It is not uncommon in wafer fabs to use one ion implanter for sequentially implanting multiple elements. This is effective, however, only if the potential for cross contamination is painstakingly controlled and monitored, particularly with phosphorus. We have validated the advantages of using in-fab TXRF analysis for implanter cross-contamination qualification.
To save cost and space in wafer fabs, ion implanters are often used to implant more than one atomic species. For example, a high-current implanter may be used to implant phosphorous in polysilicon gates, then the same system may be used to implant arsenic in source-drains. In such situations, there is a strong possibility of cross contamination [1–3] because the inside surface of the implant chamber, particularly near the wafers, is bombarded with stray ions. In addition, wafers are placed on a disc that sees the main implanting ion beam. During subsequent implants, previously deposited dopants may be sputtered off and deposit on product wafers as contamination.
Of necessity, prior to introducing a different species in the implanter, the inside of the tool has to be cleaned. Although this leads to more tool downtime, it nonetheless gives a fab the flexibility of doing more production with fewer tools.
Of course, not all cross contamination is harmful to devices, but there are a number of situations where enhanced diffusion due to contaminating ions may occur. Phosphorus-enhanced diffusion is particularly troublesome. So, it makes sense to control and monitor any implanter cleaning process to ensure the level of undesired atoms left in a wafer chamber is minimal.
Cleaning and monitoring
A widely used method for cleaning inside an implanter chamber is to bombard the wafer plane with ions of an inert element, such as argon. It has been shown that this removes most of the surface contamination via sputtering. Assuming sputtered atoms are evenly distributed over the wafer plane, we can analyze the surface of a wafer in place during this treatment to measure contamination in the implanter.
There are several wafer surface analysis techniques that provide trace element monitoring. SIMS has very low detection limits, but analysis is time consuming and tools are in a laboratory, adding delay. VPD-ICPMS is also time-consuming for sample preparation. The wafer is treated with HF vapor that dissolves the native oxide, thus collecting any impurities within this layer of native oxide. Any phosphorous implanted below this oxide layer will not be collected in the analysis, giving an inaccurate phosphorous measurement. One may choose to use test wafers with thick oxide, but this in turn raises the possibility of contamination from the extra process step.
TXRF, a fast measurement technique that does not need any sample preparation, has the required detection limits . Measurement time is ~20 min for a 5-point measurement/wafer and the analytical beam samples enough depth of the wafer to account for all phosphorous in the sampling area. This is also a nondestructive method, and since TXRF tools are inside fabs, test wafers can be reclaimed for further use.
Figure 1. P:Si values for pre- and post-implant wafers.
Measurement reliability and repeatability
Measurement reliability and repeatability are the primary requirements for using an analytical tool in wafer fab production. Therefore, our initial work focused on collecting data to establish baseline levels.
TXRF measures x-ray emissions of elements in silicon test wafers. The energy of the Ka line of phosphorus is 2.015keV and is next to the matrix silicon Ka line of 1.740keV. As a result, the phosphorus Ka peak is influenced by the silicon peak. Depending on the test wafer's history and vendor, the silicon peak varies from one batch to another and makes it hard to monitor on a daily basis. Referencing the work of others, we adopted the ratio of phosphorus and silicon peaks instead of the phosphorus peak alone. This proved to be an adequate choice for our work.
SPC chart for phosphorous
The main variables in TXRF measurements are angle of incidence of the x-ray beam, time, and x-ray intensity. Time can be controlled accurately, but angle and beam intensity may change slightly from run to run. To look at long-term repeatability of measurements, we selected a wafer with low-level phosphorous contamination and then measured three points on the wafer about once per day using two TXRF tools. (The wafer had been used as a standard reference wafer for a half year.)
This data gave us a P:Si ratio of 0.011 over the tested period. Further, our SPC chart of TXRF measurements on this wafer shows a very stable trend, demonstrating the reproducibility of TXRF over a long period of time. Our data also gave us a way to cross-link our two TXRF tools to ensure measurement accuracy.
When we looked at pre- and post-phosphorus implant data (Fig. 1), we saw that clean prime wafers constantly showed a TXRF value of P:Si = 0.004, while the post-phosphorus wafers showed variable results depending on the level of contamination. We put some of the test wafers through an RCA clean and analyzed them again. These wafers also yielded a P:Si close to 0.004, demonstrating that the base-level reading for our starting wafers is P:Si = 0.004 (i.e., the cleaning proved that added phosphorous was sputtered onto the wafer surface rather than implanted). Our result shows this is an effective method to remove surface phosphorus contamination from wafers.
Figure 2. Correlation of phosphorous counts between TXRF and VPD-ICPMS. (The horizontal axis is the recorded dose of the 0.1keV p+ implant.)
TXRF correlated to VPD-ICPMS
To further establish TXRF as a valid technique for phosphorous monitoring at low levels, we decided to correlate our measurements with VPD-ICPMS.
We selected a set of four wafers and implanted them with four different dosages of phosphorous at very low energy, so that the phosphorus was mainly on the wafer surface. We then compared silicon and phosphorus counts via TXRF and VPD-ICPMS (Fig. 2). In the figure, phosphorus implant dose is the dialed-in implant dose, and TXRF P:Si (cps) is the TXRF value in counts-per-second.
It is clear that both TXRF and VPD-ICPMS responded linearly to the phosphorus contamination levels over the tested range, although the detected amount varied. VPD-ICPMS appears to respond to the phosphorus dose accurately at lower doses, but records less when the implanted dose is increased. We believe this is caused by the reduction of actual implant dose by resputtering of the earlier implant since the phosphorus implant has an energy of 0.1keV. As a result, the real accumulated dose on the wafer surface is less than the dialed-in 5¥1013/cm2 dose. TXRF showed the same trend, but the increased amount is only two-fold while the implanted dose is increased ten times.
Figure 3. Radial distribution of P:Si on a post-phosphorus wafer.
We also ran a split lot of wafers with intentional phosphorus contamination; we implanted product wafers with an additional 0.1keV p+ implant, before source-drain extension implants. Wafer electrical tests showed no obvious device parameter shift with a phosphorus dose up to 3¥1013/cm2; thus, 1¥1013/cm2 phosphorus contamination definitely does not have an effect. But we found that with 5¥1013, there are some changes in device performance. So we concluded that if we set the control limit to 1¥1013 of phosphorus contamination (with Ar+ 40keV 1x1015 implant as the sputtering carrier), we would be safe with source-drain extension implants, which are the most sensitive implant operations. Based on the TXRF data of the companion wafers with intentional phosphorus contamination implants, 3¥1013 phosphorus on a wafer's surface gives P:Si = 0.01. We set up the control limit on the P:Si chart to 0.01, since the product did not show sensitivity up to this contamination level.
Monitoring in operation
After finishing phosphorus implants on an ion implanter, the implanter's beamline and end-station sections are cleaned with a high-dose implant of argon that sputters off residual phosphorus. We then place a test wafer on the implant disk along with dummy wafers. These wafers are then implanted with a 3¥1015/cm2 argon dose. The test wafer is then measured with TXRF.
Our control policy prevents the implanter from being used to process device wafers if the P:Si Ka value is >0.01. If this occurs, the implanter is cleaned and tested again, until the value is <0.01.
This whole procedure assumes that argon sputtering knocks off phosphorous atoms lodged on implanters' surfaces during normal implant operation, which is supported by our experimental data. Figure 3 shows the P:Si ratio for four locations on one wafer, revealing that the center has the lowest concentration with concentration increasing radially. We also confirmed the radial phosphorus distribution, which we observed with repeated measurements, with SIMS analysis.
Our work has demonstrated that in-fab TXRF can be used to effectively monitor the cross-contamination levels of implanters after phosphorus doping and the effect of ion beam cleaning. Phosphorus cross contamination is thus controlled and the possibility of phosphorus-enhanced diffusion is reduced to a minimal level. The ratio of Ka peak values of phosphorus to silicon from test wafers is a reliable indicator rather than just the phosphorus peak count. However, others will need to establish control limits based on product sensitivity.
We acknowledge AMD FAB25 management for support of this project.
We also thank Darcy Hall for collecting implant information, Laurence Kohler for helpful discussions, Bouavanh Phommachanh for TXRF data collection, Evelyn Ferrero for VPD-ICPMS analysis, and Liying Wu for SIMS analysis.
1. R.B. Fair, W.G. Meyer, ASTM STP 804, ed. D.C. Gupta, ASTM, 1983, p. 290.
2. S.S. Todorov et al., Proc. of Intern. Conf. on Ion Implant. Technol., p. 715, 2000.
3..J. Xu, H. Lee, Proc. of Intern. Conf. on Ion Implantation Technology, p. 151, 1998.
4..R. Klockenkämper, Total-reflection X-ray Fluorescence Analysis, John Wiley & Sons Inc., New York, 1997.
Zhiyong Zhao received his PhD from the University of North Texas. He is a member of the technical staff at Advanced Micro Devices Inc. (AMD), 5204 E. Ben White Blvd., Austin, TX 78743; ph 512/602-2208, fax 512/602-5299, e-mail firstname.lastname@example.org.
Amiya R. Ghatak-Roy received his PhD from Texas A&M University. He is a member of the technical staff at AMD.
Tim Z. Hossain received his PhD from the U. of Kentucky. He is a senior member of the technical staff and a section manager at AMD.
Zhiyong Zhao, Amiya R. Ghatak-Roy, Tim Z. Hossain, Advanced Micro Devices Inc., Austin, Texas