The Business Landscape for Memristor Electronics

In 2008 researchers at Hewlett Packard announced the physical realization of a new type of electronic circuit element called a memristor (i.e. memory resistor) which was theoretically predicted in 1971 by Leon Chua and which has the capability to form a new type of “transistor-less” non-volatile memory. However, in actuality, many different companies including Samsung, Micron Technology, Axon Technology, and Unity Semiconductor have been working with materials having the exact same characteristics of memristors but had simply not known the connection between their materials and the memristor theory due to the previous obscurity of Chua’s original papers. This article explores the landscape of memristors electronics from a business perspective focusing on the applications, materials, and companies involved in the commercialization of memristor materials.


Note: A categorized database containing the patent data used in developing some of the statistics below is available free upon request by sending an E-mail to bmouttet@gmu.edu with “memristor patent database” in the title.

 

Introduction

 

Over the past 40 years semiconductor electronics has been driven by Moore’s Law which predicts an exponential reduction in the size of transistors over time providing more functionality for electronic devices at less cost. However, due to physical limitations of scalability and manufacturing cost, Moore’s Law cannot continue forever and the optimization of conventional silicon transistors will inevitably come to an end. The Semiconductor Industry Roadmap forecasts further improvements in the next few years but past 2012 the future of electronics is beginning to appear uncertain. Furthermore, device electronics are evolving to require more adaptability in function such as cell phones incorporating the functions of internet, music, and video provider and video game interfaces including voice and image recognition capabilities. The degree of adaptability required to achieve these functions is gradually moving beyond what conventional computer architectures can provide. It is becoming clear that new materials other than silicon and new electronics architectures based on such new materials will be necessary to provide continued advancement in electronic functionality over the next few decades. Memristors may be the key to such new materials and architectures.
 
The memristor was originally theorized by a UC Berkeley professor named Leon Chua in 1971 as a new fundamental circuit element. It is not a small thing to suggest a new electronic circuit element since there are only a very small number of fundamental electronic components such as the resistor, capacitor, and transistor which, in a variety of combinations, form the basis of all of the electronic and computational devices in the world. In fact not since the development of the transistor by Bell Labs in the late 1940’s has a new fundamental electronic circuit element been developed (although many, many variations and improvement of transistors have been developed over the past 60 years). Like the transistor, the memristor behaves as a switch but, unlike the transistor, the memristor is a 2-terminal rather than a three-terminal device. A comparison of the attributes of the transistor vs. the memristor is given by Table 1 below. For further details on memristors Youtube has a 6-minute introduction video by Stan Williams (a senior researcher at HPLabs). Wikipedia also provide an information resource here and my other knols provide information on the mathematical analysis and a variety of applications of memristors. 
 
TABLE 1 
 

Transistor

Memristor

3-terminal switching device with an input electrode (e.g. source), an output electrode (e.g. drain), and a control electrode (e.g. gate)

2-terminal device with one of the electrodes acting either as a control electrode or a source electrode depending on the voltage magnitude.

Requires a power source to retain a data state

Does not require a power source to retain a data state

Stores data by electron charge

Stores data by resistance state

Scalable by reducing the lateral length and width dimensions between the input and output electrodes

Scalable by reducing the thickness of the memristor materials

Capable of performing analog or digital electronic functions depending on applied bias voltages

Capable of performing analog or digital electronic functions depending on particular material used for memristor

Fabrication requires optical lithography

Fabrication by optical lithography but alternative (potentially cheaper) mass production techniques such as nanoimprint lithography and self assembly have also been implemented

 

Applications

 

The three main areas of application currently under development for memristor electronics are non-volatile memory, logic/computation, and neuromorphics. Table 2 provides data on the issued U.S. patent count among the leading patent holders in each of these areas.
TABLE 2

# U.S.  patents claiming memristor application

 

Non-volatile Memory

Logic/

Computation

Neuromorphics

AMD

14

0

0

Axon Tech.

20

1

0

Energy Conversion Devices

28

5

2

Hewlett Packard

49

10

1

Micron Tech.

241

0

0

Samsung

18

0

0

Sharp

41

0

0

Unity Semi.

54

0

0

 

 

Non-volatile Memory

As evident from the patent data, non-volatile memrory is the dominant area being pursued for memristor technology. Of course most of the companies listed (with the exception of Hewlett Packard) do not refer to their memory in terms of the memristor and rather use a variety of acronyms (i.e. RRAM or ReRAM, CBRAM, PRAM, etc.) to distinguish their particular memory design. While these acronyms do represent real distinctions in terms of the materials used or the mechanism of resistance switching employed, the materials are still all memristors because they all share the same characteristic voltage-induced resistance switching behavior covered by the mathematical memristor model of Chua.

Flash memory currently dominates the semiconductor memory market. However, each memory cell of flash requires at least one transistor meaning that flash design is highly susceptible to an end to Moore’s law. On the other hand, memristor memory design is often based on a crossbar architecture which does not require transistors in the memory cells. Although transistors are still necessary for the read/write circuitry, the total number of transistors for a million memory cells can be on the order of thousands instead of millions and the potential for addressing trillions of memory cells exists using only millions (instead of trillions) of transistors. Another fundamental limitation to conventional memory architectures is Von Neumann’s bottleneck which makes it more difficult to locate information as memory density increases. Memristors offer a way to overcome this hurdle since they can integrate memory and processing functions in a common circuit architecture providing a de-segregation between processing circuitry and data storage circuitry (see my other knol for some examples).

At the 2008 Memristor Symposium at UC Berkeley HP projected that they would be producing memristor memory in 2012. Unity Semiconductor has projected their first memory products would be released in 2011. Axon Technology has licensed their version of memristor memory to Micron Technology in 2002 and to Infineon in 2004 

 
 

Logic/Computation

 
The uses of memristor technology for logic and computational electronics is less well developed than for memory architectures but the seeds of innovation in this area are currently being sown. Memristors appear particularly important to the areas of reconfigurable computing architectures such as FPGAs in which the arrangement between arrays of basic logic gates can be altered by reprogramming the wiring interconnections. Memristors may be ideal to improve the integration density and reconfigurability of such systems. In addition, since some memristor materials are capable of tunablity in their resistance state they can provide new types of analog computational systems which may find uses in modeling probabilistic systems (e.g. weather, stock market, biosystems) more efficiently than purely binary logic-based processors. Hewlett-Packard researchers have developed many of the central ideas in using memristors for reconfigurable computing although Stanford Ovshinsky (formerly of Energy Conversion Devices) has also developed some key ideas on using the analog properties of memristors for computation. 
 

 

Neuromorphic Electronics 

 
Neuromorphics has been defined in terms of electronic analog circuits that mimic neuro-biological architectures. Since the early papers of Leon Chua it was noted that the equations of the memristor were closely related to the Hodgkin-Huxley model describing the behavior of neural cells. Since memristors integrate aspects of both memory storage and signal processing in a similar manner to neural synapses they may be ideal to create a synthetic electronic system similar to the human brain capable of handling applications such as pattern recognition and adaptive control of robotics better than what is achievable with modern computer architectures. While so far few patents have been issued involving memristors applied to this area, several research groups are beginning to pursue this approach and several publications from HPLabs (Greg Snider) and other researchers (M Di Ventra, Y.V. Pershin) are exploring the possibilites in this area. Some other potential applications of memristors related to this area are discussed in my other knol
 
 
 

Materials

 
The memristor material discussed in the HP paper "The missing memristor found" was oxygen depleted titanium dioxide (TiOx). However, as pointed out by Leon Chua in the 2008 Memristor Symposium, the material used by HP is one of a variety of materials exhibiting a zero-crossing hysteresis curve in their current vs. voltage relationship which characterizes memristors. The memristive effects of many of these materials have been known for decades and provide a wide range of diversity and selectivity in the choice of memristor material for a particular application. Although the different memristor materials have their respective merits and possess differences in terms of their underlying physics each material share the same resistance switching properties possessed by memristors. One recent scientific review of these different materials is found in the article "Resistive non-volatile memory devices" by Rainer Waser. Table 3 provides some data on where different companies are focusing their patents for memristor related materials.
 
It is noted that the types of materials listed are not the only materials exhibiting memristive effects and some newly discovered memristive materials based on amorphous silicon and nanoparticle assemblies are being contemplated. 
 
 TABLE 3

# U.S. patents claiming memristor material

Binary Oxide

Metallization

Cell

Perovskite

Phase Change

 

Molecular or Polymer

AMD

0

0

0

0

12

Axon Tech.

0

20

0

0

0

Energy Conversion Devices

0

0

0

25

0

Hewlett  Packard

0

0

0

0

36

Micron Tech.

0

0

0

106

0

Samsung

4

0

0

10

0

Sharp

0

0

18

0

0

Unity Semi.

1

0

5

0

0

 

Binary Oxide 

The discovery of the resistive switching properties of binary oxide compounds actually predates the original memristor paper of Chua and can be dated back to 1967 when the article "New conduction and reversible memory phenomena in thin insulating films" was published in the Proceedings of the Royal Society. The phenomena was originally observed in silicon oxide formed with nanometer scale thickness and including gold ions. Silicon oxides have been a key insulating material used in the fabricating of electronics devices over the past 40 years. However, there is now a search for new thin oxide films having improved insulating properties to assist in the further scaling of transistor designs. This search may be very beneficial to this class of memristor material since research in new binary oxide compounds may benefit both the conventional semiconductor industry and the upcoming memristor industry. Furthermore, the manufacturing resources used in semiconductor processes may require less modification in comparison with other memristor materials and provide a smooth transition as semiconductor electronics is converted to memristor electronics over the next few decades.
 
While Samsung has some patents covering variations of binary oxide memristors there is still signifigant freedom to operate based on early published patents which are now expired such as US Patent 4,839,700 which teaches a variety of binary oxides such as WO3, Ir2O3, MoO3, ZrO2, and RhO2 adjusted to have memristive properties. A variety of other memristor variations based on TiO, CuO, NiO, ZrO, and HfO materials have been under experimental investigation for the past several years. Hewlett-Packard has been focusing on a TiO variation of the memristor but does not yet have any patents covering their materials or the manufacture. 
 
Some criticism of certain binary oxides have included poor endurance under repeated cycling and insufficient retention times. There have been efforts to improve these properties such as by Cu doping and the formation of bilayer structures including variable oxidation states but research is still ongoing. 
 

Metallization Cell 

Axon Technology has pioneered this approach occasionally referred to in terms of conductive bridge RAM (CBRAM) and have very strong patent protection over the materials involved.  The memristive effect is due to the formation of metallic filaments which interconnect two electrodes separated by an electrolytic material. The metallic filaments can be broken or reformed depending on the polarity of an applied voltage. It is noted that there may be some overlap between the materials used for this technique and some other material systems based on filament formation or fast ion conduction and since Axon has some relatively early broad patents (going back to 1996) they may have a stronger position than other larger companies such as Hewlett-Packard who lack patent protection for the materials they are using.
 
The materials exhibiting metal filment formation can exhibit fast switching times and large ratios of the high to low resistance state but may have a limit in the tunability of the resistance state so that truly analog memristors having a gradation of resistance states are more difficult to achieve. This may be disadvantagous to some analog electronics applications. 
 
 

Perovskite

Perovskite materials are based on a variety of ternary oxides including PCMO, SrTiO3, SrZrO3, and BaTiO3. These types of materials were originally studied in 2000 by a group at IBM and independently by another group at the University of Houston. These types of materials appear to have variable resistances which are more easily tunable via pulse number modulation which may make these materials more attractive for analog memristor electronics than the metallization cell or binary oxide materials. In addition, these materials have exhibited capacitive hysteresis effects in addition to the resistance hysteresis effects which may produce further applications.
 
Unity Semiconductor and Sharp have focused their non-volatile memory efforts on these materials. One potential drawback is that the high to low resistance ratio tends to be lower than in the metal oxide and metallization cell materials which may make scalability difficult.
 
 

Phase Change

The resistance switching properies of phase change materials have been studied since the 1960's and are based on chalcogenide glasses such as GeSbTe. Stanford Ovshinsky is a pioneer in this area and has been patenting innovations in these materials for the past 40 years assigned to the company Energy Conversion Devices (ECD). In the past several years Micron Technology has surpassed ECD in patents focusing on these materials. Samsung also has an interest but in more recent years (based on their patent filings) have extended their focus to include the metal oxide variety of memristors.
 
There have been several drawbacks to phase change memory in terms of long term stability and power consumption in the past but the biggest defect in comparison to the oxide materials may be the compatibility with the conventional semiconductor fabrication which could make widespread acceptance difficult. However, Intel and STMicroelectronics have begun development of phase change RAM and have shipped prototypes in 2008 indicating that these hurdles are being overcome. Numonyx was founded by Intel and STMicroelectronics in 2008 and appears to have a joint development deal with Samsung for product release in 2010.
 
 

Molecular/Polymer

Molecular and polymer materials have been investigated by Hewlett-Packard and Advanced Micro Devices as the basis for new types of non-volatile memory. HP has been working with molecular systems called rotaxane which are thought to exhibit a resistance switching effect based on a mechanical reconfiguration of the molecule. AMD has been focusing on ionic molecular and polymer materials which also produce resistance switching behavior and may have superior analog memristive properties than other materials. 
 
Temperature stability and compatibility with existing semiconductor materials and manufacturing methods are some limitations to the use of these molecules and polymers. However, ultimate scalability toward molecular electronics may require the use of the materials developed by HP or AMD.  
 
 

Companies

 
Note: The following list is not comprehensive of all of the companies that are working with materials having memristive properties. The selection of companies is primarily based on those that have patents related to materials or electronics which appear highly relevant to memristor electronics.


Advanced Micro Devices

AMD's memristor related patents focus mostly on polymeric or molecular materials used in non-volatile memory. AMD began filing patents in this area in 2001 with most of the patents being limited to molecular or polymeric materials. It is notable that AMD's flash memory interest is based on a partial ownership of Spansion, which declared bankruptcy in early 2009. It is also interesting to note that AMD assigned a continuation of one of their broader memristor-related patents (US 7,183,141) to Spansion having claims focusing on a fuse/anti-fuse memory, and which is not necessarily limited to polymeric or molecular memristors but may read on other materials such as binary oxides. However, this patent may have a validity issue in light of earlier prior art from Axon Technology and the University of Houston such as US 6,204,139.
 
 

Axon Technologies

Axon Technology has perhaps the strongest patent portfolio related to memristor technology due to their early patent priority and the breadth of their claims. While the total number of patents held is less than companies such as Micron Technology and Hewlett Packard, Axon has several basic patents such as (US 5,761,115) covering filament formation and ionic effects in resistance switching memory materials. Although originally applied to chalcogenide based materials, a similar effect appears in the memristive binary oxides used by HP and others and Axon's patents are not limited to chalcogenides. This may provide Axon with a very strong licensing position as memristor technology further develops.


Energy Conversion Devices

Energy Conversion Devices was founded by Stanford Ovshinsky who since the 1960's has pioneered many basic discoveries in the electronic properties of amorphous materials. While most of ECD's memristor-related patents are based on phase change materials it is notable that several of Ovshinsky's patents such as US 6,671,710 teach using the tunability of the resistance states of these materials for new types of computational architectures. In 2007 Ovshinsky left ECD to form a new venture called Ovshinsky Innovation but it is unclear as to whether further memristor-related development by this company or ECD will be pursued.

 

Hewlett Packard


Since the late 1990's HPLabs has been focusing on the development of a new type of molecular electronics capable of providing reconfigurable circuit architectures based on a nanoscale crossbar switch architecture. Up until around 2006 they were primarily focusing on molecular structures such as rotaxane as the switching material but have more recently switched to the binary oxide TiOx as their basic material. Despite having received much publicity due to the paper "The missing memristor found" HP's patent portfolio actually does not include protection for the materials they are using. However, it is notable that Samsung does hold a patent (US 7,417,271) covering the materials that HP is using for their prototypes so HP may need to take a license from Samsung to commercialize memristors using the TiO/TiOx materials. HP does have a few memristor related patents not limited to molecular structures (e.g. US 6,891,744, US 7,443,711) but it is notable that the validity of these patent's claims appears questionable when considering earlier prior art such as US 4,839,700 from CalTech and US 6,204,139 from the University of Houston.

At the
2008 Memristor Symposium HP researchers have suggested that their memristor memory would be released in 2012 and discussed new applications of memristor technology outside of pure non-volatile memory such as in programmable logic structures and neuromorphics but it is unclear whether these other applications would develop into products.

 

Micron Technology


Micron Technology is by far the holder of the largest number of patents related to memristors with activity going back to 1995 apparently being influenced by the basic work of Ovshinsky (based on their patent disclosures). While most of their patents are focused on the phase change type materials for resistive memory many of their claims are not limited to this type and can "read on" some of the more recently developed thin film oxide materials. Since Micron is one of the world's largest semiconductor memory companies it may have the largest stake in the development of a memristor based memory as a competitor to its existing memory markets and may use its large patent portfolio to protect its current market position and block out competitors rather than to create a memristor product. 
 

Samsung

Samsung started to focus on resistive switching memory architectures in 2003 and has obtained several basic patents for both phase change and binary oxide memristors. Samsung is notable to be the first to receive some broad patent coverage to binary oxide memristors some of which cover the materials used by Hewlett-Packard (see for example US 7,417,271). Like Micron, Samsung has a large stake in DRAM and Flash memory and may use its memristor patents to protect its market share and block out competitors rather than to create a memristor product.
 
 

Sharp

Sharp had supported some of the early research by the University of Houston in perovskite-based memristive materials and have been patenting perovskite-based memory since 2001. It is unclear at this point whether Sharp plans to commercialize their patents.
 
 

Unity Semiconductor

Unity Semiconductor is a start-up company based on a perovskite non-volatile memory system which plans to release a 1st generation 64GB memristor memory device in 2011 followed by a 256GB and 1TB in the 2nd and 3rd generations. The founder and chairman Darrell Rinerson was a VP in the flash division of Micron and a co-inventor of many of Unity's basic patents. Unity appears to have a very strong executive team with a technical advisory board including some of the original pioneers involved in the development of perovskite memory. Unity has been filing patents for their technology since 2002 and appears to have a very strong portfolio with the exception of a few overly broad patents such as US 6,870,755 in which some earlier patents from Ovshinsky or HP may invalidate some of the claims (e.g. 6,128,214).

Comments