Interferences in Atomic Absorption

Modern clinical and environmental labs run graphite-furnace AAS with Zeeman background correction as a matter of course for trace lead, cadmium, and chromium in blood and urine, precisely because biological matrices are protein- and salt-rich and would otherwise produce large, variable background absorbance that a deuterium arc (limited by its own lamp geometry at longer wavelengths) cannot always fully track. Matrix modifiers — chemical analogues of releasing/protective agents adapted for furnace chemistry, such as palladium/magnesium nitrate mixtures — are added routinely to stabilize volatile analytes like As, Se, and Cd during the char (ashing) step, preventing their loss before atomization, a furnace-specific cousin of the phosphate/lanthanum problem worked through above. Ionization buffering with cesium or potassium chloride remains standard practice for flame AAS determination of alkali and alkaline-earth elements in fertilizer extracts, cement digests, and geological samples, where sample-to-sample variation in total dissolved solids would otherwise make a single external calibration meaningless. Standard additions is the default calibration strategy whenever a lab cannot obtain or synthesize a matrix-matched blank — for example, characterizing trace metals in an uncommon industrial brine, a used lubricating oil, or a pharmaceutical excipient with an unknown but proprietary bulk composition. The interference framework built here — spectral, chemical, ionization, matrix — transfers directly, with modification, to Inductively Coupled Plasma Spectrometry, where the much higher plasma temperature (roughly 6000–10,000 K versus a flame's 2000–3000 K) suppresses most chemical interference (refractory compounds dissociate readily) and most ionization interference (the plasma's own enormous electron/ion density from argon itself swamps any sample contribution), but introduces new spectral challenges from the plasma's rich emission-line background and from polyatomic ion interferences in ICP-mass spectrometry (e.g., ArCl⁺ overlapping ⁷⁵As⁺). Understanding why AAS interferences arise from specific, nameable equilibria — rather than treating "matrix effect" as an unexplained fudge factor — is exactly the diagnostic habit that lets an analyst reason correctly about which interferences survive, worsen, or disappear when moving to a hotter, higher-electron-density atom source, and it is grounded throughout in the atomization chemistry covered under Atomization Methods for AAS.

:::quiz[In the derivation β = K_i/(K_i + [e⁻]), adding a large excess of Cs⁺ to standards and samples works because it:]

• Removes all electrons from the flame, forcing β → 0
• Reacts chemically with the analyte to prevent ionization
• Fixes [e⁻] at a high, common value dominated by Cs, clamping β to the same small value in both matrices ✓
• Increases K_i for the analyte directly

Ask the tutor: How would the ionization-suppression correction and the standard-additions correction need to be combined if a sample suffers from both an ionization interference and a proportional viscosity-driven transport interference at the same time?

Atomic absorption spectrometry measures the population of free, ground-state atoms of one element sitting in the optical path of a hollow-cathode lamp beam, inside a flame or graphite furnace atom cell. Beer's law, $A = \varepsilon b N_0$, only holds if every photon absorbed at the analytical wavelength comes from the analyte's own resonance transition, and if the number density $N_0$ of ground-state free atoms is a fixed, reproducible fraction of the total analyte introduced. Interferences are exactly the ways this idealization breaks: extra species absorb or scatter light that is not analyte absorption (spectral interference), or the equilibrium chemistry inside the atom cell diverts analyte away from free ground-state atoms (chemical and ionization interference), or the physical process of getting sample into the atom cell differs between standard and unknown (matrix interference). Spectral interference has two flavors. The first is genuine spectral overlap — a rare but real problem where another element's absorption line lies within the bandpass of the monochromator and coincides closely enough with the analyte line (e.g., the overlap of Eu 324.753 nm with a nearby Cu line region, or the classic V/Al overlap at 308.2 nm). The second, far more common, is background absorption/scattering: unvaporized salt particles, molecular species like alkali halides, and combustion products absorb broadly across a range of wavelengths or scatter light out of the optical path entirely, both of which the detector reads as "absorbance" that has nothing to do with ground-state analyte atoms. Because flame and furnace conditions change with matrix load (a seawater digest versus a dilute aqueous standard), background absorption is usually far larger in real samples than in the calibration curve, producing a strongly positive bias unless corrected. Chemical interference arises when the analyte forms a thermally stable compound in the atomizer — most classically, phosphate suppresses calcium absorption by forming refractory calcium pyrophosphate, $\mathrm{Ca_3(PO_4)_2}$-type species, that resist thermal dissociation to free Ca atoms at flame temperature. The fix works by mass-action competition: adding a large excess of a "releasing agent" like La³⁺ or Sr²⁺ preferentially binds the phosphate, releasing calcium to atomize normally. Ionization interference is a separate equilibrium — at flame temperatures (2000–3000 K), easily ionized elements (Na, K, Cs, Rb — low first ionization energy) partially ionize, $\mathrm{M} \rightleftharpoons \mathrm{M^+} + e^-$, and since M⁺ does not absorb at the neutral atom's resonance wavelength, the sample's own electron density (which depends on everything else easily ionized in the matrix) shifts this equilibrium and changes how much analyte survives as neutral atoms.

Matrix/physical interference is the least "chemical" of the four: it is a mismatch in nebulization efficiency, viscosity, surface tension, or droplet size distribution between standards and samples (organic solvents nebulize differently than water; a concentrated brine has different viscosity than a dilute standard) or, in furnace AAS, a difference in drying/ashing/atomization ramp behavior. These physical differences change how much analyte per unit time actually reaches the atom cell, independent of any chemistry happening once it's there.

The absorbance actually measured is the sum of true atomic absorbance and every non-atomic contribution:

where:

• $A_{\text{measured}}$ — raw absorbance read by the instrument at the analytical wavelength
• $A_{\text{atomic}}$ — absorbance from ground-state analyte atoms (the quantity Beer's law actually describes)
• $A_{\text{background}}$ — absorbance from scattering (particles) and molecular/continuum absorption (unrelated species)

Reading: every background-correction technique in AAS exists to measure $A_{\text{background}}$ separately and subtract it, recovering $A_{\text{atomic}}$ for use in the calibration. The degree of atomization — the fraction of total analyte present as free, ground-state gas atoms — governs chemical interference:

where:

• $\alpha_{\text{atom}}$ — atomization efficiency (dimensionless fraction, ≤ 1)
• $N_0$ — number density of free ground-state atoms actually present in the optical path
• $C_{\text{total}}$ — total analyte concentration introduced into the atomizer, atomized or not

Reading: chemical interference reduces $\alpha_{\text{atom}}$ below its calibration-standard value by tying up analyte in a stable compound (e.g., calcium pyrophosphate) that never dissociates to free atoms at flame temperature; a releasing agent restores $\alpha_{\text{atom}}$ toward 1. The ionization equilibrium is a mass-action balance identical in form to a weak-acid dissociation, with the electron concentration playing the role of a common-ion:

where:

• $K_i$ — ionization equilibrium constant at the flame temperature (from the Saha equation, strongly dependent on ionization energy and $T$)
• $[\mathrm{M^+}]$, $[\mathrm{M}]$ — concentrations (number densities) of ionized and neutral analyte in the flame gas
• $[e^-]$ — free electron number density in the flame, set by the total easily-ionized-element load of sample plus any added buffer

Reading: the fraction of analyte ionized, $\beta = [\mathrm{M^+}]/([\mathrm{M}]+[\mathrm{M^+}])$, depends on $[e^-]$; since $A_{\text{atomic}} \propto [\mathrm{M}]$, any matrix constituent that changes $[e^-]$ (e.g., Na, K in real samples but absent from clean standards) changes measured absorbance even at identical total analyte concentration.

Example 1 — Foundational: recovering true atomic absorbance from raw and background readings. A furnace AAS instrument with a deuterium-arc background corrector measures a Pb sample at 283.3 nm. The single-beam (uncorrected) absorbance reading is $A_{\text{measured}} = 0.446$. Immediately after, the deuterium lamp channel (which sees the same optical path but at a wavelength/bandpass where Pb itself does not absorb, representing the broadband background) reads $A_{D_2} = 0.087$. Using the additive model,

If the calibration curve (built from aqueous Pb standards with negligible background, so $A_{\text{atomic}} \approx A_{\text{measured}}$ for those points) has slope $k = 0.0215\ \text{L}\ \mu\text{g}^{-1}$ (from a plot of $A$ vs. concentration in µg/L, intercept ≈ 0), the corrected concentration is

Had the analyst ignored the background and used $A_{\text{measured}} = 0.446$ directly, the reported concentration would have been $0.446/0.0215 = 20.7\ \mu\text{g/L}$ — a 24% high bias, large enough to move a drinking-water sample from compliant (below the 15 µg/L action level, after accounting for the true 16.7 µg/L which is itself already over) into a badly overstated exceedance, illustrating why background correction is not optional bookkeeping but the difference between a defensible and an indefensible result. (Reporting here: $C_x = 16.7\ \mu\text{g/L}$, three significant figures, consistent with the input absorbances given to three figures.)

Limiting-case check: if the background absorbance were zero ($A_{D_2} \to 0$, a genuinely clean matrix), $A_{\text{atomic}} \to A_{\text{measured}}$ exactly, and the correction becomes a no-op — consistent with background correction being a pure subtraction that vanishes when there is nothing to subtract.

Example 2 — Applied: chemical interference from phosphate on calcium, and the releasing-agent fix. A clinical lab measures serum calcium by flame AAS at 422.7 nm. A 5.00 mg/L aqueous Ca standard (no phosphate) gives $A = 0.284$. A serum-matrix sample expected to contain 5.00 mg/L Ca (confirmed independently by a reference method) but also containing 45 mg/L phosphate as $\mathrm{PO_4^{3-}}$ gives only $A = 0.156$ under the same flame conditions, with no background correction issue (confirmed by a separate deuterium-arc check showing negligible scatter/molecular absorption at this wavelength). The atomization efficiency ratio is

meaning only 54.9% of the calcium atomizes in the presence of 45 mg/L phosphate, versus the standard — a classic chemical interference from thermally stable calcium phosphate/pyrophosphate species that resist dissociation in the flame. The fix: add lanthanum chloride to both standards and samples to a final concentration of 2000 mg/L La³⁺ (a large excess relative to the phosphate). Lanthanum preferentially binds phosphate (forming LaPO₄, more thermally and chemically favored to form than the calcium analogue under these conditions), releasing Ca to atomize freely. Repeating the measurement with 2000 mg/L La³⁺ added uniformly:

• Standard + La: $A = 0.281$ (matches the La-free standard within noise, confirming La itself doesn't interfere with Ca)
• Serum sample + La: $A = 0.278$

Now

i.e., 98.9% recovery — the chemical interference has been suppressed to within normal measurement noise (a few percent), and the calcium concentration read off the La-buffered calibration curve is now trustworthy:

consistent with the reference value within the ±3% typical precision of flame AAS.

Limiting-case check: in the limit of zero phosphate (no interferent), $\alpha_{\text{sample}}/\alpha_{\text{std}} \to 1$ even without La, and adding La then has no effect on the ratio — consistent with a releasing agent doing nothing when there's no competing equilibrium to release the analyte from.

Example 3 — Exam-level: disentangling simultaneous ionization suppression and a matrix viscosity effect, and choosing the correct calibration strategy. An analyst is asked to determine potassium in an industrial brine (high Na⁺ and Cl⁻, viscosity roughly double that of water) by flame AAS at 766.5 nm, a resonance line highly susceptible to ionization interference because K has a low first ionization energy (4.34 eV). Four measurements are made, all diluted 1:100 before analysis to bring K into range and reduce viscosity effects: 1. Aqueous K standard (no additives), 2.00 mg/L: $A = 0.512$ 2. Diluted brine sample (nominal K ≈ 2.00 mg/L by prior estimate), no additives: $A = 0.398$ 3. Same aqueous K standard (2.00 mg/L) + 2000 mg/L CsCl ionization buffer: $A = 0.487$ 4. Same diluted brine sample + 2000 mg/L CsCl: $A = 0.481$

Step 1 — diagnose the direction of the unbuffered discrepancy. Comparing (1) and (2): the brine reads lower than the standard at nominally equal K concentration, $0.398/0.512 = 0.777$, a 22% low bias. This is consistent with the brine's high Na⁺/Cl⁻ load — Na ionizes in the flame and raises the ambient electron density $[e^-]$ well above that of the clean aqueous K standard. From $\beta = K_i/(K_i+[e^-])$, a higher $[e^-]$ actually suppresses K ionization (pushes $\beta_K$ down, favoring neutral K) — so ionization alone would predict the brine reads higher, not lower. This tells the analyst that a second effect — most likely the brine's elevated viscosity reducing nebulization/aspiration efficiency (a matrix/physical interference) — is also acting, and dominates in the opposite direction.

Step 2 — use the Cs-buffered data to isolate the ionization piece. Comparing (1) and (3): adding Cs buffer to the standard itself changes its reading from 0.512 to 0.487, a decrease of $0.487/0.512 = 0.951$. This is surprising at first (Cs buffering usually protects K, not suppresses it) but is explained correctly as follows: the unbuffered aqueous standard already has very low $[e^-]$ (only K's own trace self-ionization), so K in the plain standard is already close to its minimum possible ionized fraction — nearly fully neutral. Adding Cs buffer does not change $\beta_K$ much (it's already near its ceiling), but it does add 2000 mg/L of dissolved solids that measurably changes flame temperature and nebulization slightly, producing a small physical/matrix effect of its own, here a 4.9% decrease. (This is why even ionization buffers must be added to standards and samples identically — they are not chemically "free.")

Step 3 — compare the Cs-buffered standard and Cs-buffered sample. Now both (3) and (4) contain the same 2000 mg/L Cs buffer, so both have $[e^-]$ dominated by the Cs, clamped to essentially the same value — the ionization interference is neutralized in both. The ratio

shows only a 1.2% residual discrepancy — within ordinary measurement precision for flame AAS (typically 1–3% RSD). This confirms that once ionization is buffered out, the viscosity-driven transport difference that looked large in the unbuffered comparison (22% low) was in fact largely an artifact of comparing across an ionization-sensitive line without a buffer; buffering also happens to standardize nebulization somewhat by dominating the total dissolved solids, so it can mask small physical effects too.

Step 4 — report and justify. The correct concentration is read from the Cs-buffered standard curve:

The analyst reports 1.98 mg/L K and documents that ionization buffering was required because the line (766.5 nm) is resonance-sensitive, that both standards and sample received identical Cs treatment, and that the residual 1.2% discrepancy is within method precision — no further correction (e.g., standard additions) is warranted, though it would have been the fallback if the residual discrepancy had remained large (which would indicate an uncorrected chemical or spectral interference rather than ionization/physical effects).

Limiting-case check: if the brine's Na content were reduced toward zero (a dilute, low-TDS sample), the ratio in Step 1 would approach 1 even without Cs buffering, and the entire correction procedure would become unnecessary — consistent with ionization and matrix interference both vanishing in the limit of a matrix approaching pure water.

The systematic study of AAS interferences is inseparable from the instrument's invention and refinement, most associated with the Australian physicist Alan Walsh, working at the CSIRO Division of Chemical Physics in Melbourne in the 1950s. Walsh recognized that atomic emission spectroscopy — the dominant technique at the time — suffered from severe matrix-dependent excitation effects because flame temperature and composition strongly influenced the excited-state populations that emission depends on. His insight was that absorption by ground-state atoms, using a hollow-cathode lamp emitting the analyte's own sharp resonance lines as the source, would be far less sensitive to flame excitation conditions, because the vast majority of atoms in a flame remain in the ground state regardless of temperature (a consequence of Boltzmann population statistics at flame temperatures for typical resonance transition energies). This is precisely why AAS is comparatively robust to emission-type interferences but remains fully exposed to atomization, ionization, and background problems, which Walsh's own group and others (notably Boris L'vov, who developed the graphite furnace atomizer in the Soviet Union in parallel) spent the following decades characterizing and correcting.