Common 2D NMR Artefacts

Introduction

A 2D NMR contour map seems pictorially definite: one spot appears to connect two resonances. Yet a spot can be a true transfer, an expected diagonal signal, a second molecule at the same coordinates, or a processing defect. Reliable interpretation asks three questions of every important feature: is it physically allowed by the pulse sequence, is it reproducible above local noise, and does it agree with independent observations? This page explains common failure modes before they turn into false structures.

Core explanation

In a homonuclear COSY spectrum, the diagonal contains a response from each proton to itself. It often resembles a one-dimensional proton spectrum drawn from the lower left to upper right. A true COSY cross-peak appears away from that diagonal and usually has a symmetric counterpart reflected across it, because the two proton axes describe the same set of resonances. Strong diagonal tails can spill into nearby weak cross-peaks, especially when two shifts are similar. The diagonal itself is not evidence that a proton couples to another proton.

Peak overlap is a deeper ambiguity. Two protons with almost identical shifts can occupy one coordinate on a proton axis while having different carbon partners in HSQC. A COSY cross-peak at that shared shift may belong to either proton, or both. If a one-dimensional multiplet is crowded, integration and splitting can be misleading as well. Comparing the carbon axis of HSQC, changing solvent or temperature, and recording a selective experiment can separate assignments. The map is a projection of many molecules and nuclei onto limited coordinates, not an atom-labelled photograph.

In frequency-switched or phase-sensitive experiments, correct phasing matters. Absorptive peaks should be compact; dispersive tails or mixed positive and negative lobes may create phantom contours near strong peaks. In edited HSQC, an apparent change of sign should only be interpreted with the known pulse sequence, phase convention and reference. Baseline roll and aggressive apodisation may alter whether weak spots cross the contour threshold. A low contour setting reveals noise as well as chemistry; a high one hides weak but genuine correlations. Inspect the raw or minimally processed data when a faint peak is decisive.

Vertical stripes along the indirect dimension often come from t1 noise : a strong signal changes amplitude or phase between successive increments because of temperature drift, pulse instability, convection or sample degradation. Unlike a genuine isolated cross-peak, the stripe persists over a range of the indirect axis. Truncation of short time-domain data can produce oscillating side lobes around a strong line. Limited spectral width can fold a resonance back into the displayed range, called aliasing, making the apparent chemical shift wrong unless the sweep width and carrier are considered. Solvent and water suppression can create additional nearby distortions.

Finally, not every unexpected peak is instrumental. Strong coupling, long-range scalar couplings, chemical exchange, and NOE transfer can produce real signals that violate a simplified textbook prediction. In HMBC, correlations are filtered and optimised for a range of long-range J values, not for an invariant number of bonds. In NOESY, exchange peaks and cross-relaxation can coexist. A primary ACS study of two-dimensional NMR acquisition discusses how experiment design and processing create characteristic artefacts; a structural assignment should be tested against its actual acquisition settings.

Step-by-step reasoning

1. Identify the two axes and the transfer pathway before naming a spot. 2. Locate diagonal responses and expected symmetry in a homonuclear spectrum. 3. Check whether the feature is localised or part of a vertical streak or tail. 4. Compare independent 1D, HSQC, COSY or HMBC evidence for the proposed assignment. 5. Inspect processing phase, contour threshold and sweep-width settings if a weak correlation changes the structure.

Visual explanation

Picture a COSY square with a bright diagonal. A real pair of coupled protons gives two matched off-diagonal islands on opposite sides of it. A diagonal tail looks like a smudge continuous with a bright diagonal island. A t1 noise artefact resembles a line running through several unrelated positions. The shapes and symmetry give clues before any bond is drawn.

Real-world analogy

Imagine a photograph of a road at night. A car is a compact light; a long exposure smear is not several cars. Two cars at the same location can merge into one light. Changing exposure or taking a second photograph helps distinguish the scene. NMR processing and complementary experiments serve the same role.

Real-world example

When a natural-product candidate relies on a weak HMBC spot to join two fragments, the spectroscopist checks another proton-to-carbon path, a different HMBC optimisation or a one-dimensional selective experiment. The aquatolide reassessment in the Journal of Organic Chemistry illustrates why ordinary proton data and well-supported correlations must agree, and why absence of an HMBC signal alone is inconclusive.

Why?

Why do artefacts matter more in the second dimension? A dataset is assembled from many acquisitions at incremented indirect times. Drift between those acquisitions can map into the extra frequency axis, and display thresholds can make a continuous imperfection look like separate cross-peaks.

Common misconception

"A cross-peak above noise must be chemically meaningful." Reproducibility and consistency matter too. Some errors can produce strong, repeatable-looking spots, while real long-range correlations may be weak or absent because the experiment is tuned to a different coupling range.

Worked example

Question: A COSY map has peaks at proton shifts 2.1 and 4.0 ppm and a faint contour at the coordinate (2.1, 4.0). No matching contour appears at (4.0, 2.1); the faint feature is part of a vertical stripe through many y-values. Should the two protons be assigned as coupled?

Reasoning: An ordinary COSY cross-peak between two resolved protons should usually have a mirror partner. A stripe rather than a compact spot suggests t1 noise from an unstable strong signal. The missing mirror and stripe shape together make the feature suspect. Check raw scans, process with different thresholds and seek independent evidence such as multiplet coupling before declaring a bond network.

Answer: No coupling assignment is justified from this feature alone.

Quick check

1. Does a diagonal COSY peak link two different protons? Answer: No; it is the self-response of one resonance. A streak extending through many coordinates on the indirect frequency axis instead suggests t1 noise.

Exam focus

State the expected geometry of a genuine homonuclear cross-peak, then explain overlap, phasing, t1 noise and aliasing separately. If a proposed structure hinges on one suspect contour, describe a realistic independent check rather than simply lowering the contour threshold.

Advanced insight

In nonuniformly sampled experiments, the reconstruction algorithm and sampling schedule can generate side lobes that resemble weak correlations; inspecting acquisition and reconstruction settings is part of validation. Strong coupling can make COSY cross-peak patterns unusually complex even with perfect processing. These cases require simulation or an alternative experiment rather than automatic deletion of every unusual feature.

Summary

Diagonal peaks, overlap, streaks, phase errors, truncation and aliasing can distort 2D NMR maps. Shape, symmetry, raw-data inspection and agreement with other experiments distinguish many artefacts from genuine transfer signals. A structural claim should rest on a consistent network of evidence rather than one attractive contour.

Practice questions

1. What is the expected partner of a COSY cross-peak at (δA, δB)? Answer: A counterpart at (δB, δA), subject to overlap and display sensitivity. 2. Why can HSQC help resolve an ambiguous proton overlap? Answer: Protons with similar shifts can be attached to carbons with different shifts and therefore separate along the carbon axis. 3. What acquisition detail can reveal a folded resonance? Answer: The spectral width and carrier position, together with the true one-dimensional shift. 4. Why should one not infer no bond from a missing HMBC spot? Answer: The relevant long-range coupling may be too weak or outside the delay range optimised by that experiment.