Researchers have developed a novel technique for aligning 3D semiconductor chips by using lasers to create a hologram through specially designed metalenses on the chips. This innovation could significantly reduce the production costs of 2D semiconductor chips, facilitate the creation of 3D photonic and electronic chips, and potentially lead to the development of affordable, compact sensors.
A team from the University of Massachusetts Amherst has created a groundbreaking method to align 3D semiconductor chips. This process involves directing a laser through concentric metalenses on the chips to generate a hologram. Their findings, published in Nature Communications, hold promise for reducing the costs associated with manufacturing 2D semiconductor chips, making way for advanced 3D photonic and electronic chips, and could also introduce other inexpensive, compact sensors into the market.
Semiconductor chips are pivotal in enabling electronic devices to manage, store, and transmit information. The performance of these chips depends on the arrangement of their components. However, the existing 2D structure has hit its technological ceiling, and many experts view 3D integration as the key to future advancements.
Creating a 3D chip entails stacking multiple 2D chips while ensuring their layers are aligned to within tens of nanometers (to put it into perspective, one millimeter equals 1 million nanometers). This alignment must occur in three dimensions: forward and backward, side to side, and the spacing between the chips (the x, y, and z axes).
Amir Arbabi, an associate professor of electrical and computer engineering at UMass Amherst and one of the lead authors of the study, explains, “The conventional method for aligning two layers involves using a microscope to find specific markers, like corners or crosshairs, and attempting to match them up.”
However, microscope-based techniques are inadequate for fabricating these 3D chips. Maryam Ghahremani, a doctoral candidate and the paper’s lead author, emphasizes, “A microscope can’t simultaneously focus on both sets of crosshairs because of the considerable gap between the layers, which measures hundreds of microns. The necessary focus adjustments can result in the chips shifting and becoming misaligned.” She adds that the smallest features detectable are constrained by the diffraction limit, which is around 200 nanometers.
The innovative alignment technique conceived by Arbabi and his team operates without moving parts and can detect misalignments between distant layers with greater precision. Initially targeting a precision of 100 nanometers, their method achieved detections down to 0.017 nm in the side-to-side alignment (x and y axes) and 0.134 nm for the spacing between chips (z-axis).
Arbabi notes, “Imagine you have two objects. By analyzing the light passing through them, we can identify if one has shifted by an atomic distance relative to the other,” far exceeding their initial target. The human eye can perceive errors nearly as small as a few nanometers, while computers can detect even tinier discrepancies.
To realize this, the researchers incorporated alignment markers crafted from concentric metalenses on the semiconductor chips. When laser light passes through these markers on both chips, it generates two overlapping holograms. “This interference pattern reveals whether the chips are aligned, the direction of any misalignment, and its magnitude,” explains Ghahremani.
Arbabi remarks that “[Chip alignment] poses a significant, costly challenge for companies that manufacture semiconductor tools.” He believes their method addresses a key obstacle in the production process. Additionally, reducing costs could enhance access to this technology for smaller startups aiming to innovate in the semiconductor field.
Furthermore, Arbabi highlights that this technique can also be utilized to create displacement sensors capable of measuring various physical quantities. “Many physical measurements can be interpreted as displacements, requiring just a simple laser and a camera,” he explains. For example, “a pressure sensor could track the movement of a membrane.” In principle, this method can also monitor any activity involving movement, such as vibrations, heat changes, or acceleration.
