The Felkin-Anh Model: Predicting Facial Selectivity at a Carbonyl

From Cram's rule to a conformational model

The problem is old and concrete: add a nucleophile to a carbonyl bearing a stereocenter one bond away, and you generate a new stereocenter. Which diastereomer dominates? In 1952 Donald Cram offered the first predictive picture — Cram's rule — by placing the carbonyl oxygen anti to the largest α-substituent (the "open-chain" model) and letting the nucleophile approach past the smallest group. It worked often enough to be useful, but it rested on a questionable premise: that the reactive conformation has the carbonyl eclipsed by the medium group, which is not the ground-state minimum.

In 1968, Marc Chérest, Hugh Felkin, and Nicole Prudent (Tetrahedron Letters 9, 2199) reframed the geometry. Their central insight: in the transition state, the α-substituents are staggered — not eclipsed — relative to the C=O and the forming Nu–C bond (torsional strain about the C(α)–C(carbonyl) bond is minimized), and the largest group (L) sits roughly perpendicular to the carbonyl plane, anti to the trajectory of attack. Torsional strain between the forming C–Nu bond and the α-C–H/C–R bonds, not simple steric bulk beside the oxygen, is what the molecule minimizes. This staggered picture immediately explained why selectivities were higher with bulkier nucleophiles and bulkier L groups than Cram's model predicted.

Felkin's model made a bold, falsifiable claim about conformation. But it left an obvious question: why should the largest group prefer the perpendicular position, and why should heteroatoms behave anomalously? Answering that required orbitals.

Anh and Eisenstein: the orbital backbone

In 1977, Nguyên Trong Anh and Odile Eisenstein (Nouveau Journal de Chimie 1, 61) subjected the Felkin conformations to ab initio and extended-Hückel calculations on model systems such as the addition of hydride to 2-chloropropanal. Two results transformed the rule into what we now call the Felkin-Anh model.

First, they showed the nucleophile does not attack head-on. It follows the Bürgi-Dunitz trajectory — the Nu···C=O angle opening to about 105–107° (originally reported as 105 ± 5°, so it is a soft, substrate-dependent range rather than a fixed constant), tilted away from the oxygen and toward the R group on the other side — established by Bürgi, Dunitz, and Shefter from crystallographic surveys of amino-ketone geometries (1973–74). Because attack is offset, the group placed anti to the incoming nucleophile experiences the least interaction, so the bulkiest substituent belongs there.

Second, and more subtly, Anh identified an electronic preference. The best stabilization occurs when the σ* antibonding orbital of the C–L bond in the perpendicular position is aligned parallel to the forming Nu–C bond (and to the π* of the carbonyl). This hyperconjugative overlap — donation of the incipient σ(Nu–C) electron density into σ*_C–L — lowers the transition-state energy. The consequence is decisive: the group with the best σ-acceptor ability, not merely the largest, prefers the perpendicular site. For an ordinary alkyl substrate size and σ*-acceptor ranking coincide, but for an α-heteroatom substrate they can point in different directions — which is exactly where the model earns its keep.

Polar Felkin-Anh: when electronics beat sterics

The most important refinement is the polar Felkin-Anh rule for substrates bearing an α-electron-withdrawing group — a chloride, fluoride, alkoxide, or other electronegative substituent. Because σ*_C–X for an electronegative X is a low-lying, excellent acceptor, that group is placed perpendicular regardless of its steric size, so its σ* can align with the forming bond. The nucleophile then attacks anti to the C–X bond.

• Steric Felkin-Anh: perpendicular position taken by the largest alkyl/aryl group (L); the drive is torsional/steric.
• Polar Felkin-Anh: perpendicular position taken by the best σ*-acceptor (α-EWG); the drive is electronic hyperconjugation.

This is not a cosmetic distinction. For an α-alkoxy aldehyde such as 2-(benzyloxy)propanal, the polar Felkin-Anh model (OBn perpendicular, attack anti to OBn) predicts the 1,2-anti product under non-chelating conditions (the newly formed carbinol center and the α-benzyloxy center bear a 1,2-relationship; the '3,4' label is only appropriate when this aldehyde is part of a longer numbered chain, such as an aldol adduct). Switch to a chelating Lewis acid — MgBr₂, ZnCl₂, or a titanium(IV) species — and the oxygen instead bridges to the metal, locking a rigid five-membered chelate that inverts the facial bias to give the chelation (anti-Felkin) product. The two models thus predict opposite diastereomers, and the observed sense of induction becomes a direct readout of whether chelation is operative. Reetz's systematic studies in the 1980s made this dichotomy a practical stereocontrol switch in synthesis.

A worked example: hydride reduction of 2-phenylpropanal

Take 2-phenylpropanal, (±)-PhCH(CH₃)CHO, and reduce it with LiAlH₄. The α-carbon carries H, CH₃, and C₆H₅. Rank the substituents: the phenyl group is the bulkiest, so on straightforward steric Felkin-Anh grounds it takes the perpendicular position (a modest hyperconjugative contribution from its π system may reinforce this, but C(sp²)–Ph is not a low-lying σ*-acceptor like C–Cl or C–OR, so size is the primary driver here). The medium group (CH₃) sits anti-periplanar-ish to the carbonyl oxygen region, and the small group (H) points inward toward the incoming hydride, which arrives anti to phenyl at the Bürgi-Dunitz angle.

The result is a preference for the Felkin diastereomer of 2-phenyl-1-propanol on the order of ~2:1 to 3:1 — modest, because a lone hydride is small and torsional discrimination is limited. Now swap to a bulkier hydride. With a sterically demanding reagent such as L-Selectride (lithium tri-sec-butylborohydride) or with a Grignard delivering a large R group, the same conformational bias is enforced far more strictly and diastereomeric ratios climb well past 10:1. This monotonic "bigger nucleophile → higher dr" trend is a fingerprint of Felkin-Anh control and a direct test that distinguishes it from a chelation pathway, where dr often depends instead on Lewis-acid loading and temperature.

Contrast this with the classic Cornforth substrate 2-chloropropanal. There the low-lying σ*_C–Cl demands the perpendicular slot; hydride attacks anti to chlorine, and the polar Felkin-Anh prediction matches experiment even though Cl is not the bulkiest group. Placing chlorine perpendicular by size alone would fail; placing it there by σ*-acceptor strength succeeds.

Limits, competing models, and honest caveats

The Felkin-Anh model is a transition-state argument, so it lives or dies by the Curtin-Hammett situation: the reactive rotamers interconvert far faster than they react, and product ratios reflect ΔΔG‡ between competing transition states, not ground-state populations. When that assumption breaks — very low temperature, restricted rotation, or a rigid ring — the naive prediction can fail.

• Cornforth vs. Felkin-Anh: For α-halo and α-alkoxy carbonyls, the older Cornforth model (1959) places the C–X dipole anti to the C=O dipole to minimize electrostatic repulsion, then attacks anti to X. This frequently predicts the same major product as polar Felkin-Anh, and which effect truly dominates was debated for decades. Modern computations (e.g., work by Houk, Cee, and Evans in the 2000s) indicate that for many α-halo/alkoxy aldehydes the electrostatic (Cornforth-type) contribution can rival or exceed the hyperconjugative one — so "polar Felkin-Anh" is best read as a shorthand for a transition state with the EWG perpendicular, whatever its precise physical origin.
• Chelation override: As noted, chelating metals invert selectivity, so the model must always be applied with the reaction conditions in view.
• 1,3- and higher induction: Felkin-Anh addresses 1,2-relationships. Remote (β, 1,3) stereocenters follow Evans-type electrostatic and dipole models rather than α-hyperconjugation.

It is also worth stating plainly that the perpendicular σ*-alignment picture, while widely taught, is an approximation to a continuum of transition-state geometries. The essential, robust content is staggered rather than eclipsed approach along the Bürgi-Dunitz vector; the precise weighting of steric, torsional, and electronic terms is substrate-specific and is where DFT now does the quantitative bookkeeping.

Why it matters: the workhorse of acyclic stereocontrol

Before robust catalytic asymmetric methods matured, essentially every total synthesis that strung together stereocenters on an open chain — polyketides, polypropionates, macrolides — leaned on substrate-controlled additions rationalized by Felkin-Anh and its polar and chelation variants. Choosing a protecting group that can or cannot chelate (a benzyl ether versus a silyl ether, say) became a deliberate lever to flip diastereoselectivity, precisely because the two models predict opposite faces.

The model also sharpened how chemists think about reagent-controlled versus substrate-controlled reactions. When a chiral reagent's intrinsic facial bias reinforces the substrate's Felkin-Anh preference, you get a matched pair and excellent selectivity; when they oppose, a mismatched pair, and the outcome hangs on which effect is stronger. Masamune and Roush codified this matched/mismatched language in the 1980s, and Felkin-Anh supplies the substrate half of the ledger.

Today the ideas live on in mechanistic reasoning far beyond hydride reductions: aldol additions, allylations, cuprate 1,2-additions, and enzymatic ketoreductions are all discussed through the same lens of a staggered, Bürgi-Dunitz approach with the best σ-acceptor perpendicular. Felkin and Anh did not merely give chemists a mnemonic — they connected a synthetic prediction to frontier-orbital overlap, and that bridge from stereochemical outcome to electronic structure is the model's lasting contribution.

Felkin-Anh vs. the Cornforth / chelation alternatives for predicting the major diastereomer of nucleophilic addition to an α-substituted carbonyl.

Feature	Felkin-Anh	Cornforth / Chelation
Controlling factor	Hyperconjugative σ*C–L stabilization + minimized Nu/R eclipsing	Cornforth: C=O and C–X dipoles anti; Chelation: metal bridges C=O and α-heteroatom
Perpendicular group	Largest σ*-acceptor / bulkiest substituent (L)	Cornforth: the α-electronegative X; Chelation: heteroatom locked into a 5-membered ring
Best applied to	α-alkyl and α-EWG (polar Felkin-Anh) aldehydes/ketones	Cornforth: α-halo/alkoxy w/o chelation; Chelation: α-OBn/OMOM + Zn, Mg, Ti Lewis acids
Nucleophile attacks	anti to L, along Bürgi-Dunitz trajectory	anti to X (Cornforth) or anti to the smaller face of the rigid chelate
Diastereomer favored	Usually '1,2-anti' / Felkin product (a '3,4' label applies only when the aldehyde is embedded in a longer chain, e.g. an aldol adduct)	Often the opposite ('anti-Felkin' / chelation) product