The soon-to-be-ubiquitous radio frequency identity (RFID) tag consists of a Si chip (less than 1 mm(2)) attached to an antenna on a plastic sheet. Each tag needs to have a unique, electronically readable identity number. Currently, the preferred direction of industry appears to be to fabricate all chips the same, and then program the unique identity into the chip electrically in a permanent or semi-permanent fashion. This may make the tags reusable and permit the identity number to be written by the customer. On the other hand, writing the identity into the chips permanently, during fabrication has advantages; mainly, low cost, less real estate, and less vulnerability to tampering or accidental loss of identity. Permanent programming of the chip is currently done using selective laser cutting of metal connections on the chip, but has not been widely adopted because of cost and reliability. We propose an alternate method whereby transistors are substituted for the metal links and selectively focused-ion-beam (FIB) implanted to permanently turn some of them off. Moreover, we have demonstrated that a low ion dose (similar to 10(12)/cm(2)) is sufficient, so that the throughput is calculated to be practical. For post-fabrication implantation, the region above the transistor channel must be thinned to 100 to 200 nm to use reasonable ion energies. (Of course, in-process implantation is also an option before the passivation layer is applied.) We have used n-type metal oxide semiconductor (NMOS) and p-type metal oxide semiconductor (PMOS) test transistors fabricated at Berkeley and etched away the oxide and part of the poly over the gate. This process did not alter the transistor characteristics. The transistors were implanted with a series of doses and energies using both FIB and broad-beam implantation (Ar, As, and B with broad-beam and Ga, As, and B with FIB). The required dose to turn transistors off is of order 10(12)/cm(2) as long as the ions have sufficient energy to alter the gate oxide and/or the channel. In PMOS transistors there is a strong shift in threshold voltage as well as a decrease in drain current, while in NMOS the main effect is the latter. Since Ga+ ions work, a robust FIB system can be used to implant the RFID tags. Moreover, neither high resolution nor high placement accuracy are needed. We only need to implant a given transistor without implanting its neighbors. Assuming a 200 mm wafer with 30 000 chips, 120 transistors per chip, 2 mu m(2) gate area, and a 100 pA beam current results in an implant time of 6 min/wafer. Using a write-on-the-fly scheme the wafer would have to be scanned at 6 cm/s. Since 50 mu s are needed to implant each transistor, deflection and blanking times should not be limitations. (c) 2005 American Vacuum Society.