Plasma experts devise computer programs with potential to reduce microchip costs and bolster manufacturing

Crafted from silicon, the same material abundant in sand and adorned with intricate designs, microchips serve as the backbone of modern technology, empowering smartphones, enhancing household appliances, and facilitating the operation of vehicles. At the U.S. Department of Energy’s (DOE) Princeton Plasma Physics Laboratory (PPPL), scientists are spearheading the development of computer simulation codes poised to surpass existing techniques, thus aiding in the production of microchips utilizing plasma, the electrically charged state of matter also harnessed in fusion research.

These innovative codes hold the promise of enhancing manufacturing efficiency and potentially sparking a revival of the chip industry in the United States.

“Given the indispensable role of microchip-equipped devices in our daily lives, the manufacturing processes and locations of these devices are crucial matters of national security,” remarked Igor Kaganovich, a principal research physicist who heads the low-temperature modeling group at PPPL.

“By furnishing robust and accurate simulation tools capable of predicting plasma behavior and expediting the manufacturing and design cycle of silicon chips, we can potentially propel the U.S. into a leadership position in this realm and sustain it for years to come.”

Accelerating Progress

One significant research endeavor at PPPL revolves around reducing the computational time required for computers to simulate microchip plasma reactors. This advancement would enable private industries to deploy more sophisticated and precise simulations widely, thereby contributing to their efforts to lower microchip production costs.

“Companies aspire to leverage simulations to enhance their processes, but the computational demands are typically formidable,” noted Andrew Tasman Powis, co-author of the study published in Physics of Plasmas and computational research associate at PPPL. “We’re striving to counteract this trend.”

Traditionally, physicists seek simulations that faithfully reproduce plasma behavior, generating virtual representations that unveil the complexities of plasma dynamics with exquisite precision. Achieving this entails employing algorithms—programs governed by specific rules—that simulate plasma across short time intervals and within small spatial volumes.

However, such detailed simulations often entail prolonged computing times, with computations spanning days or weeks—a duration and expense untenable for companies seeking to optimize their microchip manufacturing processes.

To address this challenge, researchers revisited the annals of plasma physics, uncovering algorithms developed in the 1980s that showed promise in expediting the simulation of microchip plasma systems. When tested, these algorithms demonstrated the ability to model such systems in significantly less time with only a marginal reduction in accuracy.

In essence, researchers found that they could attain reliable simulations despite modeling plasma particles across larger spatial domains and employing longer time intervals.

“This development holds significance as it could save companies both time and money,” emphasized Haomin Sun, the lead researcher of the study and a former graduate student in Princeton University’s Program in Plasma Physics, based at PPPL.

“This implies that, with equivalent computational resources, more simulations can be generated. Increased simulations not only facilitate process enhancement in manufacturing but also deepen our understanding of physics in general. With our limited resources, we can make more discoveries.”

Related research spearheaded by Powis lends further credence to this prospect. In a study published in Physics of Plasmas, Powis validated that computer codes can generate precise models of plasma particles while utilizing virtual “cells” or spatial volumes that surpass a conventional measure in plasma physics known as the Debye length.

This advancement implies that the codes can effectively employ fewer cells, thereby reducing the computational burden of simulations and enhancing performance.

“This development is encouraging as reducing the number of cells could mitigate the computational cost of the simulation and hence enhance performance,” Powis affirmed.

The algorithms can simulate “capacitively coupled plasma reactors,” devices instrumental in generating the plasma employed by engineers to etch intricate patterns on silicon wafers. These microscopically thin pathways constitute the microcircuitry essential for microchip functionality.

“Our interest lies in modeling this process to gain insights into controlling plasma properties, forecasting their behavior in novel systems, and projecting etching properties to enhance the process,” Powis elaborated.

The team intends to further validate these algorithms by incorporating the effects of diverse wall and electrode materials.

“We aim to bolster confidence in these algorithms, ensuring the accuracy of results,” Powis emphasized.

Recognizing and Overcoming Inherent Limitations

Another research endeavor focuses on rectifying errors that may arise in plasma simulations due to the inherent limitations of simulation methods themselves, which often model fewer plasma particles than are present in real plasma.

“In an ideal scenario, simulating plasma would entail tracking each individual particle’s trajectory in real-time,” explained Sierra Jubin, a graduate student in the Princeton Program in Plasma Physics and lead author of a study featured in Physics of Plasmas. “However, given finite computing resources, achieving this level of detail is unfeasible.”

To circumvent this challenge, researchers devise code to represent millions of particles as a single composite particle, simplifying the computational load but also amplifying the interactions of these virtual conglomerates. Consequently, changes in the distribution of particle velocities occur more rapidly than they would in natural plasma, a phenomenon known as thermalization. Essentially, the simulation deviates from reality.

“This discrepancy poses a challenge because failing to address it means our simulations fail to accurately reflect phenomena as they unfold in the real world,” Jubin elaborated.