Student Paper
Analysis of the Biasing Conditions and Latching Operation for
Si/SiGe Resonant Interband Tunnel Diode Based Tunneling SRAM
S. Sudirgo
a
, D.J. Pawlik
a
, S.L. Rommel
a
, S.K. Kurinec
a
, P.E. Thompson
b
, and P.R. Berger
c
a
Rochester Institute of Technology, Department of Microelectronic Engineering, Rochester, NY 14623,
b
Naval Research Laboratory, Code 6812, Washington, DC 20375-5347 ,
c
The Ohio State University, Department of Electrical Engineering, Columbus, OH 43210
A memory cell utilizing negative differential resistance (NDR) characteristic of tunnel diodes
was first proposed by Goto et al. in 1960 [1]. The inherently fast tunneling phenomenon and low power
consumption makes this type of memory architecture attractive. Combined with the possibility for
vertical integration of the tunnel diode directly atop of the source/drain region of the FET, a compact
design can be achieved. The concept was further refined by Van der Wagt et al. to achieve low stand-by
power by utilizing diode junction capacitance to lower peak and valley current densities. Utilizing
InP/InGaAs resonant tunnel diode (RTD), a nA operation was realized [2]. The realization in Si-based
system, however, is still limited. Utilizing p
+
Si/oxynitride/n
+
poly Si tunnel diodes, Morimoto et al.
demonstrated a low voltage T-SRAM [3]. The performance was limited by the poor performance of the
tunnel diode and the lack of capability to scale current density and size. In recent work by the authors,
Si/SiGe resonant interband tunnel diodes (RITDs) have been successfully integrated with CMOS
devices. A low voltage monostable-bistable logic element (MOBILE) latch has been demonstrated [4].
The current density of a Si/SiGe RITD is controlled by adjusting the SiGe i-layer thickness, making it
suitable for circuit integration with CMOS [5]. In this work, Si/SiGe RITD-based T-SRAM cell is
bread-boarded and its latching mechanisms have been investigated extensively.
As shown in Fig. 1(a), each T-SRAM cell consists of two tunnel diodes connected in series with
a current manipulator, NFET, connected into the sense node. In this experiment, Si/SiGe RITDs with 6
nm i-layer grown on top of p
+
implanted wells with PVCR of 2.25 and peak current density of 2.15
kA/cm
2
were used. During the stand-by (SB) mode, NFET is off, and V
DD
is fixed at 1.0V, as illustrated
in Fig. 1(b). RITD1 and RITD2 function as the driver and load, respectively, resulting in folded current-
voltage characteristics as depicted in Fig. 2. The intersections the drive and load curves indicate two
stable latching states, stand-by low (SBL) and high (SBH) at 246 mV and 746 mV, respectively. These
two states represent logic low and high, respectively. It is important to note that the intersection at the
negative differential resistance (NDR) region is unstable and is not be used as a latching point.
To latch into logic high, a current has to be supplied into the sense node through the NFET by
applying a bias of 2.0V to both the drain and gate. Fig. 1(c) depicts the circuit diagram during this write
high (WH) cycle. By activating the NFET, a current path parallel to RITD2 is formed, elevating the
overall current that passes through the load diode as shown in Fig. 3. At the same time, current injection
into the middle node also causes a decrease in drive current. As a result, the potential at the sense node
(V
SN
) changes abruptly from point A, 246 mV, to B, 802 mV. By restoring the stand-by biasing
conditions, i.e. shut-off the NFET, V
SN
restores to the nearest stable latching state at 746 mV, point C.
In the write to logic low (WL) cycle, current is subtracted or drained from the sense node by
simply applying a bias to the gate of the NFET and grounding the bit node. Since there is a potential
difference between sense and bit node, current will flow out of the sense node as illustrated in Fig. 1(d).
In other words, a current path parallel to RITD1 is created, causing a rise in the drive current. Unlike the
write high cycle, the load characteristic is undisturbed during WL cycle. As a result, V
SN
changes
suddenly from point C, 746 mV, to D, 199 mV (Fig. 4). Restoration to the stand-by conditions shifts the
latching point back to A, 246 mV. The corresponding time diagram of write and read cycles is given in
Fig. 5. Slight discrepancies between the values from I-V curves and transient diagram is due to parasitic
resistance present in the test setup.
In conclusion, a prototype of Si/SiGe RITD-based T-SRAM has been presented. Load line
analysis is performed to understand the latching mechanisms during the write and read cycles. This
demonstration will lead to realization of fully-integrated Si/SiGe RITD/NMOS T-SRAM. Ultra-low
power tunneling SRAM can be achieved by utilizing low current density RITDs.
1-4244-0084-8/05/$20.00/©2005 IEEE