In the past, quadrupole inductively coupled plasma mass spectrometry (ICP-MS) analysis has been sufficient for many low-level measurements, although interfering molecular species may compromise the accuracy of such results. At masses below mass-to-charge 80, ICP-MS has unit mass resolution, which makes the analyzer subject to several isotopic and polyatomic interferences at such mass levels. As the complexity of the matrix being analyzed increases, this phenomenon becomes an even greater concern. Combining the vapor phase decomposition (VPD) or drop scan etching sample preparation technique with magnetic sector high-resolution ICP-MS (HR-ICP-MS) can overcome this interference resolution problem.

Capable of detecting a range of elements, HR-ICP-MS not only offers low detection limits and a high degree of accuracy, it is less expensive than other surface techniques and can determine cleaning process efficiencies, chamber integrity, film integrity, and surface metallic contamination levels. After a brief discussion of surface analysis techniques and applications, this article reports on a study that demonstrates the detection capabilities of the combined VPD HR-ICP-MS technique for several critical semiconductor elements.

Several techniques can be employed to identify wafer surface contamination and its source. However, the VPD technique combined with ICP-MS or total reflection x-ray spectrometry (TXRF) are the current methods of choice for ultratrace detection of surface metallic contamination.² With VPD TXRF, elements ranging from sulfur through uranium with low detection limits and masses below 28 amu with higher detection limits can be analyzed. VPD ICP-MS offers better detection limits than TXRF for most analytes and also permits the analysis of critical semiconductor elements such as lithium, boron, aluminum, sodium, magnesium, and phosphorus. Moreover, VPD ICP-MS allows for direct calibration with certified standards. However, several groups have been working toward reconciling the standardization issues associated with TXRF instrumentation.³

The VPD ICP-MS process has been used in various research projects. In one study, it was used to determine the effects of plasma etching on silicon substrates. Before the analysis was conducted, it was generally believed that the primary concerns for the plasma etching process were wafer surface damage and crystal lattice disordering, physical effects that could be repaired by removing the damaged layer through hydrofluoric acid etch. The VPD ICP-MS tests indicated that etchers are also sources of heavy-metal contamination that could lead to a reduction in capacitor retention time and increased junction diode leakage.⁴

The VPD technique also is widely used for optimizing wet cleans by confirming the efficiency of contamination removal. An optimal wet clean results in a sufficient reduction of particle and metallic contamination levels while minimizing processing times and chemical and water use, leading to lower manufacturing costs. As much as 1500–2000 gal of ultrapure water are used to process each 200-mm wafer.⁵ As the industry continues its transition to 300-mm technologies, that amount will tend to increase. While efforts are being made to reduce processing times, water and chemical use, and the number of cleaning steps, it will be necessary to verify that contamination levels are not increasing.

In the study reported here, all samples were prepared using VPD in a Class 10 cleanroom. The VPD scanning solution was collected in high-purity Teflon vessels that were precleaned with 10-ppt-grade chemicals.

All quantification was performed using the method of standard additions. For the trace elemental analyses of wafer surfaces, the VPD scanning solution was analyzed with an Element2 HR-ICP-MS from Thermo Finnigan MAT (Bremen, Germany) equipped with a microflow nebulizer. Of interest were the identification of polyatomic interferences, the resolution of such interferences, and the verification of high results. Microsampling technologies were used in an attempt to lower detection limits.

Table I shows the operating conditions for HR-ICP-MS that were found to be optimal for this project. Using cool (600-W) plasma enabled the investigators to perform analyses under a single set of plasma conditions while still obtaining good detection limits for typical hot plasma elements such as copper, nickel, and zinc. However, in certain experiments, 1200-W plasma tunes were used as well.

Detection limits are a function of instrument sensitivity and stability, contamination levels, interference in the sampling and analysis processes, and the dilution factor of the sample. Sodium (Na), magnesium (Mg), aluminum (Al), calcium (Ca), and zinc (Zn) are common environmental contaminants. Several polyatomic species interfere with the detection of phosphorus (P), potassium (K), Ca, and iron (Fe). Arsenic (As) can be difficult to ionize and suffers from interferences from argon chloride. Microsampling techniques allow for a reduction in detection limits by decreasing the sample size. Not all of the analytes listed are soluble in the same VPD solution. Consequently, in the study discussed here, different kinds of VPD solutions were used to obtain the results.

Table II lists element detection limits that were achieved using the HR-ICP-MS instrument equipped with a 50-µl PFA microflow nebulizer and 750 µl of scanning solution. Both 600-W and normal 1200-W plasma tunes were utilized to optimize the detection of the various analytes. In addition, several kinds of VPD solutions were used in the analyses because the elements listed are not all soluble in the same solution.⁶

The comparable data in Table III show that using a 20-µl PFA microflow nebulizer resulted in even lower detection limits than were achieved with the 50-µl unit. To keep the dilution factor at a minimum, 100 µl of scanning solution were used in these analyses. While lower detection limits may be possible with the 50-µl nebulizer, attempts to use that unit with a solution volume of <100 µl led to inconsistent scanning and recovery results.

Copper metallization in processes such as damascene interconnect architecture are increasingly being used in IC manufacturing. While copper is useful for optimizing device performance, it also can be a source of contamination that decreases minority carrier lifetimes, roughens the wafer surface as a result of fast oxidation, and leads to increased dielectric leakage through ionic and atomic diffusion mechanisms.⁷ Therefore, the accurate and timely measurement of contamination resulting from copper processing is critical to device yields.

The rough surface of the wafer backside serves as a gettering agent for metals and is an obstacle to TXRF analysis, since rough surfaces scatter and reflect x-rays.⁸ In contrast, VPD ICP-MS and VPD HR-ICP-MS sample preparation does not require a flat or smooth surface, making these combination techniques well suited for the identification of both front- and backside copper contamination.

The accuracy of copper or other contamination measurements taken in a high-silicon matrix can be affected by polyatomic interferences. Because VPD scanning solutions may contain Si levels of between a few hundred and a few thousand parts per million, polyatomic Si interferences may be of concern.⁹ However, Figure 1 illustrates how using the appropriate resolution can overcome Si interference and enable an accurate quantification of copper contaminants. The cool plasma conditions listed in Table I were used for these analyses, along with a medium (R = 4000) resolution setting. Comparable results with low (R = 300) resolution were 439 and 100 ppt for ⁶⁵Cu and ⁶³Cu, respectively, which would suggest that an interference was causing falsely high readings.

Used to create thin films on the wafer surface, CVD processes are performed in chambers that are made out of ²⁷Al, which makes it necessary to monitor postprocess aluminum contamination on wafer surface layers. For example, aluminum particles from a CVD chamber may appear in BPSG, a boron- and phosphorus-doped silicon dioxide insulating layer, but analysis of this film poses several challenges. The interference issues associated with samples that contain high levels of silicon apply in this case, and it is also difficult to quantify aluminum in a matrix with high levels of boron, since there was a concern that ¹¹B¹⁶O may interfere with ²⁷Al. To address this issue, an experiment was devised to determine if boron oxide interfered with aluminum analysis in BPSG films when analyzed by quadrupole VPD ICP-MS.

To show the sensitivity of the HR-ICP-MS instrument, a DI-water sample spiked with 75 ppt of ²⁷Al was analyzed, using cool (600-W) plasma and medium (R = 4000) resolution. Those results are shown in Figure 2. Next, a low-resolution (R = 300) analysis of a BPSG film was performed in hot (1200-W) plasma. As shown in Figure 3, this analysis resulted in a false high result for ²⁷Al of 19 ppb, indicating that there was interference from ¹¹B¹⁶O. To resolve the aluminum peak from the ¹¹B¹⁶O peak, another HR-ICP-MS analysis was run, using a 1200-W plasma and medium resolution. This analysis, shown in Figure 4, resulted in a concentration of 7 ppb ²⁷Al.

Figure 2: Analysis of a 75-ppt 27Al spike in a DI-water background performed with cool plasma and medium resolution. This test result shows the capability of the HR-ICP-MS instrument to quantify 27Al at low levels.

Figure 3: Analytical results showing boron oxide interference with 27Al in a BPSG matrix. The analysis was done using hot plasma and low resolution.

Figure 4: Resolution of boron oxide from 27Al. The analysis was done using hot plasma and medium resolution.

Because it is standard practice at Micron Technology (Boise, ID) to analyze ²⁷Al with a quadrupole ICP-MS in cool plasma, an additional test was run to determine if boron interference was an issue under cool plasma conditions. The resulting data suggest that ¹¹B¹⁶O does not form in cool plasma. As seen in Figure 5, the same sample that yielded a result of 7 ppb ²⁷Al in hot plasma exhibited a concentration of 6 ppb ²⁷Al in cool plasma using medium resolution. To further investigate the theory that the boron oxide will not form in cool plasma, analyses of a different BPSG sample were performed in low-, medium-, and high-resolution modes using cool plasma. The resulting ²⁷Al concentration levels of 2666, 2636, and 2658 ppt, respectively, confirmed the hypothesis.

Figure 5: Medium-resolution scan of the sample depicted in Figure 4, performed using cool plasma. This result indicates that boron oxide does not form in cool plasma.

Diffusion chambers are also sources of aluminum contamination, which in this case may be difficult to analyze accurately because the ²⁷Al peak resides so close to the ²⁸Si peak in a high-silicon matrix. Figure 6 indicates that when HR-ICP-MS is used with cool or hot plasma, even a large ²⁸Si peak will not overlap with the ²⁷Al peak. In the hot plasma analysis, the ²⁸Si peak overranged the detector, but no resolution of aluminum and silicon was needed under either condition.

During ion implantation of boron and phosphorus, common dopants used in the IC fabrication process, the surface of the wafer is bombarded, creating the potential for residual contamination. Traditionally, identification of such nonmetallic contaminants has been accomplished using ion chromatography and surface techniques such as secondary ionization mass spectrometry and Auger electron spectroscopy.^10,11 Boron also can be measured by ICP-MS or HR-ICP-MS, but accurate determination of trace levels of phosphorus can be a problem. Phosphorus requires hot plasma conditions for ionization, but ¹⁴N¹⁶O¹H forms under such conditions and creates interference that may prevent the quantification of ³¹P. However, the low-level quantification of ³¹P in an oxide film is attainable by using HR-ICP-MS with medium (R = 4000) resolution. Other low-level analytical results for ³¹P in oxide films that were achieved in this study included 1.5 x 10¹⁰ atoms/cm² for a 1000-Å thermal oxide film and 2.2 x 10¹¹ and 3.6 x 10¹² atoms/cm² for similar films implanted with arsenic and phosphorus, respectively.

VPD HR-ICP-MS also is capable of determining low-level phosphorus carryover from phosphoric acid processing. Such data can be helpful in decreasing wafer rinse times, which can lead to reductions in water consumption and wafer processing times.¹² The results in Table IV show the efficiency of various rinse times for the removal of the phosphoric acid residue from the entire wafer surface.

As the semiconductor industry continues to evolve to technology nodes with increasingly stringent contamination limits, the need for advanced analytical capabilities will increase. VPD, when combined with HR-ICP-MS, achieves timely results and interference resolution while offering superior detection limits for a large range of detectable analytes. The detection limits for VPD HR-ICP-MS are in the range of 10⁶ atoms/cm² for several critical elements. The detection of such elements as Al, P, K, Ca, Fe, and Cu is obscured by several matrix-induced polyatomic interferences, compromising the analytical quantitation capabilities of a quadrupole ICP-MS. However, HR-ICP-MS systems in medium or high-resolution mode can confirm interfering peaks and achieve accurate results.

Developed to meet the need for new techniques, the combination of VPD and magnetic sector HR-ICP-MS offers accurate and timely results, interference resolution, and detection limits that exceed those required by The International Technology Roadmap for Semiconductors. Suitable for use with many analytes, the technique achieves detection limits in the range of 10⁶ atoms/cm² for several critical elements. In the cases of aluminum, phosphorus, potassium, calcium, iron, and copper, where several matrix-induced polyatomic interferences can compromise the analytical quantitation capabilities of a quadrupole ICP-MS, HR-ICP-MS systems in medium- or high-resolution modes enable investigators to confirm interfering peaks and ensure accurate results.