of the I-V characteristics i.e., between voltage bias of -0.75V to -2V.The homogeneity, repeatability and stability was investigated (see Supporting Information).The scientific graphing and data analysis was done using Origin.A linear regression model was used to fit the I-V curve and the TLM model data.

Figure 1 .
Figure 1.Modulation doping of silicon nanowires.a) Schematic of 2D crystal lattice of Si/SiO 2 illustrating modulation doping by Al acceptors in the near SiO 2 as created by rapid thermal annealing.Electrons are able to tunnel from the silicon to the unoccupied acceptor state, leaving behind a free hole.b) Top-view SEM image of silicon nanowire structure with a schematic horizontal cross-sectional view of the different materials used to demonstrate modulation doping in this work.c) Semi-logarithmic current-voltage (IV) characteristics of the nanowires were measured in vacuum treated with no Al 2 O 3 ALD cycles, 2, 6, and 15 Al 2 O 3 ALD cycles, respectively.Si channels had a length of 3.5 μm and a width of 500 nm.d) Total wire resistance as a function of Si channel length and width for different numbers of Al 2 O 3 ALD cycles.

Figure 2 .
Figure 2. Model of transport physics where E c,S , E v,S , and E F,S indicate the conduction band edge, valence band edge, and Fermi level of silicon respectively, E vac is the vacuum level and ϕ s is the work function of silicon, E F and  NiSi x is the Fermi energy of NiSi x and its work function.The potential barrier height is given by qϕ B , the barrier width is given by the Debye length W d as manifested by the majority carrier density in Si, and the electron affinity of Si is .a) Schematic energy band diagram of silicide-silicon interface along the channel for the reference undoped silicon nanowire.Field emission is indicated as the dominant current contribution after a voltage is applied.b) I-V characteristics in linear scale for reference undoped nanowire.c) Schematic energy band diagram of silicide-silicon interface along the channel in a heavy modulation-doped device showing the changed Fermi level in silicon and barrier height lowering for zero applied bias voltage.Thermionic emission is indicated as the dominant current contribution after a voltage is applied (W d in Si is not shown due to being extremely thin).d) I-V characteristics in linear scale for a modulation-doped nanowire treated with 15 Al 2 O 3 ALD cycles.A linear fitting function is given as a guide to the eye.

Figure 3 .
Figure 3. Side-wall transfer length method (STLM).a) Schematic of the STLM structure indicating contacts C1 to C6, I a and I b are currents measured between the two contacts.The inset indicates the schematic illustration of the current flow in the channel for a pair of contacts with an effective crosssectional area of w x t, where w is the width and t is the thickness of the nanowire channel.b) SEM image of a fabricated STLM structure.c) Enlarged view of (b) indicated by the box.The figure shows a Si channel with a width of 500 nm as well as the diffused NiSi x metal contacts extending into the channel.

Figure 4 .
Figure 4. Separation of contact and channel resistance effects.a) Equivalent circuit illustration of the test structure for a pair of adjacent contacts.The influence of modulation doping on contact resistance is indicated.b) TLM plot showing the resistance measured for different lengths and different Al 2 O 3 ALD cycle counts.c) The resistivity of silicon as a function of the number of Al 2 O 3 ALD cycles extracted from nine different TLM structures.

Table 1 .
Summary of contact resistance, specific contact resistivity, the resistivity of silicon and the barrier height for samples processed with different numbers of Al 2 O 3 ALD cycles.