The experiment, featuring the small red glow of a BEC trapped in infrared laser beams (credit: Stuart Hay, ANU)

Australian physicists have used an online optimization process based on machine learning to produce effective Bose-Einstein condensates (BECs) in a fraction of the time it would normally take the researchers.

A BEC is a state of matter of a dilute gas of atoms trapped in a laser beam and cooled to temperatures just above absolute zero. BECs are extremely sensitive to external disturbances, which makes them ideal for research into quantum phenomena or for making very precise measurements such as tiny changes in the Earth’s magnetic field or gravity.

The experiment, developed by physicists from ANU, University of Adelaide and UNSW ADFA, demonstrated that “machine-learning online optimization” can discover optimized condensation methods “with less experiments than a competing optimization method and provide insight into which parameters are important in achieving condensation,” the physicists explain in an open-access paper in the Nature group journal Scientific Reports.

Faster, cheaper than a physicist

Optical dipole trap used in the experiment, showing the three laser beams and the condensate (red-yellow oval in blue square) (credit: P. B. Wigley et al./Scientific Reports)

The team cooled the gas to around 5 microkelvin. To further cool down the trapped gas (containing about 40 million rubidium atoms) to on the order of nanokelvin*, they then handed control of the three laser beams** over to the machine-learning program.

The physicists were surprised by the clever methods the system came up with to create a BEC, like changing one laser’s power up and down, and compensating with another laser.

“I didn’t expect the machine could learn to do the experiment itself, from scratch, in under an hour,” said co-lead researcher Paul Wigley from ANU Research School of Physics and Engineering. “A simple computer program would have taken longer than the age of the universe to run through all the combinations and work this out.”

Wigley suggested that one could make a working device to measure gravity that you could take in the back of a car, and the AI would automatically recalibrate and fix itself.

“It’s cheaper than taking a physicist everywhere with you,” he said.

* Billionth of a degree above absolute zero — where a phase transition occurs, and a macroscopic number of atoms start to occupy the same quantum state, called Bose-Einstein condensation.

** The 1064 nm beam is controlled by varying the current to the laser, while the 1090 nm beam is controlled using the current and a waveplate rotation stage combined with a polarizing beamsplitter to provide additional power attenuation while maintaining mode stability.

Abstract of Fast machine-learning online optimization of ultra-cold-atom experiments

We apply an online optimization process based on machine learning to the production of Bose-Einstein condensates (BEC). BEC is typically created with an exponential evaporation ramp that is optimal for ergodic dynamics with two-body s-wave interactions and no other loss rates, but likely sub-optimal for real experiments. Through repeated machine-controlled scientific experimentation and observations our ‘learner’ discovers an optimal evaporation ramp for BEC production. In contrast to previous work, our learner uses a Gaussian process to develop a statistical model of the relationship between the parameters it controls and the quality of the BEC produced. We demonstrate that the Gaussian process machine learner is able to discover a ramp that produces high quality BECs in 10 times fewer iterations than a previously used online optimization technique. Furthermore, we show the internal model developed can be used to determine which parameters are essential in BEC creation and which are unimportant, providing insight into the optimization process of the system.

Inhibition of  media septum GABA neurons during rapid eye movement (REM) sleep reduces theta rhythm (a characteristic of REM sleep). Schematic of the in vivo recording configuration: an optic fiber delivered orange laser light to the media septum part of the brain, allowing for optogenetic inhibition of media septum GABA neurons while recording the local field potential signal from electrodes implanted in hippocampus area CA1. (credit: Richard Boyce et al./Science)

A study published in the journal Science by researchers at the Douglas Mental Health University Institute at McGill University and the University of Bern provides the first evidence that rapid eye movement (REM) sleep — the phase where dreams appear — is directly involved in memory formation (at least in mice).

“We already knew that newly acquired information is stored into different types of memories, spatial or emotional, before being consolidated or integrated,” says Sylvain Williams, a researcher and professor of psychiatry at McGill*. “How the brain performs this process has remained unclear until now. We were able to prove for the first time that REM sleep (dreaming) is indeed critical for normal spatial memory formation in mice,” said Williams.

Dream quest

Hundreds of previous studies have tried unsuccessfully to isolate neural activity during REM sleep using traditional experimental methods. In this new study, the researchers instead used optogenetics, which enables scientists to precisely target a population of neurons and control its activity by light.

“We chose to target [GABA neurons in the media septum] that regulate the activity of the hippocampus, a structure that is critical for memory formation during wakefulness and is known as the ‘GPS system’ of the brain,” Williams says.

To test the long-term spatial memory of mice, the scientists trained the rodents to spot a new object placed in a controlled environment where two objects of similar shape and volume stand. Spontaneously, mice spend more time exploring a novel object than a familiar one, showing their use of learning and recall.

Shining orange laser light on media septum (MS) GABA neurons during REM sleep reduces frequency and power (purple section) of neuron signals in dorsal CA1 area of hippocampus (credit: Richard Boyce et al./Science)

When these mice were in REM sleep, however, the researchers used light pulses to turn off their memory-associated neurons to determine if it affects their memory consolidation. The next day, the same rodents did not succeed the spatial memory task learned on the previous day. Compared to the control group, their memory seemed erased, or at least impaired.

“Silencing the same neurons for similar durations outside of REM episodes had no effect on memory. This indicates that neuronal activity specifically during REM sleep is required for normal memory consolidation,” says the study’s lead author, Richard Boyce, a PhD student.

Implications for brain disease

REM sleep is understood to be a critical component of sleep in all mammals, including humans. Poor sleep quality is increasingly associated with the onset of various brain disorders such as Alzheimer’s and Parkinson’s disease.

In particular, REM sleep is often significantly perturbed in Alzheimer’s diseases (AD), and results from this study suggest that disruption of REM sleep may contribute directly to memory impairments observed in AD, the researchers say.

This work was partly funded by the Canadian Institutes of Health Research (CIHR), the Natural Science and Engineering Research Council of Canada (NSERC), a postdoctoral fellowship from Fonds de la recherche en Santé du Québec (FRSQ) and an Alexander Graham Bell Canada Graduate scholarship (NSERC).

* Williams’ team is also part of the CIUSSS de l’Ouest-de-l’Île-de-Montréal research network. Williams co-authored the study with Antoine Adamantidis, a researcher at the University of Bern’s Department of Clinical Research and at the Sleep Wake Epilepsy Center of the Bern University Hospital.

Abstract of Causal evidence for the role of REM sleep theta rhythm in contextual memory consolidation

Rapid eye movement sleep (REMS) has been linked with spatial and emotional memory consolidation. However, establishing direct causality between neural activity during REMS and memory consolidation has proven difficult because of the transient nature of REMS and significant caveats associated with REMS deprivation techniques. In mice, we optogenetically silenced medial septum γ-aminobutyric acid–releasing (MS^GABA) neurons, allowing for temporally precise attenuation of the memory-associated theta rhythm during REMS without disturbing sleeping behavior. REMS-specific optogenetic silencing of MS^GABA neurons selectively during a REMS critical window after learning erased subsequent novel object place recognition and impaired fear-conditioned contextual memory. Silencing MS^GABA neurons for similar durations outside REMS episodes had no effect on memory. These results demonstrate that MS^GABA neuronal activity specifically during REMS is required for normal memory consolidation.

For the first time, scientists at IBM Research have demonstrated reliably storing 3 bits of data per cell using a relatively new memory technology known as phase-change memory (PCM). In this photo, the experimental multi-bit PCM chip used by IBM scientists is connected to a standard integrated circuit board. The chip consists of a 2 × 2 Mcell array with a 4- bank interleaved architecture. The memory array size is 2 × 1000 μm × 800 μm. The PCM cells are based on doped-chalcogenide alloy and were integrated into the prototype chip, serving as a characterization vehicle in 90 nm CMOS baseline technology. (credit: IBM Research)

Scientists at IBM Research have demonstrated — for the first time (today, May 17), at the IEEE International Memory Workshop in Paris — reliably storing 3 bits of data per cell in a 64k-cell array in a memory chip*, using a relatively new memory technology known as phase-change memory (PCM). Previously, scientists at IBM and elsewhere successfully demonstrated the ability to store only 1 bit per cell in PCM.

The current memory landscape includes DRAM, hard disk drives, and flash. But in the last several years, PCM has attracted the industry’s attention as a potential universal memory technology based on its combination of read/write speed, endurance, non-volatility and density. For example, PCM doesn’t lose data when powered off, unlike DRAM, and the technology can endure at least 10 million write cycles, compared to an average flash USB stick, which tops out at 3,000 write cycles.

IBM suggests this research breakthrough provides fast and easy storage to capture the exponential growth of data from mobile devices and the Internet of Things.

Scientists have long been searching for a universal, non-volatile memory technology with performance far superior to flash — today’s most ubiquitous non-volatile memory technology. The benefits of such a memory technology would allow computers and servers to boot instantaneously and would significantly enhance the overall performance of IT systems. A promising contender is PCM, which can write and retrieve data 100 times faster than Flash and enable high storage capacities, and like flash, not lose data when the power is turned off. Unlike flash, PCM is also very durable and can endure at least 10 million write cycles, compared to current enterprise-class flash at 30,000 cycles or consumer-class flash at 3,000 cycles. While 3,000 cycles will outlive many consumer devices, 30,000 cycles are orders of magnitude too low to be suitable for enterprise applications. (credit: IBM Research)

IBM scientists envision standalone PCM as well as hybrid applications that combine PCM and flash storage, with PCM as an extremely fast cache. For example, a mobile phone’s operating system could be stored in PCM, enabling the phone to launch in a few seconds. In the enterprise space, entire databases could be stored in PCM for blazing fast query processing in time-critical online applications, such as financial transactions.

Machine learning algorithms using large datasets will also see a speed boost by reducing the latency overhead when reading the data between iterations.

How PCM Works: answering the grand challenge of combining properties of DRAM and flash

PCM materials exhibit two stable states: the amorphous (without a clearly defined structure) and crystalline (with structure) phases (low and high electrical conductivity, respectively).

To store a “0″ or a “1″ bit on a PCM cell, a high or medium electrical current is applied to the material. A “0″ can be programmed to be written in the amorphous phase and a “1″ in the crystalline phase, or vice versa. Then to read the bit back, a low voltage is applied. This is how re-writable (but slower) Blu-ray discs** store videos.

Phase-change memory (PCM) is one of the most promising candidates for next-generation non-volatile memory technology. The cross-sectional tunneling electron microscopy (TEM) image of a mushroom-type PCM cell is shown in this photo. The cell consists of a layer of phase-change material, such as germanium antimony telluride (GST), sandwiched between a bottom and a top electrode. In this architecture, the bottom electrode has a radius (denoted as rE ) of approx. 15 nm and is fabricated by sub-lithographic means. The top electrode has a radius in excess of 100 nm and the thickness of the phase change layer is approx. 100 nm. A transistor or a diode is typically employed as the access device. (credit: IBM Research — Zurich)

“Phase change memory is the first instantiation of a universal memory with properties of both DRAM and flash, thus answering one of the grand challenges of our industry,” said Haris Pozidis, PhD., an author of the workshop paper and the manager of non-volatile memory research at IBM Research–Zurich. “Reaching 3 bits per cell is a significant milestone because at this density, the cost of PCM will be significantly less than DRAM and closer to flash.”

To achieve multi-bit storage, IBM scientists have developed two innovative enabling technologies: a set of drift-immune cell-state metrics and drift-tolerant coding and detection schemes***. “Combined, these advancements address the key challenges of multi-bit PCM, including drift, variability, temperature sensitivity, and endurance cycling,” said IBM FellowEvangelos Eleftheriou, PhD.

IBM scientists have also demonstrated, for the first time, phase-change memory attached to POWER8-based servers.

* At elevated temperatures and after 1 million endurance cycles.

*** “Blu-ray Disc” is owned by the Blu-ray Disc Association (BDA)

*** More specifically, the new cell-state metrics measure a physical property of the PCM cell that remains stable over time, and are thus insensitive to drift, which affects the stability of the cell’s electrical conductivity with time. To provide additional robustness of the stored data in a cell over ambient temperature fluctuations, a novel coding and detection scheme is employed. This scheme adaptively modifies the level thresholds that are used to detect the cell’s stored data so that they follow variations due to temperature change. As a result, the cell state can be read reliably over long time periods after the memory is programmed, thus offering non-volatility.

The experimental multi-bit PCM chip used by IBM scientists is connected to a standard integrated circuit board. The chip consists of a 2 × 2 Mcell array with a four-bank interleaved architecture. The memory array size is 2 × 1000 μm × 800 μm. The PCM cells are based on doped-chalcogenide alloy and were integrated into the prototype chip serving as a characterization vehicle in 90 nm CMOS baseline technology.

Abstract of Multilevel-Cell Phase Change Memory: A Viable Technology

In order for any non-volatile memory (NVM) to be considered a viable technology, its reliability should be verified at the array level. In particular, properties such as high endurance and at least moderate data retention are considered essential. Phase-change memory (PCM) is one such NVM technology that possesses highly desirable features and has reached an advanced level of maturity through intensive research and development in the past decade. Multilevel-cell (MLC) capability, i.e., storage of two bits per cell or more, is not only desirable as it reduces the effective cost per storage capacity, but a necessary feature for the competitiveness of PCM against the incumbent technologies, namely DRAM and Flash memory. MLC storage in PCM, however, is seriously challenged by phenomena such as cell variability, intrinsic noise, and resistance drift. We present a collection of advanced circuit-level solutions to the above challenges, and demonstrate the viability of MLC PCM at the array level. Notably, we demonstrate reliable storage and moderate data retention of 2 bits/cell PCM, on a 64 k cell array, at elevated temperatures and after 1 million SET/RESET endurance cycles. Under similar operating conditions, we also show feasibility of 3 bits/cell PCM, for the first time ever.