RSS icon
Twitter icon
Facebook icon
Vimeo icon
YouTube icon

Experimental SQUID Microscope Gets to Gigahertz

Gus Vlahacos mounts a microchip on the SQUID microscope’s sample stage.

Gus Vlahacos mounts a microchip on the SQUID microscope’s sample stage.

Suppose you needed to find the location of an electrical fault in a microchip running at a couple billion cycles per second. How would you do it? Answer: You could drill very small holes into the chip so that tiny voltage sensitive probes could reach in and measure what was happening at different places -- a slow and destructive process.

But now JQI Fellow Fred Wellstood and grad student Constantine “Gus” Vlahacos have devised a way to do the job without drilling holes or contacting the chip. They use a specially configured SQUID microscope inspired in part by their research in quantum computing. If the technique is fully successful, it could also be used to measure magnetic and electrical properties of newly discovered materials of urgent interest to science and technology, including carbon nanotubes and graphene, a one-atom-thick sheet of carbon atoms in a chicken-wire pattern.

At the heart of their microscope is a SQUID (superconducting quantum interference device), an electrical device that relies on a weird but useful property of superconductors: If two superconductors are placed together and separated by a thin layer of insulator, a current can flow between the two superconductors without any voltage -- up to a maximum “critical current” that depends very sensitively on the ambient magnetic field.

So the presence of an applied magnetic field, such as those produced when currents run through pathways in a microchip, changes the critical current, which can be measured by finding the maximum current that can flow through the SQUID without a voltage appearing.

This behavior, and the fact that the devices are operated at just a few degrees above absolute zero temperature, make SQUIDs exquisitely sensitive detectors of, for example, the extremely small signals produced by nerve cells in the human brain. As a result, SQUID microscopes, which scan across a sample to make images of small variations in magnetic field have been deployed for failure analysis in integrated circuits. But existing commercial systems use an external feedback loop to monitor the SQUID’s response, and that limits the response time -- and hence the sampling speed or “bandwidth.”

The Wellstood-Vlahacos design differs in two main respects. First, it does not rely on a feedback loop. Second, it is coupled to a system that shoots timed 400-picosecond (a thousandth of a billionth of a second) pulses of current to the SQUID. If there is no magnetic field emanating from the sample, the pulse is not quite large enough to cause the SQUID to generate a voltage. But if a field is present (perhaps caused by a short circuit in the chip being studied), a voltage will be triggered: The interaction of pulse and field causes the current to rise above the critical current and a voltage “on” signal appears across the SQUID.

“The SQUID will switch to the on-state only if the current pulse exceeds the critical current of the SQUID,” Wellstood says. “Otherwise it will stay at zero volts. Since the critical current depends very sensitively on the magnetic field applied to the SQUID, this gives us information about the magnetic field in a specific very short time interval.”

And, of course, the device is designed to measure over very small space intervals as well: The microscope uses a niobium SQUID that is only about 30 micrometers wide. The SQUID is mounted in a vacuum chamber and cooled to around 4.2 Kelvin, while the sample is in air at room temperature.

The SQUID and sample are separated by a very thin sapphire window that keeps air and heat way from the SQUID while allowing the SQUID to see the magnetic flux from the sample below.

“Of course, all we learn from one measurement is whether the critical current is larger or smaller than the current pulse we sent down. To pin down the precise value of the critical current, we need to make repeated measurements with different pulse sizes,” Wellstood explains.

“Since this requires repeated sampling, our technique only works with repetitive signals that allow us to repeatedly access the same’ part of the signal.” Repetitive is putting it mildly: Vlahacos recently completed a scan on a 10 mm-by-10 mm area. It took 11.5 hours.

But the microscope’s resolution and response speed are striking. The images at below show changes in the magnetic field of a pulse propagating down a 200-micrometer wide microstrip test line at a frequency of 2 gigahertz, which is typical of the clock speed of many current processor chips. The time interval between images is 1/20th of a nanosecond (billionth of a second).