Imprimir Republish


The high relief of memory

Team from Pernambuco clarifies an essential phenomenon in computers

EDUARDO CESARHard disk: rugosity between metallic layers determines the capacity for storing informationEDUARDO CESAR

Computers carry out one of their basic skills – storing information – without one managing to understand exactly how. But there has been an advance. A group of researchers from the Federal University of Pernambuco (UFPE) has explained an essential phenomenon for the new read-write heads from hard disks to function; although these were discovered 40 years ago and have been in use for three, they have survived with scant theoretic grounding.

The team coordinated by Sergio Rezende, from the Physics Department of the UFPE, proved that rugosity is crucially important in the interaction between the metallic layers of the head that reads the disk – it is this interaction that determines, directly, how computers perform. The discovery makes it possible to understand better the phenomenon that makes it feasible to condense information, at a time when all the world over memories with more capacity, more speed and lower consumption of energy are being sought.

Like a top
The tip of the read/write head currently used, called technically magnetoresistant heads, is a set of metallic layers of a few atoms. As in a sandwich, they are intercalated: to a first layer, made of ferromagnetic material (usually, an alloy of iron and nickel or cobalt), is added a second, of antiferromagnetic material (usually, nickel oxide), on so on successively. On the surface and in the inside of each layer, the electrons act like a top: they spin in a clockwise direction, or in the opposite direction.

From this rotating movement of the electron, which the physicists call spin, the magnetic field – of magnetization – is born, which makes it possible to store the information from the computer’s permanent memory, recorded in each point of the tracks of the hard disks, which are also made out of layers of magnetic materials. In each point, magnetization in one direction represents a 0 bit, and in the other a 1 bit – this is binary code, from which any text, design or image is built up. Then the information in the fast memory, called RAM (random access memory), are stored in semiconductors in the form of a positive or negative electric charge, representing 0 or 1.

Limits expanded
For a better understanding of the research of this group from Pernambuco, it is inevitable to turn to a bit of history and to a few more explanations. Until the 90’s, the reading of the hard disk was carried out by a physical process called magnetic induction: the bit magnetization produced an electric current in the read head, later processed by other devices of the computer. It was a limited resource, because the storage area had to be big, to be interpreted by the read head.

In the 90’s, there was a leap forward, with the new read heads, called magnetoresistive, and with them it was possible to reduce the area and so to increase the capacity for storing information on the hard disks. Incidentally, another Brazilian took part in this technological breakthrough: Mário Baibich, a professor at the Institute of Physics of the Federal University of Rio Grande do Sul (UFRGS). In 1989, when he was in France, Baibich discovered giant magnetoresistance, the phenomenon that gave rise to the current hard disks read/write heads and that associates the resistance of the material with the magnetic field created by the information bit on the hard disk.

On corroborating the importance of rugosity, Rezende, in a way, arrived at the results pursued by the theoretical researchers from IBM, the company that developed the first hard disk in 1956 and has not stopped looking for a way of expanding the memory of these devices.

In this time, the knowledge built up about the properties of the ferromagnetic and antiferromagnetic layers has made it possible for the capacity for storage to expand 3 million times, from the original 2 kbits per square inch on the first disk to 20 gigabits per square inch in the most recent version, from 1999. Even so, there were only hypotheses about how the interaction between the layers worked – a phenomenon known as exchange bias, or polarization by interchange – and there continued to be a lack of theoretical or experimental grounding.

“Theoretical calculations by IBM’s Alex Malozemoff showed in 1987 that rugosity could be responsible for the enormous reduction of the exchange bias magnetic field, originating in the interaction of the surface electrons between the ferromagnetic and antiferromagnetic layers”, Rezende comments. “Our experiment and the theoretical model we drew up prove that this really is the case”. The group from Pernambuco presented the experimental measures in November last year, at the Annual Conference of Magnetism, held at Seattle, United States, and published the theoretical studies in two recent articles, one in March this year in the Physical Review B, and the other in April in the Journal of Applied Physics.

Although it was not held to be so important, rugosity between the layers was already known. It is a property that is still unavoidable in the magnetic layers that form the permanent memory in computers, since at the moment it is impossible to produce an absolutely flat surface. What to a simple glance seems to be perfectly smooth, on the atomic scale takes on contours that recall the peaks and valleys of a mountain range. Two layers are like mountain ranges fitting each other perfectly, since the manufacturing process eliminates any empty space between them.

The studies by Rezende’s team detail with precision the way that rugosity – the variation in the contour – interferes with the interaction between the layers. It is because of these peaks and valleys that the electrons from the ferromagnetic layer interact at times with electrons with a spin in one direction (located, shall we say, in the peak of the mountain range), and at times with a spin in the other direction (in the valley) of the opposite layer. The problem is that rugosity causes unwanted phenomena, which physicists call complex behaviors, with their origins in physical disorders, known as frustrations, aggravated by variations in temperature. “These phenomena have been observed for years, and there was speculation about their causes, but not any proof, as there is now”, Rezende comments.

Without losses
Now on the macroscopic plane, rugosity also interferes by reducing the magnetic interaction between the layers and, in the last instance, in the performance of the read/write head. Its effect is not to be belittled: the effective field that the antiferromagnetic film creates over the magnetization of the ferromagnetic film is 100 to 1000 times smaller than the field estimated for a perfectly flat surface. “The challenge now is to eliminate rugosity”, says Rezende. “If, using an as yet unknown method, someone manages to make a double ferromagnetic and antiferromagnetic layer with a perfectly flat interface, the field that holds the magnetization of the ferromagnetic layer will be much larger”. As a consequence, the computer will work better.

Understanding the interaction between the layers may attain still more refined applications. Exchange bias, together with another phenomenon, called magnetic tunneling, is one of the essential concepts in the construction of a magnetic and lasting RAM memory, no longer a volatile one like the one used today in semiconductors. Two years ago, IBM announced a prototype of this new memory, with which something quite simple was intended: a computer switched off abruptly would no longer lose the information on the screen, and when switched on again would resume the work at exactly the same point where it was.