Friday, July 26, 2013
Abstract
Carbon
nanotubes exhibit outstanding structural, mechanical and electronic properties,
which make them ideal wires for molecular electronics. Nanotechnology proposes
using nano-scale carbon structures as the basis for a memory device. Recent developments
in nanodevices presented a new approach for a highly integrated, ultra fast, nonvolatile
Random Access Memory (RAM) based on carbon nanotubes. A single-wall carbon
nanotube would contain a charged buckyball. That buckyball will stick tightly
to one end of the tube or the other. The bit value of the device is assigned depending
on which side of the tube the ball is. The result is a high-speed, non-volatile
bit of memory. A number of schemes have been proposed for the interconnection
of these devices and examine some of the known electrical issues. This paper
consists of the basic issues involved in building a RAM out of carbon nano-structures.
An attempt has been made to simulate the various problems and challenges to
frame the carbon nanotube memory devices.
Introduction:
If the DRAM
industry is to continue with its exponential rate of density
improvement,
it seems likely that there will need to be a radical change in the construction
of memory devices at some point. Certainly quantum-dot devices have possibilities
in this role. A different possibility is in the construction of a nanometer-sized
memory device based on the self-assembly of buckyballs inside of carbon
nanotubes. This “bucky shuttle” memory offers nonvolatility and terahertz
switching speeds. Also, each bit could require as little as two square
nanometers. The possibilities of such SWNT in RAM are quite interesting.
Current dynamic RAM requires both a transistor & capacitor (the transistor
stores or releases a charge from the capacitor). And it’s dynamic because the
capacitor, which loses charge very
quickly,
needs to be refreshed constantly. The Carbon nanotubes are nonvolatile—that is,
once it’s in an “on” or “off” state, it stays there, so there’s no refreshing
involved (saving processing time as well as work).
The angle
that the nanotubes bend when “on” is LESS than the buckling angle for Carbon nanotubes.
With current Static RAM, 4-6 transistors are needed. 2 inverters (each with 2 transistors)
with the input of one going to the output of the other store a bit, and 2 more for
the read & write lines. Whereas a single nanotube can have several
junctions, each storing a bit. The switching time (determined by the time to
move the upper nanotube between the bitable states) is 100Ghz (MUCH faster than
current processors). Also the storage capacity is incredible—116 gigs per
square cm
Fullerene
nanotubes
Carbon atoms can form a number of very
different structures; two of the better known is diamond and graphite. A new
carbon structure, the buckyball, was discovered in 1985. Soon after, the
nanotube was discovered. These carbon structures, collectively known as
fullerenes, have been of great interest to the physics and chemistry communities.
Graphite
consists of sheets of carbon atoms in a hexagonal arrangement (see Figure
1).The sheets are very loosely connected to each other. Taking a single sheet
of graphite, cutting a long narrow strip and rolling it into a long, narrow
tube would be a nanotube.
The ends of
the tubes usually form caps, as the dangling atoms will be receptive to forming
bonds with their neighbors. The resulting structure is shown in (Figure 2). The
electrical properties of the newly created nanotube depend upon the exact angle
at which the graphite was cut. A cut along one of the edges of the hexagons
would result in a conductive “armchair” nanotube. Other angles would result in
semi-conductors and even
Insulators.
This chiral angle can range from 0° and 30°.
A buckyball
can be thought of as the smallest of the nanotubes. It is simply the connection
of the two caps with no “tube” in between, and consists of exactly 60 carbon atoms
(see Figure 3). Its combination of hexagons and pentagons is exactly the same
as that found on a soccer ball. Generally very short nanotubes with 70, or even
80 atoms are sometimes also called buckyballs.
The
Nanomemory Device:
The
proposed nanomemory device (NMD) consists of two parts: the “capsule” which
holds the much smaller, charged “shuttle.” (Figure 4) shows an example where
the capsule is a C240 nanotube while the shuttle is a buckyball. The buckyball
contains a potassium ion (K+), which gives the shuttle its charge. The outer
dimensions of this capsule would be about 1.4nm in diameter and about 2.0nm in
length. This K+ inside of a C60 inside of a C240 (K+@C60@C240) structure is the
smallest and simplest device that is considered.
However other options such as longer capsules, which uses other nanotubes as
shuttles, as well as having many charged shuttles inside of each NMD are
promising.
The state
of the memory device is determined by the location of the shuttle: if it is on
one side of the capsule, it is treated as a ‘1’; on the other it is treated as
a ‘0’. The Vander Waals forces between the tube and the shuttle will tightly
bind the shuttle to one end of the tube or the other. There is an unstable
equilibrium point when the shuttle is in the exact middle of the capsule, but
the proposed scheme for writing to the device would prevent the shuttle from
ever coming to rest there.
Writing
to the NMD:
The
potential energy of the shuttle at various locations in the NMD is shown in (figure
5). The solid line indicates the potential energy curve when no electric field
is applied. The two potential energy wells are found when the shuttle is on one
side of the capsule or the other. These wells keep the shuttle bound to either
side of the capsule. The other two lines display the potential energy when a
two-volt potential difference is applied. When such a voltage is applied there
exists only one local minimum, and the shuttle will move to that side of the
tube. It is with this two-volt potential difference that provides means to
write to the NMD. In general the amount of voltage, which needs to be applied,
depends upon the length of the capsule. A field of 0.1 volts/c is sufficient to
move the shuttle from one side of the tube to the other.
One
important issue is how long it takes to perform a write to the NMD. Because of
the bouncing effect observable in (Figure 6) it is necessary to wait for the
buckyball to come to a stop. Generally the time to settle will
Reading
from the NMD:
Writing to
the nanomemory devices is the easy part; reading from them is much more
challenging. Somehow the state of the device must be sensed. Numbers of ways have
been proposed to perform a read. The first requires three wires to be connected
to the capsule: one on each end, and one in the middle. The position of the
buckyball is detected by examining the resistance between the middle wire of
the nanotube and the ends. A lower resistance will be found on the end that has
the shuttle. This three-wire solution has a number of problems, not the least
of which is that making a connection to the middle of a nanotube seems
difficult. However, a long capsule and shuttle would perhaps make this solution
viable.
A
device without the middle wire would be easier to fabricate. The notion of a destructive
read could be applied here. A read would then be performed in the same way as a
write. During that write some current will flow if the shuttle moves from one
side of the nanotube to the other. The total current that will flow is limited
by the amount of charge held in the shuttles. It is this type of a read that
makes a necessity to use many shuttles in our capsule to attempt to increase
the amount of the current flow. Neither of these reading schemes is
particularly satisfactory.
From
NMD to RAM:
Once the
memory device is fabricated it will still be a challenge to integrate the
devices into a large RAM cell. The two possible implementations are
"metal-wired" and "nano-wired". The metal-wired approach is
the most viable implementation. However, it is equally useful as a
stepping-stone on the path to the nano-wired device, which offers tremendous
density improvements.
Metal-wired:
The easiest
device to fabricate would replace the traditional DRAM
Capacitor
or transistor memory cell with a large number of nanomemory devices. Current VLSI
fabrication techniques could be used, but with the addition of a layer of
nanodevices. A “forest” of nanotubes has been already built, and a similar
technique could be used to create a forest of memory devices between two
conducting layers. (Figure 7a) is a representation of a 4-bit nanomemory device
(Figure 7b) shows a more detailed view of a single bit. A number of nanomemory
devices are used to make up a single bit of memory. The number of devices per
bit will depend upon the minimum line size of the lithography process used.
With a 70nm wire width there could be nearly 1,000 nanomemory devices per bit. Considering
how a write to the memory device in (Figure 7) will work: Nearly a voltage
differential of 2.0 volts will move the shuttles from one side of the capsule
to the other. In order to write a ‘0’ to bit three, a +1.0 volt potential is
applied to wire B and a - 1.0 volt potential to wire D. If all the other wires
are held at ground, only at the addressed bit will there be a strong enough
electrical field to the shuttles. A ‘1’ is written by reversing the voltages.
Writing to an entire row (or column) would be a two-stage process as the 1’s
and the 0’s would have to be written at different times.
Considering
a destructive read with a forest of nanotubes: Forest
of nanotubes move a large number of charged ions. The data in the process of
doing the read will get destroyed if those ions move. However, it can be
written back later, much as it is handled in a traditional DRAM. This
metal-wired carbon nanomemory device has a number of useful features. It is
non-volatile, the device itself switches very quickly, and it would seem to be
just as build able using 70nm lithography as it is using 350nm lithography.
Nano-wired:
In this scheme, the memory array is
made entirely out of nanomemory devices and carbon nanowires. The metal wires
are replaced by conducting nanotubes. Each bit of memory uses only a single
NMD. The logic, sense-amps and pads are made using traditional devices. It is
similar to the metal-wired proposal. Nano-wires allow for very high densities, with each bit fitting in about two square
nanometers. Laying out this network of carbon requires self-assembly techniques
well beyond anything we can do today.
Suspended nanotube device architecture.
(A) Three-dimensional view of a suspended crossbar array
showing four junctions with two elements in the ON (contact) state and two elements
in the OFF (separated) state. The substrate consists of a conducting layer
[e.g., highly doped silicon (dark gray)] that terminates in a thin dielectric layer
[e.g., SiO2 (light gray)]. The lower nanotubes are supported directly on the
dielectric film, whereas the upper nanotubes are suspended by periodic
inorganic or organic supports (gray blocks). Each nanotube is contacted by a
metal electrode (yellow blocks). (B) Top view of an n by m
device array. The nanotubes in this
view are represented by black crossing lines, and the support blocks for the suspended
SWNTs are indicated by light gray squares. The electrodes used to address the
nanotubes are indicated by yellow squares.
Conclusion
The
proposed nanomemory device is one candidate for carrying memory devices beyond
the limits of current DRAM technology. It has three important characteristics:
it is small, non-volatile, and fast. At this point the carbon nanomemory has
been simulated and buckyballs inside of nanotubes have been created, but a
working memory device does not exist. Building it will be a challenge, but
self-assembling carbon nanotechnology is an active research area with
continuous and promising advances.
