Jacobian Conjecture Disproved: Counterexample Explained
Quick answer: The Jacobian conjecture is false in dimensions three and higher; the two-variable case remains open. The conjecture claimed that a polynomial map with a constant, nonzero Jacobian determinant must have a polynomial inverse. In July 2026, Levent Alpöge announced a three-variable counterexample, crediting Anthropic’s Claude. Its determinant is always −2, yet three distinct inputs produce the same output, making an inverse impossible.
The key idea is the difference between local and global. A map can be reversible near every single input and still send two faraway inputs to the same output. Below, we check that failure by direct substitution and verify the determinant separately.
Where the Jacobian Conjecture Stands
| Number of complex variables | Status | Why |
|---|---|---|
| 1 | True | A polynomial with a nonzero constant derivative is linear, so it is invertible. |
| 2 | Still open | The three-variable example does not settle this case. |
| 3 and higher | False | The 2026 counterexample, extended by adding unchanged coordinates. |
What Is a Jacobian, in Plain Words?
Take a map \(F\) that moves points around, for example \(F(x, y) = (x + y^2,\; y)\). Near any point, a smooth map behaves like a linear map, and the Jacobian matrix collects its partial derivatives. Its determinant is the Jacobian determinant.
For maps between real coordinate spaces, the absolute value of the Jacobian determinant measures how tiny areas or volumes are scaled, and a negative sign means orientation is reversed. A determinant of 0 means the derivative collapses at least one direction. For the complex maps in this story, the essential condition is a constant, nonzero determinant, which guarantees the map can be reversed locally, near each point.
For the example above, the Jacobian matrix is \(\begin{pmatrix}1 & 2y\\ 0 & 1\end{pmatrix}\), whose determinant is 1 everywhere, and the map has a polynomial inverse: \((u, v) \mapsto (u – v^2,\; v)\). The conjecture predicted that this always happens.
What Did the Jacobian Conjecture Say?
Ott-Heinrich Keller asked in 1939: if a polynomial map from \(\mathbb{C}^n\) to itself has the same nonzero Jacobian determinant everywhere, must it have a polynomial inverse? Locally the answer is yes. The conjecture claimed it is also true globally. It appeared on Stephen Smale’s 1998 list of problems for the 21st century, and many published “proofs” turned out to be wrong.
The Counterexample
The map announced by Levent Alpöge sends \((x, y, z)\) to \((f_1, f_2, f_3)\):
\[ \begin{aligned} f_1 &= (1+xy)^3 z + y^2(1+xy)(4+3xy),\\ f_2 &= y + 3x(1+xy)^2 z + 3xy^2(4+3xy),\\ f_3 &= 2x – 3x^2 y – x^3 z. \end{aligned} \]
Its Jacobian determinant simplifies to exactly \(-2\) everywhere, so the conjecture’s hypothesis holds. Yet three different inputs give the same output:
| Input \((x, y, z)\) | Output \((f_1, f_2, f_3)\) |
|---|---|
| \((0,\; 0,\; -\tfrac14)\) | \((-\tfrac14,\; 0,\; 0)\) |
| \((1,\; -\tfrac32,\; \tfrac{13}{2})\) | \((-\tfrac14,\; 0,\; 0)\) |
| \((-1,\; \tfrac32,\; \tfrac{13}{2})\) | \((-\tfrac14,\; 0,\; 0)\) |
Work Through a Collision Yourself
At \((0, 0, -\tfrac14)\) every term with \(x\) or \(y\) vanishes, so \(f_1 = -\tfrac14\), \(f_2 = 0\), \(f_3 = 0\).
Now try \((1, -\tfrac32, \tfrac{13}{2})\). Here \(xy = -\tfrac32\), so both \(1 + xy\) and \(4 + 3xy\) equal \(-\tfrac12\). Then
\[ \begin{aligned} f_1 &= \left(-\tfrac12\right)^3\cdot\tfrac{13}{2} + \tfrac94\cdot\left(-\tfrac12\right)\left(-\tfrac12\right) = -\tfrac{13}{16} + \tfrac{9}{16} = -\tfrac14,\\ f_2 &= -\tfrac32 + 3\cdot\tfrac14\cdot\tfrac{13}{2} + 3\cdot\tfrac94\cdot\left(-\tfrac12\right) = -\tfrac{12}{8} + \tfrac{39}{8} – \tfrac{27}{8} = 0,\\ f_3 &= 2 + \tfrac92 – \tfrac{13}{2} = 0. \end{aligned} \]
Same output as before. Two different inputs with one output are enough to rule out an inverse. Separately, a symbolic calculation confirms the determinant is \(-2\) at every point. You can run both checks with this short SymPy script, which uses exact fractions:
from sympy import symbols, Matrix, Rational, simplify
x, y, z = symbols('x y z')
F = Matrix([(1+x*y)**3*z + y**2*(1+x*y)*(4+3*x*y),
y + 3*x*(1+x*y)**2*z + 3*x*y**2*(4+3*x*y),
2*x - 3*x**2*y - x**3*z])
print(simplify(F.jacobian([x, y, z]).det())) # constant determinant: -2
for p in [(0, 0, Rational(-1, 4)), (1, Rational(-3, 2), Rational(13, 2))]:
print(list(F.subs(dict(zip((x, y, z), p))))) # distinct inputs, identical output
We ran this before publishing; it prints \(-2\) and the output \((-\tfrac14, 0, 0)\) twice.
Two different meanings of “degree” appear in reports. The map has polynomial degree 7 (its components have degrees 7, 6 and 4). Separately, a typical output has exactly three inputs that reach it, so its generic fibre has three points.
Why is local invertibility not enough? Roughly, some inputs can escape to infinity while their outputs stay bounded, so the map folds space onto itself in a way no local check can see.
Who Found It, and How Was It Verified?
- The announcement: Levent Alpöge, a mathematician at Anthropic, announced the map on July 19 to 20, 2026, crediting Anthropic’s Claude, after University of Chicago mathematician Akhil Mathew suggested the problem. Exactly how the search worked has not been described in detail.
- Checks: because the claim is an explicit formula, anyone with a computer algebra system could check it. Lean formalizations appeared within a day, and an independent Isabelle/HOL verification is available in the Archive of Formal Proofs.
- Explanations: within days, mathematicians including Terence Tao and David Speyer explained why the map works, and others found infinite families of similar counterexamples. Gao’s arXiv paper collects the geometry and higher-dimensional extensions.
Does the determinant have to be 1? No. Replacing \(f_1\) with \(-f_1/2\) gives a map with determinant exactly 1 that still has the same kind of collision.
Wasn’t There Already a Counterexample?
Sergey Pinchuk’s 1994 counterexample concerned a real two-variable map whose Jacobian determinant is never zero but is not constant. It disproved a “real Jacobian conjecture,” not the complex conjecture with a constant determinant discussed here. That is why sources before 2026 still called the Jacobian conjecture open.
What Is Still Open?
The two-variable case, for maps \(\mathbb{C}^2 \to \mathbb{C}^2\). All known counterexamples need at least three variables.
A Short History
| Year | Event |
|---|---|
| 1939 | Keller poses the question. |
| 1980 | Stuart Wang proves it for maps of degree at most 2. |
| 1982 | Bass, Connell and Wright reduce it to a special cubic form in all dimensions. |
| 1994 | Pinchuk’s real counterexample with a nonconstant determinant. |
| 1998 | Smale lists it among the key problems for the 21st century. |
| July 2026 | The three-variable counterexample is announced and quickly verified. |
Watch: The Jacobian Conjecture Counterexample in Video
If you learn better by watching, these videos are worth your time.
Check Your Understanding
Five quick questions to see how much of this has stuck. Pick an answer and press Submit; you will see a short explanation either way.
Answer each question and press Submit to see your result right away.
Frequently Asked Questions
Has the Jacobian conjecture been disproved?
Yes, in three or more variables. In July 2026, Levent Alpöge announced an explicit polynomial map from 3D complex space to itself with Jacobian determinant -2 everywhere that sends three different inputs to the same output. It has been checked by computer algebra and formal proof systems. The two-variable case is still open.
Did Claude disprove the Jacobian conjecture?
The announcement by Levent Alpöge credited Anthropic’s Claude with finding the counterexample, after Akhil Mathew suggested the problem. The search process has not been described in detail, and the counterexample itself was checked independently by many mathematicians and proof systems.
Does the determinant have to equal 1?
No. The conjecture only needs a constant nonzero determinant. Replacing the first component f1 with -f1/2 turns the 2026 example into a map with determinant exactly 1 that still fails to be one-to-one.
What about Pinchuk’s counterexample?
Pinchuk’s 1994 example is a real two-variable map whose Jacobian determinant is never zero but not constant. It disproved the real Jacobian conjecture, not the complex Jacobian conjecture, which requires a constant determinant.
Can I check the counterexample myself?
Yes. Substitute (0, 0, -1/4) and (1, -3/2, 13/2) into the three formulas and both give (-1/4, 0, 0). A computer algebra system such as SymPy confirms the determinant is -2 everywhere.
Keep Exploring
- Riemann Hypothesis Solved? What’s Verified in 2026
- Navier-Stokes Solved? The 2026 Proof Claims Explained
- Jacobian Determinant (Math Dictionary)
- Determinant Calculator
Sources and Further Reading
- Archive of Formal Proofs: Formal Verification of an Explicit Counterexample to the Jacobian Conjecture
- Gao, Counterexamples to the Jacobian conjecture in dimensions greater than two (arXiv 2608.00222)
- Wikipedia: Jacobian conjecture
- Fortune: AI and the Jacobian conjecture (July 21, 2026)
- Wolfram MathWorld: Jacobian Conjecture
Related to This Article
More math articles
- Best Canadian Universities for International Students Seeking Strong Career Support
- Quotas, Eugenics, the Second Klan, and the Election of 1928
- Federalists vs. Anti-Federalists: Ratification and the Bill of Rights
- Nonresponse bias
- Electricity and Simple Circuits
- Unemployment, Migration, Family Survival, and Depression Culture
- How long is the TEAS Test?
- What Are the Types of Math?
- Sterilizer Anatomy and Cycle Phases
- Speedy and public trial


What people say about "Jacobian Conjecture Disproved: Counterexample Explained - Effortless Math"?
No one replied yet.