The Stevens Rearrangement

What the Stevens rearrangement does

Start with a quaternary ammonium salt — a nitrogen carrying four groups and a positive charge — where one of the carbons directly attached to nitrogen bears an acidifying substituent (a carbonyl, an aryl, a nitrile). Treat it with a strong base and something surprising happens: an entire group unhooks from the nitrogen and reattaches to the carbon right next to it. A four-coordinate ammonium becomes a three-coordinate tertiary amine, and a new C-C bond appears where a C-N bond used to be.

The trick is a two-part sequence:

1. Make the ylide. The base removes the most acidic proton — the one on the carbon α to the positively charged nitrogen. The result is an ammonium ylide: a carbanion sitting immediately next to the formal positive charge on nitrogen (a 1,2-zwitterion, N⁺-C⁻).
2. Migrate. One of the three remaining groups on nitrogen (the "migrating group") transfers from N to the carbanion. The nitrogen's positive charge is neutralized as its lone pair is restored, and the carbanion is neutralized as it forms the new bond. Net result: a [1,2]-shift.

    R'                       R'
    |                        |
 R–N⁺–CH₂–EWG   ──base──►  R–N⁺–C⁻(H)–EWG   ──[1,2]──►   R–N–CH(R')–EWG
    |         (–H⁺)           |         (ylide)              (rearranged amine)
    R''                       R''
                          migrating group R' jumps N → C

The exact same choreography works on sulfonium ylides (S⁺-C⁻ instead of N⁺-C⁻), giving a rearranged sulfide. Nitrogen and sulfur are the two classic Stevens heteroatoms; the sulfonium version is often cleaner because sulfur ylides are easier to form and less prone to side reactions.

The mechanism: a caged radical pair, not a concerted shift

Here is where the Stevens rearrangement earns its place in every advanced mechanism course. A naïve student draws the [1,2]-shift as a single concerted arrow — the migrating group slides across, keeping its bond to carbon the whole way. That drawing is wrong, and the reason is symmetry.

A concerted, thermal, suprafacial [1,2]-shift of a carbon group with retention is forbidden by the Woodward-Hoffmann rules. The Stevens shift is an anionic [1,2]-migration — the group moves onto a carbanion, so it is a four-electron process (unlike a carbocation 1,2-shift, which is only two electrons and is allowed suprafacial with retention). For this four-electron case, a suprafacial thermal shift would have to proceed with inversion at the migrating carbon to be allowed — geometrically impossible for a small migrating carbon. So the molecule takes a stepwise, homolytic route instead:

1. Homolysis. The C-N bond to the migrating group breaks homolytically — one electron to each fragment. This gives a radical pair: a stabilized α-amino/enolate-type radical on the ylide carbon, and a migrating-group radical (benzyl, allyl, phenacyl — always one that gives a well-stabilized radical).
2. Solvent cage. The two radicals do not fly apart. They are born touching, inside a "cage" of solvent molecules, and they recombine within that cage — typically in under a nanosecond — before either can diffuse away or tumble much.
3. Recombination. The migrating radical adds to the carbanion-radical carbon, forming the new C-C bond. The nitrogen keeps its lone pair; the product is the neutral rearranged amine.

  ylide:    N⁺–C⁻       (zwitterion)
                │  homolysis of N–CH₂Ph bond
                ▼
  radical pair: [ N–C•   •CH₂Ph ]   ← both inside one solvent cage
                │  in-cage recombination (< 1 ns)
                ▼
  product:    N–C(–CH₂Ph)         (new C–C bond)

Two independent lines of evidence nail the radical-pair mechanism:

• CIDNP. Running the reaction inside an NMR magnet shows chemically induced dynamic nuclear polarization — anomalously enhanced absorption and emission lines in the product spectrum. CIDNP is a fingerprint that can only arise from a radical-pair intermediate whose spin state is sorted by the magnetic field. A concerted or ionic path produces no CIDNP. These 1970s experiments (Lepley, Schöllkopf, and others) are the decisive proof.
• Partial retention of configuration. Put a stereocenter on the migrating carbon and the product comes out with 70-95% retention — not 100% (which concerted would demand) and not 0% (which a long-lived free radical would give). Partial retention is exactly what a tight, short-lived cage predicts: recombination usually beats racemization, but not always.

Reagents, bases, and conditions

The single requirement is a base strong enough to deprotonate α to the onium center. How strong depends entirely on how acidic that proton is:

• Activated substrates (phenacyl, ester-, or aryl-stabilized α-carbon). The classic Stevens substrates. The α-proton is acidic enough (pKa often 18-22) that ordinary hydroxide works — aqueous or alcoholic NaOH / KOH, sometimes just on gentle warming. Stevens' own 1928 examples were phenacylbenzyldimethylammonium bromide + dilute NaOH.
• Less-activated substrates. Bare alkyl α-carbons need much stronger, non-nucleophilic bases: NaH, KOtBu, n-BuLi, PhLi, or NaNH₂. Lithium amides and alkyllithiums in THF or ether at low temperature are common for modern preparative work.
• Solvent. Protic solvents (H₂O, EtOH) for the easy cases; aprotic ethereal solvents (THF, Et₂O, DME) or liquid ammonia for the strong-base cases. The solvent cage matters mechanistically, but the reaction is not fussy about which inert solvent supplies it.
• Temperature. Room temperature to gentle reflux. Notably, raising the temperature tilts the outcome toward Stevens ([1,2]) and away from the competing Sommelet-Hauser ([2,3]) — see the comparison below.

The migrating group is not a free variable: it must be one that forms a stabilized radical, because the rate-limiting step is homolysis. Benzyl, substituted benzyl, allyl, phenacyl, and other resonance-stabilized carbons migrate readily. Simple methyl or unactivated primary alkyl groups migrate poorly or not at all — there is no stabilized radical to make.

Scope, selectivity, and stereochemistry

Three selectivity questions decide what you get:

• Which group migrates? The one that gives the most stabilized radical. When several benzylic groups are present, electron-poor benzyls (which give more stable radicals via captodative or radical-stabilizing substituents) tend to migrate preferentially. A p-nitrobenzyl migrates faster than an unsubstituted benzyl.
• Retention at the migrating carbon. As above, 70-95% retention because the cage recombination outruns radical tumbling. This partial retention is the signature that distinguishes a caged radical pair from a free radical.
• Retention at the ylide (carbanion) carbon. If the carbanion carbon is a stereocenter, its configuration is also largely retained — the pyramidal carbanion/α-amino radical does not fully planarize before recombination in the cage.

The asymmetric Stevens rearrangement is possible but hard, precisely because a free radical is achiral: any enantiocontrol has to be imposed by a chiral environment (a chiral base, a chiral auxiliary on nitrogen, or a chiral counterion) that biases which face recombines. Chiral-ammonium-ylide Stevens rearrangements with modest-to-good ee have been reported, but the reaction is fundamentally more stereochemically slippery than a concerted sigmatropic shift like the Sommelet-Hauser.

Stevens [1,2] vs Sommelet-Hauser [2,3]

The Stevens rearrangement almost never travels alone. The very same benzylic ammonium ylide can instead undergo a Sommelet-Hauser [2,3]-sigmatropic rearrangement, migrating onto the ortho carbon of the aryl ring rather than onto the carbanion. Controlling which pathway wins is the central practical problem.

	Stevens rearrangement	Sommelet-Hauser rearrangement
Formal shift	[1,2]	[2,3]-sigmatropic
Mechanism	Homolytic — caged radical pair	Concerted, pericyclic (allowed)
Bond made	C-C onto the carbanion carbon	C-C onto the ortho ring carbon
Woodward-Hoffmann	Concerted [1,2] is forbidden → goes radical	[2,3] is allowed → stays concerted
Product	Rearranged α-branched benzylamine	ortho-(aminomethyl) toluene (re-aromatized)
Stereo at migrating C	Partial retention (70-95%)	High, predictable (suprafacial)
Favored by	Higher temperature, stronger base	Lower temperature, NaNH₂ / liquid NH₃
CIDNP signal	Yes (radical pair)	No (concerted)
Ring dearomatization	Not involved	Transient (aromaticity broken then restored)

The rule of thumb: cold + weak-ish amide base → [2,3] Sommelet-Hauser; warm + strong base → [1,2] Stevens. The Sommelet-Hauser is the kinetic product — its concerted [2,3] transition state has the lower activation barrier, so it wins under kinetic control at low temperature. The Stevens [1,2]-product is generally the more stable, thermodynamic product, so heating lets the selectivity flip toward Stevens.

Worked example: the original 1928 Stevens substrate

Stevens' defining experiment: phenacylbenzyldimethylammonium bromide + dilute aqueous NaOH.

   Substrate:  [ PhC(=O)–CH₂ – N⁺(CH₃)₂ – CH₂Ph ] Br⁻

   Step 1 (deprotonate the acidic phenacyl CH₂, pKa lowered by C=O):
        PhC(=O)–CH⁻ – N⁺(CH₃)₂ – CH₂Ph        (the ylide)

   Step 2 ([1,2]-shift — benzyl migrates N → C as a caged radical pair):
        PhC(=O)–CH(CH₂Ph) – N(CH₃)₂           (rearranged α-amino ketone)

• Why the phenacyl proton? The carbonyl next to the α-carbon drops its pKa dramatically, so even dilute NaOH generates the ylide (enolate-stabilized carbanion) cleanly.
• Why does benzyl migrate? Of the three groups on nitrogen (two methyls and one benzyl), only the benzyl forms a well-stabilized radical. The methyls stay put; the benzyl jumps.
• The product is a rearranged tertiary amine — 2-(dimethylamino)-1,3-diphenylpropan-1-one, an α-amino ketone (PhCO–CH(NMe₂)–CH₂Ph) — with the benzyl now on carbon instead of nitrogen. The dimethylamino group survives untouched.
• Diagnostic: run it in an NMR magnet and you see CIDNP in the product — the experimental fingerprint that later confirmed the radical pair.

Variants and real applications

• Sulfonium Stevens. Sulfonium ylides (from a sulfonium salt + base, or from a carbene + sulfide) undergo the identical [1,2]-shift to give rearranged sulfides. This is the workhorse version in synthesis because sulfur ylides are easy to generate. Metal-carbenoid routes (Cu, Rh) generate the ylide in situ from a diazo compound and a sulfide, then let it rearrange — a powerful C-C bond-forming tactic.
• Metal-catalyzed carbenoid Stevens. Rh(II) or Cu catalysts decompose an α-diazo carbonyl in the presence of a tertiary amine or sulfide. The metal carbenoid inserts to form the ammonium/sulfonium ylide, which then does the Stevens [1,2]-shift. This packages ylide generation and rearrangement into one pot and is widely used for ring-expansions and C-C bond construction.
• Ring expansion / ring contraction. When the migrating group and the ylide carbon are part of a ring, the [1,2]-shift changes the ring size. Cyclic ammonium and sulfonium ylides undergo Stevens ring expansions used to build medium-sized nitrogen and sulfur heterocycles.
• Total synthesis. Stevens-type [1,2]-shifts (often via metal carbenoids) have been used as key C-C bond-forming steps in alkaloid and natural-product syntheses where a quaternary or benzylic stereocenter must be set adjacent to nitrogen.
• The oxygen cousin. The [1,2]-Wittig rearrangement is the direct oxygen analogue: deprotonate α to an ether oxygen and a group migrates from O to the carbanion carbon (O→C) through the same solvent-caged radical pair. The related Meisenheimer rearrangement instead runs on an amine N-oxide, shifting a group from nitrogen onto oxygen (N→O). Together with the [2,3]-variants (Sommelet-Hauser, [2,3]-Wittig), the Stevens sits in a broad family of base-induced ylide/carbanion rearrangements.

Limitations and side reactions

• Sommelet-Hauser competition. For benzylic ammonium ylides, the [2,3] pathway is the ever-present rival. If you want clean Stevens product, push temperature and base strength; if you want clean Sommelet-Hauser, cool it down and use NaNH₂/NH₃.
• Hofmann elimination. Strong base on a quaternary ammonium salt with β-hydrogens can trigger E2 (Hofmann) elimination instead of ylide formation. Substrates lacking accessible β-hydrogens, or with a much more acidic α-position, avoid this.
• β-Elimination / retro-Michael and simple ylide protonation (unproductive equilibrium back to starting material) compete when the migrating group is a poor radical.
• Migratory aptitude limits scope. Only groups that form stabilized radicals migrate. You cannot Stevens-shift a plain methyl or ethyl in useful yield — there is no stabilized radical, so homolysis is too slow.
• Stereocontrol is inherently leaky. Because the intermediate is a radical pair, you get partial, not perfect, retention. If your synthesis demands a single, fully-controlled configuration at the new center, a concerted [2,3]-shift or an ionic method is a safer choice.

Historical discovery

Thomas Stephen Stevens, a Scottish chemist, reported the rearrangement in 1928 (J. Chem. Soc., 1928, 3193) while studying the behaviour of quaternary ammonium salts of phenacyl amines under base. He observed the unexpected migration of a group from nitrogen to the neighbouring carbon and characterised the rearranged amine products. For decades the mechanism was debated — concerted ion-pair versus radical.

The question was settled in the 1970s when CIDNP (chemically induced dynamic nuclear polarization) became available: several groups showed that the rearrangement produces the tell-tale polarized NMR signals that only a spin-correlated radical pair can generate. Combined with the observation of partial (not complete) retention at the migrating carbon, this locked in the modern solvent-caged radical-pair picture. The Stevens rearrangement is now a textbook case study for how CIDNP diagnoses radical intermediates and for why the Woodward-Hoffmann rules force a "forbidden" concerted shift onto a stepwise homolytic path.