Schottky contacts are the key component of Schottky barrier diodes (SBDs) and metal-semiconductor field effect transistors (MESFETs). The basic theory and knowledge of Schottky contacts are also important in understanding ohmic contacts. Formation of ohmic contacts is essential in any semiconductor devices. Both types of contacts are required for a variety of electrical characterization techniques of materials, including , DLTS, and Hall effect measurements. Several review papers on Schottky contacts [319–321] and ohmic contacts [320, 322–327] on SiC have been published.
In the case of SiC, most metals deposited on SiC work as Schottky contacts, as long as the SiC material is not heavily doped and high-temperature contact sintering is not performed. When a metal and a semiconductor are brought into contact, the Fermi levels of the two materials line up at equilibrium (zero bias). The energy band diagrams of a Schottky barrier on (a) n-type and (b) p-type semiconductor at zero bias are shown in Figure 6.69. Here, is the barrier height, the built-in potential, and (or ) the position of the Fermi level from the conduction band (or valence band) edge. From the band diagram, the following equation is satisfied:
In an ideal model, the barrier height is determined simply by the difference in the work function between the metal and the semiconductor. In actual semiconductors, however, the barrier height is affected by the surface states [8, 328]. The barrier height can be determined by several techniques, including , IPE, and XPS.
Figure 6.70 shows the forward current density versus bias voltage for Ti, Ni, and Au/n-type 4H-SiC(0001) SBDs at room temperature [329]. The current density can be expressed by [8, 328]:
where is the ideality factor and is the effective Richardson's constant for the semiconductor, given by:
Here (or ) is the majority carrier effective mass and is the Planck constant. In SBDs on n-type 4H-SiC(0001), is determined as by using an of 3 and of [329]. This value agrees with that obtained from a careful experiment [330]. Note that the effective Richardson's constant depends on the crystal face as well as on the conductivity type (n-type, p-type). Therefore, the barrier height can be extracted from the of the semilogarithmic plot shown in Figure 6.70. Figure 6.71 shows the extracted barrier height versus the ideality factor for Ni/n-type 4H-SiC SBDs fabricated by the same process. It is clear that the barrier height is severely underestimated when the ideality factor is larger than 1.10–1.15. Thus, the barrier height extracted from the can be reliable only when the ideality factor is smaller than 1.10. Even when a relatively good ideality factor of 1.05 is obtained, the extracted barrier height is clearly smaller than the real value. The ideality factor is degraded by several reasons including inhomogeneity of the barrier height and imperfections (such as contamination) at the interface [331–333].
The capacitance of the space-charge region of SBDs (n-type) per unit area is given by:
where is the donor density and the dielectric constant of the semiconductor [8, 328]. Squaring and inverting this equation, one obtains:
By plotting versus the bias voltage, the built-in potential can be determined from the of the plot. Figure 6.72 shows versus bias voltage for Ti, Ni, and Au/n-type 4H-SiC(0001) SBDs. From the built-in potential and the calculated Fermi level , the barrier height can be extracted using . The Schottky barrier heights determined from measurements are slightly higher than those extracted from measurements. This is natural – local lowering of Schottky barrier height makes a significant difference in measurements, because the current depends exponentially on the barrier height while measurements yield the value averaged over the entire contact area and are thus less affected by local variation.
Furthermore, the net donor density can be determined from the inverse slope of the plot. Again, note that this is the net doping density, not “the carrier density.” Inside the space-charge region (except near the edge of the space-charge region) all the dopants are ionized, and the width of the space-charge region determines the capacitance. This is not very important in Si owing to nearly perfect ionization of dopants at room temperature. In SiC, however, the carrier density can be much lower than the doping density, especially for p-type materials at room temperature. Thus, one has to be aware of the fact that measurements always give the net doping density. This can be easily confirmed by very little change in the density obtained from measurements on p-type SBDs at different temperatures (e.g., from RT to 300 °C or from to RT).
Another technique to determine the barrier height is IPE. In this technique, semi-transparent Schottky contacts must be prepared. Monochromatic light is projected through the contact, and the photocurrent is monitored as the wavelength of the light is changed. When the photon energy is larger than the barrier height, the square root of the photocurrent yield (photocurrent divided by the photon number) is linearly dependent on the photon energy as follows:
where is a proportionality constant. The linearity in the square root of photocurrent comes from the energy dependence of the density of states in the energy band. Figure 6.73 shows the square root of photocurrent yield versus the photon energy for Ti, Ni, and Au/n-type 4H-SiC(0001) SBDs [334]. The barrier height can be directly determined from the intercept of the plot. IPE gives the most reliable values of Schottky barrier heights, though it requires special equipment and semi-transparent Schottky contacts.
Figure 6.74 shows the barrier height versus metal work function for n-type 4H-SiC SBDs with various metals [334–336]. Data for 4H-SiC(0001), , and are plotted. For a given metal, the barrier height is slightly higher on , slightly lower on (0001), and that on is in between. This difference may be attributed to the existence of polarity-dependent dipoles at the interface and/or different distribution of surface states. The slope of this plot is 0.8–0.9, indicating that the metal/SiC interface is free from Fermi-level pinning, and is close to the Schottky–Mott limit [328, 337]. To obtain this kind of data, of course, high-quality materials and careful surface cleaning are necessary. The barrier heights can be slightly changed by employing different processes, such as surface treatment, prior to metal deposition. It is noted that the ideality factor and reproducibility are improved by annealing at relatively low temperatures (200–500 °C) after metal deposition. When the Schottky barrier height on n-type 4H-SiC is compared with that on n-type 6H-SiC [338] for a given Schottky metal, the barrier height on 4H-SiC is about 0.2 eV higher, which corresponds to the difference in the bandgaps of the two polytypes. The barrier height is also discussed in Section 7.2.
Schottky barrier heights on p-type SiC have been much less studied [339]. Figure 6.75 shows the barrier height versus the metal work function for p-type 4H-SiC(0001) SBDs with various metals. The slope of the plot is about , and the sum of Schottky barrier heights on n- and p-type SiC for a Schottky metal material is close to the bandgap :
Therefore, a metal having a small work function, such as Ti, gives a high barrier height on p-type SiC.
Unique device physics is involved in the reverse leakage current of SiC SBDs. In SiC, the electric field strength in the space-charge region can be almost 10 times higher than in space-charge regions of Si-based devices. Therefore, the band bending can be so sharp that the potential barrier can be very thin. Figure 6.76 shows the band diagram of an n-type SiC Schottky barrier under high reverse bias. In Si SBDs, the reverse leakage current is well described by a thermionic emission model, taking into account the barrier height lowering by the image force [8, 328], unless the semiconductor is heavily doped. In SiC SBDs, however, the observed leakage current density is many orders of magnitude higher than the value calculated by the thermionic emission model (taking into account barrier height lowering). At first, leakage current through crystalline defects or severe local lowering of Schottky barrier heights was suspected. However, it turns out that there is an essential reason for the relatively large leakage current in SiC SBDs. In SiC, the triangular-like potential barrier can be very thin because of the high electric field (slope of the band diagram), and the leakage current is governed by thermionic field emission (TFE) [340, 341], as shown in Figure 6.76. The leakage current density based on a TFE model can be expressed by the following equation [341]:
Here and are the tunneling mass of carriers and the electric field strength, respectively. Figure 6.77 shows an example of reverse characteristics of a Ti/4H-SiC(n-type) SBD at various temperatures [341]. The considerable increase in leakage current with increasing bias voltage, as well as the small temperature dependence, can be reproduced by the TFE model. It has been reported that the leakage current of SBDs formed on high-quality GaN(0001) can also be reproduced by the TFE model [342]. Thus, the TFE current will be dominant in SBDs on any wide-bandgap materials which exhibit high electric field strength, such as SiC, GaN, , and diamond.
When the barrier height is highly inhomogeneous, abnormal characteristics are observed [343, 344]. Such an example is shown in Figure 6.78, where forward characteristics of a Ti/SiC SBD at various temperatures are plotted [344]. At room temperature, a small shoulder-like feature is seen in the characteristics, suggesting non-ideal behavior and a non-unity ideality factor. The shoulder becomes distinct at low temperature, because the slope of the log versus plot indicated in Equation 6.26 becomes sharp at low temperature. In this case, there exist small regions where the local barrier height (,local) is considerably lower than the main Schottky barrier . The ratio of the region and the total contact region is estimated from the ratio of the almost-saturated current of the component and that of the main Schottky barrier (the “saturation” is caused by a series resistance). The ratio is approximately in Figure 6.78. Note that the ideality factor is close to unity for both the region and the main diode. When the value of the is estimated, the corresponding area (not the main area) should be employed for the calculation. Furthermore, such a region can cause considerably increased leakage under reverse bias [345, 346]. The diode can be regarded as a parallel connection of the main diode and a diode with . Thus, the reverse leakage current is governed by TFE through the region of the diode. Note that it is difficult to detect a region by or IPE measurements, since these techniques measure the average values.
The characteristics required for ohmic contacts include low contact resistivity (or specific contact resistance), surface flatness, and long-term stability. In particular, the contact resistivity is a key electrical parameter. This is the contact resistance per unit area, measured in , and the voltage drop across the contact is thus calculated by multiplying the contact resistance by the current density . The contact resistance should be negligibly small compared with the total resistance between two main terminals of devices (e.g., anode-cathode and drain-source). In the cathode contact of a SBD and the drain contact of a vertical field-effect transistor (FET), the contact can cover almost all the device area (active area). In the source contact of a vertical FET and the emitter contact of a bipolar junction transistor, however, the contact area is usually much smaller ( of the device area), meaning that very low contact resistivity in the range of is generally required.
Because of the wide bandgap of SiC, one can find very few contact metals with very low barrier heights. Figure 6.79 shows the band diagram of 4H-SiC, including the vacuum level. Since the electron affinity of 4H-SiC is about 3.8 eV [147, 176] and the bandgap 3.26 eV at room temperature, a contact metal with a work function lower than about 4 eV is ideal for an ohmic contact to n-type 4H-SiC, and that with a work function higher than about 7 eV is ideal for a contact to p-type. This rough sketch clearly shows the difficulty in formation of ohmic contacts, especially on p-type 4H-SiC.
Ohmic behavior is usually achieved by using the tunneling current through the thin potential barrier in SiC. When the tunneling current is caused primarily by electrons with energies near the Fermi level (that is, in the case of relatively low temperature or low barrier height), the tunneling is called field emission (FE). In the FE regime, the contact resistivity for an n-type semiconductor is given approximately by the following equation [8, 347]:
As the temperature increases, an appreciable number of electrons with energies above the Fermi level can contribute to the tunneling. This thermally-assisted tunneling is called thermionic field emission. In the TFE regime, the contact resistivity for an n-type semiconductor is given approximately by the following equation [8, 347]:
As seen in Equations 6.33 and 6.36, the contact resistivity is proportional to in both cases. More detailed modeling in the case of SiC is found in the literature [348, 349]. In general, the pure FE mechanism is dominant when , and TFE becomes important when [347].
Therefore, the strategy for obtaining low contact resistivity is, in principle, simple: (i) increase the surface doping density and (ii) select a metal which forms a low barrier height. In SiC, however, complicated interface reactions are involved to reduce the high barrier heights commonly observed. Figure 6.80 shows the contact resistivity versus the dopant density for barrier heights from 0.3 to 1.0 eV, calculated for the contacts on n-type 4H-SiC, taking both FE and TFE currents into account. In almost all the regime, the TFE mechanism dominates. It is clear that heavy doping is necessary to obtain a low contact resistivity of for a metal with a barrier height of 0.5 eV.
A special test structure must be prepared to measure the contact resistivity. The most common structure is shown in Figure 6.81. On the top of a rectangular mesa structure, multiple rectangular contacts are formed with varying spacing. The resistance between two terminals is measured with four probes (two for current, two for voltage), and the measured resistance is plotted against the contact spacing. Figure 6.82 shows an example of this plot for Ni/n-type 4H-SiC sintered at 1000 °C. Since the measured resistance consists of the two contact resistances and the resistance of the semiconductor, the following equation is satisfied [207]:
where is the sheet resistance of the semiconductor, the spacing between contacts, the contact width, and the contact resistance given by the -axis intercept. , the linear transfer length, can be extracted from the -axis intercept of the plot, as shown in Figure 6.81. The contact resistivity is given by:
Note that the sheet resistance can also be determined from the slope of the plot. This measurement is called the linear transfer length method (TLM). This method is based on the linear transmission line model, which takes into account the voltage and current distribution beneath contacts [207]. If the contacts exhibit non-ohmic characteristics, some modification is required [350]. With the TLM method, contact resistivities down to about can be evaluated. For more accurate measurements of the contact resistivity in the to range, so-called Kelvin structures must be formed [207].
In formation of good ohmic contacts, the metal chosen, and sintering after metal deposition are critical. In the case of SiC, one must be aware of reactions between the metal and Si or C as a function of temperature. In this sense, contact metals are classified into (i) metals which form only silicide(s) (no carbides), (ii) metals which form only carbide(s) (no silicides), and (iii) metals which form both silicide(s) and carbide(s). Therefore, a metal-Si-C phase diagram is useful for selection of contact metals, and for obtaining insights into the ohmic behavior. However, formation of silicide(s) or carbide(s) is not always sufficient to ensure ohmic behavior in SiC, and more complicated phenomena, such as vacancy formation and associated diffusion of carbon (or silicon) must occur to attain good ohmic contacts. This is evidenced by the observation that non-sintered, as-deposited metal silicide or carbide on SiC usually exhibits Schottky characteristics.
Because of the chemical inertness of SiC, the sintering process to obtain ohmic contacts is typically performed at 900–1000 °C. It is a challenge to control the structure and thickness of the chemically reacted layer at the interface. Furthermore, the reaction between the metal and SiC can lead to surface roughening [351] which makes wire bonding difficult. Thus, the contact metal must be thin enough to minimize the surface roughening, and additional thick layers of metals (both a barrier metal and an interconnecting metal) are deposited as overlays after high-temperature sintering, as shown in Figure 6.83.
So far, a variety of metals and annealing processes have been investigated for both n- and p-type SiC. Table 6.4 shows typical ohmic contacts reported in the literature. Because of the limited space, the next sections mainly describe common ohmic contacts: Ni for n-type SiC and Al/Ti for p-type SiC. For a survey of ohmic contacts on SiC, please see review papers [322–327].
Table 6.4 Typical ohmic contacts for n- and p-type SiC reported in the literature.
n-type | p-type | |
Ohmic contacts | Ni (sintered) | Al/Ti (sintered) |
Ti (sintered) | Al/Ni/Ti (sintered) | |
Al (sintered) | Al/Ti/Al (sintered) | |
Mo (sintered) | Al/Ti/Ge (sintered) | |
W (sintered) | AlSi (sintered) | |
Al/Ni (sintered) | Pt (sintered) | |
Al/Ti (sintered) | Ni (sintered) | |
Ni/Ti/Al (sintered) | Pd (sintered) | |
TiC (sintered) | Ta (sintered) | |
TiW (sintered) | Si/Co (sintered) | |
NiCr (sintered) | — |
Ni is a good Schottky contact when the sintering temperature is lower than 500 °C on lightly doped SiC. When sintering is carried out at above 700–800 °C on relatively heavily doped n-type SiC, Ni forms a good ohmic contact [322–327, 352–358]. Figure 6.84 shows the contact resistivity as a function of sintering temperature for Ni/n-type 4H-SiC. Sintering was done for 2 min by a RTP. The thickness of deposited Ni is 100 nm, and the donor density of SiC is about . The contact resistivity decreases significantly with increasing sintering temperature, and is almost saturated at about above 1000 °C. Since Ni does not form carbides, a carbon film or clusters can be formed near the interface as well as the metal surface. The carbon layer on the metal surface must be carefully removed (if formed) to ensure low contact resistivity and to improve adhesion of the subsequent layer of metals. When the Ni layer is too thin, the contact resistivity is increased. On the other hand, the surface is severely roughened when the Ni layer is too thick. Optimum Ni thickness is about 50–100 nm. Figure 6.85 shows the contact resistivity of Ni versus the donor density of 4H-SiC; sintering was performed at 1000 °C for 2 min by RTP. To obtain a low contact resistivity , the doping concentration should be increased to .
In spite of its long history, the mechanism of ohmic behavior for Ni contacts to n-type SiC is not fully understood. Physical characterization by RBS and AES showed that Ni and SiC react to form during sintering [353]. formation already occurs at a sintering temperature of 600 °C, although the resultant contact is not ohmic. To achieve good ohmic characteristics, further sintering (typically at higher temperature) is required. During this sintering, some carbon is accumulated near the interface and some carbon moves up to the surface. It is suggested that the excess carbon formed near the interface by high-temperature sintering yields a low barrier height on n-type 4H-SiC, and a role for intermediate carbon in the ohmic behavior is argued [359]. However, further basic studies are required to elucidate the ohmic mechanism.
Since Ni does not form carbides, the control of excess carbon is a critical issue. To overcome this issue, [360], Ti-based alloys [361, 362], Ta-based alloys [363], W-based alloys [364, 365], Co-based alloys [366], and several multilayer structures have been investigated. At present, however, Ni is the most popular metal for ohmic contacts employed in real device fabrication, though modifications are made in different groups.
As predicted from the contact theory, a metal with a very low barrier height can work as an ohmic contact without sintering (as-deposited ohmic contact) if the doping density of the semiconductor is sufficiently high. This is true even in n-type SiC. Al and Ti are good ohmic contacts without a sintering process if the donor density is higher than . The contact resistivity is approximately to for as-deposited contacts and it can be reduced by high-temperature sintering.
In power MOSFETs and junction field-effect transistors (JFETs), a contact, which exhibits ohmic behavior to both n-type and p-type SiC, is required to simplify the fabrication process. Such a simultaneous ohmic contact can be formed by using Ni when the p-type SiC is heavily doped [357, 358]. TiW [367] or Al/Ti/Ni [368], sintered at 900–950 °C, also form simultaneous ohmic contacts.
An Al-based metal is a good ohmic contact to p-type SiC after sintering at 900–1000 °C. Although aluminum is an effective acceptor in SiC, no direct evidence on Al doping after sintering has been reported. An obvious problem in this system is the low melting point of Al (about 630 °C). Because of severe segregation of Al during high-temperature sintering, it is difficult to form uniform ohmic contacts using a pure Al metal sintered at high temperature. To solve this problem, AlSi alloy or Al/Ti stacks have been commonly employed [369–371]. In particular, Al/Ti and its modification (e.g., Al/Ni/Ti) are the standard ohmic contacts to p-type SiC [369–376]. An optimized stack structure is Al (300 nm)/Ti (80 nm)/SiC [325, 357]. Figure 6.86 shows the contact resistivity of Al/Ti versus the acceptor density of 4H-SiC; sintering was performed at 1000 °C for 2 min by RTP. To obtain a contact resistivity , the doping concentration should be increased to . TEM observation revealed that is the main phase in contact with SiC after the sintering process [377, 378]. It is, however, not very clear how this compound contributes to the ohmic behavior. On p-type SiC, as-deposited ohmic contacts can be obtained if the acceptor concentration is extremely high . However, the contact resistivity on heavily doped p-type SiC formed by implantation is always much higher than that on heavily Al-doped epitaxial layers with the same acceptor density. Other than Al/Ti, Pd-based metals [379], Ni-based metals [380, 381], Ti-based metals [362], and an Al/Ti/Ge stack [377] have been investigated.
The long-term stability of ohmic contacts has been assessed. The contact resistivity and the surface roughness are monitored after high-temperature aging in an inert atmosphere and promising results are reported. For example, no degradation is observed after aging at 300 °C for 5000 h nor at 500 °C for 500 h for both Ni/n-SiC and Al/Ti/p-SiC contacts [323, 325, 357]. Therefore, the ohmic contacts are rather stable, and the contact reliability will not be limited by degradation of ohmic contacts themselves. Reactions between the interconnecting metal and a dielectric layer are of more concern.
Basic issues of device process technologies in SiC were described. Selective doping of donors and acceptors by ion implantation is feasible, and relatively high activation ratio values over 90% are obtained unless the implant dose is very high . Formation of heavily-doped n-type SiC is successful (sheet by high-dose phosphorus implantation), while the resistivity of heavily-doped p-type SiC formed by aluminum implantation needs further improvement. The surface roughening during high-temperature activation annealing has been greatly reduced by employing a carbon cap. The pn junctions formed by aluminum implantation into n-type SiC and np junctions formed by nitrogen or phosphorus implantation into p-type SiC exhibit good characteristics. However, a high density of extended and point defects are generated inside the implanted region as well as the tail region. Impacts of these defects on SiC device performance should be carefully investigated. To produce lightly-doped n-type SiC with high uniformity, neutron transmutation doping (NTD), which utilizes conversion from to by neutron irradiation, is promising for the future [382].
Dry etching is relatively easy in SiC, while wet etching is not a choice for device fabrication. Both fluorine and chlorine chemistries give reasonable results for SiC etching. deposited by CVD is the preferred masking material. Remaining issues include increasing the etching rate and improving the etching selectivity against the mask material. Control of etching profiles has been tried.
Control of MOS interfaces and their accurate characterization are still big challenges. In spite of extensive efforts to improve interface quality, the interface state density of 4H-SiC is still very high. Although the n-channel mobility has been improved to about on 4H-SiC(0001), on , and on and , the mobility is significantly lower in processed power MOSFETs. Both thermal oxides and deposited oxides have individual advantages and disadvantages. Importantly, the improvement of interface properties has been achieved by process optimization without solid understanding of the underlying physics. Insights into the origin of interface states as well as the physical/chemical structure of the SiC MOS interface are very limited. Therefore, many more fundamental experimental and theoretical studies are required in the future. Regarding the oxide reliability, continuous progress has been made, and a sufficiently long life at elevated temperature of 200–250 °C has been reported by several groups. However, this still needs to be proven in real MOS devices with a large chip size.
A basic process to form Schottky and ohmic contacts has been established. Owing to the lack of surface Fermi-level pinning, the Schottky barrier height can be well controlled over a wide range for both n-type and p-type SiC. Ni and Al/Ti are the standard ohmic contacts to n-type and p-type SiC, respectively. To obtain a low contact resistivity , however, sintering at 950–1000 °C and high doping density are required. The physical/chemical mechanism of ohmic behaviors is not fully understood at present.