Development of LGAD Sensor Technology on 12 inch Wafers for 3D Integration with advanced CMOS processes

In the world of particle physics, seeing is not just about capturing an image; it is about capturing a moment with impossible precision. When subatomic particles zip through a detector, they leave behind fleeting traces that vanish in a fraction of a second. To understand the universe, scientists need sensors that can record not only where a particle passed but exactly when it arrived, down to a trillionth of a second. For decades, these sensors and the electronic brains that read them have been built as separate pieces, then glued together. This separation creates a bottleneck, adding bulk and slowing down the signal. The new frontier is to fuse these two components into a single, three-dimensional stack, allowing the sensor to talk directly to the processor with almost no delay. This requires a marriage of two very different technologies: the delicate, high-speed sensors needed to catch particles, and the advanced computer chips used to process data, both manufactured on massive twelve-inch silicon wafers.

A team of researchers from Fermilab, SLAC National Accelerator Laboratory, and Lawrence Livermore National Laboratory has taken a significant step toward this goal. They successfully designed, built, and began testing a new type of ultra-fast sensor called a Low Gain Avalanche Diode, or LGAD, using a standard computer chip manufacturing process. These sensors were created on twelve-inch wafers using a sixty-five nanometer technology, a scale typically reserved for high-performance electronics rather than particle detectors. The team's work proves that it is possible to manufacture these specialized sensors alongside the complex circuits needed to read them, paving the way for a future where entire layers of sensors and processors are bonded together like sheets of glass. This integration promises to create detectors that are thinner, faster, and capable of tracking particles with unprecedented accuracy.

The project, funded by the U.S. Department of Energy, aimed to demonstrate that these advanced sensors could be made ready for a final assembly step known as hybrid bonding. In this process, a full wafer of sensors would be glued directly onto a full wafer of readout chips, creating a seamless three-dimensional structure. To prepare for this, the researchers fabricated three distinct versions of the LGAD sensor on the same silicon wafer. The first was a standard design that uses a specific arrangement of electrical charges to create a region of high energy where particles can be detected. The second version, known as an AC-coupled sensor, separates the signal from the electrode using a thin layer of insulation, which allows for better sharing of information between neighboring pixels. The third and most novel version is the Deep Junction sensor, which places the sensitive detection layer deep inside the silicon rather than near the surface. This deep placement avoids electrical problems at the edges of the pixels, allowing the sensors to be packed much closer together without losing performance.

The manufacturing process was a delicate balancing act. The team had to adapt a standard computer chip factory flow, which is designed for logic circuits, to create the specific electrical conditions needed for particle detection. They were allowed to tweak the amount of energy and the quantity of atoms injected into the silicon to tune the sensors, but the basic steps of the factory process remained unchanged. One challenge was creating the electrical connections for the AC-coupled sensors. The standard layers of insulation in the chip factory were too thick for the usual method, so the team ingeniously used the space between two different metal layers within the chip to create the necessary electrical storage. This required careful adjustment to ensure the sensors could function correctly within the strict rules of the chip design.

To predict how these sensors would behave before they were even built, the team ran detailed computer simulations. These virtual models helped them understand how the electrical fields would move through the silicon and how the sensors would respond to different voltages. The simulations showed that the sensors would need specific voltage levels to operate correctly and that the thin layer of silicon used in this process made the sensors quite sensitive to changes in that voltage. The computer models also helped the team design the protective rings around the sensors, ensuring they would not break down under high electrical stress. These simulations were crucial for planning the different versions of the sensors, allowing the researchers to test various combinations of electrical properties without wasting time and materials.

Once the wafers arrived at the laboratories, the testing phase began. The researchers first checked the basic electrical health of the sensors in the dark and under light from a red LED. They compared sensors with the special gain layer to those without it to measure how much the signal was amplified. The results showed that the sensors were working as expected, with the AC-coupled versions showing a range of amplification at voltages between 116 and 160 volts. The computer simulations had predicted these values with remarkable accuracy, usually within ten volts of the actual measurements. The team also found that the standard sensors required slightly higher voltages to operate than the AC-coupled ones, a difference caused by the specific electrical setup of the AC design.

To test how fast the sensors could react, the team fired a laser and a beam of radioactive particles at the devices. The laser pulses were extremely short, allowing the team to measure the exact moment the sensor reacted. The results were impressive. For a small sensor, the time resolution reached 40 picoseconds, which is 40 trillionths of a second. When they tested an even smaller sensor, the resolution improved to 23 picoseconds. These numbers confirm that the sensors can capture the timing of particle events with the extreme precision required for future experiments. The Deep Junction sensors, while showing promise with good timing and amplification, proved to be a bit more temperamental. They exhibited a slow drift in their electrical properties over time, making detailed measurements more difficult, but they still demonstrated the ability to detect particles effectively.

The work does not end with these initial tests. The researchers are currently bonding the sensor wafers to the readout chips to see how the two work together in a fully integrated package. They are also planning to test the sensors under intense radiation to ensure they can survive the harsh environments of future particle colliders. A new batch of wafers with slightly different electrical properties is already on the way to refine the design further. The success of this project marks a turning point in detector technology. By proving that these high-speed sensors can be made using standard computer chip processes and integrated into three-dimensional stacks, the team has opened the door to a new generation of detectors. These future devices will be capable of tracking particles in five dimensions—position, time, and angle—providing a level of detail that could reveal new secrets about the fundamental building blocks of our universe.