The basic current-voltageproperties of SiC
JFET's at 600 °C are qualitatively similar to room
temperature GaAs MESFET's employed in high-
speed digital IC's. Therefore, a resistive load direct-
coupled FET logic (DCFL) approach, with some
obvious parallels to high-yield GaAs DCFL, was
adopted to demonstrate simple 600 °C digital logic
using SiC JFET's. A simple inverter logic gate merely
requires a near-zero threshold voltage FET coupled
to a resistor.
The non-planar epitaxial gate JFET design
shown in Figure 1 was adopted in favor of a planar
ion-implanted structure, largely to alleviate the
challenging process of sufficiently activating high-
dose p+ ion implants in SiC. The two-level
interconnect approach uses oxidation-resistant
silicon nitride as the dielectric passivant along with
oxidation resistant gold for the metal interconnect.
Contrary to the depiction of Figure 1, the devices
are laid out so that both the second and third layers
of silicon nitride always covered the first via layer to
the oxygen-sensitive metal-SiC ohmic contact
interface. However, because non-optimized ohmic
contact metals were employed in this experiment,
long term 600 °C operation was not obtained.
Fabrication Process
The mesa-isolated epitaxial channel JFET
structure shown in Figure 1 was grown (starting from
an off-axis commercial p-type 6H-SiC substrate [5])
at the NASA Glenn Research Center using the
epitaxial growth system described in [6]. Proper
application of site-competition dopant control
principles was key to successfully realizing this
device structure [7]. An underlying p- buffer epilayer
around 12 microns thick and doped p-type below
5x1015 cm-3 is not depicted in Figure 1. A 0.5-0.7
_m n-channel layer doped around lx1016 cm-'3 was
grown to obtain threshold voltages acceptably close
to zero over the temperature range of interest. The
thin p* gate epilayer (0.1-0.2 #m, NA > 102o cm -3)
must be as degenerately doped as possible, to
assure sufficient electrical contact to the gate layer
as well as adequate conductivity through long
narrow gate fingers with no metal directly on top.
Site-competition enables all the key n-type and p-
type layers to be grown within a single continuous
run that minimizes undesired contamination and
dopant spikes that can arise from growth
interruption and re-initiation [8]. It also enables
better immunity to epi-reactor impurities that can
harm reproducibility as the reactor components,
especially the sample-holding susceptor, can
degrade from one growth run to the next.
Following epitaxlal growth, an aluminum p+ gate
pattern etch mask is defined and patterned by liftoff.
I Met2_
_//_Met!|! , I iMet! r.l-_
P Substrate
Figure 1: Simplified (see text) cross-sectional
representation of 6H-SiC epitaxiat JFET's used
to implement 600 °C logic gates. The shaded
regions are insulating silicon nitride dielectric.
A reactive ion etch (RIE) in 14 CHF3 : 10 02 400 W
plasma removes the p+ epilayer from unmasked
areas to expose the n-channel layer. Without
removing the first gate etch mask, a second
aluminum etch mask defining mesas for the active
n-channel areas is deposited and liftoff patterned. A
second RIE to a depth of approximately 0.7 p.m is
then carried out to isolate each device. Both
aluminum etch masks are then stripped with a wet
etch.
In preparation for ion implantation,
approximately 0.1 #m of silicon is E-beam deposited
over the entire wafer. Then, a 0.8 I_m layer of silicon
is liftoff patterned to delineate areas for high dose
nitrogen source/drain implants to facilitate ohmic
contacts to the n-channel layer. The wafers are then
sent to a commercial vendor [9] for 600 °C nitrogen
implantation of a box profile using 1.4x10_% 7x10 TM,
and 3.6x10 TM cm-2 doses at 80, 40, and 20 keV
energies, respectively. Following Ion implantation
the silicon mask layers are chemically stripped in 1
HF : 1 HNO3. The same basic implant process is
then repeated to carry out lighter-dose nitrogen
implant (3.2x10 _2,8x10 _, 1.2x10 _2,and lx1012 cm -2
dose at 33, 24, 18, and 10 keV, respectively) that
self-aligns with the etched p* gate as shown in
Figure 1. This second implant prevents complete
pinchoff of the n-layer resistors by the sub-channel
and surface depletion regions and increases
transistor source-to-gate and gate-to-drain
conductance. After implant mask stripping and
solvent/acid wafer cleaning, the nitrogen implants
are electrically activated by placing the wafer into a
closed SiC crucible (made from SiC wafer pieces)
and annealing at 1400 °C for 30 minutes in an
argon atmosphere tube furnace. The wafer is again
cleaned with acids to remove any residual oxides
that might have formed during the implant anneal.
NASA/TM--2000-209928 2