Phasechange random access memory: A scalable technology

1
Phase-change random access memory A scalable
technology

• S. Raoux, G.W. Burr, M.J. Breitwisch, C.T.
  Rettner, Y.-C. Chen, R.M. Shelby, M. Salinga, D.
  Krebs, S.-H. Chen, H.-L. Lung, C.H. Lam
• Presentation by
• Ioannis Papadopoulos
• ipapadop_at_cse.tamu.edu

2
Phase-Change Memory Overview

• Phase Change Memory (PCRAM) is based on
  phase-change materials
• At least two phases with completely different
  properties
• In the case of PCRAM, property is electrical
  conductivity differs at least 5 orders of
  magnitude between the two states
• Crystallized phase (SET) logical 1 (low
  resistance)
• Amorphous phase (RESET) logical 0 (high
  resistance)
• Change from SET to RESET short, powerful current
  pulse
• Change from RESET to SET longer, less powerful
  pulse
• Read short, weak pulse

3
PCRAM Current-Voltage Curve

• Threshold switching
• Higher conductivity after the Threshold voltage
  not completely explained
• Probably due to impact ionization, carrier
  recombination
• The reason that PCRAM is feasible and interesting
  as technology

4
WARNING!

• The rest of the paper is more of interest to
• Physists
• Chemical Engineers
• Electrical Engineers with focus on manufacturing
  and analog circuit analysis
• So I'm going to present the same content focused
  instead for
• Computer Scientists and
• Computer Engineers
• You can find the original paper on
• http//portal.acm.org/citation.cfm?id1462871.1462
  885

5
Support for PCRAM from manufacturers

• Increased system functionality means
• More megabytes
• Smaller size
• Higher speeds
• Most are reluctant to invest until sure that
• The technology works
• It scales for more than one generation
• Transition costs and is risky
• Will only be done if results are promising and
  future looks bright

6
The Scaling Problem

• It needs to be shown that the technology scales
• Scalability studies to
• Forsee roadblocks
• Estimate the far future
• What does the PCRAM cell serve? How small can it
  become?
• Critical characteristics
• SET/RESET resistance distributions
• Ability to switch between states
• Ability to read without petubation to the data
• Life cycle
• Data retention time
• Fast SET speed

7
Possible problems with PCRAM

• Voltages may end up too close (for set, reset,
  read)
• Problems with doping of material when going in
  smaller scale
• Can large currents be provided by the (current
  and future) device interfaces?
• Are they realistic or practical?
• Is PCRAM economically viable?
• SET dictates speed, RESET consumption
• Can the device be fast and energy-efficient?

8
Simulation importance

• Reducing the current for RESET depends on a lot
  of factors
• Predictive numerical simulations
• Hybrid simulations that combine results from
  different types of simulations and experiments
• Slow-thin film crystallization experiments
• Optical-pulse experiments
• Goal
• Heat more efficiently the phase change material
• Reduce the material

9
Phase change materials

• The ultimate phase-change material
• Low melting point (but not too low)
• High resistance in crystallized form (but not too
  high)
• Fast in crystallizing (but not too fast)
• Be small (but not inexistant)
• Mainly discovered by chemists
• Too many parameters but we care about two
• A material that can be used with CMOS technology
• The device to work within common operating
  temperatures

10
PCRAM evolution

• Size and dimension of materials
• Thin films (1 dimension)
• Nanowires (2 dimensions)
• Nanoparticles (3 dimensions)
• Sorry no 4th dimension but who knows...
• Crystalization time
• From 10µs down to 100ns
• Phase change material is reduced
• Lower voltages, lower currents, less power, lower
  temperatures
• Lower temperatures increase data retention time
• Different alloys have different characteristics
• Still an open research area (for chemists)

11
Proof of Scaling

• Phase change bridge
• A simple research device
• Two electrodes, a bridge of material
• Waveform generator to create PCRAM-like pulses
• Used to test new materials
• Easy to construct for rapid prototyping
• Experiments to estimate real PCRAM behaviour
• Integration into a product is another issue
• There are proposed methods though for lithography
  to support it

12
Paper Results

• It is not a research presentation paper
• It is a paper trying to prove through various
  research attempts that
• PCRAM can scale
• Is highly intergratable
• Can provide products
• The authors attempt to convince manufacturers and
  investors that
• PCRAM research can lead to competitive products
• PCRAM is as good, if not better, compared to
  Flash-Memory

13
Conclusion

• Complete analysis of the scaling behaviour
• Phase change materials can be scaled down
  considerably
• High integration can be achieved
• "Thus, in many aspects, PCRAM technology appears
  to be readily scalable to several future
  technology nodes"
• In other words, PCRAM can replace flash-memory
• Please pay attention and fund PCRAM research!!!

14
Thank you!!

• Questions??