Sam Zeluf, 22, made a 1,200-transistor chip single-handedly in his New Jersey garage. He used homemade equipment.

Zeluf cut silicon wafers, painted microscopic patterns on them using ultraviolet light, and manually dipped them in acid.

The guy built his first chip back in high school in 2018. It had 200 times fewer transistors. Zeluf himself jokes that he has outpaced the entire semiconductor industry and even Moore's law, according to which the number of transistors on a chip doubles approximately every two years.

Zeluf aims to build a chip that will compete with the 1971 Intel 4004, the first commercial microprocessor with 2300 transistors.

According to the enthusiast, society will benefit if the production of microcircuits becomes more accessible to inventors without multi-million dollar budgets.

Zeluf is inspired by YouTube videos posted by inventor and entrepreneur Jerry Ellsworth. A woman made her own thumb-sized transistors. Zeluf decided to replicate Ellsworth's design and take the next step that seemed logical to him: move from single transistors to integrated circuits: a leap that historically took about a decade.

When Zeluf started blogging about his project, some industry experts began to write to him that it was impossible. However, the family supported the guy. His father asked a semiconductor engineer friend for some advice on job safety. Zeluf was assisted by Mark Rothman, who has spent 40 years developing chips and now works for a company that develops technologies for OLED screens.

Zeluf couldn't afford billions of dollars of equipment, so he studied patents and textbooks from the 1960s and 70s, when engineers at pioneering companies like Fairchild Semiconductor made chips on conventional workbenches. “They describe methods using X-Acto blades, tape, and multiple glasses,” Zeluf says. On eBay and other sites, he found cheap equipment from the 1970s and 80s that belonged to now-closed California technology companies. Most of the equipment required repair, including the electron microscope, which cost $250,000 in the early 1990s; Zeluf bought it for only $1000 and repaired it. Now the developer uses a microscope to check his chips for defects, as well as nanostructures.

Sometimes Zeluf had to improvise. Like a real chip factory, he wanted to transfer his microscopically detailed designs to devices using a process called photolithography. This is coating the future chip with a light-sensitive material and using a device like an ultra-precise projector to burn a template that will guide subsequent processing steps. Photolithography machines are expensive—up to $150 million—so Zeluf made his own by screwing a modified conference room projector he bought from Amazon onto a microscope. He projects his designs on a tiny scale onto silicon wafers, which he coats with a UV-sensitive material.

In 2018 Zeluf developed his first chip with six transistors. This took approximately 12 hours. The Z1 used transistors with characteristics as low as 175 microns.

In late 2018, Zeluf began working at Carnegie Mellon University, hacking parts of garage equipment and studying electrical engineering. Although he says he followed security protocols, the university objected to the x-ray machine in his dorm room. During trips home, Zeluf upgraded his installation. It uses a faster switching transistor design based on wafers of crystalline silicon known as polysilicon that became dominant in the 1970s.

Zeluf rotated hand-cut half-inch squares of polysilicon, each becoming a separate chip, on a small homemade turntable at 4,000 rpm to coat them with light-sensitive material. He got a grid of 12 circuits, each containing 100 transistors; a total of 1200 transistors. Zeluf then applied the method of sintering phosphorus atoms to adjust the conductivity. Three more rounds under the photolithographic machine, interrupted by time in a vacuum chamber filled with glowing plasma to etch the polysilicon, completed the job.Today's factories produce chips in a broadly similar way, using a sequence of steps to gradually add and remove material in different parts of the design. These chips are much more complex, with billions of much smaller transistors tightly coupled to each other, and the steps are done by machines rather thand. The transistors on the second-generation Zeluf chips were about 10 times faster than those on his first, and had characteristics as low as 10 microns.

In August, Zeluf tested the Z2 by plugging it into a square, beige semiconductor analyzer made by Hewlett Packard about two decades before the boy was born. A series of slowly rising current-voltage curves on a glowing green screen signaled success. "That curve was amazing to look at," Zeluf says, "the first sign of life after you've been dipping this little piece of crystal in a beaker all day."

Now Zeluf is getting advice from veterans of the 1970s semiconductor industry. He says he's not sure what he wants to do after graduating this spring, but he'll continue to make his own chips.

Ellsworth, whose homemade transistors inspired Zeluf, says it can be helpful to ensure that high-quality chips are handcrafted. “The tools that we have today can make this affordable, and I think for some problems that makes a lot of sense,” she notes. Ellsworth believes that chip technology, which is considered obsolete by leading factories, can still be useful for engineers.

Zeluf recently upgraded his photolithographic machine to print parts around 0.3 microns, or 300 nanometers, about the size of the commercial chip industry in the mid-90s. Now he's thinking about features he could build into an Intel 4004-scale chip. "I want to take garage silicon further and open people's minds to the possibility that we can do some of these things at home," he says.

bbabo.Net