In 1959, Richard Feynman gave a visionary talk entitled "There's Plenty of Room at the Bottom". He asked the question whether it will be possible to shrink devices all the way down to the atomic level. Since he could not find any law of physics against that, he concluded that is must be possible and will be done some day. In his characteristic way he went on to muse about all the neat things that one could do on such a scale. He started out with an atomic memory. Allowing 5x5x5=125 atoms to store one bit he estimated that all printed information accumulated over centuries since the Gutenberg Bible could be stored in a cube of material 1/200" = 0.1 mm wide, which is the barely visible to the naked eye. The ultimate storage medium would store a bit in a single atom, with a few atomic spacings between bits in order to prevent coupling between them.
How far have we come forty years later? A two-dimensional version of Feynman's atomic
memory can be formed on the surface of silicon by a small amount of gold (below on the right).
It looks similar to the CD-ROM on the left, but the scale is in nanometers instead
of micrometers. Thefore, the storage density is a million times higher.
The gold triggers the formation of self-assembled tracks, which are exactly five atoms wide. Extra silicon atoms sit on top of the tracks in well-defined positions. It is suggestive to assign an extra silicon atom to a
1 and a vacancy to a 0. The minimum empty area required around each bit is 5x4=20 atoms,
4 atoms along the track and 5 atoms from one track to the next. Feynman's 1959 suggestion of
spacing the bits 5 atoms apart was right on the mark.
Reading the memory consists of a simple, line scan with a scanning tunneling microscope (STM) along the
self-assembled tracks. There is no need to search in two dimensions for the location of a bit.
The signal is highly predictable since all atoms have the same shape and sit on well-defined
lattice sites. That allows for a high level of filtering and error correction.
Writing is more difficult. While loosely-attached surface atoms can be moved around with a STM at liquid helium temperature, that is much harder to achieve that at room temperature. In order to prevent them from moving around spontaneously, it is necessary to choose atoms that are strongly bound to the surface. Pushing them around with the STM tip requires a close approach and often leads to atoms jumping over to the tip. This problem can be turned into a solution by using the STM tip to remove a silicon atom for writing a 0. The memory is pre-formatted with 1 everywhere by depositing silicon atoms onto all vacant sites.
By investigating a storage device at the single atom limit one can learn something about how today's data storage might evolve in the future. The graph below shows readout speed versus storage density, two key properties of a memory. Compared to traditional data storage in hard disks, the silicon atom memory has a very impressive density (250 Terabits per square inch), but its data rate is extremely low. As the size of a bit shrinks, less energy that can be extracted from it during readout. Therefore, one needs a longer integration time for obtaining an acceptable signal-to-noise level. Even the theoretical limit of the data rate with the best possible readout electronics (top of the shaded region) is still lower than what hard disks achieve today. In the future, we can expect a tradeoff between density and speed. Actually, the data rate of hard disks shows signs of leveling out already. The data rate will have to be recovered by by a high degree of parallelism, such as many reading heads and multiple disks.
It is interesting to compare the silicon atom memory to data storage in Nature. DNA needs 32 atoms to store one bit, which is comparable to the area of 20 atoms around each bit at the silicon surface. The data rates are similar as well (see the green dot for DNA).
Images:
Large Images:
The highest commercial storage density is achieved with magnetic hard disks, whose areal density has increased by 100 million since their invention in Feynman's days. A storage density of more than 200 Gigabits per square inch has been achieved in demos. Typical storage media consist of a combination of several metals, which segregate into magnetic particles embedded into a non-magnetic matrix that keeps them magnetically independent. A rectangle containing about a hundred particles makes up a bit. Their magnetic orientation is color coded in the figure below. When such a bit is imaged by a magnetic force microscope (on the right) the collection of these particles shows up as white or dark line, depending on the magnetic orientation.
The storage density in magnetic data storage is limited by two factors:
1) Variations in
particle size, shape, spacing, and magnetic switching currently require averaging the signal over about a
hundred particles per bit. The error limits are extremely
stringent (less than one error in 108 read/write cycles, which can be
reduced further to one error in 1012 cycles by error-correcting codes).
2) Thermal energy is able to switch a bit if the magnetic particles become too small. This superparamagnetic limit is about 4-10 nm, and today's storage media are rather close to the limit already.
For further improvements it is critical to make the particles more homogeneous, such that fewer particles can be used to store one bit. The ultimate goal is a single particle per bit, which would increase the storage density by two orders of magnitude.
For a single-particle-per-bit memory one needs to line up the particles along well-defined tracks, such that the reading head can find them quickly. The big question is, how to create such patterned storage media without expensive lithography. We are investigating two generic methods for creating linear or circular arrays of dots. They can be formed by self-assembly (such as in the atomic scale memory above), or by directed assembly on a surface patterned by EUV interference lithography at 13 nm wavelength.
Self-assembly becomes more difficult when increasing the size of the tracks from atomic dimensions to the 10 nm regime required for magnetic storage. An example is shown below, where calcium fluoride produces strings of 10 nm dots along step edges of a silicon surface. The dot array shown here has a density of
2 Tera-dots per square inch.
Such dots might serve as masks for the deposition
of magnetic dots or for the assembly of magnetic particles.
J. Viernow et al., Appl. Phys. Lett. 74, 2125 (1999).
Adam Li, et al., Phys. Rev. Lett. 85, 5380 (2000).
Directed assembly is our current method of choice for lining up magnetic particles into tracks. A new NSEC nano-center provides the infrastructure for patterning surfaces with high resolution by using synchrotron light with a wavelength of 13 nm, fifteen times shorter that used in silicon technology today. We have deposited magnetic nanoparticles selectively on chemical line patterns. The current goal is to improve the packing density by enhancing chemical selectivity. Using chemically-selective NEXAFS spectroscopy we are able to find out about the bonding of the nanoparticles to the substrate.
Supported by NSF-DMR
