The Ullmann Reaction

What the Ullmann reaction does

Take two molecules of an aryl iodide, throw in copper powder, and heat hard. The copper strips both halides away and stitches the two aromatic rings together at the carbons that used to hold the iodine. The product is a biaryl — two rings joined by a single C–C bond that rotates freely (or not, if the ring is crowded enough to lock it). This is the oldest named biaryl coupling in the toolbox, and for a century it was the only practical way to make symmetrical biaryls.

    2 Ph-I   +   2 Cu   ──Δ (≈200 °C)──→   Ph-Ph   +   2 CuI

    iodobenzene         copper              biphenyl    copper(I) iodide

The two things to notice up front. First, the copper is consumed in the classic version — it ends up as CuI, so you need at least one equivalent of copper per halide, not a catalytic pinch. Second, with two identical aryl halides you get a clean symmetrical biaryl; with two different ones you get a statistical mixture (Ar–Ar, Ar–Ar′, Ar′–Ar′), which is the reaction's central weakness and the reason cross-couplings later replaced it for unsymmetrical targets.

The mechanism, step by step

The exact pathway is still argued over — the Ullmann coupling is famous for being both ancient and mechanistically murky, because the reactive copper species are transient, heterogeneous, and hard to isolate. Two families of mechanism are proposed, and both funnel through an aryl–copper (Ar–Cu) intermediate.

The organometallic (Cu(I)/Cu(III)) cycle — the modern consensus for ligand-accelerated variants:

1. Oxidative addition. A copper(I) center inserts into the aryl C–X bond. The two electrons of the breaking C–I bond flow onto copper, so copper's oxidation state climbs by two, from Cu(I) to a fleeting aryl–copper(III) halide, Ar–Cu^III–X. This is the slow, temperature-hungry step.
2. Delivery of the second aryl. A second aryl is brought onto the same copper — either by a second oxidative addition of Ar–X, or by transmetalation from a pre-formed aryl–copper. Copper now holds both aryl groups: Ar–Cu^III–Ar.
3. Reductive elimination. The two aryls, sitting cis on the same metal, couple. Their two Cu–C bonds collapse into one new C–C bond of the biaryl, and the electrons flow back onto copper, dropping it two oxidation states to Cu(I). The product falls off; copper re-enters the cycle.

The radical (single-electron transfer) pathway — favored under harsh, ligand-free conditions:

1. Copper transfers a single electron into the C–X σ* of the aryl halide, giving an aryl-halide radical anion that fragments into an aryl radical (Ar•) plus halide.
2. The aryl radical is captured by copper (forming Ar–Cu) or couples directly with a second aryl radical / aryl–copper to make the biaryl.

    ORGANOMETALLIC VIEW (dominant with ligands):

      Cu(I)  +  Ar-I   ──oxidative addition──→   Ar-Cu(III)-I
      Ar-Cu(III)-I  +  Ar-[Cu]   ──transmetalation──→   Ar-Cu(III)-Ar
      Ar-Cu(III)-Ar   ──reductive elimination──→   Ar-Ar  +  Cu(I)

    RADICAL VIEW (harsh, ligand-free):

      Cu  +  Ar-I   ──SET──→   [Ar-I]•⁻  +  Cu⁺   →   Ar•  +  I⁻
      Ar•  +  Ar-[Cu]   ──→   Ar-Ar

The practical upshot is the same regardless of which arrow-pushing you prefer: copper breaks two aryl–halide bonds and delivers a new aryl–aryl bond, and the two electrons that leave with the biaryl were "borrowed" from copper's redox couple.

Reagents, catalyst, and conditions

The reaction comes in two eras, and they look almost nothing alike.

• Classic Ullmann (1901-style). Substrate: an aryl iodide (bromides are sluggish, chlorides essentially dead). Reagent: activated copper bronze — finely divided copper powder, often freshly reduced or "activated" with iodine or by washing to expose clean metal surface. Conditions: neat (no solvent) or in a high-boiling solvent such as nitrobenzene, DMF, or DMSO, heated to ≈ 200–260 °C for hours. Stoichiometry: ≥ 1 equivalent copper per C–I bond. Ortho substituents — even bulky nitro or ester groups — accelerate the reaction (an unusual "ortho effect") because they help pre-coordinate copper.
• Modern ligand-accelerated Cu(I) catalysis. Substrate: aryl iodides and bromides. Catalyst: a soluble copper(I) salt — CuI, CuBr, CuCl, or Cu₂O — at 1–10 mol%, paired with a chelating ligand (a 1,2-diamine such as N,N′-dimethylethylenediamine, 1,10-phenanthroline, an amino acid like proline, or a β-diketone). Base: K₃PO₄, Cs₂CO₃, or K₂CO₃. Solvent: dioxane, DMSO, DMF, or toluene. Temperature: a far gentler 80–130 °C. The ligand solubilizes copper, tunes its redox potential, and suppresses the radical side-chemistry, turning a brutal metallurgical reaction into a bench-scale catalytic one.

Why the chelating ligand matters so much: a bare copper(I) ion is a lazy, hard-to-control catalyst. Wrapping it in a bidentate nitrogen ligand raises the electron density at copper (accelerating oxidative addition), keeps it in solution (no more heterogeneous slog), and stabilizes the reactive Cu(I)/Cu(III) shuttle. Buchwald, Ma, and others turned this insight into whole families of ligands during the 2000s, resurrecting a reaction that palladium had nearly retired.

Scope, selectivity, and the homocoupling problem

The Ullmann coupling's defining trait — and its defining limitation — is that it is fundamentally a homocoupling. Feed it one aryl halide and it makes the symmetrical biaryl beautifully. Feed it two different aryl halides hoping for the cross product Ar–Ar′ and statistics fight you:

    Ar-I  +  Ar'-I   ──Cu──→   Ar-Ar   +   Ar-Ar'   +   Ar'-Ar'
                                 ~25%       ~50%         ~25%   (statistical)

You can bias a cross-coupling if the two partners have very different reactivities (one much more electron-poor and iodinated, the other bromo), but you never get the clean chemoselectivity of a modern palladium method. This is precisely the gap that Suzuki, Negishi, and Stille couplings closed: by making the two partners chemically distinct (one a halide, one an organometal), they force cross-selectivity by design.

• Symmetrical biaryls: Ullmann's home turf. Biphenyls, binaphthyls, dinitrobiphenyls, and the biaryl axes of natural products.
• Stereochemistry: the coupling makes no new stereocenter, but it does create the biaryl axis. When both rings carry bulky ortho substituents, rotation about the new C–C bond is frozen and the biaryl becomes atropisomeric (axially chiral) — the basis of ligands like BINAP and of many bioactive biaryl natural products.
• Functional-group tolerance: the classic high-temperature version is rough on sensitive groups; the modern ligand-accelerated version tolerates esters, ketones, nitriles, and free amines much better.

Ullmann vs the palladium cross-couplings

	Ullmann (classic)	Ullmann (modern Cu/ligand)	Suzuki / Negishi (Pd)
Metal	Copper (stoichiometric)	Cu(I) catalytic + ligand	Pd(0) catalytic + ligand
Coupling partners	2 × aryl halide	2 × aryl halide, or Ar–X + nucleophile	Ar–X + Ar′–[B/Zn/Sn]
Cross-selective?	No — statistical mixture	Limited	Yes — by design
Temperature	200–260 °C, often neat	80–130 °C	25–100 °C
Best halide	Aryl iodide	Aryl iodide/bromide	Aryl iodide/bromide/triflate
Metal cost	Cheap (Cu)	Cheap (Cu)	Expensive (Pd)
C–N / C–O version	Ullmann–Goldberg / Ullmann ether	Ligand-accelerated Cu amination/etherification	Buchwald–Hartwig (C–N)
Functional-group tolerance	Poor (harsh heat)	Good	Excellent
Typical use today	Historical / teaching	Cheap C–N/C–O, symmetrical biaryls	Default for unsymmetrical biaryls

Worked example: making biphenyl (and a real natural-product axis)

The textbook demonstration is the synthesis of biphenyl from iodobenzene:

    2 C₆H₅-I   +   2 Cu   ──neat, 210–230 °C, 4–6 h──→   C₆H₅-C₆H₅   +   2 CuI

    Reagents:  iodobenzene (2.0 equiv), activated Cu bronze (2.2 equiv)
    Conditions: sealed/neat, 210–230 °C, mechanical stirring
    Workup:    cool, extract product from copper salts with hot toluene, recrystallize
    Yield:     ~70–75% biphenyl (symmetrical — no cross-product problem)

A more consequential use is building ortho,ortho′-disubstituted biaryls where the classic "ortho effect" actually helps. Take 2-nitroiodobenzene: the two ortho nitro groups both accelerate the coupling and — once the biaryl forms — lock rotation about the new axis. The product, 2,2′-dinitrobiphenyl, is a workhorse precursor to 2,2′-diaminobiphenyl and thence to chiral biaryl ligands. This is the Ullmann reaction doing something no simple electrophilic route can: forging a hindered aryl–aryl bond that becomes an axis of chirality.

        NO₂                          NO₂  NO₂
         |                             \  /
    ⟨ring⟩-I   ×2   ──Cu, Δ──→     ⟨ring⟩-⟨ring⟩    (2,2′-dinitrobiphenyl)

The Ullmann–Goldberg heteroatom couplings

Ullmann's biaryl coupling has a whole family of siblings that don't join two carbons at all. Working with Ullmann, Irma Goldberg (1903–1906) showed that copper also couples an aryl halide to a nucleophile, forging a C–N or C–O bond instead of a C–C bond:

• Ullmann–Goldberg amination (C–N). Aryl halide + an amine or amide + Cu → a diarylamine (or N-aryl amide). This is the classic route to triarylamines and to the N-aryl linkages in dyes and, later, OLED hole-transport materials. Its modern, ligand-accelerated form is a cheap copper alternative to Buchwald–Hartwig amination on palladium.
• Ullmann ether synthesis (C–O). Aryl halide + a phenol (as its copper phenolate) + Cu → a diaryl ether. This is how many diaryl-ether motifs — including the aryl-ether crosslinks in natural products like vancomycin's aglycone — are assembled.

    C–N (Goldberg):   Ar-I  +  H₂N-Ar'   ──Cu, base──→   Ar-NH-Ar'   +  HI
    C–O (ether):      Ar-I  +  HO-Ar'     ──Cu, base──→   Ar-O-Ar'    +  HI

Mechanistically these run through the same Cu(I)/Cu(III) oxidative-addition / reductive-elimination logic as the biaryl coupling — the only change is that the second group delivered to copper is a nitrogen or oxygen nucleophile rather than a second aryl. Because copper is far cheaper than palladium and the diamine/amino-acid ligands are trivial to make, these Ullmann–Goldberg variants remain genuinely competitive in process chemistry today.

Limitations and side reactions

• Statistical cross-coupling. The biggest one: two different halides give a three-way mixture. Use Ullmann only when the target is symmetrical, or accept a chromatography-heavy separation.
• Brutal classic conditions. 200 °C neat destroys thermally sensitive functionality; aldehydes, some esters, and any thermally labile group won't survive the original protocol. The modern ligand version fixes most of this.
• Halide reactivity ceiling. Aryl chlorides and fluorides barely react even under forcing conditions; you generally must start from an iodide (or a bromide with modern catalysis).
• Reduction / dehalogenation. A competing side path replaces C–X with C–H (protodehalogenation), consuming starting material without coupling it — worse when the aryl radical pathway dominates and a good H-atom donor is around.
• Copper waste. The classic version generates stoichiometric copper(I) halide and copper sludge — an environmental and workup burden the catalytic modern version largely removes.

Discovery and revival

Fritz Ullmann and his co-worker Jean Bielecki reported the copper-mediated biaryl coupling in 1901. Within a few years, working alongside Ullmann, Irma Goldberg extended copper's reach to C–N bond formation (the Goldberg amination, ~1903–1906), giving chemists a cheap way to make diarylamines and diaryl ethers decades before anyone had heard of palladium catalysis. For most of the 20th century these copper reactions were the backbone of dye and pigment manufacture.

Then, from the 1970s onward, the palladium cross-couplings — Heck, Sonogashira, Negishi, Stille, Suzuki, and Buchwald–Hartwig — offered milder conditions and true cross-selectivity, and the Ullmann reaction was largely pushed aside. Its second life came around 2000, when Buchwald, Ma, Taillefer, and others showed that adding cheap chelating ligands (diamines, amino acids, diketones) to soluble copper(I) salts collapsed the temperature requirement and restored broad functional-group tolerance. The result: copper-catalyzed C–N and C–O coupling is once again a first-choice, low-cost tool in industrial process chemistry — the old metallurgical reaction reborn as a modern catalytic one.