Vapor Phase Decomposition VPD-ICP-MS

Reviewed by SEMI Gr

VPD-ICP-MS: Ultra-Trace Wafer Surface Contamination Analysis

Metal concentrations of 5 × 10⁹ atoms/cm² may affect device efficiency and yield. At these concentrations, transition metals diffuse into silicon at high temperatures and act as recombination centers, reducing minority carrier lifetime, slowing oxidation, and increasing gate oxide leakage. The VPD-ICP-MS method can detect contamination below this threshold because it dissolves the entire surface oxide and concentrates it in one drop.
MatX Lab performs VPD-ICP-MS on bare and filmed wafers from 100 mm through 300 mm.

What the VPD-ICP-MS Measures

The result is a surface concentration of metallic contamination, reported in atoms per square centimeter, integrated over the scanned area.
It is a surface measurement in the strict sense. Only the topmost oxide layer and whatever sits on it are collected, typically the top few ångströms of the wafer. Metals distributed through the bulk of a thick film are not part of the answer, and reaching them requires film-stripping preparation rather than this one.
The measurement gives no spatial information on its own. Everything collected from the scanned area arrives at the instrument as one droplet, so the output is a total, not a map.

The Three-Step VPD Process

Decomposition. The wafer sits in an enclosed chamber and is exposed to hydrofluoric acid vapor, commonly for around ten minutes. The HF dissolves the native or thermal oxide into hexafluoro silicic acid, releasing every metal held in or on that oxide. As a by-product, the wafer surface turns hydrophobic, having been hydrophilic before, and that change is what makes the next step work.

Collection. A droplet of dilute acid recovery solution, typically a mix of dilute HF and hydrogen peroxide (about 250 µL), is scanned across the hydrophobic surface. The droplet stays coherent on the now-hydrophobic wafer and picks up the decomposed oxide together with the released metals.

Analysis: The droplet is then introduced into the ICP-MS system for analysis, usually via flow injection. This is because the contaminant in the entire 300 mm wafer has been concentrated into just a few hundred microliters of solution.

Detection Limits and Elemental Range

Detection limits vary by element, wafer size, and instrument generation. Almost all elements can attain detection limits of about 10⁷ atoms/cm² on a 300 mm wafer. Other elements have detection limits of 10⁶ or even 10¹⁰ atoms/cm².

Elemental detection limits are about 30 to 41 elements per normal analysis, mainly transition metals, and the literature shows the possibility of accessing even 60 elements. Contaminants such as sodium, potassium, calcium, aluminum, chromium, iron, nickel, copper, and zinc can arise from cleaning, etching, oxide formation, and implantation, driving the need for analysis.

Light elements are a specific strength here. TXRF cannot detect low-atomic-number elements because their fluorescence is too weak, while ICP-MS handles them comfortably.

Which Wafers and Films Are Suitable

The decomposition step requires a surface that dissolves in HF vapor, so silicon with native or thermal oxide is the natural choice. Silicon nitride and other deposited dielectric films also work.

Wafers carrying thin films from ALD, CVD, PVD, and similar processes can be analyzed, with the caveat that the measurement still describes the top layer rather than the film’s bulk composition.

Non-silicon substrates such as silicon carbide and gallium arsenide can be scanned using the same droplet technique without the HF decomposition step. Recovery then depends on how tightly the contamination is bound to that surface, and you should establish efficiency rather than assume it.
Bevel-edge-only scans are possible where edge-handling contamination is the suspected source rather than the wafer face.

What the Analysis Costs You

The surface oxide is gone afterward. VPD-ICP-MS consumes the layer it measures, so you cannot return the wafer to process or re-analyze it with the same technique.

Recovery is not always complete, and the amount recovered can vary depending on how contamination is held on the surface. If an element is present in a chemically resistant form, or is trapped in a film that HF cannot completely break down, the measured concentration may be lower than the actual amount on the wafer. When the goal is to determine an absolute concentration rather than simply compare samples, a spiked wafer can be used to check how much of each target element is actually recovered from that particular surface.
Handling can contaminate and alter the measurement outcome. When the concentration is in the range of 10⁷ atoms/cm², even a tiny source can make a significant contribution. Such a source may be contamination from a fingerprint, tweezers, or a dirty cassette. Placing wafer samples in clean carriers and following controlled handling from sample receipt through measurement minimize the risk of contamination.

Reporting

Each element is reported as atoms per square centimeter. The result is shown with the method blank and the detection limit for that element, so these values appear directly in the results rather than in a footnote.

The report also records exactly which part of the sample was analyzed. This includes the size of the scanned area and whether the measurement covered the entire face, only a selected portion, or the bevel edge. The report also includes details of the recovery process, such as the composition of the recovery solution, the volume of liquid applied to the surface, and the time allowed for decomposition before analysis.

The report identifies results that fall close to the method blank. This distinction matters because a result at, or only slightly above, the blank level does not necessarily demonstrate that the element is present on the wafer at a meaningful concentration.