Assessing MIM Capacitor Defect Levels for 6 inch HBT Production Line Lance Rushing, James Zhang and Pete Zampardi* Skyworks Solutions, Inc., 2427 Hillcrest Drive, Newbury Park, California 91320 Phone: 805-480-4600, email: lance.rushing@skyworksinc.com *Formerly of Skyworks Solutions, Inc Abstract An initial assessment of a test methodology to determine MIM defect rates was investigated. Prior to this methodology the device level reliability of the MIM structure was explored. This exploration included several TDDB type studies. Ultimately, the limitations of these TDDB studies because of low sample size led to a production level capacitor voltage stress test. An initial study was performed on approximately 400,000 capacitors. I.Introduction In consumer electronics, the reduction of failure rates to 100PPM and below is an industry standard. For device level reliability, this poses a special problem as the samples needed are quite large. While it is straight forward to generate a cumulative failures plot and extract to low cumulative failures, it is difficult to prove out these low failures. By using a simple binomial expansion, the quantities needed for 100PPM are in the tens of thousands of devices. This would also be difficult to cumulate using years of product reliability. In this study, a production approach to device reliability testing is studied. Prior to this, the reliability basics of the MIM capacitor are established. These basics include the characterization of intrinsic and extrinsic reliability. This particular piece of the study is customer driven as they often believe that defects must be extrinsic failures. Whether using Poisson statistics or a thickness model, extrinsic failures are effectively a simple Weibull transformation of cumulative failures. Therefore, the extrinsic failures would have the same shape and characteristic lifetime parameters but would simply have shorter lifetimes. Defects are rarely so well behaved. II.Experimental Skyworks has a long history of performing device reliability. The MIM capacitor is no exception. The reliability of these capacitors has been explored in many ways. Historically, for device reliability, the end-of-life metrics are the most important and FIT metrics are established using product reliability tests (more appropriate for infant and random failure rates). Recently, with the drive towards lower PPM requirements, capacitor defect rates have become of interest. To understand the defect rates, we performed a variety of industry standard device level tests including size studies and ramped vs static TDDB. These tests were mostly designed to explore the extrinsic reliability, or the deviation towards lower lifetimes. A.End of life Reliability tests are typically performed with a static dc voltage close (but far enough away) to breakdown. The leakage current is monitored until catastrophic failure and the time until failure is recorded. In this way the following cumulative failures plot (Figure 1) can be generated. This plot is a linearized Weibull plot for a sampling of MIM capacitors from the latest HBT process. Figure 1: MIM Cumulative Failures Using the linear field model, this plot yields a lifetime of t0.1%>10,000hrs at use condition. In general and disregarding ill-behaved defects, this extrapolation (to low cumulative failures) should account for any extrinsic reliability concerns. B.Size vs Lifetme To support this statement, we performed a few size and parameter versus lifetime studies. Since lifetime is area dependent due to Poisson distributed defects, these tests should provide illumination as to the extrinsic lifetime.(1) The results of a simple study are shown in Figure 2. Clearly, the larger capacitors have lower lifetime. Observe that the -('(& 6ROLG 6WDWH 7HFKQRORJ\ $VVRFLDWLRQ  3DJH 