February_EDFA_Digital
edfas.org ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 20 NO. 1 10 MONOGRAIN DEFECT IN POLYSILICON GATES Thierry Parrassin and Laurent Clément, STMicroelectronics, Crolles, France thierry.parrassin@st.com ; laurent-renaud.clement@st.com EDFAAO (2018) 1:10-18 1537-0755/$19.00 ©ASM International ® INTRODUCTION There is a tricky soft electrical failure that is hard to identify in electrical test validation and, if undetected, could generate fails in products delivered to custom- ers. Such failure occurs more often at high V DD and high frequency, limiting the maximum reachable frequency. It happens only at very low levels—2 or 3 ppm—but impacts critical paths, clock trees, scan chains, ormemory cells, which can lead to zero yields. The tricky failure, exacerbated by transistor scaling, is observable mostly on technology nodes below 40 nm. The failure is caused by a defect that gives very few signatures during failure analysis localization because of its dynamic failure mode with a test sequence that is hard to set up. It does not show abnormality in parametric measurements when probing isolated transistors, because of its transient electrical acti- vation. Finally, it does not showvisible andmorphological defects in physical analysis using classical techniques such as transmission electronmicroscopy (TEM) lamellae, and it requires advanced techniques such as nanobeam diffraction [1] to be observed. However, after many failure analysis cases and using some experimental techniques, the authors are now able to recognize the symptoms of this failure, observe the defect, and limit the impact on electrical timing margins with both design and process corrections. Thedefect has been identifiedas amonograin defect in the polysilicon gate of the transistor. ELECTRICAL BEHAVIOR What is the common electrical failing factor between a scan chain working in scan mode but failing in capture mode and a memory cell failing at high voltage? In the case of themonograin issue, both are related to the clock frequency. Thedefect directly impacts the signals gatedby the clock pulse width, leading to a delay increase. For circuits with cores running at high frequency, it is difficult to check the value of internal logic states at speed directly at the chip interfaces. The I/O at the interfaces of the integrated circuit has too much capacitive load and self-inductance toproperly output high-frequency signals. Therefore, validating the logic states can be done only at low frequency. A scan chain is used to cover this test. It takes a picture of the logic state at a given moment using the same high-frequency clock that is running the core of the circuit; this is the “capture” sequence. The data are memorized into the flip-flop registers of the scan chain, as shown in Fig. 1. The data are serially shifted out of the circuit using a low-frequency clock and can be read by automatic tester equipment; this is the “scan” or “shift” sequence. To summarize, the capture sequence uses an internal clock with a small pulse width, resulting in a capture fail if there is a monograin defect, and the shift sequenceuses anexternal clockwitha largepulsewidth to allowexternal observationwithout any additional failures. A complete scan chain test sequence is alternating scan and capture modes. In the case of a memory cell failure, it is more or less the same approach. The external clock driving a memory block is self-formatted by the memory into a small clock pulse width; this is referred to as self-timing. It allows high-speed operations by shrinking the clock period to the maximum frequency allowed by the memory. It also saves dynamic consumption by reducing the memory bitline time to discharge. Like the scan chain, memories are also sensitive to the monograin defect when it affects signals gated by short clock pulse widths. FAILURE ANALYSIS AND DEFECT LOCALIZATION The failure mode of the monograin defect is very spe- cific. It is visible with transient, dynamic test sequences and only with the clock “at speed.” It is identified as a transition fault by scan diagnosis systems; static failure analysis techniques such as photoemission and static
Made with FlippingBook
RkJQdWJsaXNoZXIy MjA4MTAy