MATERIALS SCIENCE & ENGINEERING ELSEVIER
Materials Science and Engineering B34 (1995) 83-105
B
Critical review
A critical review of ohmic and rectifying contacts for silicon carbide Lisa M. Porter, Robert F. Davis Department of Materials Science and Engineering, North Carolina State University, Raleigh, NC 27695-7907, USA Received 10 May 1995
Abstract
For more than three decades, SiC has been investigated as a wide band gap semiconductor. This paper reviews ohmic and rectifying metal contacts on n- and p-type ~- and fl-SiC reported throughout that time period. Electrical characteristics, Schottky barrier heights (SBHs), thermal stability, and chemical reactions are discussed. Most metals formed very good rectifying contacts in the as-deposited condition on both n- and p-type 6H-SiC with Schottky barrier heights ~>1 eV. Low ideality factors (n < 1.1) and high breakdown voltages (> 1100 V) have been displayed in this material. The electrical properties of contacts on 3C-SiC have been more dependent on the quality of the 3C-SiC films, which have been plagued by higher defect densities than 6H-SiC. In general, a partial pinning of the Fermi level has been evidenced by positive correlations, which are less than 1 (~0.2-0.6), between the SBHs and the metal work functions. Ohmic contacts with low contact resistivities (~< 10 -5 f~ cm2), especially important for high power applications, on any of the SiC polytypes have been exceptionally difficult to achieve. Most of the ohmic contacts have relied on high doping concentrations in combination with annealing at temperatures between 800 and 1300 °C. Annealed Ni and A1 have primarily been used in ohmic contact metallizations for n- and p-type SiC, respectively. The tendency of SiC to react with metals to form carbides and/or silicides at potential device operating temperatures ( ~ 600 °C) can be a problem for potential long term applications at high temperature. These critical issues are discussed along with future perspectives for research approaches.
Keywords: Silicon carbide; Semiconductors; Ohmic metal contacts; Rectifying metal contacts
I. Introduction
In certain close-packed materials such as SiC, there exists a one-dimensional type of polymorphism called polytypism. Polytypes are alike within the closestpacked planes but differ in the stacking sequence in the dimension perpendicular to these planes. The stacking sequence of these closest-packed planes is commonly described by an ABC notation. In SiC each letter refers to a bilayer composed of individual layers of Si and C. A repetitive ABC stacking sequence yields the zinc blende structure. This is the only cubic SiC polytype and is referred to as 3C- or fl-SiC, where the 3 refers to the number of Si/C bilayers in the periodic sequence. The hexagonal (wurtzite) ( A B A B . . . ) sequence is also found in SiC. Both can also occur in more complex, intermixed forms yielding a wider range of ordered, larger period, hexagonal (H) or rhombohedral (R) structures of which 6H-SiC is the most common. In 0921-5107/95/$09.50 © 1995 - - Elsevier Science S.A. All rights reserved S S D I 0921 -5107(95)01276-1
actuality, this polytype is only 33% hexagonal. Its ABCACB stacking sequence consists of two (identical neighboring) hexagonal bilayers and four (different neighboring) cubic bilayers. All of these noncubic structures are known collectively as a-SiC. The multitude of extreme thermal and electronic properties of SiC produces numerous and novel combinations of attributes for a semiconductor material for both optoelectronic and for high-power, -frequency, -temperature, -speed, and radiation hard microelectronic devices. The great potential of (6H)-SiC as a semiconductor in such electronic device applications is challenged by the difficulty of controlling metal contact properties. These properties of the metal/SiC interfaces include uniformity and thickness of the interfacial region, stability at high temperatures ( ~ 600 °C), and most importantly, the Schottky barrier height (SBH), or the energy barrier for electrons traversing the interface.
84
L.M. Porter, R . F Davis / Materials" Science and Engineering B34 (1995) 83 105
Table 1 Selected structural, thermal, and electronic properties of 6H- and 3C-SiC at 300 K Property
6H-SiC
Ref.
3C-SiC
Ref.
Band gap Lattice parameter
Eg = 3.0 eV a = 3.08086 A c= 15.1174 A ~s = 4.75 + / - 0.10 eV (n-type, (0001) surface): 4.85 + / - 0.10 eV (p-type, (0001) surface) Zs = 3.3 eV
[31 [116]
2.3 eV a - 3.08269/k e = 7.55124/~ ~s ~ 5.2 eV
[3] [116]
Zs - 4.0 eV
[12,117] [119]
[120]
/Ln=980cm 2 (Vs) t (at n =4.0× 1016cm 3) Usat = 2.5 x 107 cm S i
E R = 4 x 106Vcm i
[122]
EB - 3 × 106Vcm i
[122]
~:1(0)= 10.03 c±(0) = 9.66 ~;11(3c)= 6.70 ~:±(oc)= 6.52 ~rT=4.9Wcm I deg-t
[123]
~:(0)= 9.72 c(,z) = 6.52
[1231
[124]
aT--4.9Wcm --Ideg ;
[124]
Work function
Electron affinity Electron mobility Saturated electron drift velocity Breakdown electric field Dielectric constant Thermal conductivity
/~n-700cm 2 (Vs) I (at n= 1.1 x 1016cm 3) vs~,t= 2.0 x 107 cm s i
[12]
calculated from [12] [118]
T h e S c h o t t k y b a r r i e r height d e t e r m i n e s the electrical b e h a v i o r o f an o h m i c o r S c h o t t k y contact. A n o h m i c contact, i m p o r t a n t for m a k i n g o u t s i d e c o m m u n i c a t i o n to a device, is defined as having: (1) a linear a n d s y m m e t r i c c u r r e n t - v o l t a g e r e l a t i o n s h i p for positive a n d negative voltages; a n d (2) negligible resistance c o m p a r e d with the b u l k o f the device. Therefore, a low S B H is necessary to create a g o o d o h m i c contact. H o w e v e r , a large S B H is necessary to create a g o o d S c h o t t k y , o r rectifying, contact. A rectifying c o n t a c t allows c u r r e n t to flow for only one sign (positive o r negative) o f v o l t a g e bias, t h e r e b y b e c o m i n g a c o m p o nent o f the active region o f the device. Because o f the critical i m p o r t a n c e o f the S B H (also written s y m b o l i c a l l y as ~B), m u c h research has been d e v o t e d to achieving an u n d e r s t a n d i n g o f the u n d e r l y ing p h e n o m e n a , its c o n t r o l , a n d the e x p e r i m e n t a l p r o cess routes by which the latter m a y b e achieved. W h e n two m a t e r i a l s c o m e in contact, the respective F e r m i levels m u s t align. In 1938 M o t t [1] defined a relationship which is n o w k n o w n as the S c h o t t k y M o t t limit for an ideal metal s e m i c o n d u c t o r contact. This relat i o n s h i p states t h a t (for an n - t y p e s e m i c o n d u c t o r ) the b a r r i e r height is equal to the w o r k function o f the metal, ~M, m i n u s the electron affinity o f the semicond u c t o r , Zs (1)B = ( ~ M - - X S
(1)
This r e l a t i o n s h i p implies t h a t c o n t r o l over the b a r r i e r height is achieved b y the choice o f metal. In practice, however, it is c o m m o n l y f o u n d t h a t ~B is either w e a k l y d e p e n d e n t on o r i n d e p e n d e n t o f ¢~M. This
calculated from Xs value
[121]
e m p i r i c a l d e v i a t i o n from the S c h o t t k y - M o t t limit led to the d e v e l o p m e n t o f a t h e o r y b a s e d on surface states r a t h e r t h a n w o r k function differences [2]. Because the surface o f a m a t e r i a l is itself an imperfection, surface states can exist within the energy gap o f the bulk s e m i c o n d u c t o r . Bardeen f o u n d t h a t the density o f surface states can be so high t h a t they p r e d e t e r m i n e the b a r r i e r height, such that ~B is c o m p l e t e l y i n d e p e n d e n t o f q~M- In this case the b a r r i e r height o r the F e r m i level is said to be ' p i n n e d ' , a n d c o n t r o l over the b a r r i e r height is lost. Usually the actual r e l a t i o n s h i p between ~B a n d ~ u falls s o m e w h e r e between the S c h o t t k y M o r t limit a n d the B a r d e e n limit. In a d d i t i o n to the electronic interactions at an 'intim a t e ' m e t a l - s e m i c o n d u c t o r contact, the chemical a n d structural interactions m u s t be considered. M o s t b a r r i e r height m o d e l s d o not address chemical b o n d s at the interface, yet few interfaces c o n t a i n m a t e r i a l s which are inert with respect to each other. T h e structure m a y be i m p o r t a n t in terms o f affecting the n u m b e r o f surface (interface) states, which in t u r n affects the v a r i a b i l i t y o f the b a r r i e r height.
2. Important properties of silicon carbide T a b l e 1 lists some o f the structural, thermal, a n d electronic p r o p e r t i e s o f c~(6H)-SiC a n d fl(3C)-SiC. T h e b a n d gaps o f the 6H a n d 3C p o l y t y p e s at r o o m temperature are ~ 3 . 0 a n d ~ 2 . 3 eV, respectively [3]. Different rating systems for s e m i c o n d u c t o r s , d e p e n d e n t on the p a r t i c u l a r o p e r a t i n g specifications, have been d e v e l o p e d
L.M. Porter, R.F. Davis / Materials Science and Engineering B34 (1995) 83-105
as "figures of merit." Johnson's figure of merit [4],
(EBVsat/TC)2 is based on high frequency and high power operation in discrete devices. Keyes' figure of merit [5], av(Vs,t/e) 1/2 considers high speed operation in devices in integrated circuits. Thus, both a high breakdown electric field, EB, and a high saturated electron drift velocity, Vsat, are desired for high frequency/power operating conditions. In the case of SiC, it is the high breakdown field which gives it a high rating. A high thermal conductivity, aT, is needed to dissipate heat quickly from small, closely-spaced devices. The conductivity of SiC is superior to that of Cu at room temperature. Baliga's figure of merit [6], (g,t/Eg3) is based on the need to minimize conduction losses in power field effect transistors (FETs). Here e is the dielectric constant of the semiconductor,/z is the electron mobility, and Eg is the band gap. Baliga later developed a figure of merit for high frequency applications which accounts for switching losses [7], (/,tEBZV~°5/(2VBI'5)), where V~ is the gate drive voltage and VB is the breakdown voltage. These four figures of merit for a few semiconductors are normalized with respect to Si in Table 2. The very high figures of merit for 6H-SiC combined with the ability to grow boules and films as well as dope the material both n- and p-type make it a distinguished wide band gap semiconductor. The reader is referred to Refs. [8-11] for detailed descriptions of properties and technological status of SiC and other wide band gap semiconductors. Two other properties of the semiconductor which are important in terms of contacts are work function and electron affinity. Pelletier et al. [12] measured under ultra high vacuum conditions the work function of nand p-type 6H-SiC as a function of temperature between 300 and 1600 K, and from these measurements the electron affinity may be calculated. The n-type material was doped with N and the p-type material with A1. Special attention was given to cleaning the surface so as not to introduce extrinsic surface states caused by impurities. The normally Si-terminated (0001) surfaces were heated to T~< 1230°C. Higher Table 2 Figures of merit calculated for selected semiconductor materials ratioed to silicon: Johnson's figure of merit (JFM), Keyes' figure of merit (KFM), Baliga's figure of merit (BFM), and Baliga's high frequency figure of merit (BHFFM) Material
JFM [115] KFM [115] BFM [7,125] BHFFM [7,125]
Si GaAs InP GaN 6H-SiC 3C-SiC Diamond
1.0 7 16 282 695 1137 8206
1.0 0.456 0.608 1.76 5.12 5.8 32.2
1.0 13 l0 910 106 40 8574
1.0 10 7 100 13 12 454
85
temperatures were not used in order to avoid the formation of excess carbon on the surface as a result of Si eyaporation. However, heat treatment at those temperatures should still leave a C-rich surface according to [13]. The work function remained constant for each dopant type within the temperature range measured. In addition, the difference in work functions between nand p-type samples (4.75 and 4.85 eV, respectively) was small. These results led to the conclusion that the Fermi level is pinned at the surface due to intrinsic surface states. However, if the surfaces were graphite terminated, as is likely for the temperatures used, these surface states may have been extrinsic in nature rather than intrinsic. The electron affinity, which is closely related to the work function, was also calculated by this group. Using the method of Frese [14], where Zs = O s - 0.5Eg, the room temperature electron affinity was calculated to be 3.7eV. However, using the intrinsic work function value of 4.80 eV and subtracting 1.5 eV, which is half of the band gap, a value of 3.3 eV is obtained. Because the 3.3 eV value follows from the measured work function of Pelletier et al., it has been used in the present research as the electron affinity of 6H-SiC to calculate predicted Schottky barrier heights. Any error in Zs simply shifts all of the theoretical barrier heights by the same amount and, therefore, does not affect the slope of • B vs. OM. While the room temperature work function was not measured for 3C-SiC, the room temperature electron affinity was reported to be 4.0 eV. Using the method stated above, this would correspond with a work function of 5.2 eV for intrinsic fl-SiC.
3. Surface chemistry As the preparation of a clean semiconductor surface is critically important to the resulting contact properties, this section briefly reviews the various cleaning procedures used on SiC, the resulting chemistry and structure of the surface, and associated problems. Preparation of a chemically-clean, atomically-ordered surface has proven to be a nontrivial problem in SiC. The surface chemistry and structure of single crystals of the e (6H) polytype were first investigated in 1975 by van Bommel et al. [13]. The surfaces were monitored with low energy electron diffraction (LEED) and Auger electron spectroscopy (AES) after exposure to various temperatures under ultra high vacuum (UHV) conditions. As a result of low temperature heating (at 250 °C) for an unspecified amount of time, the Si-terminated (0001) surface exhibited a _x/3 × ~/3 LEED pattern, while the C-terminated (0001) surface exhibited a 1 x 1 pattern. A 6x/3 × 6x/3, R30 ° pattern began to develop on the (0001) face during successive 15min anneals at 800 °C; this became the only pattern after
86
L.M. Porter, R.F. Davis / Materials Science and Engineering B34 (1995) 83 105
heat treatment at 1000 °C. As a result of annealing at 1500 °C the surface exhibited a 1 x 1 graphite phase. The (000]-) face showed 2 x 2 or 3 x 3 patterns after 800 and 1000 °C. After 1500 °C graphite rings appeared on this latter face, indicative of polycrystalline graphite. The surfaces after the 800 °C exposure showed AES C peaks of the carbide type. After 1000 °C the Si-to-C ratio decreased, and the C peak changed from carbidic to graphitic. These two results, along with the development of a peak at 20 eV, indicated the formation of graphite. The L E E D results at 250 °C were interpreted by van Bommel et al. as being representative of reconstructed surfaces with possible submonolayer impurities. The 6x/3, 30 ° pattern was attributed to the SiC substrate with a monocrystalline graphite top layer, which had rotated 30 ° with respect to the underlying SiC lattice. Thus, under the U H V conditions used by these investigators graphite begins to form at the (0001) surface as low as 800 °C. Graphite formation occurred at lower temperatures on the (000i) surface. On both faces the mechanism was reported to consist of the collapse of three C layers to form one graphite layer due to the incongruent evaporation of Si. Many orientations are possible, but apparently only one is favored on the Si-face. Following the work of van Bommel, researchers have experimented with methods to clean the SiC surface without forming a graphitic surface layer. Kaplan and Parrill [15] heated fl-SiC (100) and 6H-SiC (0001) and (000i) samples at progressively higher temperatures, typically for 3 min, in the presence of a Ga flux to volatilize the surface oxide. The oxide was reduced at temperatures as low as 850 °C; however, complete removal was a very slow process below 1050°C. In addition, temperatures which resulted in complete removal of oxygen also resulted in a decrease of the Si/C ratio. After annealing at 1185 °C for 3 rain, the Si/C ratio decreased significantly, and the data indicated the initial formation of graphite. In fl-SiC C2 x 2 structures were observed by LEED, which were in agreement with the results of Dayan [16] for comparable temperatures. The C-face of 6H-SiC displayed 1 x 1 patterns, except after extended heating at 900 °C, when some areas displayed a x/3 x x/3 reconstruction. The Si-face displayed a 3 × 3 reconstruction. The residual, tenacious oxide was thought to be located in regions of the surface where oxygen can penetrate well into the surface defects. Therefore, complete oxide removal on rough surfaces should be more difficult. Although the surface oxide was volatilized by the Ga, the temperatures required to completely remove the oxide resulted in the same problem of graphite termination. In later work [17] Kaplan heated SiC in the presence of an elemental Si flux, which removed the surface
oxide as volatilized SiO. Prior to this step and to loading into the U H V chamber, the samples were chemically cleaned successively in detergent solution, trichloroethylene, acetone, and methanol. Some samples were also dipped in hydrofluoric acid followed by deionized water, but this treatment did not noticeably affect the amount of residual oxide. The 6H-SiC (0001) and fl-SiC (111) surfaces were indistinguishable. Faint satellite spots appeared with a small amount of oxide still present. After heat treatment at 850, 950, and > 1000 °C, the LEED patterns were 3 x 3; (x/3 x x/3), R30°; and 1 x 1, respectively. The Si surface concentration decreased with temperature. Parallel electron energy loss spectroscopy (PEELS) revealed the beginning of graphite segregation at 950 °C. The 3 x 3 phase was asserted to correspond with an adsorbed Si bilayer with a high vacancy concentration. Annealing at 950 °C depleted the surface Si and caused the surface reconstruction to change to a x/3 x x/3 phase. Additional annealing at > 1000 °C resulted in continual depletion of surface Si, and the development of a corresponding 1 x 1 pattern. The 1 x 1 phase was attributed to a range of surface chemistries and defect structures. As would be expected from the above model, with the continued Si depletion the amount of graphitic C increased. The initial appearance of the 1 x 1 phase was believed to be due to the elimination of the Si adlayer resulting in defects. By comparison, the work of van Bommel exhibited a direct change from x/3 x x/3 to a graphitic state without first exhibiting a 1 x 1 phase. In order to more efficiently remove the oxide at low temperatures, metals which form vapor phase oxides with more negative free energies than O bound to Si are worth considering as an extension of the experiments with Ga and Si. A promising candidate might be A1, which readily forms vapor phase oxides (A10 and A120) and which may be easily evaporated from the surface. Other possible choices are Mg or Ca. Bozso et al. [18] investigated the Si-terminated (0001) surface of 6H-SiC crystals using X-ray photoelectron spectroscopy (XPS) and X-ray and electron excited Auger spectroscopy. Prior to these analytical studies, the crystals were dipped into concentrated aqueous HF, rinsed in distilled H20, and cleaned in acetone and propyl alcohol. The major residual contaminants left were fluorine and oxygen. After the chemical clean the samples were exposed to a combination of low angle ( ~ 80 ° angle of incidence) argon ion sputtering (500 eV) and high temperature annealing. Fluorine and oxygen remained after this treatment at 800 °C for 40 min but were removed at 1000 °C for 2 min. However, Ar was still present, but was eventually eliminated after prolonged heating. An atomically clean surface was achieved, and preferential removal of Si or C was not observed. However, possible surface damage due to the sputtering process was not reported. The resulting state
L.M. Porter, R.F. Davis / Materials Science and Engineering B34 (1995) 83-105
of the surface should be further investigated before this preparation technique is employed for metal contacts, as defect states can pin the Fermi level. If the annealing prevents surface damage, this technique may be a viable solution to the surface cleaning problem. The Si- and C-faces of a mixture of 6H and 15R polytypes were investigated after cleaning and disordering by Ar ion sputtering (similar conditions as above) and subsequent annealing [19]. In agreement with the previous study [18], the Si/C ratio remained constant. Changes did occur with heat treatment to ~ 1000 °C, which caused a decrease in the XPS C (Is) binding energy. The Si (2p and 2s) binding energies remained constant. These results indicated a change in the C chemistry. Examination of the peak areas revealed an increase in the C/Si ratio on the C-face between 630 and 830 °C and a substantial increase on both faces above 1030 °C. The C enrichment was much higher on the C-face. These results were reaffirmed by the AES analyses. Electron energy loss spectra indicated the formation of graphite at ~ 900 °C. While in the latter two studies 500 eV Ar ions were used in the sputtering process, Andreev et al. [20] sputtered oxidized (0001) and (0001) surfaces of 6H-SiC with 640 eV electrons at a flux density of 2 - 5 x 1017 e/ (cm 2 s). The sample temperature was between 700 and 1000 °C. Electrons were used instead of ions because of their low mass and, hence, lower tendency to damage the SiC surface. Auger analysis showed that this procedure resulted in an uncontaminated, stoichiometric surface. However, it was stated that graphitization occurred when the procedure was continued after the oxide was removed. Although evidence was not given, the surface was reported to have been undamaged. Considering the chemical and microstructural results along with the experimental process, this cleaning procedure may be a viable technique if the equipment is available and the parameters are carefully controlled. A chemical and thermal cleaning process [21], which would be more easily implemented for commercial devices, was developed in conjunction with the fabrication and characterization of Schottky contacts on SiC. Oxidized 6H-SiC (0001) surfaces were first etched in a 10% H F aqueous solution. Thermal oxidation was found to be a critical step for leaving a non-C-rich surface. Hydrocarbons were the main contaminant left on the surface after the etching step, as shown by the minor peak component on the high binding energy side of the C ls XPS peak (Fig. 1). After heating the samples in a vacuum system (base pressure ~ 1 5 x l0 9 Torr) at 700 °C for 15 min, these hydrocarbons were removed from the surface, while residual O and a trace amount of F were still present. As illustrated in Fig. 2, XPS was also used to determine the amount of band bending before and after deposition of various metal contacts. After these cleaning steps an
87
(b) C ls tf
f
..........Before 700°C desorption ..... After 700°0 desorption
.~/
:.:
.,"
,,"
~
~
~
.,?'
I
. . . .
295
I
. . . .
I
.
.
.
.
290 285 Binding Energy (eV)
.
.
"
280
Fig. 1. XPS C ls spectra of 6H-SiC (0001) after 10 min in 10% HF followed by a deionized water rinse. The minor peak attributed to hydrocarbons was removed after heating at 700 °C for 15 min in UHV [21].
upward band bending of a few tenths of an electronvolt was calculated, indicating the presence of surface states. Evidence for the absence of surface states in the band gap under different conditions has been reported by Parrill and Bermudez [22]. Alpha SiC, of unspecified polytype and undetermined Si or C termination, was chemically cleaned in acetone, methanol, 30s in concentrated aqueous HF, and a final rinse in methanol followed by annealing at 850-950 °C in a Si-flux. All of the oxygen was removed, but the surfaces did contain excess Si. The excess Si was removed by annealing for 1-2 min at temperatures ranging from 960 to 1160 °C. The surface showed a 1 x 1 L E E D pattern after annealing in the Si flux. Subsequent heating at ,~ 1160 °C gave a 3 x 3 reconstruction and no graphite formation according to Auger analysis. Data was collected only from the 1 x 1 surface (Si-rich surface) since 3 x 3 surfaces were not uniform. The bulk valence band was studied by using high energy photons (Mg K~, 1253.6eV) for which the photoelectron attenuation length is ~ 7 monolayers in the valence band. The
Ec EF . . . . . . . .
J -
j
Ev >o
J EC is
J
Fig. 2. Energy band diagram of 6H-SiC showing the relationship between • B and the binding energies measured by XPS. Prior to metal deposition, the value of • B was calculated to be 0.40 + 0.1 eV
[21].
88
L.M. Porter, R.F. Davis' / Materials Science and Engineering B34 (1995) 83 105
surface valence band was investigated using low energy photons (Zr M-zeta, 151.4 eV) for which the attenuation length is ~2.5 monolayers. The photoemission results showed no states within the optical band gap; however, they do suggest that surface states exist below the valence band edge. The evidence seems to indicate that certain adatom terminations (e.g. Si or O) on the (0001) or (0001) surface of a-SiC result in a reasonably low or the absence of surface states within the band gap. In the case of the partial monolayer of oxygen, there appears to be a sufficient density of surface states to reduce the index of behavior, defined as the slope of the plot of the barrier height vs. metal work function, from 1.0 but which is insufficient to actually completely pin the Fermi level (shown in later sections). An index of behavior of 1.0 for a particular semiconductor indicates that the Fermi level is unpinned, and the Schottky barrier heights follow the ideal behavior predicted by the Schottky Mott relationship. Other work [23] does indicate that there is some pinning due to intrinsic C vacancies, which vary in concentration with polytype. The barrier heights of Cr on m a n y different SiC polytypes were compared with the theoretically predicted behavior of vacancies on the surface of a semiconductor. Carbon-terminated surfaces were prepared by chemical cleaning, which was not described, and then heated in vacuum at 600 °C for 1 h. The barrier height correlated well with the degree of hexagonality for several hexagonal and rhombohedral polytypes. Although the band gap varies similarly, the Fermi level did not sit at a constant value of the band gap, as would occur if this were purely a function of band gap. As the concentration of vacancies increased, the barrier height decreased. (The barrier height of Cr/6H-SiC was 1.1-1.2 eV.) The report claims that carbon vacancies form a level at the surface in the upper half of the band gap and that a higher concentration of carbon vacancies causes a greater shift of Ev towards E c. Comparing this theory with SBH results presented later, one might conclude that intrinsic properties (vacancies) of SiC m a y cause some gap states at the surface, but the specific details of the surface and its preparation will be the final factor in determining the density of surface states and where they are located.
4. Contacts to ~-silicon carbide
4.1. Schottky contacts to or-SiC 4. I. 1. n-type a-SiC Some of the early investigations of Schottky contacts on n-type ~(6H)-SiC (summarized in Table 3) were performed in 1964 by Mead and Spitzer [24]. In this report barrier heights for metals on SiC and other
elemental G r o u p IV and I I I - V semiconductors were found to be nearly independent of work function. Although the work functions of Au and A1 differ by ~0.8 eV, their resulting barrier heights on SiC were 1.95 and 2.0 eV, respectively. It was reported that for most G r o u p IV and III V semiconductors where both vacuum cleavage and chemical preparation could be employed, the barrier heights resulting from the two preparation techniques were the same; however, it was not specified if both processes were used for SiC. Therefore, it cannot be concluded whether the results can be attributed to the surface preparation. Subsequently, Hagen [25] also found q~B on a-SiC crystals, comprised of the 6H and 15R polytypes, to be independent of qbM. The barrier height on n-type samples was 1.45 eV for Au, Ag, and A1. In addition, q~B for Au contacts on etched (0001) surfaces was the same as that on cleaved (IT00) surfaces. The former surfaces were etched in a molten s o d a - b o r a x mixture, rinsed with deionized water, treated with an H F solution, again rinsed with deionized water and kept in methanol until placed in the vacuum system. It was concluded that the Fermi level is fixed by surface states at the middle of the band gap. Notable results were reported for Au contacts thermally evaporated (base pressure < 10 6 Torr) on n-type 6H-SiC (0001) [26]. The C-terminated surfaces were cleaned with concentrated HF, etched in molten sodium peroxide, rinsed in dilute HC1 and deionized water and then dipped in methanol and dried with N2. The semilogarithmic current vs. voltage plot extends over seven orders of magnitude with an ideality factor of n = 1.07. The calculated barrier height of 1.45 eV was in agreement with that of Hagen [25] for Au on the Si-terminated (0001) surface. Platinum, Au, Ag, Ni, Cr, and A1 sputtered on 6H-SiC produced rectifying contacts [27]. Surfaces were etched in a mixture of potassium hydroxide and sodium nitrate in ratios ranging from 1:4 to 1:100 at 4 7 7 577 °C. However, the carrier type and orientation of the SiC were not indicated. Depending on the metal and the surface preparation, the band bending at the semiconductor surface ranged from 0.6 to 2.5 eV. Aluminum evaporated in vacuum (10 ~'Torr) on n-type SiC was characterized after heat treatment by another group [28]. The substrates were etched in molten K O H . The resulting oxide was then etched by dipping in H F followed by deionized water. Prior to the A1 deposition Ni was deposited on the back and annealed for 3 min at 1200 °C in an Ar ambient to create an ohmic contact. After the AI was deposited (vacuum pressure 10 6 Torr) the samples were annealed in an Ar ambient for 3 min at 660 °C (m.p. of A1), 770 and 900 °C. Before annealing, the AI contacts showed high leakage currents. After annealing, the leakage decreased ( ~ 2 × 10 2 A c m 2 at - 5 V ) , and the ideality factor
L.M. Porter, R.F. Davis / Materials Science and Engineering B34 (1995) 83 105
89
Table 3 R e c t i f y i n g c o n t a c t s o n n - t y p e ~ ( 6 H o r m i x e d ) - S i C . O n l y t h o s e studies w h i c h r e p o r t e d S B H s a r e listed h e r e f o r brevity. T h e s u r f a c e p r e p a r a t i o n s w h i c h c o n s i s t e d o f a t least a s u r f a c e o x i d e e t c h i n g step a n d a h y d r o c a r b o n r e m o v a l step b y h e a t i n g in h i g h o r u l t r a - h i g h v a c u u m w e r e r e a t e d as ' v e r y g o o d ' . T h o s e w h i c h c o n s i s t e d o f at least a c h e m i c a l c l e a n a n d a s u r f a c e o x i d e e t c h i n g step b u t l a c k e d the c a p a b i l i t y f o r h e a t i n g in h i g h v a c u u m w e r e r a t e d as ' g o o d ' . T h i s r a t i n g s y s t e m is b a s e d o n a n a l y s e s o f the SiC s u r f a c e b y X P S Metallization
Deposition method
Deposition temperature
Deposition pressure (Torr)
Annealing condition
~ a (eV)
SiC surface
Method of qba m e a s .
Origin of SiC
Surface preparation
Ref.
Au A1 Au
NR NR thermal evap.
NR NR RT
NR NR ~10 9
none none none
1.95 2.0 1.45
NR NR
C-V C-V C - V , PR
NR NR NR
NR NR
[24] [24]
cleaved in
[25]
(1i00)
vac.
Ag
thermal evap.
RT
~10 9
none
1.45
(1i00)
C V, PR
NR
AI
thermal evap.
RT
~ 10 9
none
1.45
(1100)
C - V , PR
NR
Au
thermal evap.
NR
< I0 6 b.p.
none
1.40
(000T)
AI
thermal evap.
NR
10 - 6
~ 1.7
(0001)
Ti
e-beam evap.
RT
10 io b.p.
to 900 °C, 3 min none
C - V , PR, I-V C-V
0.84-0.88
(0001)
Ti
e-beam evap.
RT
10- io b.p.
0.86-1.04
(0001)
Pd
thermal evap.
RT
10 to b.p.
700 °C, 4 0 - 6 0 min none
Au
thermal evap.
RT
10 io b.p.
none
Ag
thermal evap.
RT
10 io b.p.
none
Mn
thermal evap.
RT
10 - l ° b.p.
none
1.60-1.62 1.11 1.14-1.19 1.37-1.40 1.10 0.83-0.92 0.79 0.81
(0001) (0001) (0001) (0001) (0001) (0001) (0001)
AI
thermal evap.
RT
10 lo b.p.
none
Mg
thermal evap.
RT
10 ~o b.p.
none
Pt
sputtered
140 °C
NR
none
0.84-0.89 0.26-0.30 0.33 0.30-0.34 1.04-1.10
(000I) (0001) (0001) (0001) NR
Ti
thermal evap.
RT
10 lo b.p.
none
Ti
thermal evap.
RT
10 1o b.p.
400 °C
Ni
thermal evap.
RT
10 ~o b.p.
none
Ni
thermal evap.
RT
10 ~o b.p.
Ni
thermal evap.
RT
10 - I ° b.p.
(000T) (0001) (000I) (0001) (000i) (0001) (000i) (0001) (0001)
AI
thermal evap.
RT
10 io b.p.
400°C, 30s 600 °C, 30 s none
1.0-1.03 0.73 0.98 0.93 0.97 1.54 1.68 1.24 1.29 1.51-1.66 1.23-1.25 1.16-1.39
AI
thermal evap.
RT
10 ~o b.p.
Au
thermal evap.
RT
NR
600 °C, 30s none
0.84 0.89 0.26-0.30 1.36-1.66 0.82-1.12 1.4
(0001) (0001) (000T) (0001) (000T)
I V and C-V 1 V, C - V , and XPS I V, C V, and XPS I V, C V, and XPS I V, C - V , and XPS I-V, C-V, and XPS l - V , C V, and XPS I V, C - V , and XPS I-V
Pt
e-beam evap.
RT
10 ~o b.p.
none
1.06 1.33
(0001)
I - V and XPS
Pt
e-beam evap.
RT
10 1o b.p.
1.15-1.26
(0001)
I-V
Hf
e-beam evap.
RT
10 ~o b.p.
450 750 °C, 20 min none
0.97
(0001)
I V and C - V
Hf
e-beam evap.
RT
10 ~o b.p.
Co
e-beam evap.
NR
Co
e-beam evap.
Ni
e-beam evap.
I-V, C-V, XPS l-V, C-V I - V , C V, and XPS I-V, C - V , and XPS l - V , C V, and XPS I - V , C V, and XPS I V, C - V , and XPS I - V and XPS
1.01-0.86
(0001)
l V and C V
10 7
700 °C, 20 60 min none
0.79
(0001)
1-V and C V
RT
10 ~o b.p.
none
1.06-1.15
(0001)
RT
10 ~o b.p.
none
1.14-1.21
(0001)
l - V , C V, and XPS I - V and C V
[25]
subl.
cleaved in vac. cleaved in vac. good
[25]
Acheson
good
[28]
seeded sub.; CVD epi seeded sub.; CVD epi CVD
very good
[21,30,34]
very good
[21,30,34]
very good
[32]
CVD
very good
[32]
CVD
very good
[32]
CVD
very good
[32]
CVD
very good
[32]
CVD
very good .
[32]
seeded sub.; CVD ept seeded sub.; CVD eD seeded sub.; CVD epl seeded sub.; CVD e D seeded sub.; CVD e D seeded sub.; CVD ep~ seeded sub.; CVD epi seeded sub.; CVD epi Acbeson substr.; CVD epl seeded sub.; CVD eD seeded sub.; CVD ept seeded sub.; CVD epl seeded sub.; CVD e D seeded sub.; CVD ep~ seeded sub.; CVD epl seeded sub.; CVD e D
NR
[441
very good
[33]
very good
[33]
very good
[33]
very good
[33]
very good
[33]
very good
[33]
very good
[33]
good
[45]
very good
[34,36]
very good
[34,36]
very good
[34]
very good
[34]
good
[90,126]
very good
[35]
very good
[43]
[26]
OB = S c h o t t k y b a r r i e r height; N R = n o t r e p o r t e d ; R T = r o o m t e m p e r a t u r e ; b.p. = b a s e p r e s s u r e ; sub. = s u b s t r a t e ; epi = epilayer; I V = c u r r e n t voltage; C V = capacitance voltage; XPS = X-ray photoelectron spectroscopy.
90
L.M. Porter, R.F. Davis / Materials Science and Engineering B34 (1995) 83 105
Fig. 3. High resolution TEM image of as-deposited Ti/6H-SiC. The boxed-in region in the upper part of the image is magnified in the lower part of the image. The arrow at the interface marks the position of a step in the SiC surface. The misfit dislocations which are marked with lines, did not extend all the way to the interface. After Ref. [21].
was 2.0-2.3. The built-in voltage was 1.7 V after annealing at 900 °C. It is likely that annealing the SiC at 1200 °C prior to deposition of A1 depleted some of the surface Si, which in turn may have caused leaky contacts. Reaction of A1 with the excess C very likely occurred during post-deposition annealing and resulted in lower leakage currents. The lower temperatures should not be sufficiently high to activate A1 in SiC and create a p-type surface layer. Titanium films deposited by UHV electron beam evaporation at room temperature onto n-type 6H-SiC (0001) grew epitaxially and displayed rectifying behavior [21,29 31]. Both Ti ( a = 2 . 9 5 A , c = 4 . 6 8 / k ) and 6H-SiC (a = 3.08 A, c = 15.11 A) have hexagonal crystal structures, corresponding to a - 4 % lattice mismatch between the close packed planes. Fig. 3 shows a high resolution T E M image of this nearly ideal interface. The misfit dislocations did not extend to the interface, and therefore, the Ti film was considered pseudomorphic. Surfaces were cleaned in a solution of ethanol/hydrofluoric acid/deionized water (10:1:1) followed by a 700 °C thermal desorption in UHV. Analysis of unannealed, thin ( 4 - 1 2 / k ) Ti films by XPS indicated Ti C bonding at the Ti/SiC interface. The
leakage currents of thicker 1000 ~ ) films were as low as9xl0 8 A c m 2 a t - 1 0 V , and theideality factors were between 1.01 and 1.09. The contacts remained stable after annealing at 700 °C for 1 h, which resulted in a thin, crystalline TiC layer at the SiC interface and TisSi 3 in the remainder of the reacted region. Barrier heights before and after annealing were 0.88 and 1.04 eV, respectively. The change in barrier height is attributed to the chemical reaction, which created a new metal/SiC interface. From the results of I V, C - V , and XPS analyses, Waldrop et al. [32] have calculated the barrier heights of Pd, Au, Ag, Tb, Er, Mn, A1, and Mg deposited at room temperature on the Si- and/or C-faces of n-type 6H-SiC. The metals were deposited in UHV by evaporation from W wire baskets. Prior to the depositions the substrates were RCA cleaned, oxidized, etched in HF, and desorbed at 600 °C under UHV conditions. It was estimated that ~ 1/2 monolayer oxygen remained on the C-face and ~ 3 / 4 on the Si-face. A relatively large barrier height variation resulted. With the exclusion of the rare earth elements, in general there was an increase in q)~ with q~M. This trend is displayed in Fig. 4, where the SBHs calculated from the three techniques
L.M. Porter, R.F. Davis / Materials Science and Engineering B34 (1995) 83 105
are plotted vs. the work functions of the metals and compared to the theoretical SBHs according to the S c h o t t k y - M o t t limit. Three linear fits to the three sets of data points are also shown. The slopes, S, of these lines, often referred to as the index of interface behavior, give a quantitative indication of how closely the system follows S c h o t t k y - M o t t behavior (i.e. how close S is to 1). As is evident from Fig. 4, the index of interface behavior may depend on the technique used to measure the SBHs. In this case S varied from 0.47 to 0.63. These results give evidence for a partial pinning of the Fermi level in 6H-SiC. While not shown in the figure, the barrier heights of the rare earth elements, Er and Tb, ranged from 1.61 to 2.18 eV calculated from C - V measurements and 1.04 to 1.41 eV calculated from I - V measurements. Barrier heights of unannealed and annealed Ni, Ti, and A1 were later measured by the same group on n-type 6H-SiC [33]. These data points are also included in Fig. 4. X-ray photoelectron spectroscopy analyses indicated the formation of TiC and TiSi~ in the unannealed Ti/SiC interface region. No chemical reactions were detected on the Si-face before or after annealing Ni/SiC at 400 °C. In addition to the Ti contacts described in a previous paragraph, Porter et al. [34-36] have also studied the chemistry and microstructure along with the electrical properties and SBH's of other metal/SiC systems. The SBHs calculated from I - V , C - V , and XPS analyses of Ti, Pt, HI', Co, and Ni contacts are shown in Fig. 5. The slopes S vary from 0.12 for the data points calculated from I - V measurements to 0.40 and 0.41 for those calculated from C - V and XPS measurements, respectively. While these slopes are somewhat less than those derived from Waldrop's data, both sets of results show a significant, positive correlation between SBH and metal work function and a susceptibility to measurement technique. Differences due to processing should be minimal since similar cleaning and deposition procedures and substrate material were used in both 2.52.0~-
I
-I- theoretical •", measured (I-V) I 0 measured (XPS) 0 measured (C-V)
.-
S=l;O...--"
_=z 15o
~, 1 . 0 E
~
~
~s=047
S-0.62 (XPS)
o.5-
. - ~
s=o.6s 0-v)
0.0Metal Work Function (eV) Fig.
4. D a t a
points
of
experimentally-determined
and
theoretical
barrier heights on n-type 6H-SiC (000l) vs. work function of the metal contacts. The experimental data points are from Waldrop et al. [32,33]. The slopes, S, of the linear fits through each set of data points
are indicated.
91
2.5-l.. Zx 0 0
I
2.0>A
theoretical I measured (I-V) I measured (C-V) measured (XPS 1
.e S=1.0 , ' " .-'" o• • o
~- 1 . 5 -
-•
•1-
S=0.41
o °
.~_ 1 . 0 P_
~
2
ZX
0.5-
0.03.5
I 4.0
I 4.5 510 Metal Work Function (eV)
515
610
Fig. 5. Data points of experimentally-determined and theoretical
barrier heights on n-type 6H-SiC (0001) vs. work function of the metal contacts. The experimental data points are from Porter et al. [21,34-36,43]. The slopes, S, of the linear fits through each set of data points are indicated. studies. The main difference in the two sets of results occurs between the calculations based on I - V measurements and XPS. This result is quite interesting considering that I - V measurements are very dependent on the homogeneity of the interface [37-41]. The lower slopes in the former set of data are believed to be partly due to the particular metals studied. For example, the Pt contacts yielded a significantly lower SBH calculated from I - V measurements than that calculated from XPS analyses, a result attributed to inhomogeneity of the interface [36]. In this latter study the increase in grain size and grain orientation with the temperature of the anneal was accompanied by an increase in the SBH calculated from I - V measurements. The samples were successively annealed for 20 rain at 450, 550, 650, and 750 °C. No reaction occurred at temperatures below 750 °C. The first reaction phases, crystalline PtzSi and an amorphous C phase, formed after annealing at 750 °C. This interfacial reaction was proposed to reduce any adherence problems associated with Pt contacts. These contacts displayed low ideality factors (n < 1.1) and typical leakage currents of 5 × 10 - S A c m 2 a n d 2 × 10 8 A c r e -2 at - 1 0 V before and after annealing, respectively. Cobalt contacts [35] deposited at room temperature also formed excellent rectifying contacts with ideality factors and reverse leakage currents similar to the Pt contacts. However, annealing at 800 °C for 20 min resuited in a substantial increase in the leakage currents, while ohmic-like behavior was displayed after annealing at 1000 °C for 2 min. Analysis of the latter contacts revealed that extensive reaction occurred resulting in the formation of CoSi and graphite. This reaction resulted in a highly irregular interface, as shown in Fig. 6. Nickel, which reacts with SiC at even lower temperature [42] and is chemically similar to Co, is commonly used as an ohmic contact to n-type SiC and may also form a highly irregular interface after annealing.
92
L.M. Porter, R.F. Davis / Materials Science and Engineering B34 (1995) 83 105
Fig. 6. High resolution T E M image of Co/6H-SiC interface after annealing at 1000 °C for 2 min. After Ref. [35].
As stated previously, the contacts studied by Mead and Spitzer [24] and Hagen [25] showed no correlation between SBH and metal work function. These results give strong evidence that the barrier height mechanism in both studies was due to Fermi level pinning. The fact that the Fermi level is evidently completely pinned in these studies as opposed to the more recent studies in which partial pinning is evident may be due to different surface preparation and perhaps more importantly the quality of the substrate material. Prior to about the last 10 years,
single crystalline SiC material of a single polytype was very difficult to grow. Therefore, substrate material was likely a mixture of polytypes. It should also be noted that both the source of the SiC material and the surface preparation were similar in the experiments by Waldrop et al. [32,33] and Porter et al. [21,34 36,43]. Therefore, comparisons of the results can more easily be made than in cases where the material and its preparation differ. Table 3 summarizes the Schottky barrier heights calculated for rectifying contacts on n-type 6H-SiC. Most of these calculations were based on XPS and/or I V and C - V measurements. Only those studies which reported the SBHs were included in the table for brevity. In most cases the surface termination (i.e. Si or C) is indicated in parentheses next to the calculated SBHs. Most of the metals on the Si-face had SBHs between approximately 0.8 and 1.25 eV. On the C-face most of the SBHs were between 1.0 and 1.6 eV. This limited range is most likely associated with the fact that most transition metals have work functions within about 0.8 eV of each other. The first seven contacts listed, which represent the earlier studies, showed higher SBHs; while in the more recent studies, A1 and Mg, which have lower work functions, showed lower SBHs. Schottky diodes with very high breakdown voltages and low specific on-resistances have been fabricated with Pt [44] and Au [45], which is good news for power device applications. The Pt diodes on the Si-terminated surface of 6H-SiC yielded breakdown voltages of greater than 400 V. The Au diodes on the C-terminated surface of 6H-SiC yielded breakdown voltages of greater than 1100 V. In addition, the leakage currents of the Au diodes were very low, 4 . 0 x 10 6 A c m 2 at - 2 0 0 V and 2.1 x 10 3 A cm 2 at - 1100 V. Referring to Table 3, the higher SBHs of metals on the C-terminated surface may allow higher breakdown voltages than on the Si-terminated surface.
Table 4 Rectifying contacts on p-type ~(6H or mixed)-SiC. See Table 3 for a description of the rating system for surface preparations Metallization
Deposition method
Deposition temperature
Deposition pressure (Torr)
Anneal condition
q~B (eV)
SiC surface Method of Origin of q~B meas. SiC
A1
thermal evap.
RT
~ 10 9
none
1.36, 1.45 (1100)
PR, C V
NR
Ag
thermal evap.
RT
~10 9
none
1.15
PR
NR
(1100)
Surface preparation
Ref.
cleaved in vac. cleaved in
[25] [25]
vac,
Au
thermal evap.
RT
~10
9
none
1.07, 1.37 (1100)
PR, C V
NR
cleaved in
[25]
vac,
Au
e-beam, evap.
RT
10 io b.p.
none
1.27
(0001)
I V
NiA1
e-beam evap.
RT
10 io b.p.
none
~ 1.37
(0001)
1 V
Ni
e-beam evap.
RT
10 to b.p.
none
~ 1.36
(0001)
I V
q~B=SChottky barrier height; N R = n o t reported; R T - r o o m temperature; b . p . - b a s e V = current voltage; C - V - capacitance voltage; PR = photoresponse.
seeded sub.; '~ CVD epi seeded sub.; C V D epi seeded sub.; C V D epi
very good
[127]
very good
[43]
very good
[43]
pressure; sub. =sublimation; epi=epilayer;
I
L.M. Porter, R.F. Davis / Materials Science and Engineering B34 (1995) 83-105
4.1.2. p-type a-SiC Table 4 summarizes the limited number of Schottky contacts on p-type a-SiC found in the literature and their calculated SBHs. These studies were not as extensive as m a n y of those performed for Schottky contacts on n-type SiC. Gold and A1 on p-type samples produced SBHs of 0 . 9 - 1 . 2 5 e V and 1.30-1.47eV, respectively [25]. The higher SBH of A1 compared with Au on p-type material would be expected considering the lower work function of the former. Our measurements on p-type SiC have shown consistent differences from measurements on n-type material. The SBHs tend to be higher on p-type than on n-type 6H-SiC. While leakage currents for Au, NiA1, and Ni contacts on p-type 6H-SiC (0001) were comparable with those on n-type 6H-SiC, the ideality factors were higher (n ~ 1.1-2.1). These higher ideality factors indicate that thermionic emission is not the dominant current transport mechanism. Deep level transient spectroscopy (DLTS) will probably be used to determine whether recombination at deep levels accounts for the electrical behavior of the contacts on p-type material. 4.2. Ohmic contacts to c~-SiC 4.2.1. n-type e - S i C The earliest work found in the literature on ohmic contacts to n-type a-SiC was performed in 1970 on Cr and Cr alloys [46]. The remainder of the work has been confined within the past few years. In the former case the main concerns were to form ohmic contacts which were both ductile and oxidation resistant and had low resistance. The contacts described were Cr or Cr alloys (at least 5 wt.%) with Fe, Ni, or Fe and Ni. To form ohmic contacts the metals were heated to above their respective melting temperatures (1500-1900 °C). These high temperatures assured a significant reaction with the SiC, and very likely created a graded, highly-defective interface associated with the ohmic behavior. Due to the processing requirements and the confinements of size in today's devices, lower processing temperatures are required. Although lower annealing temperatures have been used in the more recent work, in most cases the temperatures have been above 900 °C. The exceptions to this last statement have been in those cases where the SiC was very heavily doped [47,48]. The two exceptions listed in Table 5 along with the other ohmic contacts are the TiN contacts reported by Glass et al. [49,50] and the TiW contacts reported by Crofton et al. [51]. In both cases the ohmic behavior is likely associated with a thin insulating layer. The TiN contacts were deposited by ion assisted reactive evaporation and annealed at 600 °C for 30 min. They were found by XPS
93
to contain a thin (5-15 A) layer of silicon nitride at the interface, creating a metal insulator semiconductor (MIS) structure. While TiN has a low work function (favorable for an ohmic contact on n-type material), ohmic behavior did not occur in the absence of ion assisted deposition, a process associated with the formation of the S i - N layer. This insulating layer was critical to the creation of ohmic contacts but not solely responsible for this behavior, as Pt contacts (high work function) with a S i - N interlayer resulted in rectifying characteristics. It was stated that the insulating layer may act to passivate surface states, resulting in an electron energy band relationship which behaves according to the Schottky Mott limit. Because the insulating layer was so thin, electrons would be able to tunnel through it. The Ti W (10:90 wt.%) contacts formed ohmic contacts after annealing at 600 °C for 5 min following exposure of the SiC surface to an oxygen plasma. Ohmic behavior did not result without the oxygen plasma exposure. It is likely that this process step resulted in the formation of a thin SiO2 layer at the surface, which may be responsible for the ohmic behavior by forming a MIS structure similar to that for the TiN contacts. Annealed Ni has been the most widely used metal for ohmic contacts to n-type SiC. As shown in Table 5, this process generally involves annealing at temperatures above 900 °C. While Crofton et al. reported in an earlier paper [51] a room temperature contact resistivity in the mid 10 2 ~-~cm 2 range (5 x 1018 c m - 3 c a r r . conc.), they were later able to lower this value to less than 9 x 10 - 6 ~ cm 2 (7 9 x 1018 c m - 3 c a r r . conc.) [52]. The substantial decrease in the calculated contact resistivity was partly attributed by the first author (private communication) to a combination of the higher carrier concentration and of the changes in the measurement technique, the annealing conditions, and the processing systems. Comparably low values (1.7-6 x 10 5 ~ c m 2) were also reported for Ni on 6H-SiC with a 2000/k thick, highly-doped ( 1 - 3 x 10 I9 c m 3) 3C-SiC interlayer [53]. The authors attribute the low contact resistivity to the smaller band gap of the 3C polytype. To determine whether the 3C-SiC layer does in fact reduce the barrier height at the 6H-SiC surface, the energy band relationship at the interface should be measured. The lowest values reported to date of contact resistivities of ~< 1 x 10 6 ~ c m 2 [54] have been achieved for annealed (1000°C/5min) Ni on very highly doped (4.5 x 1020 cm 3) n + 6H-SiC (000i). The n + 6H-SiC films were grown by liquid phase epitaxy (LPE) on Acheson-grown 6H-SiC substrates. Silicon nitride was contained in the melt to achieve the notably high N doping levels. Prior to the annealing step, the contact resistivities were greater than 1 x 10 4f~ c m 2.
L.M. Porter, R.F. Davis / Materials Science and Engineering B34 (1995) 83 105
94
Table 5
O h m i c c o n t a c t s on n-type c~(6H or mixed)-SiC. See T a b l e 3 f o r a description o f the rating system for surface preparations Metallization Deposition method
Deposition Deposition Annealing temperature pressure condition
p , ( f l c m 2)
SiC carr. conc. (cm
3)
(Torr) Cr Ni
melting e-beam
NR
SiC Method surface of p~. mess.
Origin of SiC
Surface Ref. preparation
NR
10 3 b.p.
melting
NR
NR
NR
NR
NR
(~>2130 °C) 100M °C/20 s
1.7 x 10 4
4.5 × 1017
(0001) T L M
[128]
600°C/30
4 x 10 2
~1 x 10 ~s
(0001) T L M
seeded sub.; N R C V D epi seeded sub. very good
4.7x10 ~
(0001) circular
seeded sub.: 0 2 plasma C V D epi seeded sub.; good C V D epi seeded sub.; 02 plasma C V D epi
[51]
Lely
[59]
evap. TiN
ion-assisted 3 5 0 ° C
10
io b.p.
[461
[49,50]
min
e-beam evap. TiW
sputtering
RT
10 ~' b.p.
O2plasma+
7.8x10
4
600 °C/5 min Ni
e-beam
RT
10 6
950°C/5min
TLM mid 10 2
4.7 x 10 ~a
(0001) four-
evap. Ni C r
sputtering
point RT
10 6 b.p.
9 5 0 ° C / 5 rain
1.8 × 10 ~
4.7 x 10 ~8
(0001) circular TLM
(60 40 wt.%) W
thermal
NR
NR
TiW
sputtering
RT
10 6 b.p.
~ lxl0
4 3x1018
lxl0~,~
(0001) four-
1200 1600°C
5x10
750°C/5min
I x 10 2 5 × 10 4 1 x 10 J7 I x 10 ~'; (000i) point ~ 8 x [0 4 7 8 x I(11~ (0001) circular
evap.
TLM Ti Mo
thermal evap.
RT
sputtering
10 7 b.p.
NR NR
none none
lxl0
2 <2x10
~ I x 10 4
52x
1018 1 × 10 ~° (0001) circular
> l x 10 I~)
TLM (000l) fotlrpoint
Ta
sputtering
10 7 b.p.
N.R.
none
~ 1 x 10 4
> 1 x 10 a~J
and T L M (0001) four-
resistive
NR
NR
100M °C/30 s
evap.
< 1.7 x 10 5
1 2 x 10 as
< 6 x 10 5
Ni
e-beam
RT
10 ('
Ni
evap. sputtering
NR
10 5 b.p.
[51]
seeded sub.: good C V D epi seeded sub. good
[1291
seeded sub.: good
[48]
[47]
C V D epi
seeded sub.: good C V D epi
[48]
and T L M (0001) Cox and (0001) Strack
Lely
[53]
seeded sub.; good C V D epi seeded sub. good
[52]
point Ni/3C-SiC
varied
[51]
very good
950 °C/2 min
< 5 × 10 6
7 9 x 1018
1050°C/5
10 3 10 4
9 . 8 x 1017
(0001) T L M (000i) (0001) T L M
10 3 10 4
9.8 X 1017
(0001) T L M
seeded sub.
good
[1301
1 x 10 ~'
4.5 x 1020
(0001) contact
LPE
good
[541
4.5 x 102"
area (000i) contact area
LPE
good
[541
[130]
min W/Ti/Ni
sputtering
NR
10 s b.p.
1050°C/5
min Ni
thermal
Ti-AI
evap. thermal evap.
NR
NR
1000 °C/5
min NR
NR
1000°C/5
< 1 x 10 3
min
N R = not reported; R T = r o o m temperature: b . p . = base pressure; s u b . = sublimation: epi = epilayer.
4.2.2. p-type e - S i C It is difficult to form ohmic contacts to p-type material by reducing the SBH because o f the large band gap and work function o f SiC. Aluminum is conventionally used in pure form or as an alloy to create ohmic contacts. Annealing the contacts causes A1 to diffuse into the SiC, resulting in an enhanced p-type carrier concentration near the surface. A higher carrier concentration corresponds to a narrower depletion region through which holes can effectively tunnel. However, the low melting point and the oxidation characteristics o f AI make processing the contacts difficult. The melting point can be increased by using A1 alloys (e.g. A1-Ti [55 57]), but the extremely high thermodynamic driving force for oxidation o f A1 places strict requirements on processing and passivating layers.
Table 6 lists a chronology of the various ohmic contacts to p-type c~-SiC. Boron, which is a deep level acceptor in SiC, and A1 were used in the earliest contacts of this type to SiC [58]. The contacts were made by fusing pellets of A1-Si ( ~ 1:1) or Si with a few percent B to the SiC. This process required temperatures o f 1700 °C for the A I - S i and from 1700 to above 2000 °C for the Si-B. It was believed that the enhanced p-type concentration at the SiC surface was formed by recrystallization from solution rather than diffusion of A1 into the SiC. While the temperatures required for these contacts are extremely high, these materials may be more desirable as interlayers, as Si has a significantly smaller band gap than SiC. Hall [58] also investigated W as an ohmic contact because of its similar thermal expansion coefficient to
L.M. Porter, R.F. Davis/Materials Science and Engineering B34 (1995) 83-105
95
Table 6 Ohmic contacts to p-type ~(6H or mixed)-SiC. Multi-layered contacts are designated with slashes to separate the distinct layers; layers at the surface to the interface with SiC proceed from left to right. See Table 3 for a description of the rating system for surface preparations Metallization
Deposition Deposition Deposition Annealing method temperature pressure condition
p~ ( ~ cm 2)
SiC carr. conc.
SiC
Method
(cm
surface
of p~
3)
Origin of SiC
Surface Ref. preparation
meas.
(Torr) Al-Si (~1:1)
melting
NR
1700°C
NR
NR
(0001)
--
Lely
NR
[58]
S i - B (a few
melting
NR
1700 2000
NR
NR
(0001)
--
Lely
NR
[58]
%B)
°C
W
melting
NR
Cu-Ti
melting
I0
5
1900 °C
NR
NR
(0001)
--
Lely
NR
[58]
> 8 8 0 °C
NR
NR
NR
--
Carborun-
good
[131]
good
[131]
good
[55]
good
[51]
varied
[59]
(71:29 at.%) AI-Si
d u m Co. melting
--
10 5
(88.7:I 1.3 at.%) A1-Ti
900-1000
NR
NR
NR
--
°C NR
NR
NR
950
d u m Co.
°C/5
NR
NR
(0001)
--
min AI
e-beam
RT
10 6
evap. W / A u W / W / A I sputtering
700 °C/10
NR
1800
seeded sub.;
LPE epi 1.7 x 10 3
1.8 x 1018
(0001)
TLM
min NR
Carborun-
seeded sub.;
CVD epi 2 5 x 10 4
NR
°C/120 s
(0001)
four-
and
point
Lely
(oooT) AI Ti
sputtering
RT
10 6 b . p .
1000°C/5
2.9×10
2 1.5)<10
5
5xl0tS
2x1019
(0001)
min
circular TLM
fourpoint and T L M fourpoint and T L M fourpoint and T L M
seeded sub.; good C V D epi seeded sub.; good CVD epi
[57]
seeded sub.; good CVD epi
[48]
seeded sub.; good CVD epi
[48]
Mo
sputtering
NR
10 " v b.p.
none
2 x 10 4
> l × l019
(0001)
Ta
sputtering
NR
10 7 b.p.
none
7 x 10 4
> 1 x 1019
{000l)
Ti
sputtering
NR
10 7 b.p.
none
3 × 10 4
> 1 × l0 ~9
(0001)
AI
sputtering
NR
l0
800°C/10
10 2 - 1 0
-~
8 × 1018
(0001)
TLM
seeded sub.
good
[130]
800 8 5 0 ° C 10 2 - 1 0
3
8 x 1018
(0001)
TLM
seeded sub.
good
[130]
NR
(0001)
TLM
Lely sub.;
very good [53]
5 b.p.
[48]
min W/Pt/At
sputtering
NR
10 5 b.p.
/10 min Ti/AI/C-
e-beam
SiC
evap./
Pt
LPCVD e-beam
NR
thermal
evap.
2 - 3 x 10 s
min RT
10
m b.p.
450 850
L P E epi NR
>1 x 10 Is
NR
°C/20 min
evap. Ti/AI
950 °C/2
NR
NR
NR
1000 °C/5
seeded sub.; very good [60] C V D epi
NR
NR
(000i)
min
Acheson
good
[54]
sub.; L P E
epi N R = not reported; R T = r o o m temperature; b . p . = base pressure; s u b . = sublimation; epi = epilayer.
SiC. However, annealing these contacts resulted in even greater penetration depth than for the Si alloys. T w o eutectics, C u - T i and A1 Si, were later found to be ohmic after annealing at > 8 8 0 °C and 9 0 0 - 1 0 0 0 °C, respectively [13]. These annealing temperatures are above the eutectic melting points and resulted in penetration o f the contacts in the SiC o f up to 1 pm. Similar penetration depths occurred for the W/AuW/W/A1 contacts [59]. These extensive chemical interactions would be devastating in many of today's devices. As in the case of n-type SiC discussed above, the need for annealing can be reduced or eliminated by heavily doping the surface to create a p + layer in p-type SiC. As-deposited Mo, Ta, and Ti contacts on p + SiC epitaxial films grown by C V D yielded ohmic
contacts [48]. Contact resistivities measured by Kuphal's 4-point method were in the 10 -4 ~ c m 2 range, which was reported to be a conservative value since the method does not account for nonuniform current densities. The patent by Glass et al. [60] is based on the relationship that the contact resistance decreases with increasing doping concentration and decreasing SBH. When tunneling dominates the current transport, as occurs for high doping concentrations and a finite barrier, the following proportionality holds R c oc
exp(OB/x/N)
(2)
where Rc is the specific contact resistance, OB is the Schottky barrier height, and N is the carrier concen-
L.M. Porter, R.F. Davis
96
Materials' Science and Engineering B34 (1995) 83 105
Table 7 Rectifying contacts on n-type fl-SiC. The deposition and annealing conditions, calculated Schottky barrier heights, and SiC information are listed for most of the contact metallizations. See Table 3 for a description of the rating system for surface preparations Metallization
Deposition method
Deposition temperature
Deposition pressure (Torr)
Annealing condition
qbB(eV)
SiC surface
Method of OB means.
Origin of SiC
Surface Ref. preparation
Au
e-beam evap. thermal evap. thermal evap. thermal evap. thermal evap. thermal evap. thermal evap. thermal evap. thermal evap. thermal evap. thermal evap. thermal evap.
NR
10 7
1.2
(100)
C V
CVD
good
[62]
RT
10 ¢' b.p.
none; to 500 °C/1 h 900°C/3 5 m i n
NR
(100)
CVD
good
[70]
RT
10 ~' b.p.
none
1.2
(100)
C V
CVD
good
[70]
~RT
10 m b.p.
none
1.4
(100)
XPS
CVD
Si-rich
[67]
RT
10 m b.p.
none
0.95, 0.92
(100)
XPS, C V
CVD
very good
[63]
RT
10 m b.p.
none
0.78, 0.87
(100)
XPS, C V
CVD
very good
[63]
RT
10 m b.p.
none
0.69, 0.73
(100)
XPS, C V
CVD
very good
[63]
RT
10 m b.p.
none
0.53
(100)
XPS
CVD
very good
[63]
RT
10 m b.p.
none
0.40
(100)
XPS
CVD
very good
[63]
RT
10 m b.p.
none
0.35
(100)
XPS
CVD
very good
[63]
RT
10 m b.p.
none
0.16
(100)
XPS
CVD
very good
[63]
RT
10 6 b.p.
none
CVD
good
[64]
sputtering
NR
NR
none
carbonization of Si
good
[71]
AI Au A1 Pd Au Co Ti Ag Tb AI Au
vicinal & off-axis
(100) Ni
NR
N R = not reported; RT = room temperature; b.p. = base pressure; XPS = X-ray photoelectron spectroscopy; C V = capacitance voltage.
tration [61]. To reduce the SBH on p-type material, a high work function metal should be used. Although empirical evidence indicates that there is a partial pinning of the Fermi level in 6H-SiC, the SBH has some dependence on metal work function, as shown in Figs. 4 and 5. In this invention ohmic contacts were fabricated by depositing Pt on p-type SiC with a p-t- layer at the surface. Both as-deposited contacts and contacts annealed at temperatures to 850 °C displayed ohmic behavior. Platinum has several advantageous properties, including its high melting point and oxidation resistance; however, even its high work function (5.65 eV) is not enough to eliminate the SBH on p-type SiC. For this reason the carrier concentration at the SiC surface was increased by ion implantation of A1 (at 10kV, 6 x 1015cm 2 dose or 50kV, 2 x 1015 cm 2 dose; 600 °C substrate) and annealing at 1500 °C prior to metallization. 5. Contacts to fl-silicon carbide
5.1. Schottky contacts to /]-SIC With the advent of the ability to grow //-SIC thin films in the 1980s came a thrust of research on metal
contacts to this material. However, large concentrations of stacking faults, associated with the low stacking fault energy of this polytype, as well as twins and threading dislocations occurring in heteroepitaxial films have precluded the achievement o f the predicted electrical properties, both within the semiconductor substrate and at the metal/SiC interface. Table 7 lists the rectifying contacts on n-type //-SIC found in the literature. No rectifying contacts on p-type fl-SiC have been reported. Several of these studies have included Au [62 64], which has a high work function and is nonreactive with Si and C. Ioannou et al. [62] deposited Au films on chemically cleaned substrates by electron beam evaporation at approximately 1 )< l0 7 Tort. The cleaning procedure, which consisted of degreasing with solvents and etching in HF, should have removed almost all of the O, while some hydrocarbon contamination could have adsorbed to the samples during their brief exposure to air while loading in the vacuum chamber. The as-deposited contacts displayed ideality factors of 1.5 (linear region > 6 decades) and soft breakdown at 8 - 1 0 V ( l e a k a g e s 2 . 5 × 10 7 A c r e 2 at 1V). A Schottky barrier height of 1.2 eV was
L.M. Porter, R.F. Davis /Materials Science and Engineering B34 (1995) 83 105
calculated from C - V measurements. After annealing at 300 °C for 1 h, the ideality factors and SBH decreased slightly, while the reverse leakage currents increased. Subsequent annealing at 500 °C for 1.5 h resulted in further increases in the reverse leakage currents. Ohmic behavior ensued after annealing at 700 °C for 30 , i n Auger analyses showed significant diffusion of Si into the Au film after the latter annealing step. As the substrates are believed to have been relatively clean, it is quite likely that the inferior Schottky characteristics (in comparison to 6H-SiC) and degradation after low temperature annealing were associated with defects in the fl-SiC films. The study of Das et al. [64] gives strong evidence for the degrading effects of defects on electrical properties in /7-SIC substrates. Gold films were thermally evaporated onto fl-SiC films grown on both nominal on-axis (100) and off-axis ([100] oriented 2 ° 4 ° toward [011]) Si. A previous study [65] had shown the presence of antiphase domain boundaries (APBs) in films grown on the on-axis (100) substrates. The values of both the ideality factors and the reverse leakage currents were greater for the contacts on the on-axis films. In addition, current transport was determined to be dominated by space charge limited current rather than thermionic emission, which has been dominant in many 6H-SiC contacts [66]. As listed in Table 7, Waldrop and Grant [63] measured the SBHs of several metals on (100) fl-SiC by XPS and C - V measurements. The metals were deposited at room temperature in UHV after chemical and thermal cleaning which left ~ 1 monolayer of O as the only detectable contaminant by XPS. The calculated and theoretical barrier heights are plotted vs. the metal work functions in Fig. 7. Both sets of data (XPS and C V) show a positive correlation between SBH and metal work function. The slopes, S, of the linear fits to the two sets of data points were 0.26 and 1.52
1.0-
s = 1.o. , #
(C-V) 0.5-
.E
"r
0.0-
E .''
-0.5-
.Q- theoretical I ; measured (XPS) measured (C-V
• • •-•
-1 . 0 2.5
30
315
20
.15
510
5!5
Metal Work Function (eV)
Fig. 7. Data points of experimentally-determined and theoretical barrier heights on n-type 3C-SiC (100) vs. work function of the metal contacts. The experimental data points are from Waldrop and Grant [63]. The slopes, S, of the linear fits through each set of data points are indicated.
97
from XPS and C - V measurements, respectively. Although the slope from the data points calculated from C - V measurements is very high, the small number of data points over a small range of work function values introduces a higher susceptibility to variation. If the SBH of Tb (@M = 3.0 eV) is neglected, S for the XPS set of data becomes 0.58, which is more than twice as high as S for the complete set of data. Thus, it is again shown that the index of behavior, S, can be very susceptible to both the particular metal contacts and the measurement technique. As for 6H-SiC, all results indicate some positive correlation between SBH and metal work function. From XPS analyses Bermudez [67] found an upward band bending of 1.5 eV at the (100) fl-SiC surface after annealing in UHV at 1150°C for 2 , i n followed by exposure to a Si flux, which resulted in a Si rich surface. The surface states, which are indicated by the band bending, may have been enhanced by the excess Si; however, the high degree of covalency of SiC makes it likely that some surface states are intrinsic to this material [68]. After depositing 9 A of AI (sample temperature <200 °C), the band bending was reduced by 0.1 eV, yielding a SBH of 1.4 eV. No pronounced interdiffusion was detected after "flash annealing" the 9/~ films to 1050 °C. This result suggests that longer annealing times at that temperature are necessary to activate AI in any p-type SiC polytype for the fabrication of ohmic contacts from A1 or AI alloys. 5.2. Ohmic' contacts to ~7-SIC 5.2.1. n-type fl-SiC In addition to 6H-SiC, annealed (900-1250 °C) Ni has also been commonly used for ohmic contacts to fl-SiC [69-72]. Steckl and Su [71] employed as-deposited and annealed (900 °C for 3 - 5 , i n ) Ni for rectifying and ohmic contacts, respectively, in one device. However, Cho et al. [72] reported Ni to be ohmic on 10mY 10JScm 3 / - S I C in the as-deposited case. The ohmic behavior in the latter case is likely to be the result of a high concentration of defects in the SiC films (owing to its high work function, one would not expect Ni to be ohmic on any n-type SiC polytype); thus, the substrate quality is very critical to the contact properties. Daimon et al. [69] and Edmond et al. [70] found both annealed Ni and as-deposited AI to be ohmic on n-type fl-SiC (100) with low carrier concentrations (5× 1016 l x 10~7cm-3). The latter study reported high contact resistivities for both contacts, as might be expected for the low doping levels in the substrates. The relatively low work function of A1 probably contributed to the formation of a low SBH. Annealing the A1 at 900°C for 3 - 5 , i n resulted in rectification. This change was attributed to the creation of a p - n junction by diffusion of AI into the SiC.
L.M. Porter, R.F. Davis / Materials Science and Engineering B34 (1995) 83 105
98 Table 8 Ohmic contacts on n-type//-SIC.
M u l t i - l a y e r c o n t a c t s a r e l i s t e d in s e q u e n c e (left t o r i g h t ) f r o m t h e t o p m o s t
SiC. See Table 3 for a description Metallization
l a y e r to t h e l a y e r in c o n t a c t w i t h t h e
of the rating system for surace preparations
Deposition
Deposition
Deposition
Annealing
p~ at R T (f~
SiC carr.
SiC
Method of p~ Origin of
Surface
method
temperature
pressure
condition
cm:)
cone.
surface
meas.
SiC
preparation
(Torr)
(cm
Ref.
~/
Ni
NR
RT
NR
930 °C/3 min
NR
6 x 1() [¢' I x 10 Iv
(100)
CVD
fair
[69]
AI
NR
RT
NR
as-deposited
NR
6 × 10 It'
(100)
CVD
fair
[69]
1 x l0 w
Au
e-beam
NR
10 7
700 ° C / 3 0 m i n
NR
5 5
IO n' IO w
(100)
NR
CVD
good
[62]
RT
10 ~' b.p.
as-deposited
1.6 x 10
5
I0'"
(100)
three-contact
CVD
good
[70]
CVD
good
[70]
CVD
good
[7(I]
CVI)
good
[70]
CVD
good
[70]
CVD
sputtered w/
[86]
evap. A1
thermal evap.
and extrapolation
Ni
thermal evap.
RT
thermal
RT
10 (' b.p.
1250°C5min
1 . 4 x 10
5
10 ~r'
(100)
three-contact and extrapolation
Cr
l0
(' b.p.
1250 °C,5 rain
7.0 x 10 2
5 × 1() 16
(100)
evap.
three-contact and extrapolation
Au Ta (9% 3 at.%)
thermal evap.
RT
10 (' b.p.
1250°C5min
3 . 0 x 10 2
TaSi 2
sputtering
RT
NR
850 °C,'5 rain
2.0 x 10 2
5 x 10 I(,
(100)
Ag
MBE
RT
UHV
1000 °C
NR
NR
(100)
5 × I() I('
(lO0)
three-contact and extrapolation three-contact and extrapolation
500 eV A r ions: no W
sputtering
RT
10 7
850°C/30min
2 . 4 × 10
2 × l0 w
(100)
contact area
CVD
contaminants sputtered
Ti
sputtering
RT
10 v b.p.
300 °C/30 90
7.6 9.2 x 10 ~
I() v~ I() ~
(10(/)
four-point
CVD
good
[75]
10
nrin as-dep, to 600
1.5 × 10 :
10 ~
I(# s
(100)
four-point
CVD
good
[75]
°C/10min
2.3 x 10
10 " b.p.
1000°C/10s ~
1.1 × 10 4
I() ~v I() ~s
(100)
four-point
CVD
good
[75]
10 v b.p.
450 °C/390 min 1000°C'10s+
3 . 9 x l0
I() Iv 1() I~
(100)
four-point
CVD
good
[75]
6.2 4.0 × 10 2
1017 10/~
(100)
extrapolation
CVD
NR
[72]
NR
NR
NR
W TiSi: WSi:
sputtering sputtering sputtering
RT RT RT
7
b.p.
i
4
[74]
450 °C, 390 min Mo
e-beam
NR
10 ~' b.p.
evap. Ni
as-dep, and 1200 °C/6(I
sputtering
NR
NR
min
900 °C/3 5
carboniza- good
rain Au/Pt/Ti
sputtering
NR
l0
Au/Pt/W
sputtering
NR
Au/Pt/TiN/Ti Pt/TiW, Ti
sputtering sputtering
NR NR
v b.p.
[71]
tion of Si I.l x 10 4
l0 u' 10 r
(100)
four-point
CVD
sputtered
[76]
10 7 b.p.
operated at 650 °C,,I h operated at
2.0 x 10 4
10 I(' l017
(100)
four-point
CVD
sputtered
[76]
l(I 7 b.p.
65O °@8 h operated at
1.4 x l0
l0 I(' 1017 (100)
four-point
CVD
sputtered
[76]
10 v b.p.
650 °@31 h operated at
2.6 x I0 4
1() I~' l0 w
(100)
four-point
CVD
sputtered
[76]
10
4
NR NR
10 v b.p.
lO 17 IO Is
(100)
Ta
sputtenng sputtering
650 °C/3 h as-deposited
10 7 b.p.
as-dep, to
7 x 10 7
5 × I() I'~
(100)
circular T L M circular T L M
CVD CVD
good fair
[10(I] [77]
Re
sputtering
NR
l(I 7 b . p .
1000 °C/I h as-dep, t o 9 0 0 ° C /30 min
4.3 x 10 ~' 1 × 1 0 "* l x I0 s
5 × 1 0 v;
(100)
circular T L M
CVD
t~fir
[771
Pt
sputtering
NR
10 v b.p.
as-dep, to 5 0 0 ° C /30 rain
6 × 1 0 <~ 1 × 10 5
5 × 1() I~
(100)
circular T L M
CVD
fair
[77]
W
NR = not reported;
RT = room temperature;
b.p. = base pressure.
3
L.M. Porter, R.F. Davis / Materials Science and Engineering B34 (1995) 83-105
Table 8 lists these studies along with several other ohmic contact metallization schemes. While most of these contacts were annealed above ~ 8 0 0 °C, there does not seem to be any simple formula for creating an ohmic contact to n-type fl-SiC. One of the metals (Ni) is a silicide former, m a n y (Cr, Ta, W, Ti, Mo) are both silicide and carbide formers, and some (Ag and Au) are neither carbide nor silicide formers [73]. In addition, the work functions of m a n y of the silicides are quite similar to the pure metals; thus, it is not believed that the product phases in general have the intrinsic properties to create an ohmic contact (i.e. as-deposited). While low work function metals may form relatively low barrier contacts, the data suggests that the formation of most of the ohmic contacts to n-type fl-SiC has been associated with the creation of defects (e.g. vacancies) after annealing and possibly aided by defects present within the SiC films. Gold formed ohmic contacts after annealing at 700 °C for 30 min [62]. Auger analyses showed that the annealing resulted in Si diffusing into the Au film. In this case the ohmic behavior is very likely associated with the creation of Si vacancies rather than the formation of any new interface compounds. Many of the studies listed in Table 8 reported contact resistivities. Edmond et al. [70] reported room temperature values on lightly doped SiC (5 x 1016 cm 3) to be between 2.0 x 10 2 f~ c m 2 and 1.6 x 10 i f~ c m 2, with TaSi2 yielding the lowest value. Some improvement in the TaSi2 contacts was observed with heating at 400 °C. Geib et al. [74] estimated contact resistivities of 2.4 x 10 1 and 8.0 x 10 2f~ cm 2 (2 x 1017 cm 3 carrier conc.) for W contacts at 23 °C and 900 °C, respectively. In comparison Chaudhry et al. [75] calculated a contact resistivity of 1.5 x 10 2 f~ cm 2 for as-deposited W contacts o n 10 j7 t o 1018 cm 3 fl-SiC by the four point probe method. This value decreased to 6.1 x l 0 - 3 ~ c m 2 after annealing at 300 °C for 30 min. The contact resistivity of Ti contacts on the same material decreased significantly from 1.69 ~ cm 2 after annealing at 300°C for 3 0 m i n to 7.6 x 10 3 ~ c m 2 after 90 min. The decrease in both cases was attributed to the dissolution of an interfacial oxide layer. Annealing the Ti contacts at 600 °C for 10 min resulted in a slight increase to 9.2 x 10 3 ~-~c m 2. The authors stated that this increase may be due to the formation of a high resistivity TIC0.6 phase. Recent work [21,31] has shown that a thin layer (2 3 nm) of TiC1 , forms at the interface with 6H-SiC after annealing Ti/SiC at 700 °C for 20 60min. It is possible that the formation of a thin carbide layer resulted in a slight increase in the contact resistivity; however, a significant increase due to the higher resistivity of the carbide is unlikely considering its thickness. In addition, the much smaller work function of TiC is more likely to cause a reduction in the SBH. In fact, one study [76] attributes an observed
99
reduction in the contact resistivity of Ti contacts (annealed at 650 °C for 3 h) to the formation of TiC. The silicides of these metals (TiSi 2 and WSi2) yielded lower contact resistivities than the pure metals and displayed better thermal stability [75]. The lowest reported values for these metallizations were 1.1 x 1 0 - 4 ~ c m 2 for TiSi2 annealed at 450°C for 390rain and 2.0 x 10 4 f ~ c m 2 for WSi 2 annealed at 450 °C for 10 min. A correlation between a high contact resistivity and an interfacial oxide layer was observed, emphasizing the importance of surface cleaning and processing conditions on good contact properties. Substantially lower contact resistivities were reported by Chen et al. [77] for Ta, Re, and Pt contacts. In the latter study the metals were deposited on both unintentionally-doped ( ~ 10 IT cm 3) and N-implanted (5 X l019 c m - 3 ) n-type (001) fl-SiC. On the unintentionally-doped samples the Ta and Re contacts were ohmic with contact resistivities of 5 x 10 5 f~ cm 2 and 4 x 10 4 ~-~c m 2, respectively, calculated using the T L M method. The contact resistivity of Ta increased to 1 x 10 - 4 ~ c m 2 after annealing at 500 °C for 30 min; that for Re decreased to 2.5 x 1 0 - 4 f ~ c m 2. Both of these contacts became rectifying after annealing 900 °C for 30 min. The Pt contacts were not ohmic before or after heat treatment (to 1000 °C) on these samples. However, all three metals were ohmic in the as-deposited condition on the implanted samples. The contact resistivities for as-deposited Ta, Re, and Pt on these latter samples were 7 x 10 7, l x 10 4, and 6 x 10 - 6 ~ c m 2, respectively. The Pt contacts became rectifying after annelaing at 900 °C for 30 min. After annealing at 1000 °C for 1 h, the contact resistivity of the Ta contacts increased to 4.3 x 10 6 f~ c m 2. These contacts remained ohmic after heating to 1100 °C for 30 min, but interfacial voids formed, resulting in erratic I V measurements. The Re contacts were annealed to 900 °C for 30 min, which resulted in a decrease in the contact resistivity from the as-deposited value to 1 x 10 5 ~ c m 2. Shor et al. [76] investigated multilevel metallization schemes based on Ti and W for high temperature (650 750 °C) operation (see Table 8). This type of study is very important for extended high temperature operation of electronic or optoelectronic devices, yet very little has been published on this subject. The main concerns include the reactivity and diffusivity between the contact metallization and the SiC substrate and the oxidation of the contacts. Prior to the I - V measurements taken at room and elevated temperature, the contacts were annealed at 650 and 750 °C for various times. The most promising metallization scheme was Au/Pt/TiN/Ti, which remained ohmic throughout operation at 650 °C for 31 h after which it became rectifying. The reported contact resistivity of the asdeposited contacts (1016-101Vcm-3carr. conc.) was
L.M. Porter, R.F. Davis
100
Materials Science and Engineering B34 (19951 83 105
Table 9 O h m i c c o n t a c t s on p-type/']-SIC. M u l t i - l a y e r c o n t a c t s are listed in sequence (left to right) from the t o p m o s t layer to the layer in contact with the SiC. All of the /]-SIC (1001 films were grown on Si by CVD. See T a b l e 3 for a description of the r a t i n g system for surface p r e p a r a t i o n s Metallization
Deposition method
Deposition temperature
Deposition pressure (Tow)
Annealing condition
p~. at R T (f~ cm 21
SiC carr. cone. (cm ~)
AI Si (89:11 wrY/o) Au Ta-A1 (91:2:7 at.%)
NR
RT
NR
930 °C,'3 min
NR
thermal evap.
RT
10 (' b.p.
1200°C, ' 30 min
4.7 x 10 ~
I x 10/r' 4 x 1017 1 x 10 u'
sputtering/ thermal evap. thermal evap.
RT
10 ~' b.p.
1200 °C,' 30 min
2.0 x 10 ~
I x 10 ~6
RT
10 ~' b.p.
8811°C:3 min
3.1 x 10 2
I × 10 t~'
NR
10 (' b.p.
as-dep, to 7 0 0 ° C :15min
4.1 x 10 2 2.8 x 10 -'
ll) ~7 10 ~
TaSi./AI
AI
Ni
e-beam evap.
M e t h o d of p~ mess.
three-contact and extrapolation three-contact and extrapolation three-contact and extrapolation extrapolation
Surface preparation
Ref.
fair
[691
good
[70]
good
[70]
good
[70]
NR
[721
N R = not reported: R T - r o o m t e m p e r a t u r e ; b.p. - base pressure
1.4 x 10 4~'~cm2. This value was reduced by 65% (4.9 x 10 s fl c m 2) after annealing at 650 °C for 1 h. It was believed that the TiN layer acted as a good diffusion barrier, thereby improving its thermal stability. While the authors suggest that a high defect density in the fl-SiC films may have contributed to a higher contact resistivity than may be achieved on better quality material, we believe that these same defects may also contribute to the actual ohmicity of the as-deposited Ti and W contacts. The Au/Pt/W contacts also displayed relatively good thermal stability. Considering the facts that W has a high melting point and TiN is a good diffusion barrier, it would be interesting to see if the thermal stability could be improved by combining both of these materials in the same metallization scheme.
5.2.2. p-type ~l-SiC The small number of studies on ohmic contacts on p-type /l-SiC are listed in Table 9. All but one of the contact metallizations contains AI. As mentioned in a previous section, Al is used because it is a p-type dopant in SiC and therefore can be used to highly dope the surface of the SiC, narrowing the depletion region and allowing carriers to tunnel through the barrier. The contact resistivities of A1 annealed at 880 °C for 3 m i n and TaSi2/A1 and Au Ta AI annealed at 1200°C for 30min were 3.1 x 10 2 2 . 0 x 10 ~, and 4.7 x 10 1 f~ cm 2, respectively on I x 10 ~6 cm 3 /l-SiC. This high annealing temperature should be enough to substantially increase the doping level in the substrate; thus, a higher carrier concentration in the substrate may not significantly improve the contact characteristics. However, the very high driving force for oxidation of AI could have resulted in A1 diffusing away from the interface to react with O. This problem should be considered with any contact metallization involving A1.
It is believed that the variation in, and sometimes conflicting, results regarding contacts on fl-SiC may be attributed to two main causes: (1) differences in crystalline quality of the //-SiC films, considering that the defect densities are much higher than in 6H-SiC; and (2) differences in the states of the surface prior to metal deposition. For example, in one study [72] Ni was ohmic as deposited on p-type (1017 l0 Is cm 3) /]-SIC. After annealing at 700 °C (15 min?), X-ray diffraction indicated the formation of Ni2Si, and the contact resistivity decreased from 4.1 × 10 2 to 2.8 × 10 2f~cm 2. These results indicate that the ohmic behavior was not caused by the formation of Ni silicide; however, it was likely associated with a high concentration of defects near the SiC surface, which probably increased upon annealing.
6. Discussion
6.1. Suit/hoe states and Fermi level pinning Throughout this paper the characteristics of many contacts on n- and p-type c~(6H)- and //-SIC were reported. Considering the contacts in their unreacted forms, Figs. 4, 5 and 7 show a positive correlation between the SBHs and the work functions of the metals on both ~- and /]-SIC as the Schottky Mott limit would predict. However, these empirically determined correlations were significantly less than 1.0, indicating a partial pinning of the Fermi level attributed to surface states. A partial pinning of the Fermi level in SiC agrees with the work published by Kurtin et al.
[68]. For many decades it has been known that Fermi level pinning is more common in covalent materials than in
L.M. Porter, R.F. Davis / Materials Science and Engineering B34 (1995) 83-I05
ionic materials [78,79]. Pioneering work was performed by Kurtin et al. [68] who plotted results [80,81] of (I)B vs. XM, where XM is the electronegativity of the metal, for several metals on Si, GaSe, and SiO2 substrates ~. The values of S for Si, GaSe, and SiO2 were ~0.05, 0.6, and 1.0, respectively. The slopes increased with the degree of ionicity of the substrate, as defined by the difference in the electronegativities, AX, of the two components [82]. A very important trend was revealed in a plot [68] of S vs. AX, for many semiconductors including the three above. The plot, which is represented in Fig. 8, shows a region of low S ( < ~ 0.1) for semiconductors with A X < 0 . 6 and a region of high S (>0.9) for semiconductors with A X > 0.8 connected by an abrupt transition region (0.6 < A X < 0 . 8 ) . Carbon and silicon have electronegativities of 2.55 and 1.90, respectively, which corresponds to A X = 0.65 for SiC. The data indicates that Fermi level pinning dominates Schottky barrier formation in covalent materials, plays a variable role in materials with a mixture of ionic and covalent bonding (0.6 < A X < 0 . 8 ) , and does not determine the SBH in the more ionic materials. According to Fig. 8, SiC is in the transition between strong Fermi level pinning (surface states) and no Fermi level pinning (no surface states), which is in good agreement with the empirical results presented in this paper (e.g. Figs. 4, 5 and 7). Kolomiets et al. [83] attributed the frequency dependence of the capacitance of SiC on relaxation (filling time 10 2 - 1 0 6 S ) of surface states. The c u r r e n t voltage characteristics of metal/SiC contacts were independent of the metal work functions. In addition, very recent work [84] has also shown direct evidence for the presence of surface states on 6H-SiC (0001). After etching for 10 min in 10% H F and annealing at 960 °C
I 0 --
AJN • ZnO A.1203SrTiO
ZnS
SnO~
sio5
0.8
,o~d] GaS
0.6
• CdS Se
0.4
Ga203
SiCfOxI ZnSe
• ~/q
0.2
~e
CIIFe • , GaA
0.0
I 0.0 0.2
I 0.4
I 0.6
s~
I 0.8
I 1.0
I 1.2
I 1.4
I 1.6
I 1.8
I 2.0
I 2.2 2.4
AX Fig. 8. Index of interface behavior, S, plotted for various semiconductors as a function of the electronegativity difference in their elemental components. The index of interface behavior is defined as the slope in a plot of Schottky barrier height vs. metal work function. After Ref. [68]. The electronegativity has been shown to have a linear relationship with the work function (W. Gordy and W.J.O. Thomas, J. Chem. Phys, 24 (1956) 439-444). However, the work function is dependent on the crystal plane, whereas electronegativity is an atomic property.
101
in a silane flux, both ultra-violet photoelectron spectroscopy (UPS) and XPS analyses indicated that surface states were present within the band gap. These surface states, however, were removed after hydrogen plasma exposure yielded a H terminated surface. Therefore, it is possible to passivate the surface states with H. If these states could remain passivated after depositing metal contacts, more control over the SBH, and, hence, the electrical properties, could be achieved.
6.2. Reactivity of SiC Extensions of the coavalent vs. ionic theory [68] have also included heats of formation, AHf, of the semiconductors. Brillson [85] plotted the S values taken from Mead [79] (and references therein) vs. AHf for a variety of semiconductors. The plot showed the same type of transition from covalent to ionic as for S vs. AX. Covalent semiconductors have low heats of formation and are less stable against chemical reaction. It has been shown through many of the studies discussed in this paper that SiC readily reacts with most metal contacts. However, the critical reaction temperatures are usually greater than those for Si. These reactions are very important to the resulting electrical characteristics, especially for operation at elevated temperatures. Even those metals such as Au which do not form silicides or carbides have shown interdiffusion with SiC [62], while Ag has evaporated from the SiC surface at temperatures above 600 °C [86]. This may mean that diffusion barriers will be a necessary part of contact metallizations for long term operation at high temperatures. Although SiC consists of 50 at.% Si, the reaction characteristics of SiC with metals can be very different from the metals with Si. In addition, it is difficult to predict the reaction products in material systems having three or more components and in the process of attaining thermodynamic equilibrium, e.g. a metal and SiC chemically bonded at an interface and heated to the extent that chemical interdiffusion occurs. Thermodynamic data can yield the equilibrium phase fields; however, the diffusion path is controlled by the diffusion coefficients of each component in the pertinent phases. It is the ratio of the diffusivities which determines the interface compositions, and accordingly the diffusion path [87]. It is possible that reactions in certain metal/SiC systems are limited by the dissociation of SiC because of the strong Si C bond. Evidence in support of this statement is provided by the initial formation of metalrich silicides, for example between SiC and Ni [42,88,89], Co [35,42,90], and Pd [88,91], which in most cases were followed by the formation of less metal-rich silicides. Other metals such as Ti [21,92-95], Mo [96], Fe [97,98], Nb [99], and W [100] formed both carbides
102
L.M. Porter, R.F. Davis
Materials Science and Engineering B34 (1995) 83 105
and silicides after annealing At temperatures between 570 and 1200 °C, the product phases in Ti/SiC were reported to be TiC [101,102], TisSi3 [103], or both TiC1 ,. and TisSi3 [21,92 95]. These results were similar in the Mo/SiC [96] system annealed at 1200 °C for 1 h and W/SiC [100] system annealed at 950 1100 °C for 60s in which MesSi 3 and Me2C ( M e = m e t a l ) formed. It is immediately apparent that these product phases are very different than if these same metals reacted with Si. Not only is the temperature important, but the thickness of the films and the amount of time are very important determinants of the resulting phases in metal/SiC systems. 6.3. Future perspectives
There are still important questions left unanswered regarding SiC contacts. However, important strides have been achieved which have provided opportunities for further research. Because of the nature of SiC, good Schottky contacts with SBHs of approximately l eV can be easily fabricated. However, these same properties make it difficult to form low barrier ohmic contacts. It is the ability to create ohmic contacts with low contact resistivities ( ~< 10 6 f~ cm 2) which will be one of the major challenges facing the SiC community in the foreseeable future. Perhaps it would be wise to examine solutions to similar problems in GaAs. The practical difficulties of creating good ohmic and/or Schottky contacts to GaAs has led to recent efforts in barrier height engineering. The tendency of the Femri level to pin near midgap in this material led several investigators to try to decrease the band bending at the surface by using doped interlayers of Ge and Si [104 109] and to increase it by using interlayers of NiA1 and AlAs [109-114]. To decrease the barrier height on GaAs, Chambers [109] deposited thin semiconductor interlayers of n+Si, n + G e and GaSe,.As~ , with a smaller band gap and a favorable band lineup with the substrate. Chambers and Loebs [108] grew thin epitaxial overlayers of undoped and As-doped (n +) Si on n- and p-type GaAs (001)-(2 x 4) to determine their influence on surface band bending. X-ray photoelectron spectroscopy (XPS) measurements indicated that the Fermi level remained pinned after depositing undoped Si (at ~0.6 eV below the conduction band for n-GaAs and --~0.5 eV above the valence band for p-GaAs). However, depositing n Si onto n-GaAs (p-GaAs) resulted in a significant reduction (increase) in band bending ( ~ 0 . 3 e V for n-GaAs and ~0.8 eV for p-GaAs). The authors concluded that the Fermi level was not unpinned, which would be associated with an elimination of interface states. Intead, it was believed that the reduction in band bending could be attributed to charge transfer from the highly-doped Si interlayer.
Similarly, it would be very desirable to decrease the band bending at the SiC surface so that low barrier height contacts could be produced. This might be achieved by depositing doped semiconducting interlayers with favorable band lineups between the conduction or valence bands on n-type or p-type SiC, respectively. A similar but alternate approach is also advocated. This would involve depositing thin insulating interlayers followed by low (on n-type) or high (on p-type) work function metals as an extension of the investigations of TiN [49,50] and TiW [51] contacts discussed in Section 4.2. In addition to a potentially reduced SBH, a good diffusion barrier between the metal and the SiC may be a beneficial result. Because the long term operation of SiC devices is an important goal, diffusion barriers are a legitimate consideration. The reactions of SiC with many metals at potential device operating temperatures ( ~ 6 0 0 °C) can destroy a device. Therefore, it is important to develop contacts which are stable at high temperatures. However, very few studies have reported the thermal stability of contacts at elevated temperatures for extended periods of time. This is another important issue to be addressed by the SiC community. As a result of the research which has been performed for the past few decades, a good understanding of the limitations of metal contacts to SiC has been gained. We must now use these achievements as the foundation for further improvements in contacts if the potential of SiC devices is to be realized.
7. Conclusions
Schottky and ohmic contacts on n- and p-type ~- and //-SIC have been reviewed and discussed in terms of their electrical properties, Schottky barrier heights, thermal stability, and reaction characteristics. There is significant evidence for the presence of surface states in SiC. However, if the surface is maintained relatively clean, the density of surface states is not sufficiently high to completely pin the Fermi level. Positive correlations ( g 0 . 2 - 0 . 6 ) between the SBHs on both ~- and fl-SiC and the work functions of the metals were displayed, The partial pinning of the Fermi level is believed to be associated with the percent ionicity of SiC in agreement with the results of Kurtin et al. [68]. Most metals in the as-deposited condition form very good rectifying contacts on both n- and p-type 6H-SiC with SBHs of ~>1 eV. Low leakage currents, low ideality factors, and high breakdown voltages have been displayed. The quality of contacts on [i-SiC seems to be limited by the quality of the /t-SiC films. On both polytypes it has been difficult to fabricate ohmic contacts with low contact resistivities. Heavily doping the surface has shown some success and may be
L.M. Porter, R.F. Davis / Materials Science and Engineering B34 (1995) 83-I05
a viable technique for many devices. Alternative approaches using semiconducting or insulating interlayers should also be investigated to reduce the barrier and, correspondingly, the contact resistance. The reactivity of SiC with metals to form silicides and/or carbides can be a serious problem for long term operation of devices at high temperatures. These reactions can ultimately destroy a device. For these reasons diffusion barriers should be investigated for contact metallizations. There is also a need for testing contacts for extended operation at high temperatures. These critical issues must be addressed and transferred to industry if we are to push the envelope of SiC device operating conditions and take advantage of the extreme thermal and electronic properties of SiC.
[14] [15] [16] El7] [18] [19] [20] [21] [22] [23] [24] [25] [26]
Acknowledgments We would like to express our sincere gratitude to J.S. Bow, M.J. Kim, and R.W. Carpenter of Arizona State University for contributing the TEM images. We would like to thank Cree Research, Inc. for providing the 6H-SiC wafers on which the research at N C S U was based; S. Tanaka for taking valuable time to translate literature; and M.C. Benjamin, S.W. King, and R.J. Nemanich for contributing unpublished data regarding surface states on SiC. We would also like to thank K. Das, R. Kaplan, J.R. Waldrop, S.A. Chambers, J. Crofton, and D. Alok for valuable discussions and R.S. Kern for contributing helpful information. We also thank many of the authors referenced in this paper for sending us copies of their publications.
[27]
[28] [29]
[30]
[31] [32] [33]
References [1] N.F. Mott, Proc. Cambr. Philos. Soc., 34 (1938) 568. [2] J. Bardeen, Phys. Rev., 71 (1947) 717. [3] W.J. Choyke, Optical properties of polytypes of SiC: interband absorption, and luminescence of nitrogen-exciton complexes (presented at the International Conference on Silicon Carbide, University Park, Pennsylvania, 1968), in H.K. Henisch and R. Roy (eds.), Materials Research Bulletin, Vol. 4, Pergamon, 1968, pp. S141 S152. [4] E.O. Johnson, RCA Rev., 26 (1965) 163 177. [5] R.W. Keyes, Proe. IEEE, 60 (1972) 225. [6] B.J. Baliga, J. Appl. Phys., 53 (1982) 1759-1764. [7] B.J. Baliga, 1EEE Electron Device Lett., I0 (1989) 455-457. [8] R.F. Davis, Z. Sitar, B.E. Williams, H.S. Kong, H.J. Kim, J.W. Palmour, J.A. Edmond, J. Ryu, J.T. Glass and C.H. Carter Jr., Mater. Sci. Eng., BI (1988) 77 104. [9] R.F. Davis, G. Kelner, M. Shut, J.W. Palmour and J.A. Edmond, Proe. IEEE, 79 (1991) 677 701. [10] R.F. Davis, Phys. B, 185(1993) 1-15. [11] J.H. Edgar, J. Mater. Res., 7 (1992) 235 252. [12] J. Pelletier, D. Gervais and C. Pomot, J. Appl. Phys., 55(1984) 994 1002. [13] A.J. van Bommel, J.E. Crombeen and A. van Tooren, Surf Sci., 48 (1975) 463 472.
[34]
[35] [36] [37] [38] [39] [40] [41] [42] [43] [44] [45]
103
K.W. Frese, Jr., J. Vae. Sci. Technol., 16 (1979) 1042-1044. R. Kaplan and T.M. Parrill, Surf Sci., 165 (1986) L45 L52. M. Dayan, J. Vac. Sci. Technol. A, 3 (1985) 361. R. Kaplan, Surf Sci., 215 (1989) 111-134. F. Bozso, L. Muehlhoff, M. Trenary, W.J. Choyke and J.T. Yates, Jr., J. Vae. Sci. Technol. A, 2 (1984) 1271 1274. L. Muehlhoff, W.J. Choyke, M.J. Bozack and J.T. Yates, Jr., J. Appl. Phys., 60 (1986) 2842-2853. A.N. Andreev, M.M. Anikin, A.L. Syrkin and V.E. Chelnokov, Semicond., 28 (1994) 577 579. L.M. Porter, J.S. Bow, R.C. Glass, M.J. Kim, R.W. Carpenter and R.F. Davis, J. Mater. Res., 10 (1995) 668 679. T.M. Parrill and V.M. Bermudez, Solid State Commun., 63 (1987) 231 235. R.G. Verenchikova, V.1. Sankin and E.I. Radovanova, Soy. Phys. Semicond., 17 (1983) 1123-1125. C.A. Mead and W.G. Spitzer, Phys. Rev., 134 (1964) A713 A716. S.H. Hagen, J. Appl. Phys., 39 (1968) 1458 1461. S.Y. Wu and R.B. Campbell, Solid-State Electron., 17 (1974) 683-687. L.A. Kosyachenko, E.F. Kukhto and V.M. Sklyarchuk, Translated from Zhurnal Prikladnoi Spektroskopii, 41 (1984) 615 620. K. Yasuda, T. Hayakawa and M. Saji, IEEE Trans. Eleetron Devices, ED-34 (1987) 2002 2008. L.M. Spellman, R.C. Glass, R.F. Davis, T.P. Humphreys, H. Jeon, R.J. Nemanich, S. Chevacharoenkul and N.R. Parikh, Mat. Res. Soc. Symp. Proe., 221 (1991) 99 104. L.M. Spellman, R.C. Glass, R.F. Davis, T.P. Humphreys, R.J. Nemanich, K. Das and S. Chevacharoenkul, Electrical characterization of epitaxial titanium contacts to ~(6H) silicon carbide (presented at Amorphous and Crystalline Silicon Carbide IV, Santa Clara, CA, 1992), in C.Y. Yang, M.M. Rahman and G.L. Harris (eds.), Springer Proceedings in Physics, Vol. 71, Springer, 1992, pp. 417-422. J.S. Bow, L.M. Porter, M.J. Kim, R.W. Carpenter and R.F. Davis, Ultramicroseopy, 52 (1993) 289 296. J.R. Waldrop, R.W. Grant, Y.C. Wang and R.F. Davis, J. Appl. Phys., 72 (1992) 4757 4760. J.R. Waldrop and R.W. Grant, Appl. Phys. Lett., 62 (1993) 2685 2687. L.M. Porter, R.C. Glass, R.F. Davis, J.S. Bow, M.J. Kim and R.W. Carpenter, Mat. Res. Soc. Symp. Proc., 282 (1993) 471 477. L.M. Porter, J.S. Bow, M.J. Kim, R.W. Carpenter and R.F. Davis, J. Mater. Res., I0 (1995) 26 33. L.M. Porter, J.S. Bow, M.J. Kim, R.W. Carpenter and R.F. Davis, J. Mater. Res., in press (1995). I. Ohdomari and K.N. Tu, J. Appl. Phys., 51 (1980) 3735 3739. J.L. Freeouf, T.N. Jackson, S.E. Laux and J.M. Woodall, J. Vac. Sci. Teehnol., 21 (1982) 570 573. J.M. Woodall and J.L. Freeouf, J. Vac. Sci. Technol., 21 (1982) 574 576. R.T. Tung, Appl. Phys. Lett., 58 (1991) 2821-2823. J.H. Werner and H.H. Guttler, J. Appl. Phys., 69 (1991) 1522 1533. M. Nathan and J.S. Ahearn, J. Appl. Phys., 70(1991) 811 820. L.M. Porter, unpublished results. M. Bhatnagar, P.K. McLarty and B.J. Baliga, IEEE Electron Device Lett., 13 (1992) 501 503. T. Urushidani, S. Kobayashi, T. Kimoto and H. Matsunami, High-voltage Au/6H-SiC Schottky barrier diodes (presented at the Silicon Carbide and Related Materials Conference, Washington, DC, 1993), in M.G. Spencer, R.P. Devaty, J.A. Edmond, M.A. Khan, R. Kaplan and M. Rahman (eds.), Institute of Physies ConJerence Series, Vol. 137, Institute of Physics, 1993, pp. 471 474.
104
L.M. Porter, R.F. Davis / Materials Science and Engineerhlg B34 (1995) 83 105
[46] A. Addamiano, Semiconductive Crystals of Silicon Carbide with Improved Chromium-Containing Electrical Contacts, US Patent No. 3 510 733 (1970). [47] D. Alok, B.J. Baliga and P.K. McLarty, IEDM Technical DiRest, IEDM 1993 (1993) 69l 694. [48] J.B. Petit, P.G. Neudeck, C.S. Salupo, D.J. Larkin and J.A. Powell, Metal contacts to n- and p-type 6H-SiC: electrical characteristics and high-temperature stability (presented at the Silicon Carbide and Related Materials Conference, Washington DC, 1993), in M.G. Spencer, R.P. Devaty, J.A. Edmond, M.A. Khan, R. Kaptan and M. Rahman (eds.), h~stitute of Physies Col{/'erence Series, Vol. 137, Institute of Physics, 1993, pp. 679 682. [49] R.C. Glass, L.M. Spellman and R.F. Davis, Appl. Phys. Lett., 59 (1991) 2868 2870. [50] R.C. Glass, L.M. Spellman, S. Tanaka and R.F. Davis, J. Vac. Sci. Technol. A, 10 (1992) 1625 I630. [51] J. Crofton, J.M. Ferrero, P.A. Barnes. J.R. Williams, M.J. Bozack, C.C. Tin, C.D. Ellis, J.A. Spitznagel and P.G. Mcmullin. Metallization studies on epitaxial 6H-SiC, in C.Y. Yang, M.M. Rahman and G.L. Harris (eds.), Amorphous and Crystalline Silicon Carbide 1V, Springer, Berlin, 1992, pp. 176 182. [52] J. Crofton, P.G. McMullin, J.R. Williams and M.J. Bozack, Trans. 2nd High Temperature Electronics Co111erence, Charlotte, NC, 1994. [53] V.A. Dmitriev, K. Irvine and M. Spencer, Appl. Phys. Lett., 64 (1994) 318 320. [54] T. Uemoto, Jpn. J. App/. Phys., 34 (1995) L7 L9. [55] T. Nakata, K. Koga, Y. Matsushita, Y. Ueda and T. Niina, Single crystal growth of 6H-SiC by a vacuum sublimation method and blue LEDs (presented at Amorphous and Crystalline Silicon Carbide and Related Materials II, 1989), ira M.M. Rahman, C.Y.-W. Yang and G.L. Harris (eds.), Sprblger Proceedings h~ Physics, Voh 43, Springer, Berlin, 1989, pp. 26 34. [56] A. Suzuki, Y. Fujii, H. Saito, Y. Tajima and K. Furukawa, J. Cryst. Growth, 115 (1991) 623 627. [57] J. Crofton, P.A. Barnes, J.R. Williams and J.A. Edmond, Appl. Phys. Lett., 62 (1993) 384 386. [58] R.N. Hall, J. Appl. Phys., 29 (1958) 914 917. [59] M.M. Anikin, M.G. Rastegaeva, A i . Syrkin and I.V. Chuiko, Ohmic contacts to silicon carbide devices, in G.L. Harris, M.G. Spencer and C.Y. Yang (eds.), Amorphous and Ci3'stalline Silieon Carbide II1, Springer Proceedings in Physics. Vol. 56, Springer, Berlin, 1992, pp. 183 189. [60] R.C. Glass, J.W. Palmour, R.F. Davis and L.M. Porter, Method of forming ohmic contacts to p-type wide bandgap semiconductors and resulting ohmic contact structure, US Patent No. 5 323 022 (1994). [61] S.M. Sze, Physics qf Semiconductor Devices, 2nd edn., Wiley, New York, 1981. [62] D.E. loannou, N.A. Papanicolaou and P.E. Nordquist Jr., IEEE Trans. Electron Det:., ED-34 (1987) 1694 1699. [63] J.R. Waldrop and R.W. Grant, Appl. PIo,s. Lett., 56 (1990) 557 559. [64] K. Das, H.S. Kong, J.B. Petit, J.W. Bumgarner, L.G. Matus and R.F. Davis, J. Eleetrochem. Sot., 137 (1990) 1598 1603. [65] H. Kong, H.J. Kim, J.A. Edmond, J.W. Palmour, J. Ryu, C.H. Carter, Jr., J.T. Glass and R.F. Davis, Mat. Res. Soe. Syrup. Proc., 97 (1987) 233 245. [66] L.M. Porter, Chemistry, microstructure, and electrical properties and their relationships to the Schottky barrier heights at interfaces between metals and single crystalline, n-type, alpha (6H) silicon carbide, Ph.D. Thesis, North Carolina State University, 1993. [67] V.M. Bermudez, J. Appl. Phys., 63 (1988) 4951 4959.
[68] S. Kurtin, T.C. McGill and C.A. Mead, Physic. Rev. Lett., 22 (1969) 1433 1436. [69] H. Daimon, M. Yamanaka, E. Sakuma, S. Misawa and S. Yoshida, Jlm. J. Appl. Phys., 25 (1986) L592 L594. [70] J.A. Edmond, J. Ryu, J.T. Glass and R.F. Davis, J. Electrochem. Sot., 135 (1988) 359 362. [71] A.J. Steckl and J.N. Su, IEDM (1993) 695 698. [72] H.J. Cho, C.S. Hwang, W. Bang and H.J. Kim, Effect of reaction products in monocrystalline /t-SiC~metal contact on contact resistivity (presented at the Silicon Carbide and Related Materials Conference, Washington, DC, 1993), in M.G. Spencer, R.P. Devaty, J.A. Edmond. M.A. Khan, R. Kaplan and M. Rahman (eds.), blstitute q[ Physics Co111brence Series, Vol. 137, Institute of Physics, 1993, p. 663. [73] T.B. Massalski, H. Okamoto, P.R. Subramaniara and L. Kacprzak (eds.), Binary Alloy Phase Diagrams, Vol. 2, ASM International, Materials Park, OH, 1990. [74] K.M. Geib, J.E. Mahan and C.W. Wilmsem W/SiC contact resistance at elevated temperatures (presented at Amorphous and Crystalline Silicon Carbide and Related Materials II, 1989), in M.M. Rahman, C.Y.-W. Yang and G i . Harris (eds.), Springer Proceedings in Physics, Vol. 43. Springer, 1989, pp. 224 228. [75] M.I. Chaudhry, W.B. Berry and M.V. Zeller, Int. J. Electronics, 71 (1991)439 444. [76] J.S. Shor, R.A. Weber, L.G. Provost, D. Goldstein and A.D. Kurtz, J. Eleetrochem. Soe., 141 (1994) 579 581. [77] J.S. Chen, A. Bachli, M.-A. Nicolet, L. Baud, C. Jaussaud and R. Madar, Mater. Sci. Eng. B, 29 (1995) 185 189. [78] C.A. Mead, Appl. Phys. Lett., 6 (1965) 103 104. [79] C.A. Mead, Solid-State Electron., 9 (1966) 1023 1033. [80] B.E. Deal, E.H. Snow and C.A. Mead, J. Ph)'s. CTrem. Solids, 27 (1966) 1873. [81] M.J. Turner and E.H. Rhoderick, Solid-State Electron., II (1968) 291 300. [82] L. Pauling, Tire Nature q f tire (Tremica/Bond, 3rd edn., Cornell University Press, Ithaca, NY, 1967. [83] B.T. Kolomiets, A.Y. Karachentsev and V.V. Spevak, Soy. Phys. Semicond., 7 (1974) 872 875. [84] M.C. Benjamin, S.W. King, R.J. Nemanich and R.F. Davis, unpublished results (1995). [85] L.J. Brillson, Phys. Rer. B, 18 (1978) 2431 2446. [86] D.W. Niles, H. Hochst, G.W. Zajac, T.H. Fleisch, B.C. Johnson and J.M. Meesc, J. Vae. Sci. Technol. A, 6 (1988) 1584 1588. [87] G.R. Purdy, D.H. Weidel and J.S. Kirkaldy, Trans. Met. Sot., 230 (1964) 1025 1034. [88] C.S. Pai, C.M. Hanson, and S.S. Lau, J. Appl. Phys., 57(1985) 618 619. [89] H. Hochst, D.W. Niles, G.W. Zajac, T.H. Gleisch, B.C. Johnson and J.M. Meese, J. Vac. Sei. Technol. B, 6 (1988) 1320 1325. [90] N. Lundberg, C.-M. Zetterling and M. Ostling, Appl. Sur/i Sei., 73 (1993) 316 321. [91] V.M. Bermudez, Applic. SurJl Sei., 17 (1983) 12 22. [92] M. Backhaus-Ricoult, Physicochemical processes at metal ceramic interfaces, in M. Ruhle, A.G. Evans, M.F. Ashby and J.P. Hirth (eds.), Metal Ceramic hire(laces, Vol. 4, AetaScripta Metallurgica Proceedings Series, Pergamon, New York, 1990, pp. 79 92. [93] C.G. Rhodes and R.A. Spruling, Fiber matrix reaction zone growth kinetics in SiC-reinforced Ti-6A1-4V as studied by transmission electron microscopy, in J.R. Vinson and M. Taya (eds.), Recent Advances in Composites in tire United States and Jpaan, Vol. 864, A S T M Spee. Teeh. Publ., American Society for Testing and Materials, Philadelphia, PA, 1985, pp. 585 599. [94] M. Backhaus-Ricoult, Ber. Bunsenges. Phys. Chem., 93 (I989) 1277 1281.
L.M. Porter, R.F. Darts / Materials Science and Engineering B34 (1995) 83-105
[95] I. Gotman, E.Y. Gutmanas and P. Mogilevsky, J. Mater. Res., 8 (1993) 2725 2733. [96] S. Hara, K. Suzuki, A. Furuya, Y. Matsui, T. Ueno, I. Ohdomari, S. Misawa, E. Sakuma, S. Yoshida, Y. Ueda and S. Suzuki, Jpn. J. Appl. Phys., 29 (1990) L394-L397. [97] R. Kaplan, P.H. Klein and A. Addamiano, J. Appl. Phys., 58 (1985) 321 326. [98] K.M. Geib, C.W. Wilmsen, J.E. Maban and M.C. Bost, J. Appl. Pbys., 61 (1987) 5299 5302. [99] D.L. Yaney and A. Joshi, J. Mater. Res., 5 (1990) 2197 2208. [100] L. Baud, C. Jaussaud, R. Madar, C. Bernard, J.S. Chen and M.A. Nicolet, Mater. Sci. Eng. B, 29 (1995) 126-130. [101] M.B. Chamberlain, Thin Solid Films, 72 (1980) 305 311. [102] J.J. Bellina, Jr. and M.V. Zeller, Mat. Res. Soc. Symp. Proc., 97 (1987) 265 270. [103] M. Nathan and J.S. Ahearn, Mater. Sci. Eng. A, 126 (1990) 225 230, [104] R. Stall, C.E.C. Wood, K. Board and L.F. Eastman, Electron. Lett., 15 (1979) 800~-801. [105] R.W. Grant and J.R. Waldrop, J. Vae. Sci. Technol. B, 5(1987) 1015 1019. [106] S.A. Chambers and T.J. Irwin, Phys. Rev. B, 38 (1988) 7484 7492. [107] S.A. Chambers and T.J. Irwin, Phys. Rev. B, 38 (1988) 7858 7861. [108] S.A. Chambers and V.A. Loebs, Phys. Rev. B, 47 (1992) 9513-9522. [109] S.A. Chambers, J. Vac. Sei. Technol. A, 11 (1993) 860-868. [110] S.A. Chambers, J. Vac. Sci. Technol. B, 7 (1989) 737-741. [111] S.A. Chambers, V.A. Loebs and D.H. Doyle, J. Vae. Sei. Technol. B, 8 (1990) 985 989. [112] S.A. Chambers and V.A. Loebs, J. Vac. Sci. Teehnol. A, 8(1990) 2074-2078. [113] S.A. Chambers and V.A. Loebs, J. Vac. Sci. Technol. B, 8(1990) 724 729. [114] S.A. Chambers and V.A. Loebs, J. Vac. Sci. Technol. A, 10(1991) 1940-1945. [115] R.F. Davis, Recent advances regarding the definition of the atomic environment, film growth and microelectronic device development in silicon carbide, in R. Freer (ed.), The Physics and Chemistry of Carbides, Nitrides and Borides, Kluwer, Dortrecht, 1990, pp. 589 623.
105
[116] Y.M. Tairov and V.F. Tsvetkov, in P. Krishna (ed.), Progress in Crystal Growth and Characterization, Vol. 7, Pergamon, New York, 1983, pp. 111-162. [117] H.R. Philipp and E.A. Taft, Intrinsic optical absorption in single crystal silicon carbide (presented at the Conference on Silicon Carbide, Boston, MA, 1959), in J.R. O'Connor and J. Smiltens (eds.), Silicon Carbide: A High Temperature Semiconductor, Pergamon, New York, 1959, pp. 366 370. [118] M, Schadt, G. Pensl, R.P. Devaty, W.J. Choyke and R. Stein, Appl. Phys. Lett., 65 (1994) 3120 3122. [119] W.E. Nelson, F.A. Halden and A. Rosengreen, J. Appl. Phys., 37 (1966) 333-336. [120] W. v. Muench and E. Pettenpaul, J. Appl. Phys., 48 (1977) 4823 4825. [121] D.K. Ferry, Phys. Rev. B, 12 (1975) 2361. [122] M. Bhatnagar and B.J. Baliga, IEEE Trans. Electron Deviees, 40 (1993) 645-655. [123] L. Patrick and W.J. Choyke, Physic. Rev. B, 2 (1970) 2255 2256. [124] G.A. Slack, J. Appl. Phys., 35 (1964) 3460 3466. [125] T.P. Chow and R. Tyagi, 1EEE Trans. Electron Dev., 41 (1994) 1481 1483. [126] N. Lundberg and M. Ostling, Appl. Phys. Lett., 63 (1993) 3O69 3071. [127] L.M. Porter, R.F. Davis, J.S. Bow, M.J. Kim and R.W. Carpenter, Trans. 2nd Internat. High Temperature Electronics Conference, Charlotte, NC, 1994. [128] G. Kelner, S. Binari, M. Shur and J.W. Palmour, Electron. Lett., 27(1991) 1038 1040. [129] J. Crofton, J.R. Williams, M.J. Bozack and P.A. Barnes, A TiW high-temperature ohmic contact to n-type 6H-SiC (presented at the Silicon Carbide and Related Materials Conference, Washington, DC, 1993), in M.G. Spencer, R.P. Devaty, J.A. Edmond, M.A. Khan, R. Kaplan, and M. Rahman (eds.), Institute q[ Physics Conference Series, Vol. 137, Institute of Physics, 1993, pp. 719-722. [130] S. Adams, C. Severt, J. Leonard and S. Liu, Trans. 2nd lnternat. High Temperature Electronics Con[erence, Charlotte, NC, 1994. [131] J.S. Shier, J. Appl. Phys., 41 (1970) 771-773.