While 10 years ago SEMI deemed that 10-ppb purity levels were adequate for many chemicals, today 100-ppt levels are typical, and for some of the more critical materials, 10-ppt levels are being proposed. A guideline for 30–32% hydrogen peroxide (H₂O₂) of 100 ppt for 18 trace elements is in its final stages of approval, after which it will go into the Book of SEMI Standards (BOSS) as a Grade 4 specification.^1,2 However, because H₂O₂ is widely used in semiconductor manufacturing, a next-generation (Grade 5) purity level of 10 ppt for 21 elements has already been proposed.

Traditionally, inductively coupled plasma mass spectrometry (ICP-MS) has been the technique of choice for ultralow trace-element determinations in high-purity chemicals. It has been used
successfully to determine impurities at the 100-ppt (Grade 4) level in hydrogen peroxide using SEMI methodology. Unfortunately, it does not have the capability to detect all 21 trace metals specified at the proposed H₂O₂ purity level of 10 ppt, even when various background reduction techniques, such as the cool plasma approach, are implemented.

To meet that challenge, a new approach to improving ICP-MS detection limits has been developed. Known as dynamic reaction cell (DRC) technology, this approach utilizes ion-molecule chemistry to eliminate many argon- and solvent-based interferences, thereby improving ICP-MS detection limits and background equivalent concentrations (BECs) for the problematic elements.³ BEC is defined as the apparent concentration for the background signal based on the sensitivity of an element at a specified mass. The lower the BEC value, the more easily a signal generated by the element can be discerned from the background. It is widely believed that the BEC is a more accurate indicator than detection limits for measuring the performance of an ICP-MS system for each element of interest.

After reviewing the use of hydrogen peroxide in the semiconductor industry, this article describes the dynamic reaction cell technology in detail and explains how its use leads to detection limit and BEC improvements. Data collected by Solvay Interox, a global supplier of high-purity chemicals based in Brussels, are presented showing that detection limits and spike recoveries at the Grade 5 level in 31% hydrogen peroxide have been achieved using DRC ICP-MS and the methodology specified by SEMI.

Because hydrogen peroxide comes in contact with wafers during several steps in the semiconductor manufacturing process, it must be extremely clean. It is usually mixed with other chemicals, such as ammonium hydroxide (NH₄OH), hydrochloric acid (HCl), and sulfuric acid (H₂SO₄), which are used to dissolve metals from the wafer surface or to build a silicon dioxide layer on top of the silicon substrate. Since extremely large quantities of these chemical mixtures are used, there has always been a very stringent purity requirement for the H₂O₂ supplied to the industry. In addition, whenever SEMI seeks to lower contamination levels in order to improve device yields, hydrogen peroxide is always in the first group of chemicals to be evaluated. For these reasons, SEMI is investigating the possibility of lowering the maximum contamination level from 100 to 10 ppt for 21 elements.

Figure 1: ICP-MS sensitivity for a selected group of elements in different HNO3 concentrations under (a) cool plasma conditions (800 W of RF power, 1.5 L/min of nebulizer gas) and (b) hot plasma conditions (1600 W of RF power, 1 L/min of nebulizer gas).

Solvay Interox was among the first chemical suppliers to utilize ICP-MS for monitoring ultralow trace-metal contamination levels. However, the company soon realized that even under cool plasma conditions, there were a number of critical elements, such as iron, calcium, and potassium, that ICP-MS could not detect at very low levels. Many other chemical suppliers that used ICP-MS also discovered its limitations. Some of them tried to overcome those limitations by using simple preconcentration methods, while others invested in such detection-enhancing techniques as electrothermal vaporization ICP-MS, magnetic sector–high resolution ICP-MS, and, more recently, collision cell ICP-MS.^4–6 But although these alternative approaches offered some improvement, their users still could not achieve the 10-ppt-level contamination data for all 21 elements being proposed by SEMI.

When DRC ICP-MS was introduced in 1999 and data were presented on its successful use with semiconductor process chemicals, Solvay Interox decided to investigate the technology's potential for the analysis of Grade 5 H₂O₂. The results of an early study had revealed that for many key elements, particularly boron, the background levels on the DRC were significantly lower than those with traditional ICP-MS instrumentation. After lengthy evaluation, the chemical supplier invested in a DRC-based instrument, the ELAN 6100 DRC from PerkinElmer SCIEX Instruments (Concord, ON, Canada), in early 2001. The data generated by that instrumentation are presented later in this article.

When a traditional quadrupole ICP-MS instrument operates under normal plasma conditions (typically 1000 W of RF power and 1.0 L/min of nebulizer gas flow), argon ions combine with matrix- and solvent-based ions (hydrogen, oxygen, chlorine, nitrogen, sulfur, carbon, etc.) to generate polyatomic spectral species that interfere with the analysis of many analyte ions. Table I lists some typical examples. There are several approaches to reduce such spectral interferences. The first major breakthrough came in the mid-1990s with the commercialization of the cool, or cold, plasma approach. This technology, first reported in the literature in the late 1980s, uses a reduced-temperature plasma to minimize the formation of the troublesome polyatomic ions.⁷

Figure 2: Schematic illustrating how DRC technology reduces the spectral interference of 40Ar16O on 56Fe.

Under cool plasma conditions (500–800 W of RF power and a 1.5–1.8-L/min nebulizer gas flow), ionization reactions in the plasma are changed in such a manner that many of the typical polyatomic interferences are reduced. The result is that detection limits are improved for a number of critical semiconductor elements, including iron, potassium, and calcium.⁸ However, because the cool plasma contains much less energy than normal, high-temperature plasma, the instrument's sensitivity for the majority of elements can be severely affected by the sample matrix. This inherent weakness of the cool plasma approach can be seen by comparing Figure 1a, which shows the sensitivity for a selected group of elements in varying concentrations of nitric acid under cool plasma conditions, and Figure 1b, which shows the sensitivity for the same group of elements under hot plasma conditions. With the cool plasma, analyte sensitivity decreased dramatically as the acid concentration increased, whereas under hot plasma conditions, the sensitivity for most of the elements varied only slightly with changes in acid concentration.

The variation in instrument performance under cool plasma conditions may make it necessary to use standard additions or matrix matching to achieve satisfactory results. Additionally, obtaining the most accurate results for a multielement analysis often requires the use of two sets of operating conditions—one for the elements that can be detected at lower levels under cool plasma conditions and another for the other elements—which can be a very time-consuming and cumbersome process.

Figure 3: Spectral scans of varying concentrations of iron. Virtually no signal intensity can be seen at mass 56 for doubly DI water (DDIW), which indicates there is very little background interference.

Another major development in reducing ICP-MS polyatomic spectral interferences employs a multipole (typically a hexapole or octapole) housed within a cell that is placed upstream of the analyzer quadrupole.⁹ The flow of gas, typically hydrogen or helium, within this cell stimulates ion-molecule collisions that convert the ions associated with spectral interference to noninterfering species. After passing through the collision cell, the sample is directed to the quadrupole analyzer for normal mass separation. Although this technology initially showed great promise for use with process chemicals, it did not reduce background levels low enough to significantly improve detection limits for such problematic elements as iron, potassium, and calcium over the cool plasma approach.

The need to continue to improve ICP-MS detection limits for semiconductor-related elements led to the development of DRC technology.¹⁰ Similar in appearance to the collision cell, the dynamic reaction cell is a pressurized multipole positioned upstream of the analyzer quadrupole. However, this is where the similarity of the two techniques ends. In DRC technology, the upstream cell contains a quadrupole instead of a hexapole or octapole, and a reactive gas such as ammonia is bled into this cell, which causes ion-molecule interactions. By a number of different mechanisms, which are predominantly reactions and not just collisions, the gaseous molecules convert the potentially interfering ions into species that will not interfere with the analyte during the subsequent conventional mass separation in the analyzer quadrupole.

The advantage of using a quadrupole in the reaction cell is that the stability regions are much better defined than with a hexapole or octapole, allowing precise control of the chemical reactions that take place within the cell. Unwanted reaction by-products that could potentially lead to new interferences are prevented via optimization of the quadrupole's electrical fields. This cleansing process, known as chemical resolution, means that every time a sample enters the reaction cell, the bandpass of the quadrupole is optimized for the specific interference problem associated with the first analyte of interest; it is then changed on the fly for the next analyte.

This chemical resolution process is depicted schematically in Figure 2 for one of the most common problems in ICP-MS, the polyatomic interference of ⁴⁰Ar¹⁶O on the determination of ⁵⁶Fe. In this example, the reaction gas (NH₃) that is bled into the reaction cell reacts with the ⁴⁰Ar¹⁶O in the sample matrix to form new species that will not interfere with ⁵⁶Fe, but does not react with the ⁵⁶Fe. When the flow of the reaction gas is optimized, the chemical cleansing process is so thorough that the background from the ⁴⁰Ar¹⁶O is virtually eliminated. This is exemplified in Figure 3, which shows a spectral scan of varying concentrations of iron. It can be seen that the background signal contribution at mass 56 for DI water is less than 10 counts per second. If no reaction gas had been used in the cell, the background would have been five to six orders of magnitude higher. Figure 4 indicates that with the ultralow ⁴⁰Ar¹⁶O background, DRC ICP-MS can achieve a detection limit for ⁵⁶Fe of <0.2 ppt, which is significantly lower than that achieved for ICP-MS with a collision cell.

Figure 4: ICP-MS iron detection as a function of gas flow in a reduction cell. By optimizing the flow of reaction gas, the 40Ar16O background can be reduced to such a level that a detection limit of <0.2 ppt is achievable.

In addition to its effectiveness for iron analysis, the DRC approach has achieved background reductions for other common polyatomic ions that cause spectral interference, including ⁴⁰Ar on the determination of ⁴⁰Ca and ³⁸Ar¹H on the determination of ³⁹K. Using the technology, detection limits for both these elements can be reduced to better than 1 ppt.

Unlike the cool plasma approach, DRC ICP-MS uses a normal-temperature plasma, which means that elements with high ionization potential can be determined with no sacrifice in sensitivity. Therefore, many of the other problematic interferences, such as ⁴⁰Ar⁴⁰Ar on ⁸⁰Se and ⁴⁰Ar³⁵Cl on ⁷⁵As, are relatively straightforward to overcome. DRC also can be used without a reaction gas for the determination of elements that do not suffer from polyatomic spectral interferences. In that mode, the cell is used as an ion-focusing guide rather than to carry out ion-molecule reactions. Table II presents some typical multielement detection limits and BECs achieved with the technology. The data for the elements in red were obtained using a reaction gas, while the data for the elements in black were obtained with the ICP-MS instrument in standard mode. All results were achieved using a 1-second integration time; further detection limit improvements may be achieved using longer integration times in the single-element mode.

The conversion of a proposed SEMI guideline for process chemical purity to an industry standard depends on the generation of analytical data showing that it is possible to achieve a spike recovery between 75 and 125% at 50% of the proposed specification using SEMI methodology and existing instrumentation. Only when this capability is established will a guideline be published in the BOSS as a specification.

For any instrumental technique to comply with SEMI requirements, the analytical data must meet very stringent validation criteria, which are based on analyzing two unspiked samples and two spiked samples. Table III lists the data-reporting requirements.¹¹ The data generated are then evaluated based on certain performance criteria, which are summarized in Table IV.

The early spike recovery and detection level data generated for Grade 5 H₂O₂ at Solvay Interox have revealed that DRC ICP-MS technology can satisfy these SEMI requirements.¹³ The ELAN 6100 instrument, which has been installed in a Class 100 cleanroom, was used to generate these data. The main instrumental parameters are shown in Table V, and the reaction cell gas flows for each element are listed in Table VI. The instrument's spray chamber is quartz cyclonic, the microconcentric nebulizer is made from PFA, the sample injector is quartz, and the interface cones are platinum.

Calibration was performed with a 100-ppt multielement standard made by diluting a commercial stock solution in ultrapure water (UPW) and adding 1% nitric acid (HNO₃). The 31% ultra-high-purity H₂O₂ was injected directly into the instrument using a PFA-100 microconcentric nebulizer (ESI; Omaha, NE). The method detection limit (MDL) was calculated from linear regression of a calibration curve generated from two sets of H₂O₂ samples spiked with 0, 5, 10, 20, 50, and 100 ppt of each analyte element.

A summary of the data for the 21 elements included in the proposed Grade 5 specification is given in Table VII. (Although a UPW blank also was analyzed, the data in this table have not been blank subtracted.) It can be seen that the criteria for success listed in Table IV were met for every element. The enhanced detection capability of the DRC ICP-MS technology enabled all the elements to be measured at the 10-ppt level, a performance that could not have been achieved with conventional ICP-MS technology. Although this study has focused on hydrogen peroxide, the results are typical of those for additional process chemicals that other researchers have analyzed with the DRC technology.¹⁴

To keep pace with the increasingly stringent requirements of the semiconductor industry, analytical instrumentation must undergo continual improvement. In the area of high-purity process chemicals, ICP-MS is routinely used for trace-element analysis. However, to reach next-generation purity levels, the detection capabilities of ICP-MS must be improved. This study focused on one such detection-enhancing approach called DRC technology, in which the sample passes through a cell in which ion-molecule reactions occur. By the process of chemical resolution, these reactions convert interfering ions into species that do not interfere with the analysis, thereby improving the spectrometer's detection limits. The results reported in this article illustrate the ability of DRC ICP-MS technology to analyze hydrogen peroxide at the proposed SEMI Grade 5 purity level.