Determination of Carbon Content in Monocrystalline Silicon Wafers: Principles, Methods, and Industry

Carbon is one of the most critical impurities in monocrystalline silicon wafers, directly influencing crystal integrity, gate oxide quality, device yield, and solar cell efficiency. Accurate determination of carbon content—most commonly by Fourier transform infrared (FTIR) spectroscopy in accordance with SEMI MF1391 / ASTM F1391—has become a routine requirement for incoming material inspection, crystal growth control, and product certification in the semiconductor and photovoltaic industries. This article explains the significance of carbon measurement, applicable sample types, key testing items, mainstream analytical methods, and answers to frequently asked questions.

Technical Overview

Carbon enters monocrystalline silicon primarily through the polysilicon feedstock, graphite components of the Czochralski (CZ) hot zone, crucible graphite susceptors, and the ambient atmosphere during crystal growth. Because silicon has a carbon solubility limit of roughly 3×10¹⁷ atoms/cm³ near the melting point, excess carbon precipitates as silicon carbide (SiC) particles during cooling, creating structural defects within the crystal lattice.

In silicon, carbon exists in two principal forms: substitutional carbon atoms occupying silicon lattice sites, and precipitated carbon in the form of SiC inclusions. Most routine quality control targets substitutional carbon, as it acts as a nucleation center for oxygen precipitation and microdefect formation. Elevated carbon levels are associated with several reliability and performance issues:

• Formation of swirl defects and SiC precipitates that degrade gate oxide integrity in MOS devices
• Enhanced or uncontrolled oxygen precipitation during high-temperature thermal processing
• Reduced minority carrier lifetime and conversion efficiency in photovoltaic cells
• Increased wafer warpage, slip, and mechanical fracture risk during device fabrication
• Loss of pattern fidelity in advanced lithography processes due to bulk microdefects

Carbon content is expressed either in atoms per cubic centimeter (at/cm³) or in parts per million atomic (ppma). Since the atomic density of silicon is approximately 5×10²² atoms/cm³, 1 ppma corresponds to 5×10¹⁶ at/cm³. Typical specification limits range from ≤1×10¹⁶ at/cm³ (0.2 ppma) for advanced semiconductor-grade material to several ×10¹⁷ at/cm³ for photovoltaic-grade wafers, depending on the end application and customer requirements.

Testing Samples

Polished Monocrystalline Silicon Wafers

Wafers prepared with double-side polished surfaces are the standard specimen for infrared transmission measurement. Surface roughness causes scattering and distorts the absorption baseline, so as-cut or lapped wafers must be re-polished before FTIR testing. Specimen thickness is typically around 2 mm for maximum sensitivity per the reference method, although thinner production wafers can be measured with reduced detection capability. The thickness must be determined accurately, since it enters directly into the calculation of the absorption coefficient.

Crystal Ingot and Rod Sections

For process control during crystal growth, coupons are sliced from the seed end, tang end, and body of the ingot, then ground and double-side polished. This allows axial carbon distribution to be profiled and hot-zone contamination sources to be identified. Float-zone (FZ) crystals may be similarly sampled to confirm inherently low carbon levels.

Polysilicon Feedstock and Reclaimed Material

Feedstock chips or chunks can be evaluated either by melting them into a small FZ crystal that is then sectioned and polished, or by bulk analytical techniques such as combustion infrared absorption or glow discharge mass spectrometry, which do not require optically polished transmission specimens.

Reference (Blank) Specimens

The infrared method is differential: it requires a reference specimen of float-zone silicon with carbon content below the detection limit and a thickness matched to the test specimen. This blank compensates for the intrinsic two-phonon lattice absorption of silicon in the region of the carbon band, which is essential for accurate results.

Testing Items

Standard carbon evaluation services for monocrystalline silicon materials typically include the following measurable items:

• Substitutional carbon concentration (Cs) in at/cm³ or ppma — the primary quality control parameter
• Total carbon content, including precipitated SiC, determined by SIMS, GDMS, combustion analysis, or activation analysis
• Interstitial oxygen concentration (Oi), routinely co-measured at the 1107 cm⁻¹ band on the same specimen
• Axial or radial carbon distribution profiles across an ingot for growth-process characterization
• Conformity assessment against customer specifications or relevant SEMI silicon material specifications

Testing Methods

Fourier Transform Infrared (FTIR) Absorption Spectroscopy

FTIR is the industry-standard technique for substitutional carbon, performed per SEMI MF1391 (historically ASTM F1391) and equivalent national standards. Substitutional carbon produces a characteristic localized vibrational absorption band at approximately 605 cm⁻¹ (about 16.5 μm wavelength). Because the silicon lattice itself absorbs strongly in this spectral region, the measurement is performed in transmission mode against a matched float-zone reference blank, and the net absorption coefficient is calculated from the peak height at 605 cm⁻¹ minus the baseline at a nearby reference point (typically 730 cm⁻¹ or 640 cm⁻¹ as specified in the standard). The carbon concentration is then obtained by multiplying the net absorption coefficient by the conversion coefficient defined in the standard.

Key practical requirements for reliable FTIR results include:

• Double-side polished specimen with accurately known thickness
• Instrument purging with dry nitrogen or dry air to remove atmospheric water vapor and carbon dioxide interference
• Measurement at room temperature (approximately 300 K) with adequate spectral resolution and signal averaging
• Use of a carbon-free reference specimen of matching thickness and surface finish

The practical working range is approximately 5×10¹⁵ at/cm³ (0.1 ppma) to the carbon solubility limit, with a typical detection limit on the order of 5×10¹⁵ at/cm³ for standard 2 mm thick specimens. Interstitial oxygen can be determined simultaneously from the 1107 cm⁻¹ band per SEMI MF1188, providing a complete impurity picture from a single measurement.

Secondary Ion Mass Spectrometry (SIMS)

SIMS offers much lower detection limits (down to the 10¹⁴–10¹⁵ at/cm³ range) and measures total carbon, including precipitates. It is used for depth profiling, near-surface carbon evaluation, epitaxial layer qualification, and certification of reference materials, requiring carbon-implanted calibration standards for quantification.

Glow Discharge Mass Spectrometry (GDMS)

GDMS provides bulk trace-element analysis of silicon, including carbon, with excellent multi-element capability. It is well suited to feedstock qualification and lot characterization where total impurity screening is required.

Combustion Infrared Absorption and Activation Analysis

High-temperature combustion infrared analyzers determine total carbon in silicon chunks and feedstock by burning the sample and detecting evolved carbon oxides. Charged particle activation analysis (CPAA) serves as an absolute reference method for total carbon and is used primarily to certify standard reference materials rather than for routine production control.

Applications

Carbon content determination supports quality assurance and process optimization throughout the silicon value chain:

• Incoming material inspection of monocrystalline wafers by chipmakers, cell manufacturers, and module assemblers
• Hot-zone and process control in CZ crystal growth, including evaluation of graphite furniture, crucibles, and purge gas purity
• Polysilicon feedstock procurement and supplier qualification
• Research and development of FZ silicon, neutron transmutation doped (NTD) material, and advanced substrates
• Failure analysis of wafer breakage, yield loss, gate oxide breakdown, and lifetime degradation incidents
• Lot release and certification testing performed by accredited third-party laboratories against SEMI specifications and customer-defined limits, with laboratories commonly operating under ISO/IEC 17025 quality systems

FAQ

Why does the FTIR method measure substitutional carbon rather than total carbon?

The 605 cm⁻¹ infrared absorption band arises only from carbon atoms substituting on silicon lattice sites. Carbon that has already precipitated as silicon carbide is not infrared-active at this band. When total carbon is required, complementary techniques such as SIMS, GDMS, combustion analysis, or activation analysis must be used.

What is the detection limit of the FTIR method?

With standard 2 mm thick, double-side polished specimens and a matched float-zone reference, the practical detection limit is on the order of 5×10¹⁵ at/cm³ (approximately 0.1 ppma). Thinner production wafers exhibit proportionally higher detection limits due to the reduced optical path length.

Can as-cut or textured wafers be tested directly?

No. Rough, as-cut, lapped, or textured surfaces scatter infrared radiation and distort the absorption baseline. Specimens must be double-side polished, and their thickness must be accurately measured, before valid FTIR results can be obtained.

Which standards govern carbon measurement in silicon?

The primary method is SEMI MF1391, which superseded ASTM F1391, covering the determination of substitutional atomic carbon in silicon by infrared absorption. Companion standards such as SEMI MF1188 cover interstitial oxygen, and various GB/T national standards define equivalent procedures for the Chinese market. Testing laboratories typically hold ISO/IEC 17025 accreditation for these methods.

How do I convert between ppma and atoms per cubic centimeter?

Silicon contains approximately 5×10²² atoms/cm³, so 1 ppma equals 5×10¹⁶ at/cm³. For example, a reported carbon content of 1×10¹⁷ at/cm³ corresponds to 2 ppma. Reporting in both units avoids ambiguity when comparing against different specification sheets.

What causes high carbon levels in CZ silicon, and how can they be reduced?

The dominant sources are carbon-contaminated polysilicon feedstock and erosion of graphite crucible susceptor and heater components in the hot zone, along with inadequate gas purging. Corrective actions include tighter feedstock specifications, high-purity graphite or coated components, optimized argon purge flow, and routine carbon monitoring of ingot seed and tail sections.