A Case Study in a 100 zyxwvu X Reduction in Sodium Ions in a zyxw 0.8m BiCMOS Process using Triangular Voltage Sweep zyxw Larry Anderson, Suketu Parikh, Samuel Nagalingam, and Chris Haidinyak Silicon Systems, Inc. 2300 Delaware Avenue Santa Cruz, CA 95060 (408)429-4680; FAX: (408)425-4301 email: larry.anderson@scz.ssil .com Abstract-This case study shows how to use Triangular Volt- age Sweep (TVS) to reduce Na and K in the backend of a triple-metal BiCMOS process from 10l2 ions/cm2to 1O1O ions/cm2. TVS is com- pared to Bias Temperature Stress (BTS) techniques. While Capaci- tance-Voltage plots are good monitors for bulk contamination (metal, deposited oxide, etc.), data is presented which shows that TVS is superior for detecting surface-introduced mobile ions (photoresist, solvent strip, etc.). Process integration techniques and issues in the reduction of mobile ions are discussed. When TVS structures are put on product-like wafers, the mobile ions can be measured accurately and repeatably within 10 minutes of completing the process step- no alloy is required! Finally, the use of Ammonium Fluoride solu- tions to reduce the surface mobile ions are discussed. BEFORE: DESCRIPTION OF PROBLEM Positive mobile ionic contamination (PMIC), most frequently sodium ions, were present in a 380 nm field oxide causing unaccept- able BTS instability of AV,= 12V in a new BiCMOS technology which does not use BPSG nor PSG. Standard CV techniques de- scribed below did not catch the problem; nor could analysis of Na levels in the Solvent Strip solution zyxwvutsrqp (<< 100 ppb). BTS measure- ments are slower, requiring additional processing and multiple repli- cate measurements. This case study will explain how we reduced the Na levels quickly and reliably by combining WLR process integra- tion and optimized Triangular Voltage Sweep method (TVS), and why TVS worked while CV plots did not. AFTER: WHERE WE zyxwvut ARE Now We now have a TVS method to measure the mobile ions as fast as 5 minutes after we complete Solvent Resist Strip, with an order of magnitude greater sensitivity than BTS techniques ( lo9 vs 1Olo cm-2). WLR is moving upstream into the fab for quick response. We lowered the PMIC by over two orders of magnitude, from 5.0 x 1OI2to less than 3.0 x 1Olo Na+/cmz ( AV~ < 1V). By use of TVS methods and structures, we were able to optimize the Multilevel Metal process with only minor changes in the Solvent Resist Strip process. See Figure 1 for a sample TVS plot. BICMOS mTAL-0 TO METAL- 1 PROCESS MODULE DESCRIPTION This advanced BiCMOS process (primarily optimized for Bi- polar as opposed to most BiCMOS processes that are optimized for CMOS), has recessed Field Oxide, and no BPSG nor PSG before Metal-1. Therefore, we require less Na than other processes in the backend MLM (Multilevel Metalization). The process sequence from Pt to Metal-1 is shown in Figure 2. Figure 1. Example of TVS Measurement After Hole Mask * Preclean>>PlatinumDep>>Pt RTP SintewPt Strip Metal4 TiW DepmMO AICutARC>>MO M a s k M O AlCu EtchMesist Strip>>MO TiW Wet Etch * Hole Mask (Defines TiW Fuses and TiW Staps)>>Wet Metal Etch>>Solvent Strip Resist Strip * Undoped Plasma Teos Dep>>SOG>>Undoped Plasma Teos Dep Contact MasbXontact Etch>>Contact Resist Strip Etch>>Resist StriwAllov-1 * Metal4 TiW Dep>>Ml AICu+ARC>>Ml M a s k M 1 Figure 2. Process Sequence THE TVS TECHNIQUE AND WLR The TVS technique is well documented [I-31, and its use is quietly increasing at WLR to prevent PMIC yield crashes related to MLM. While BTS with area capacitors works well for bulk oxide and metal mobile charge (e,,,), measurement cannot be done until after additional processing including alloy-this is time consuming. Different ionic species like sodium, potassium, and hydrogen cannot be differentiated. Detection is limited to about 5.OE+10 cmb2. 0-7803~2705~51961$5.00° 1996 IEEE 45