May_EDFA_Digital

edfas.org ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 21 NO. 2 4 EDFAAO (2019) 2:4-7 1537-0755/$19.00 ©ASM International ® TIVA MEASUREMENTS WITH VISIBLE AND 1064-nm LASERS Paiboon Tangyunyong and Andrea Rodarte Sandia National Laboratories, Albuquerque, New Mexico ptangyu@sandia.gov INTRODUCTION Thermally-induced voltage alteration (TIVA) [1] and light-induced voltage alteration (LIVA) [2] are widely used laser-based failure analysis techniques for localizing defects in integrated circuits (ICs). Conventional practice is to use lasers with wavelengths (e.g., 1340 nm) larger than the band gap wavelength of silicon (~1120 nm) to create localized heating in TIVA measurements. For LIVA, either 1064-nm or visible (e.g., 532-nm) lasers are used; thewavelengths of these lasers are smaller than the band gap wavelength of silicon, enabling creation of electron- hole pairs and generation of photocurrent at the electri- cal junctions. This articledemonstrates that 1064-nmor visible lasers can also be used to generate localized heating for TIVA measurements. Rationales are provided and examples shown for using visible and 1064-nm lasers in TIVA mea- surements. In some cases, visible and 1064-nm lasers may be preferable to a 1340-nm laser due to enhanced TIVA signals and better spatial resolution. COMPARISION OF TIVA AND LIVA SIGNALS FOR VARIOUS LASER WAVELENGTHS Figure 1 shows a comparison of TIVA and LIVA signals obtained with 1340-nm (Fig. 1a), 1064-nm (Fig. 1b), and 532-nm (Fig. 1c) lasers on a failed SRAM device (Fig. 1d). For the 1340-nm laser, the local- ized heating of the laser generates discrete TIVA signals (bright spots) in the memory arrays. Three of the strongest TIVA signals are denoted by the yellow arrows in Fig. 1a. In the case of the 1064-nm laser, both localized heating and photocurrent genera- tion occur, resulting in both TIVA and LIVA signals. The TIVA signals (bright spots) generated fromthe 1064-nm laser are very similar to those generatedwith the 1340-nm laser. The LIVA effects from the 1064-nm laser create a dark band (denoted by a red solid arrow) within the memory arrays and several smaller dark bands near the input/output (I/O) pins of the device; one of the LIVA signals near the I/O pin is denoted by a red dashed arrow. For the 532-nm laser, LIVA signals are the dominant signals, as evidenced by a strong dark band within the memory arrays (denoted by a red solid arrow) and several smaller dark bands near the I/O pins (red dashed arrow) of the device. The LIVA signals generated fromthe 532-nm laser are very similar to those generatedwith the 1064-nm laser. In contrast, theTIVA signals generatedby the 532-nm laser are much weaker with only three faint bright spots (yellow arrows) observed in Fig. 1c. The main takeaway from Fig. 1 is that both TIVA and LIVA signals are generated fromboth 1064-nmand532-nm Fig. 1 Comparison of TIVA images of a failed SRAMdevice from three different lasers: (a) 1340-nm (top left), (b) 1064-nm (bottom left), and (c) 532-nm laser (top right). The reflected-light image of the device is shown on the bottom right (d). (a) (b) (c) (d)