8/20/2019 MeRAM

1/26

A Seminar Report On

MAGNETOELECTRIC RANDOM

ACCESS MEMORY(MeRAM)

Submitted in partial fulfilment for theDegree of Bachelor of Technology in

Electronics And Communication

Submitted byAlaknanda Agarwal

Under the guidance of Prof. Kiran Dange

USHA MITTAL INSTITUTE OF TECHNOLOGYSNDT WOMEN’S UNIVERSITY

SANTACRUZ(W)MUMBAI2015-2016

1

8/20/2019 MeRAM

2/26

CERTIFICATE

This is to certify that   Alaknanda Agarwal   has completed the Seminarreport on the topic “Magnetoelectric Random Access Memory - MeRAM” satisfactorily in partial fulfillment for the B.Tech Degree in  Electronics AndCommunication under the guidance of   Prof. Kiran Dange  during the year2015 as prescribed by Usha Mittal Institute of Technology, SNDT Women’sUniversity, Santacruz(w), Mumbai.

Guide Head Of DepartmentProf. Kiran Dange Santoshi Pote

PrincipalDr. Sanjay Pawar

Examiner 1 Examiner 2

2

8/20/2019 MeRAM

3/26

ACKNOWLEDGEMENT

I take great pleasure in expressing my gratitude to all those who have contributedto and motivated me during my project work. A report of this magnitude wouldnot have been come into reality without the able guidance, support and wishes of all those who stood by us in its development. I wish to give my special thanks toProf. Kiran Dange , for her timely advice and guidance. Hence, with deep senseof gratitude, I would like to acknowledge the inspiring guidance of our teacher. Iacknowledge the department for the availability of lab facilities and library.

DATE: Alaknanda Agarwal

3

8/20/2019 MeRAM

4/26

Abstract

A computer consists of two key components: a central processing unit to performcalculations and logical operations, and a memory bank to store instructions anddata. This is the basic recipe.No existing form of memory is good at everything. So to move instructions anddata as fast as possible, engineers have had to compromise. Todays computers use

a smorgasbord of different memory technologies, exploiting the best parts of each.This is far from ideal. One of the biggest energy drains comes from shuttling dataaround the CPU and all the levels of memory that surround it. Also, our fastestmemories lose their data if power isnt continuously supplied to them. And thememories that are most compactthose that can store large numbers of bits in asmall areaare slow, which is the main reason it takes so long for our gadgets andcomputers to wake up from sleep.Those of us who work on alternative memories have long sought a way past theselimitations. Weve dreamed of a memory that must be fast, to minimize the delaysassociated with reading and writing data, and it must consume little power every

time it is used. We should be able to manufacture it on the same chip as a CPU,allowing us to put it close to the computational action, and be able to make itdense enough to compete with existing memories on cost. At the same time, wewould like it to be nonvolatileableto retain data without having to continuouslydraw powerso it can be shut down when it isnt needed.A new type of memory being developed is one of the most promising candi-dates: aform of magnetic memory called magnetoelectric random access memory(MeRAM).

8/20/2019 MeRAM

5/26

Contents

Abstract   1

1 Introduction and Brief History   4

1.1 Random access memory (RAM)   . . . . . . . . . . . . . . . . . . . . 51.1.1 Static RAM (SRAM)  . . . . . . . . . . . . . . . . . . . . . . 51.1.2 Dynamic random-access memory(DRAM)   . . . . . . . . . . 51.1.3 Ferroelectric RAM(FRAM)   . . . . . . . . . . . . . . . . . . 61.1.4 Resistive random-access memory(RRAM)   . . . . . . . . . . 61.1.5 Magnetoresistive RAM(MRAM) . . . . . . . . . . . . . . . . 6

1.2 Read Only Memory (ROM)   . . . . . . . . . . . . . . . . . . . . . . 61.2.1 Masked ROM   . . . . . . . . . . . . . . . . . . . . . . . . . . 81.2.2 Programmable Read Only Memory(PROM)   . . . . . . . . . 81.2.3 Erasable Programmable ROM (EPROM) . . . . . . . . . . . 8

1.2.4 Electrically Erasable PROM (EEPROM)   . . . . . . . . . . . 81.2.5 Flash memory  . . . . . . . . . . . . . . . . . . . . . . . . . . 8

2 Working   92.1 Magnetoresistive RAM (MRAM)   . . . . . . . . . . . . . . . . . . . 9

2.1.1 Working principle  . . . . . . . . . . . . . . . . . . . . . . . . 102.1.2 Internal Structure of MRAM   . . . . . . . . . . . . . . . . . 102.1.3 Working of Classical MRAM . . . . . . . . . . . . . . . . . . 122.1.4 Working of Spin Transfer Torque magnetic RAM (STT-RAM)   132.1.5 Disadvantages of MRAM  . . . . . . . . . . . . . . . . . . . . 14

2.2 Magnetoelectric Random Access Memory (MeRAM) . . . . . . . . . 14

3 Applications and future scope of MeRAM   18

4 Comparitive Study   194.1 MRAM vs. MeRAM   . . . . . . . . . . . . . . . . . . . . . . . . . . 204.2 Advantages of MeRAM over other memories   . . . . . . . . . . . . . 21

1

8/20/2019 MeRAM

6/26

5 References   22

2

8/20/2019 MeRAM

7/26

List of Figures

1.1 Internal Structure of ROM - an example.   . . . . . . . . . . . . . . . 7

2.1 Magnetic Junction Tunnel.   . . . . . . . . . . . . . . . . . . . . . . . 102.2 Internal structure of MRAM.   . . . . . . . . . . . . . . . . . . . . . 11

2.3 Write operation in MRAM.   . . . . . . . . . . . . . . . . . . . . . . 132.4 MeRAM. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

4.1 Comparision of MRAM with previous permissions. . . . . . . . . . . 194.2 A Schematic of MeRAM   . . . . . . . . . . . . . . . . . . . . . . . . 20

3

8/20/2019 MeRAM

8/26

Chapter 1

Introduction and Brief History

A   ll storage devices are characterized with the following features:

•   Speed

•   Volatility

•  Access method

•   Portability

•  Cost and capacity

A bit is the smallest unit of data in a computer and can either store a 0 or 1.Bytesare typically a sequence of eight bits put together to create a single computer al-phabetical or numerical character.More often referred to in larger multiples, bytesmay appear as Kilobytes (1,024 bytes), Megabytes (1,048,576 bytes), GigaBytes(1,073,741,824).Bytes are used to quantify the amount of data digitally stored (ondisks, tapes) or transmitted (over the internet), and are also used to measure thememory and document size.The Term Computer Memory is defined as one or more sets of chips that storeData/program instructions, either temporarily or permanently. It is the criticalprocessing component in any computer. The PCs use several different types.They

are :1. Main Memory / Primary Memory units - The RAM and ROM are examples of main memory; this is the memory the microprocessor uses in executing and storingprograms. It should be fast enough to keep up with the execution speed of themicroprocessor, and thus it should be random access memory.2. Secondary Memory/Auxiliary Memory - Also termed as auxiliary or backupstorage, it is typically used as a supplement to main storage. It is much cheaper

4

8/20/2019 MeRAM

9/26

than the main storage and stores large amount of data and instructions perma-nently. Hardware devices like magnetic tapes and disks fall under this category.Computers memory can be classified into two types RAM and ROM.

1.1 Random access memory (RAM)

is the best known form of computer memory. RAM is considered ”random ac-cess” because you can access any memory cell directly. It allows data items tobe accessed (read or written) in almost the same amount of time irrespective of the physical location of data inside the memory. Random access memory is the

temporary memory of a computer. RAM is normally associated with volatile typesof memory (such as DRAM memory modules), where stored information is lost if power is removed, although many efforts have been made to develop non-volatileRAM chips.the different types of RAM are:

1.1.1 Static RAM (SRAM)

SRAM is a random access memory (RAM) is made up of flip flops, and it storesthe bit as a voltage. Each memory cell requires six transistors, thus thie memorychip has low density but high speed.This memory is more expensive and consumes

more power than DRAM described below. SRAM is used for a computer’scachememory in high speed processors(Intel 486 and Pentium),and as part of therandomaccess memory digital-to-analog converteron a video card. In addition, high-speedcache is also included external to the processor to improve the system performance.

1.1.2 Dynamic random-access memory(DRAM)

DRAM is a type ofrandom-access memorythat is made up of MOS transistor gates.It stores eachbitof data in a separatecapacitor within anintegrated circuit as a

charge. The capacitor can be either charged or discharged; these two states aretaken to represent the two values of a bit, conventionally called 0 and 1. Sinceeven ”nonconducting” transistors always leak a small amount, the capacitors willslowly discharge, and the information eventually fades unless the capacitor chargeisrefreshedperiodically. Because of this refresh requirement, it is adynamicmem-ory. The advantage of DRAM is its structural simplicity: only onetransistoranda capacitor are required per bit, compared to four or six transistors in SRAM.

5

8/20/2019 MeRAM

10/26

This allows DRAM to reach very highdensities. As this form of memory is lessexpensive to produce than static RAM, it is the predominant form of computermemory used in modern computers. Also, its power consumption is lower than

SRAM.

1.1.3 Ferroelectric RAM(FRAM)

FRAM is arandom-access memorysimilar in construction toDRAMbut uses afer-roelectriclayer instead of adielectriclayer to achieve non-volatility. It is a memorytechnology that combines the best of Flash and SRAM. It is non-volatile likeFlash, but offers fast and low power writes. FRAM stores data as a polarization of a ferroelectric material (Lead-Zirkonate-Titanate). As an electric field is applied,

dipoles shift in a crystalline structure to store information.

1.1.4 Resistive random-access memory(RRAM)

RRAM is a type ofnon-volatile(NV)random-access(RAM) computer memory thatworks by changing the resistance across adielectricsolid-state material. Althoughcommonly anticipated as a replacement technology forflash memory, the cost ben-efit and performance benefit of RRAM have not been obvious enough to mostcompanies to proceed with the replacement.

1.1.5 Magnetoresistive RAM(MRAM)

MRAM is a form of non-volatile RAM memory technology that uses magneticcharges to store data instead of electric charges.In the early 1990s the companyHoneywell conceived a new class of Magnetoresistence memory devices which of-fered promise for high density, random access and nonvolatile memory.

1.2 Read Only Memory (ROM)The ROM is a non volatile memory; it retains stored information even if thepower is turned off. This memory is used for programs and data that need notbe altered. As the name suggests, the information can be read only, which meansthat once a bit pattern is stored, it is permanent or at least semi-permanent. Thepermanent group includes two types of memory - masked ROM and PROM. The

6

8/20/2019 MeRAM

11/26

Figure 1.1: Internal Structure of ROM - an example.

semi-permanent group includes two types - EPROM and EEPROM.The concept underlying the ROM can be explained with the diodes arranged ina matrix format, as shown in the figure. The horizontal lines are connected tothe vertical lines only through the diodes and not where they appear to cross inthe diagram. Each of the horizontal lines can be viewed as a register with binaryaddresses ranging from 000 to 111; information is stored by the diodes in theregisters as 0s or 1s. The presence of a diode stores 1, and its absence stores 0.When a register is selected, the voltage of that line goes high, and the output lines,where the diodes are connected, go high. For example, when the memory register111 is selected, the data byte 0111 1000 (78H) can be read at the data lines D7-D0.

The diode representation is a simplified version of the actual MOSFET memorycell. The manufacturer of the ROM designs the MOSFET matrix according to theinformation to be stored; therefore, information is permanently recorded in theROM.

7

8/20/2019 MeRAM

12/26

1.2.1 Masked ROM

In this ROM, a bit pattern is permanently recorded by masking and metalization

process. Memory manufacturers are generally equipped to do this process. It is anexpensive and specialized process, but economical for large production quantities.

1.2.2 Programmable Read Only Memory(PROM)

This memory has nichrome or polysilicon wires arranged in a matrix; these wirescan be functionally viewed as diodes or fuses. This memory can be programmedby the user using a special PROM programmer that selectively burns the fusesaccording to the bit pattern to be stored. The process is known as burning thePROM, and the information stored is permanent.

1.2.3 Erasable Programmable ROM (EPROM)

This memory stores a bit by charging the floating gate of an FET. Information isstored by using an EPROM programmer, which applies high voltages to charge thegate. All the information can be erased by exposing the chip to ultraviolet lightthrough its quartz window, and the chip can be reprogrammed. The disadvantagesof EPROM are - it must be taken out of the circuit to erase it, the entire chipmust be erased, and the erasing process takes 15-20 minutes.

1.2.4 Electrically Erasable PROM (EEPROM)It is functionally similar to EPROM, except that information can be altered byusing electrical signals at the register level rather than erasing all the information.This has an advantage in field and remote control applications. In microprocessorsystems, software update is a common occurence. If EEPROMs are to be used inthe systems, they can be updated from a central computer by using a remote link.However, this memory is expensive as compared to EPROM or flash memory.

1.2.5 Flash memory

This is a major variation of EEPROM that is becoming popular. The majordifference between the flash memory and EEPROM is in the erasure procedure:the EEPROM can be erased at a register level, but the flash memory must beerased either in its entirety or at the sector level. These memory chips can beerased and programmed at least a million times. This memory is ideally suited forlow power systems.

8

8/20/2019 MeRAM

13/26

Chapter 2

Working

W   eve seen earlier the different types of memories before MRAM, their advan-tages and disadvantages. SRAM is the fastest memory, but costly and volatile.DRAM is also volatile and also slower than SRAM. Other non volatile memorieslike flash, are very dense and hence cheap, but very slow, even slower than DRAM.It also requires large voltages to operate. These shortcomings have led memoryresearchers to consider alternative, nonvolatile memories that use spin. Spin is abasic quantum-mechanical property of subatomic particles, such as electrons. In-magnetic materials such as iron, cobalt, and nickel, it is the spins of the electronsthat give those metals their overall magnetic propertiesthat is, their north andsouth poles. It is what gives the material its magnetization.MeRAM, which stands for magnetoelectric RAM, is an improvement on one ad-vanced RAM technology known as MRAM, or Magnetoresistive RAM. Thus, wewill first understand the working of MRAM memory and then see what changeshave been made in the working of MeRAM from that of MRAM.

2.1 Magnetoresistive RAM (MRAM)

MRAM is a memory (RAM) technology that uses electron spin to store information(based on Spintronics). MRAM has been called ”the ideal memory”, potentiallycombining the density of DRAM with the speed of SRAM and non-volatility of 

FLASH memory or hard disk, and all this while consuming a very low amount of power. MRAM can resist high radiation, and can operate in extreme temperatureconditions, very suited for military and space applications.

9

8/20/2019 MeRAM

14/26

2.1.1 Working principle

MRAM works on the principle Magnetoresistance. The change in resistance with

the magnetic state of the device is an effect known as Magnetoresistance. Theresistance with which the magnetic state is associated is measured on the basis of the magnetic moment of the inherent layers of the MRAM.

2.1.2 Internal Structure of MRAM

MRAM is constituted of various storage elements called Magnetic Tunnel Junc-tions (MTJ) integrated with CMOS processing.

Magnetic Tunnel Junctions - A magnetic tunnel junction (MTJ) can beconsidered as a spintronic device since it is composed of two ferromagneticmaterials, such as nickel, cobalt or iron, separated by an ultrathin layerof dielectric insulator. It exhibits two resistances, low or high dependingon the relative direction of ferromagnet magnetizations, parallel (P) or an-tiparallel (AP), respectively. The insulating layer is so thin that electronscan tunnel through the barrier if a bias voltage is applied between thetwo metal electrodes. In MTJs the tunneling current depends on the rel-ative orientation of magnetizations of the two ferromagnetic layers, whichcan be changed by an applied magnetic field. This phenomenon is calledtunnel magnetoresistance (TMR). Thus, Each MTJ is composed of twolayers (ferromagnetic plates), fixed and free separated by a thin dielectric

material. In the fixed layer the magnetic polarity of the electrons in it is

Figure 2.1: Magnetic Junction Tunnel.

fixed. Whereas, in the free Layer, magnetic polarity is subject to changein accordance with the magnetic field which is the resultant of the appliedcurrent.

10

8/20/2019 MeRAM

15/26

Each MRAM memory element is connected to a transistor that performsthe function of read addressing for an array of memory elements. To writea memory state, i.e. to reverse the magnetic moment direction in the stor-

age layer, in a two dimensional array of memory elements, an x-y grid of conducting wires are laid over and under the memory elements, as shownin Figure 3. For writing a selected element, the two corresponding con-ducting wires are activated with current pulses, generating a magnetic fieldalong the long-axis (x-component of the field) and a magnetic field alongthe short-axis (y-component) simultaneously.

Figure 2.2: Internal structure of MRAM.

11

8/20/2019 MeRAM

16/26

2.1.3 Working of Classical MRAM

In the junction, one of the magnetic layers is pinned, which means the direction of 

its magnetization is fixed to serve as a reference. The other magnetic layer, whichis referred to as the free layer, is where information is stored. The free layersmagnetization can be switched so that its either oriented in the same directionas the pinned layer or 180degrees in the other direction.The MTJ device has alow resistance when the magnetic moment of the free layer is parallel to the fixedlayer and a high resistance when the free layer moment is oriented anti-parallelto the fixed layer moment. The orientation of this free layer affects how readilycurrent can quantum mechanically tunnel across the device, through the insulatingbarrier. So the value of the resistance of the device indicates the orientation of the free-layer magnetization, and thus whether the bit is 0 or 1. The data isstored as a magnetic state rather than a charge, and sensed by measuring theresistance without disturbing the magnetic state. The magnetic polarization doesnot leak away with time like charge does, so the information is stored even whenthe power is turned off.Switching the magnetic polarization between the two statesdoes not involve actual movement of electrons or atoms, and thus no known wear-out mechanism exists.The data is stored by switching the polarity of one of the ferromagnetic plates inthe spin valve to represent either a 0 or a 1 in conventional computer logic. Thecurrent pulses are passed through a digit line and a bit line, writing only the bitat the cross point of those two lines.

12

8/20/2019 MeRAM

17/26

Figure 2.3: Write operation in MRAM.

The data may then be read by utilizing the fact that the electrical resistance

changes due to the orientation of the fields on the two plates of the MRAM bit.The target bits isolation transistor is turned on to bias the MTJ, and the resultingcurrent is compared to a reference to determine if the resistance state is low orhigh using the sense line.Smaller transistors have to contend with higher leakage currents.Leakage currenttranslates into static power consumption, the power that a chip consumes evenwhen it is sitting and doing nothing. This power limits greatly the miniaturizationand improvement of electronic devices. Beyond electrical charge, devices based onspintronics attract a broad attention and show performance advantages in manyaspects.

2.1.4 Working of Spin Transfer Torque magnetic RAM(STT-RAM)

•  Read OperationSTT-MRAM, reading is carried out by measuring the electrical resistance of the cell. A particular cell is selected by powering an associated transistor

13

8/20/2019 MeRAM

18/26

that switches current from a supply line through the cell to ground. Due tothe magnetic tunnel effect, the electrical resistance of the cell changes dueto the orientation of the fields in the two plates. By measuring the resulting

current, the resistance inside any particular cell can be determined, andfrom this the polarity of the writable plate. If the two plates have the samepolarity, it is typically considered meaning a 0, while if the two plates are of opposite polarity denotes a 1. Since the read operation is non-destructive, itinvolves the sensing of the cell and no write-back. Therefore, it only takes acouple of nanoseconds, and makes STTMRAM a promising working memory.

•   Write Operation STT-MRAM writes data to the cells using spin-aligned(polarized) electrons to directly torque the magnetic state. To rotate thedirection of the free layer and write data, certain level of current needs to be

applied to the STT-MRAM cell. If write current is higher than this criticalcurrent, free layer will rotate, and write operation is completed. In practice,the STT-MRAM write operation requires 2.6x more operating time than itsread operation, which is one of the critical points to be addressed while re-placing DRAM. This writing process is what makes the current STTMRAMcell so large for most part.

2.1.5 Disadvantages of MRAM

1. Power Efficiency - Writing in to MRAM requires substantial current. Chang-ing data in magnetic fields requires a large power.

2. Size - Half select Problem: Induced field overlaps adjacent cells over a smallarea.

3. Manufacturing - As chips get smaller individual circuits hold less charge. Ithas been found experimentally that the resistance of the magnetic devicevaries exponentially with the thickness.

4. Cost - It is comparatively higher than the other memories.

2.2 Magnetoelectric Random Access Memory (MeRAM)Weve dreamed of creating a single, universal memory that could do everything welland could therefore replace the many kinds of memory we have now. It must befast, to minimize the delays associated with reading and writing data, and it mustconsume little power every time it is used. We should be able to manufacture it onthe same chip as a CPU, allowing us to put it close to the computational action,

14

8/20/2019 MeRAM

19/26

and be able to make it dense enough to compete with existing memories on cost. Atthe same time, we would like it to be nonvolatileableto retain data without havingto continuously draw powerso it can be shut down when it isnt needed. One of the

most promising candidates is aform of magnetic memory called magnetoelectricrandom access memory (MeRAM). With MeRAM technology, its possible to createlogic that is its own form of memoryaswitch that can both perform computationsand remember. This switch would retain its state even when its powered down.Such nonvolatile logic may one day spur a full overhaul of the microprocessor,allowing us to build chips that can very quickly shut off parts that arent being usedin order to save energy, freeze their state if they lose power, and remember exactlywhat they were last doing the instant you turn them back on. In our lab werefond of calling this new form of instant-on electronics Instantonics, and we think itcould dramatically enhance the speed and battery life of computers, tablets, and

smartphones. It could also provide a massive boost to some of the most memory-intensive computing tasks, such as video and multimedia signal processing, patternrecognition, virtual reality, and machine learning.

Figure 2.4: MeRAM.

Both spin transfer torque MRAM (STT-MRAM) and magnetoelectric RAM(MeRAM) can use the same basic architecture to store data in theorientation of electron spin. Each bit in an array can be accessed at the intersection of two linesof electrodesa source line and a bit line. Athird electrodethe word lineis used tocontrol voltage supplied to the bit. A single bit of information can be stored inthe free layer of each magnetic tunnel junction. In STT-MRAM, current flows

15

8/20/2019 MeRAM

20/26

directly through the junction in order to write the bit. MeRAM, which boasts athicker insulating layer, does not permit current to flow as readily and instead usesvoltage-associated effects to change the state of the bit. In each case, the magneti-

zation of the free layer can be flipped [inset, right]. When the magnetization of thefree layer and that of the fixed layer (which serves as a reference) point in the samedirection, resistance is relatively low. It is higher when the two magnetizationspoint in opposite directions. Note that a MeRAM transistor is much smaller thanan STT-MRAM transistor, since it does not have to provide as large a current. Asa result MeRAM cells are smaller overall, and arrays of them can be made moredense.STT-MRAM does not allow much room for improvement in energy efficiency. Thereason is a fundamentalone: Thedevice is essentially a wire. As current is driventhrough it, energy is lost to heat. And as with a wire, the narrower the device

gets, the higher its resistance. Also, each memory cell in STT-MRAM needs atransistor to drive the write current through the device. Because the transistorneeds to provide relatively large currents, it cant easily be shrunk. So althoughSTT-MRAM is picking up steam, its memory cells will have to stay fairly largeper-haps three to five times the size of DRAM cells.It turns out, though, that many of the limitations of MRAM and STT-MRAMcan be avoided by designing a device that uses voltage instead of current to switchmagnetization. About 10 years ago, theoristsbegan exploringwhat would happenif you tried to use a metallic magnetic material to make a very thin electric-field-controlled structureabout a nanometer, orless than 10 atoms, thick. This thickness

is still greater than the distance an electric field can penetrate into a metal beforebeing largely scrambled. But its thin enough that small changes at the surface of the material can have a big impact on the films overall properties, including thenatural inclination of the spins to align in a particular direction.The magnetoelectric RAM, or MeRAM, is one of these thin-film spintronic mem-ories. In many ways its similar to STT-MRAM. For one thing, it can be madewith the same cobalt-iron alloys. The main change is in the device structure andengineering of its interfaces. The layer that acts as the bit is very thin. Andthe insulating layer that in STT-MRAM would pass a current is made thicker, sothat very little current can flow. This change in structure essentially turns themagnetic tunnel junction into a capacitiveinstead of a resistivedevice. When avoltage is applied across this capacitor, the resulting electric field alters the mag-netic properties of the devices free layer. This will change how strongly the bitsspin is locked into an up or down orientation. The spin direction will begin toswing, and if the voltage pulse is cut off at the right time, the bits spin will endup pointing 180degrees opposite its initial orientation. A small magnetic field orcurrent can also be used to nudge the spin into the desired final state.

16

8/20/2019 MeRAM

21/26

The benefits of this arrangement are impressive. Because MeRAM does not re-quire a large current to switch, its transistor can be made much smaller than inSTT-MRAM, resulting in much denser arrays. Data can be written in less than a

nanosecond, using just a tenth of the energy needed to write to STT-MRAM. Asthe devices get smaller and materials improve, we expect to be able to further cutdown on the energy by as much as a factor of 100. Taking full advantage of thesetechnologies will require a willingness to move beyond the strange and somewhatinelegant mix of logic and memory weve relied on for decades. But once we getover the shock of pushing aside the status quo, we will find we can accomplishgreat things.

”The ability to switch nanoscale magnets using voltages is an exciting and fast-growing area of research in magnetism,” Pedram Khalili, one of MeRAMs develop-

ers, said. ”This work presents new insights into questions such as how to controlthe switching direction using voltage pulses, how to ensure that devices will workwithout needing external magnetic fields, and how to integrate them into high-density memory arrays. ”Once developed into a product,” he added, ”MeRAM’sadvantage over competing technologies will not be limited to its lower power dis-sipation, but equally importantly, it may allow for extremely dense MRAM. Thiscan open up new application areas where low cost and high capacity are the mainconstraints.”

Said Juan G. Alzate, another developer: ”The recent announcement of the first

commercial chips for STT-RAM also opens the door for MeRAM, since our devicesshare a very similar set of materials and fabrication processes, maintaining com-patibility with the current logic circuit technology of STT-RAM while alleviatingthe constraints on power and density.”

17

8/20/2019 MeRAM

22/26

Chapter 3

Applications and future scope of MeRAM

1. MeRAM has great potential to be used in future memory chips for almostall electronic applications, including smart-phones, tablets, computers andmicroprocessors, as well as for data storage, like the solid-state disks used incomputers and large data centers.

2. Chips that can run at extraordinarily low power could be very useful indevices that are physically difficult to access and therefore cant be easilyrecharged. Examples include medical implant chips and sensors in hard-to-reach places, such as at high elevations, in space, below ground, underwater,

or in environments otherwise dangerous to humans. With vast amounts of low-power, on-chip memory, these devices could also be substantially betterat particular memory-intensive computing tasks, such as pattern recogni-tion and machine learning. Defense and space applications will benefit frominstant-on systems, which will mean power failure, radiation, and other dis-ruptions will be far less likely to result in a loss of critical information.

3. It will open up new application areas where low cost and high capacity arethe main constraints.

18

8/20/2019 MeRAM

23/26

Chapter 4

Comparitive Study

Figure 4.1: Comparision of MRAM with previous permissions.

19

8/20/2019 MeRAM

24/26

4.1 MRAM vs. MeRAM

1. MRAM today uses spin transfer torque (STT) technology, which depends

on the magnetic property of electrons referred to as spin in addition totheir charge. STT utilizes an electric current to move electrons to write datainto the memory. Yet while STT is superior in many respects to competingmemory technologies, its electric currentbased write mechanism still requiresa certain amount of power, which means that it generates heat when datais written into it. MeRAM, replaces STT’s electric current with voltage towrite data into the memory. This eliminates the need to move large numbersof electrons through wires and instead uses voltage the difference in electricalpotential to switch the magnetic bits and write information into the memory.This has resulted in computer memory that generates much less heat, making

it 10 to 1,000 times more energy-efficient.

2. MRAM does not have great data density, meaning that less information canbe stored on a chip of it. its memory capacity is limited by how close to eachother bits of data can be physically placed, a process which itself is limitedby the currents required to write information. MeRAM fixes this problem byusing a different method to write information to chips, which makes MeRAMfive times more efficient when it comes to data density.

Figure 4.2: A Schematic of MeRAM

3. The low bit capacity(density) of MRAM translates into a relatively large costper bit, limiting STT’s range of applications. Since the MeRAM is more thanfive-times as dense as MRAM, with more bits of information stored in thesame physical area, it also brings down the cost per bit.When it comes to RAM, there are many types of RAM which have seen thelight of day - SRAM, DRAM, DDRAM. But as various as these types may

20

8/20/2019 MeRAM

25/26

seem, they all have at least one thing in common: volatility.almost all RAMloses its memory as soon as it loses power. In fact, this is part of the reasonwhy hard drives exist RAM is not suitable for long term storage, because it

always has to be connected to a power source.

4.2 Advantages of MeRAM over other memories

1. MeRAM and its predecessor are exceptions to the principle of volatility.MeRAM is actually a type of Non-Volatile RAM. This means that thechips are capable of holding memory even while the system is powered down.Thus, when a computer system with non-volatile memory experienced powerloss, unsaved data wouldn’t necessarily disappear like it would in regular

computers. Using MeRAM, a computer could potentially boot back up intoits original state, all information left intact.

2. Also,it won’t be difficult to get MeRAM on the market, since it is con-structed similarly to MRAM. Therefore, manufacturers already producingMRAM could potentially start creating the new standard without entirelynew facilities.

21

8/20/2019 MeRAM

26/26

Chapter 5

References

1. http://vr-zone.com/articles/researchers-claim-non-volatile-meram-is-better-and-more-power-efficient-than-current-memory/18441.html

2. http://study.com/academy/lesson/what-is-random-access-memory-ram-definition-history-quiz.html

3. http://dce.kar.nic.in/new

4. http://newsroom.ucla.edu/releases/ucla-engineers-have-developed-241538

5. Spintronics based Computing by Springer edited by Weisheng Zhao, Guil-laume Prenat

6. http://courses.daiict.ac.in/pluginfile.php/5827/mod resource/content/0/presentation/pres N

7. http://www.ti.com/sitesearch/docs/universalsearch.tsp?searchTerm=mram

8. http://www.mram-info.com/introduction

9. http://www.nve.com/MRAM-video.php

10. http://physics.usask.ca/ chang/homepage/Spintronics/Spintronics.html

11. http://spectrum.ieee.org/semiconductors/memory/spin-memory-shows-its-might

12. http://www.cs.utah.edu/thememoryforum/jin.pdf 

13. http://www.nve-spintronics.com/mram-operation.php

22