The present invention concerns a ferroelectric or electret volumetric
memory device, wherein a ferroelectric or electret memory material is provided in
sandwich between first and second electrode layers respectively comprising first
and second parallel stripe-like electrodes forming word lines and bit lines of a
matrix-addressable memory array, wherein word lines and bit lines of the array are
oriented substantially at right angles to each other, wherein memory cells are defined
in volumes of memory material sandwiched between respective crossings word lines
and bit lines, and wherein a plurality of memory arrays are provided in at least
one stack such that the at least one stack of memory arrays realizes the memory
device with a volumetric configuration.
In a wider sense the present invention also concerns data storage
and/or processing devices based on ferroelectric thin films.
No prior art of direct relevance regarding braiding/folding as taught
in the present invention has been identified. However, a brief general background
shall be given to illustrate the present state of the art, put the present invention
into context and highlight the objects of the invention:
Memory chips have the advantage over conventional magnetic, optical
and other mechanical storage devices of being capable of very fast read and write
operations. Furthermore, they are solid state, have reasonably low power consumption
and may offer high transfer speeds. The disadvantage is their limited capacity to
store information, and a relatively high production cost relative to the storage
capacity. Due to scaling problems and limited area, typically restricted to less
than 1 cm2/chip, this situation is not likely to be much altered in the
A solid state memory concept which circumvents the limitations described
above has been developed based on hybrid silicon/polymer chips. The approach includes
stacking thin layers of polymeric memory films on silicon substrates and accessing
the passively addressed memory layers via the substrate circuitry. The problem with
this solution, however, is that the number of memory layers in the stack typically
is limited to 8 - 16 layers. Increasing this number is technically possible, but
generally not practically viable for most mass market applications. Negative factors
in this connection include extra overhead and real estate costs for driver circuitry
(decoders and sense amplifiers in particular); reduced yield due to increased number
of processing steps; problems associated with planarization when the number of memory
layers becomes larger then the range mentioned; and a larger number of processing
steps increasing the risk that underlying polymer layers are negatively affected,
with a reduced functionality as a result.
There is also an unbalance in the hybrid memory concept to the extent
that producing the silicon part is complex and requires advanced (albeit standard)
fabrication processing, while building the memory stack in itself is a very simple
low cost procedure, which potentially could be done outside fabrication facility,
with non-lithographic tools. However, when these stacks are built on silicon the
factors listed above combine to make this more costly and capacity limiting than
desirable, e.g. it does become more cost-effective to use two or several chips to
achieve the same capacity.
Also, the procedure used to deposit memory films on silicon is in
practice limited to simple spin coating. This deposition technique has several advantages,
but may also introduce unwanted side effects, like creating a larger than desirable
internal stress, problems in controlling the film morphology and uniformity, etc.
One procedure used to improve morphology is stretching of films, which cannot be
applied in the hybrid case, another is to anneal films under high pressure which
is not very applicable either (when spin coating and rigid substrates like silicon
Because of the area restrictions related to a silicon-based device,
the only feasible patterning approach is standard photomicrolithography i.e. providing
a high resolution line pitch. This excludes low cost, non-lithographic patterning
tools like inkjet printing and micropatterning.
Another problem related to the hybrid concept is packaging, especially
soldering, which requires temperatures much higher than the melting temperature
of polymers (more than 60°C higher). The polymer does not necessarily destruct when
exposed to higher temperatures than its melting temperature, but a rework (reanneal)
is required to bring back the film properties. More problematic is what happens
to the electrode/film interface, which easily is destroyed when the polymer enters
the liquid phase. This represents is substantial problem when multilayer stacks
Film properties are also seriously affected by electrode application,
e.g. the top electrode deposition may have negative effects on the bottom electrode
interface, e.g. by kicking off undesired ion transport, which may initiate a fatigue
process in the polymer films. Morphological chain defects may also be induced.
In regard of the above-mentioned disadvantages, it is thus a first
object of the present invention to provide novel architectures for solid state thin-film-based
devices whereby the effective area available for data storage and/or processing
can be made large through stacking of individual layers in a dense volumetric structure.
It is a second object of the present invention to prescribe how such
stacking can be achieved in a practical manner while at the same time providing
individual addressability for locations inside the stack through a limited number
of electrical connections that are accessible from the outside of the stack.
It is a third object of the present invention to provide stacks containing
a multiplicity of matrices, where each matrix contains a large number of thin-film
cells that can be individually connected via passive matrix addressing.
It is a fourth object of the present invention to provide individual
stacks in the form of modular units suitable for integration into devices with specialized
functions and/or into larger units that add the capacities of two or more separate
It is a fifth object of the present invention to apply the stacking
concept to the manufacturing of data storage and/or processing devices that contain
sub-units demanding mutually incompatible process steps.
The above objects as well as further features and advantages are realized
with a memory device which according to the invention is characterized in that a
stack of memory arrays is formed with two or more ribbon-like structures being folded
and/or braided into each other, each ribbon-like structure comprising a flexible
substrate of non-conducting material, that first and second electrode layers respectively,
are provided on each surface of the substrate, such that the electrode layers each
comprises the parallel stripe-like electrodes provided extending along the ribbon-like
structure and a layer of memory material covers one of the electrode layers thereof
and extends uninterrupted between the edges of the ribbon-like structure, that each
memory array of the stack is formed by overlapping portions of a pair of adjacent
ribbon-like structures folded and/or braided such that they cross in substantially
orthogonal relationship, and that the word lines and the memory layer of a memory
array in a stack are contained in the first ribbon-like structure of a pair of adjacent
structures of this kind and the bit lines contained in the second ribbon-like structure
Further features and advantages of the present invention are disclosed
by the dependent claims.
The invention shall now be explained in greater detail with reference
to preferred embodiments and the appended drawing figures, of which
- fig. 1a shows a longitudinal cross-section of ribbon-like structure as used
in the present invention,
- fig. 1b a transversal cross-section thereof,
- fig. 1c two ribbon-like structures crossing and contacting each other, forming
a memory array in the overlap area,
- fig. 2a an example of a stack of ribbon-like structures according to the invention,
- fig. 2b an example of an embodiment of the present invention with two stacks
similar to the one in fig. 2a,
- figs. 3a, 3b cross sections of respectively the first and second stacks in fig.
- fig. 4a schematically a ribbon-like structure with connection areas or contact
fields on the front and back, and
- fig. 4b how the ribbon-like structures may be provided to form stacked memory
arrays according to another embodiment of the present invention.
According to the invention, shown in figs. 1a and 1b, there is provided
a ribbon with a carrier substrate 3 of a flexible material (e.g. polymer or metal)
coated on one surface with a thin layer of memory film 1 on top of parallel electrodes
2 (in the longitudal direction), while the opposite surface of the ribbon either
has a similar structure or a layer of parallel electrodes 4 only, the whole embodiment
thus constituting a ribbon-like structure R (henceforth termed only as "ribbon")
as used in the present invention.
The memory film 1 has addressing, charge storing (bistability) and/or
switchability capabilities allowing memory matrices to be passively addressed and
memory cells to be constituted by the memory film 1 being sandwiched between crossing
electrodes 2;4 by a suitable arrangement of two or more ribbons contacting each
other and oriented mutually at an angle of about 90°.
According to a preferred embodiment of the invention two or more ribbons
R are stacked together, such that at each interface portion a memory array M is
created in the adjoining ribbons and represented by top and bottom electrodes 2;4
and the memory film 1 in between. This is illustrated in fig.1c, showing a crossing
between two ribbons Rk and Rk+1 which may constitute part
of a larger stack. The ribbons R may have an arbitrary width and are mutually oriented
in 90° angles with respect to each other such that the resulting stack has a square/cubic
(chip-like) shape. Fig. 2a shows how ribbon R2 is stacked against ribbon R1, ribbon
R3 against ribbon R2, and so forth, up to ribbon R10 stacked against ribbon R9.
In fig. 2a the odd-numbered ribbons R1,...R9 form a first subset or
assembly X1, the even-numbered ribbons R2,...R10 a second assembly X2
oriented perpendicularly to the first. The resulting memory arrays form a stack
The ribbon overlaps are laminated together through an anneal process,
e.g. under high vacuum/high pressure. By using a similar structure, i.e. memory
film against memory film, compatibility problems related to the lamination process
may be substantially reduced.
In an alternative embodiment the ribbons are folded according to a
plurality of patterns, including concertina-wise, oval, circular/ring or twisted,
as e.g. shown in fig. 2b. "Woven" threads may be made, which again can be used to
make fabrics, etc. It will be possible to build large area structures in this way,
thus enabling integration of memory into other devices, like into the casing of
mobile phones, as a "top coating" on curved or other surfaces, etc.
The ribbons R1,...R9 arranged in a first subset or assembly X1
and the ribbons R2,...R10 arranged in a second subset or assembly X2
are applied in a memory device according to the invention. The embodiment of such
a memory device as shown in fig. 2b forms two distinct stacked structures S 1, S2
as indicated by the boxes with stitched outlines. Each of the ribbons R1-R10 is
bent such that pairs of odd and even numbered ribbons are permuted in the stack
S2 with respect to the succession of ribbons in stack S1. This implies e.g. that
the lowermost ribbon R9 in stack S1 is bent upwards to pair with the ribbon R10,
while e.g. ribbon R1 which is paired with ribbon R10 in the stack S 1 now is bent
downwards to pair with ribbon R2 in stack S2.
If the embodiment shown in fig. 2b is regarded as a plan view of the
arrangement of the ribbons in the device according to the invention, it will by
persons skilled in the art readily be understood that the combination of displacements
and staggering shall provide a significant contribution with regard to minimizing
capacitive crosstalk or other undesired couplings between the separate memory arrays
M forming each stack. Not only can each individual memory cell in a memory array
in a stack be addressed, i.e. written or read, without unwanted disturbances which
seriously reduce the signal/noise ratio of e.g. an output signal, but in addition
a parallel addressing of all memory cells in an array and if desired all memory
arrays in a stack will be possible while simultaneously still keeping any disturbing
influences at a minimum.
The arrangement of the stacks S1, S2 in the memory device according
to the invention is shown to better advantage in cross sections in respectively
figs. 3a, and 3b. In stack S1 (fig. 3a) the first ribbon R9 is provided orthogonally
to the second ribbon R2 and second electrodes 2 in ribbon R9 now can be considered
as word lines in a memory array M1 with the bit lines provided by the electrodes
4 in ribbon R2 and so on. In other words a memory array M comprises and is formed
by a portion of respectively adjacent pairs of ribbons Rk, Rk+1
in a stack S. The stack S2 as shown in fig. 3b appears similar to the stack S1 in
fig. 3a, but with the directions of adjacent pairs of ribbons R now rotated by 90°
such that the orthogonal crossing between successive ribbons in the stack S2 is
retained. It will be seen from fig. 2b that the first assembly X1 of
odd-numbered ribbons R1....R9 and the second assembly X2 of even-numbered
ribbons R2...R10 each could be followed by a similar adjacent assembly of ribbons
provided in the lateral direction (i.e. in a side-by-side arrangement) with a corresponding
orientation such that additional stacked structures S could be formed in the regions
where the assemblies X1, X2 intersect. The additional assemblies
of ribbons can also be provided in a staggered arrangement if so desired. Moreover
it can also be seen from a contemplation of the arrangement in fig. 2b that the
direction of staggering could inverted between neighbour stacks of memory arrays,
i.e. for instance of ribbon R9 in stack S1 would still pair with ribbon R10 in stack
S2, but now in the projected position of ribbon R1 in stack S2, while of course
ribbon R1 then still would pair with ribbon R2 in stack S2, but now in the projected
position as shown for ribbon 9 in fig. 2b, and so on. The implication of this is
of course that similar considerations would also be applied to the stagger of the
ribbons in the even-numbered assembly X2 of ribbons R2,...R10. The distance
as measured by the length of the electrodes 2,4 and ribbons R between the memory
arrays M in one stack S 1 and the memory arrays M in the other stack S2 would then
At the end of the ribbons R there can be provided connecting and contact
means 5, as shown in figs. 4a and 4b. This would e.g. allow the electrodes 2,4 to
be passively connected to pads on an underlying, not shown silicon chip, in which
case sufficient redundancy is required to allow for a certain degree of misalignement.
Alternatively or in addition there is provided for some (e.g. decoder/router) circuitry
based on thin-film transistors (TFT) at the electrode ends, reducing the number
of contacting points to facilitate a more robust connection. Such robustification
may not only allow for a much denser electrode pattern, and thus increased storage
density, but also allow folded (and packaged) memory stacks to be connected to the
silicon chips (or pads connected to the silicon chip(s) by the end user, thus opening
up for very low cost add-on memory modules.
A further enhancement of this concept would be to provide all driver
circuitry, also including sensing circuitry, and required to operate the memory,
at the ribbon ends. This would turn into the folded memory stack completely self-sustained
Yet another enhancement would be to distribute the required circuitry
evenly over the ribbons, directly in contact with (at the sides of) each individual
memory matrix, as row and column drivers/decoders, and just contacting every one
of these to a common bus/traffic coordinator on the ribbon(s) and then communicate
with external hardware via hardwired or wireless contact(s).
When silicon or silicon/TFT circuitry on a supporting substrate is
used, ribbon surfaces can be attached to silicon driver circuitry by bending one
surface end over the other as shown in fig. 4b before attachment to contact pads
on the underlying, not shown device substrate surface. If all driver circuitry is
built on the ribbon(s) such bending is not required.
The resulting memory stack S built in this manner represents novel
approaches and solutions to the problems discussed in the introduction. What basically
takes place from an architectural standpoint is that because each memory array M
is built on an individual substrate, the challenges are mostly reduced to those
related to building single layer memories. This shall involve simple modular sub-units
represented by the individual ribbons, which can be manufactured in specialized
manufacturing equipment before being assembled in scalable fashion into stacks.
The concept allows the use of a very large number of stackable memory
ribbons, the only restriction being access to silicon real estate in the silicon
or hybrid silicon/TFT circuitry case, a restriction which does not exists in the
"all circuitry on the ribbon (all-TFT)" case. This translates directly into a very
large storage capacity, or arbitrarily large storage capacity in the all-TFT cases.
Because the approach is close to processing a single layer memory,
most, if not all, the process and temperature compatibility issues related to multilayer
Similarly, by avoiding the deposition of top electrodes directly onto
the memory films, possible negative effects of such procedure can also be avoided.
A further positive effect on film morphology is the possibility to utilize stretching
of the films before baking, thus ensuring a more orderly crystalline structure.
Alternative deposition techniques to spin coating, like dip-coating/doctor blading/meniscus
coating, may also have positive influence on film morphology.
Because the available area is large, a much more relaxed patterning
process can be implemented, allowing non-lithographic tools and true reel-to-reel
processing to be realised. This in its turn appreciably shall reduce the production
The large feature sizes that can be utilized will also improve the
signal/noise ratios with respect to cell signals, simply because cell sizes are
so much larger. This will allow more variation in film thickness etc., thus reducing
potential problems related to processing memory structures on flexible substrates.
High temperature packaging is facilitated in cases where devices are
built on silicon chips, since the silicon part can be processed and soldered before
the polymer is attached.