Limits of Detection and Quantification

Introduction

A detector may register a small signal that is difficult to distinguish from background. Even when presence is plausible, assigning an accurate numerical concentration can be harder. Detection and quantification limits describe different performance thresholds under specified rules. Their values belong to a complete method and sample matrix, not just an instrument brochure.

Core explanation

Measure blanks through the relevant preparation and instrument process to characterise background response and its variability. A simple conceptual approach compares sample signal with the distribution of blank signals and requires enough separation to control false positives and false negatives. In a linear calibration y = a + bc, the blank's response scatter σy converts approximately to concentration scatter σc = σy/b. A steeper slope improves sensitivity in concentration units, while larger blank noise worsens it.

Rules of thumb sometimes express detection as a few times blank standard deviation divided by slope and quantification as a larger multiple, often around ten times. These are approximations, not universal definitions. The numerical factors depend on the decision criterion, number of measurements, desired error rates and whether blank and low-level sample responses have equal variance.

The limit of detection, LOD, concerns whether a low analyte level can be distinguished from blank by a stated procedure. The limit of quantification, LOQ, concerns whether that level can be measured with acceptable precision and bias under a stated criterion. A result between LOD and LOQ may be reported as detected but not reliably quantified, rather than as a highly precise concentration. A signal below LOD is not proof the analyte is absent; it means the method cannot establish its presence at that level under the decision rule.

Matrix matters. A clean solvent blank may be quiet while a real food extract produces interfering peaks or ion suppression. A method detection limit should reflect full extraction and analysis of realistic blank-like matrix when possible. Background contamination from reagents or containers can dominate at trace levels, and carryover from a high sample can make the next blank appear positive. Processing more sample may improve detection only until matrix and blank contributions grow too.

Sensitivity and selectivity differ. A detector can respond strongly to analyte but also to an interferent. Its steep calibration slope does not yield a low practical detection limit if the background signal is variable or indistinguishable. The analyst should specify analyte, matrix, method and decision rule with any reported LOD or LOQ.

Step-by-step reasoning

1. Define the analyte, matrix and complete method whose limit is sought. 2. Measure independent blanks and low-level samples through preparation and detection. 3. Assess blank distribution, calibration slope and relevant false-decision criteria. 4. Validate a concentration above which quantitation meets predefined bias and precision needs. 5. Report low results according to the method rule, without calling non-detection absence.

Visual explanation

Draw overlapping bell-like distributions for blank response and a low-concentration sample response. Mark a decision threshold between them and show that overlap creates false-positive and false-negative possibilities. Draw a calibration line with a steep and a shallow slope; the same vertical noise corresponds to a smaller concentration interval on the steep line.

Real-world analogy

Hearing a whisper in a noisy room has two questions: can you tell that someone spoke, and can you repeat their exact words reliably? The first resembles detection; the second resembles quantification. Turning up microphone sensitivity helps only if it does not amplify background noise and competing voices equally.

Real-world example

A water laboratory measures a pesticide by LC–MS. The instrument detects a weak ion signal in a sample, but a matrix-matched low-level control shows high variation. The result can be flagged as detected below the validated quantification limit. Reporting “0.001237 mg L⁻¹” would imply unjustified precision despite the many digits the software can produce.

Why?

Why does blank variability influence LOD? A signal only slightly above one blank reading may be ordinary background fluctuation. The broader the blank distribution, the higher a threshold must be to keep false detections acceptably rare, all else equal.

Common misconception

“Below LOD means zero concentration” is false; the analyte may be present but indistinguishable from method background. “A universal 3σ or 10σ formula defines every LOD or LOQ” ignores calibration, replicate design, matrix and decision risks.

Worked example

Suppose calibration slope is 50 signal units per mg L⁻¹ and blank standard deviation is 5 signal units. The blank scatter corresponds to 5/50 = 0.10 mg L⁻¹. A simple three-times-noise teaching estimate would be around 0.30 mg L⁻¹ for detection, while a ten-times-noise estimate would be around 1.0 mg L⁻¹ for quantification. These are illustrative approximations; a validated method must state and test its actual rules.

Quick check

1. Can a numerical value below LOQ but above LOD be treated as a fully reliable concentration? Answer: Usually not. Presence may be supported, but the method has not met its stated criteria for quantitative accuracy or precision at that level.

Exam focus

Define LOD and LOQ by different decisions, relate blank noise and slope to concentration limits, and state the method/matrix basis. Avoid interpreting a nondetect as analyte-free or a software-generated decimal as validated quantitation.

Advanced insight

Detection is a decision-theory problem involving rates of false positives and false negatives. A stricter threshold reduces false positives but can increase false negatives. The suitable balance depends on the use of the result, and the calibration model's uncertainty near the blank can matter as much as simple signal noise.

Summary

LOD concerns distinguishing a low analyte level from background; LOQ concerns measuring it with acceptable performance. Both depend on the complete method, blank variation, calibration sensitivity and sample matrix. Low-level results should be reported according to validated decision rules rather than equated with exact zero or unsupported precision.

Practice questions

1. If calibration slope doubles while blank noise stays the same, what happens to a simple noise/slope concentration limit? Answer: It is approximately halved, because the same signal variation corresponds to a smaller concentration difference.

2. Why process matrix blanks rather than only pure solvent blanks? Answer: Real matrix preparation can add contamination or interference and change noise, so solvent-only blanks may overstate method capability.

3. What does “not detected” mean under a defined analytical method? Answer: The measured signal did not satisfy that method's presence decision criterion; it does not establish that the analyte concentration is exactly zero.