COSY 2D NMR: Reading Off-Diagonal Cross-Peaks for Connectivity

What COSY Is and Where It Applies

COSY converts a 1D ¹H spectrum into a two-dimensional map with two chemical-shift axes, F1 and F2, plotting the same proton frequencies on each. The result is a square with two features:

• Diagonal peaks — one for every resonance, lying on the F1 = F2 line; this is just the ordinary 1D spectrum projected onto the diagonal.
• Cross-peaks — off-diagonal spots that appear only between protons connected by a scalar J-coupling.

Because vicinal (³J) and geminal (²J) couplings dominate, a cross-peak at (δ_A, δ_X) almost always means A and X sit on adjacent, bonded atoms — typically two or three bonds apart. That single fact is enormously powerful: it lets a chemist walk along a carbon skeleton, hopping from one CH group to its neighbor, reconstructing an entire spin system. COSY is the workhorse of small-molecule structure elucidation, natural-product characterization, and the first assignment pass in protein and nucleic-acid NMR, where overlapping multiplets make 1D analysis hopeless.

The Pulse Sequence and How Cross-Peaks Arise

The basic COSY sequence is disarmingly short: relaxation delay – 90° – t1 – 90° – acquire(t2). Two 90° pulses bracket an incremented evolution time t1.

• Preparation: the first 90° pulse tips equilibrium magnetization into the transverse plane.
• Evolution (t1): each spin precesses at its own chemical shift and its coupling evolves; the signal becomes labeled (frequency-encoded) by t1.
• Mixing (2nd 90°): this pulse is the crux. For J-coupled spins, it transfers coherence from spin A to its coupling partner X — antiphase magnetization on A becomes observable magnetization on X.
• Detection (t2): the signal now oscillates at X's frequency during t2 but carries A's frequency memory from t1.

The experiment is repeated for hundreds of incremented t1 values, and a double Fourier transform (over t1 and t2) builds the 2D map. Magnetization that stayed on the same spin gives a diagonal peak; magnetization transferred A→X gives a cross-peak at (δ_A, δ_X), with its mirror at (δ_X, δ_A). No coupling, no transfer, no cross-peak — that is why the off-diagonal spots are the connectivity readout.

Key Quantities and a Worked Reading

The information in a cross-peak lives in two coordinates and its fine structure. Consider ethyl acetate, CH₃–CH₂–O–C(=O)–CH₃:

• The triplet CH₃ at δ 1.26 ppm and the quartet OCH₂ at δ 4.12 ppm are vicinal, ³J ≈ 7.1 Hz. COSY shows a strong cross-peak at (1.26, 4.12) and its symmetry partner (4.12, 1.26).
• The acetyl CH₃ at δ 2.04 ppm is a singlet — no coupled neighbor — so it gives a diagonal peak but no cross-peaks. Its isolation is itself diagnostic.

Vicinal coupling follows the Karplus relation: ³J = A·cos²θ + B·cosθ + C, where θ is the H–C–C–H dihedral angle and A, B, C are empirical constants (roughly A ≈ 7, B ≈ −1, C ≈ 5 Hz for the classic curve). This predicts ³J ≈ 9–12 Hz at θ = 180° (anti), a minimum near 0–2 Hz at θ ≈ 90°, and ~8–11 Hz at θ = 0° (eclipsed). Cross-peak intensity scales with sin(πJt1)·sin(πJt2), so very small couplings (<1 Hz) produce weak or missing cross-peaks — a genuine limit of the method.

How It's Measured and Run in Practice

A modern COSY on a 400–600 MHz spectrometer is fast — often 5–15 minutes for a concentrated small molecule. Practical parameters and choices:

• Spectral windows: typically the same ~10 ppm in F1 and F2; the resulting spectrum is symmetric about the diagonal, and symmetrization is applied to reject t1 noise (streaks along F1).
• Increments: 128–512 t1 points; more increments buy F1 resolution at the cost of time. Digital resolution in F1 is usually coarser than F2, so cross-peak fine structure is read along F2.
• Gradient selection now replaces lengthy phase cycling, giving clean spectra in a single scan per increment.

Interpretation is a walk: pick a resolved proton on the diagonal, trace horizontally (or vertically) to a cross-peak, then drop to the diagonal to find its coupled partner, and repeat. Chaining these hops assigns a full spin system — an ethyl, an isopropyl, a sugar ring, an amino-acid side chain. Magnitude-mode COSY (the classic) is robust but has broad, star-shaped peaks; DQF-COSY is preferred when you need sharp, phase-sensitive peaks and want to measure J values from the antiphase multiplet splittings.

COSY Versus Its Relatives

COSY belongs to a family of 2D experiments, and choosing the right one hinges on what a cross-peak means:

• DQF-COSY adds a third 90° pulse and a double-quantum filter. It suppresses singlets (uncoupled protons and solvent), sharpens the diagonal from broad dispersion to absorption lineshape, and lets you extract accurate J couplings — at a cost of roughly half the sensitivity.
• TOCSY replaces the single mixing pulse with an isotropic mixing (spin-lock) period, relaying magnetization along an entire coupled network. One TOCSY strip reveals a whole spin system, not just nearest neighbors.
• NOESY looks identical in layout but its cross-peaks report through-space dipolar proximity (<5 Å), not bonds — essential for stereochemistry, but a trap if confused with COSY.
• HSQC / HMBC are heteronuclear: they correlate ¹H to ¹³C over one bond (HSQC) or two-to-three bonds (HMBC), bridging quaternary carbons and heteroatoms that ¹H-only COSY cannot cross.

In a real assignment, these are used together: HSQC pins each H to its C, COSY chains adjacent CH units, and HMBC/NOESY stitch fragments across gaps.

History, Significance, and Limits

COSY is the experiment that launched multidimensional NMR. Jean Jeener proposed the two-pulse, double-Fourier idea in a 1971 lecture at the Ampère Summer School in Yugoslavia; it was first implemented and published by Walter Aue, Enrico Bartholdi, and Richard R. Ernst in 1976 ("Two-dimensional spectroscopy. Application to nuclear magnetic resonance," J. Chem. Phys. 64, 2229). Ernst received the 1991 Nobel Prize in Chemistry for developing high-resolution 2D NMR; the same conceptual leap underpins Kurt Wüthrich's Nobel-winning protein structure work, which relied on COSY/NOESY networks to assign resonances.

Its limits are real: cross-peaks vanish when J is very small (long-range or near-orthogonal Karplus geometry), overlap on the crowded diagonal can hide short-range correlations, and it says nothing about carbons bearing no protons. It also cannot cross heteroatoms or quaternary centers — a saturated chain interrupted by an O, N, or C(=O) breaks the COSY walk, which is exactly where HMBC takes over. Still, for tracing who-is-bonded-to-whom among protons, the off-diagonal cross-peak remains the single most direct answer in all of NMR.

COSY versus its close 2D NMR cousins — what each cross-peak proves

Experiment	Correlates	Cross-peak reports	Typical use
COSY	¹H ↔ ¹H	²J/³J scalar coupling (adjacent H)	Trace spin systems, vicinal connectivity
DQF-COSY	¹H ↔ ¹H	Same, singlets suppressed, pure absorption	Clean diagonal, measure J from fine structure
TOCSY	¹H ↔ ¹H	Whole coupled network (relayed)	Identify entire spin system at once
NOESY	¹H ↔ ¹H	Through-space (<5 Å), NOT bonds	Stereochemistry, conformation, folding
HSQC	¹H ↔ ¹³C (1 bond)	Which C bears which H	C–H assignment, backbone building
HMBC	¹H ↔ ¹³C (2–3 bond)	Long-range C–H, across quaternaries	Link fragments over heteroatoms