> Full-text translation from Chinese. The original remains authoritative for its research claims; this translation is not independent mathematical verification.

# ACTM Round 3
## Exponential Matrices, Multiplier Intersections, and the One-Dimensional Gap of Four Exponentials
### Exponential Matrices, Multiplier Intersections, and the One-Dimensional Gap of Four Exponentials

**Research programme:** Algebraic Combinatorial Transcendence Methodology (ACTM)  
**Authors:** Neo.K; Aletheia (GPT-5.6 Sol)  
**Institution:** EveMissLab / 一言諾科技有限公司  
**Date:** 2026-09-30  
**Version:** Round 3 / v0.1  
**Document type:** Formal research note / calibration round for matrices and simultaneous transcendence  
**Status statement:** The Six Exponentials Theorem is a known theorem; the Four Exponentials Conjecture remains an open problem as of 2026-09-30. The multiplier-space and intersection-dimension language proposed in this round is primarily a structural ACTM reformulation and research interface; it does not claim to resolve the Four Exponentials Conjecture.

---

## Abstract

ACTM Round 1 establishes:

$$
\text{algebraic seed rank}
\longrightarrow
\text{transcendence degree}.
$$

Round 2 establishes:

$$
\text{algebraic scalar action}
\longrightarrow
\text{descent / escape boundary}.
$$

Round 3 enters simultaneous transcendence and studies the exponential matrix:

$$
M_{ij}
=
e^{x_i y_j}.
$$

If all $M_{ij}$ are algebraic numbers, then all exponents:

$$
z_{ij}=x_i y_j
$$

lie in the algebraic logarithm locus:

$$
\mathcal L_{\mathrm{alg}}
=
\left\{
z\in\mathbb C:
e^z\in\overline{\mathbb Q}^{\times}
\right\}.
$$

On the other hand:

$$
Z=(z_{ij})=xy^T
$$

must be a rank-one matrix.

The Six Exponentials Theorem tells us that if

$$
x_1,\ldots,x_d
$$

and

$$
y_1,\ldots,y_\ell
$$

are each linearly independent over $\mathbb Q$, and:

$$
d\ell>d+\ell,
$$

then it is impossible for all:

$$
e^{x_i y_j}
$$

to be algebraic numbers.

This round rewrites the condition as:

$$
d\ell-d-\ell
=
(d-1)(\ell-1)-1.
$$

Thus Six Exponentials excludes the cases in which the rank-one determinantal locus has codimension at least $2$; the only nondegenerate boundary size is:

$$
2\times2,
$$

Here the rank-one locus is defined by a single quadratic determinant relation:

$$
z_{11}z_{22}-z_{12}z_{21}=0
$$

as displayed above. This is precisely the unresolved boundary of the Four Exponentials Conjecture.

This round further fixes:

$$
0\neq\lambda\in\mathcal L_{\mathrm{alg}}
$$

and defines the multiplier space:

$$
\mathfrak M_\lambda
=
\left\{
c\in\mathbb C:
c\lambda\in\mathcal L_{\mathrm{alg}}
\right\}.
$$

It is a $\mathbb Q$-vector space. Round 2 has already implied:

$$
\boxed{
\mathfrak M_\lambda
\cap
\overline{\mathbb Q}
=
\mathbb Q.
}
$$

For:

$$
b\in
\mathfrak M_\lambda\setminus\mathbb Q
$$

define the multiplier intersection:

$$
W_\lambda(b)
=
\mathfrak M_\lambda
\cap
b^{-1}\mathfrak M_\lambda.
$$

Then the Six Exponentials Theorem equivalently gives:

$$
\boxed{
1
\le
\dim_{\mathbb Q}W_\lambda(b)
\le
2.
}
$$

Whereas the Four Exponentials Conjecture is equivalent to:

$$
\boxed{
\dim_{\mathbb Q}W_\lambda(b)=1
}
$$

that is:

$$
\boxed{
W_\lambda(b)=\mathbb Q.
}
$$

Thus, from the ACTM perspective, the gap between Six and Four Exponentials is compressed into:

$$
\boxed{
2
\longrightarrow
1
}
$$

a multiplier-intersection dimension gap.

---

# 1. Exponential Matrices

Let:

$$
x=
(x_1,\ldots,x_d)^T
\in
\mathbb C^d,
$$

and:

$$
y=
(y_1,\ldots,y_\ell)^T
\in
\mathbb C^\ell.
$$

Define the exponent matrix:

$$
Z
=
xy^T.
$$

Thus:

$$
z_{ij}
=
x_i y_j.
$$

Its complex matrix rank satisfies:

$$
\boxed{
\operatorname{rank}_{\mathbb C}Z=1
}
$$

provided that:

$$
x\neq0,
\qquad
y\neq0.
$$

The exponential output matrix is:

$$
E(Z)
=
\left(
e^{z_{ij}}
\right).
$$

---

# 2. Simultaneous Algebraic Descent

If:

$$
e^{z_{ij}}
\in
\overline{\mathbb Q}^{\times}
$$

holds for all $i,j$, then:

$$
z_{ij}
\in
\mathcal L_{\mathrm{alg}}.
$$

Thus the simultaneous transcendence problem can be rewritten as:

> Can arrangements of certain special low matrix ranks exist within $\mathcal L_{\mathrm{alg}}$?

In particular:

$$
\boxed{
Z\in
M_{d\times\ell}
(\mathcal L_{\mathrm{alg}})
}
$$

and at the same time:

$$
\boxed{
\operatorname{rank}_{\mathbb C}Z=1.
}
$$

---

# 3. Six Exponentials Theorem

The classical Six Exponentials Theorem can be stated as follows:

If:

$$
x_1,\ldots,x_d
$$

are linearly independent over $\mathbb Q$,

$$
y_1,\ldots,y_\ell
$$

are also linearly independent over $\mathbb Q$, and:

$$
d\ell>d+\ell,
$$

then at least one:

$$
e^{x_i y_j}
$$

is transcendental.

Equivalently:

> There is no rank-one matrix whose entries all belong to $\mathcal L_{\mathrm{alg}}$
>
> $$
> Z=xy^T
> $$
>
> for which $x$ and $y$ each have full $\mathbb Q$-rank and $d\ell>d+\ell$.

---

# 4. Geometric Counting: Why $2\times3$ and $2\times2$ Differ

All:

$$
d\times\ell
$$

matrices form the ambient space:

$$
M_{d\times\ell}(\mathbb C)
$$

whose dimension is:

$$
d\ell.
$$

Nonzero rank-one matrices are parametrised by:

$$
Z=xy^T
$$

the expression above.

$x$ has $d$ parameters and $y$ has $\ell$ parameters, but there is a scaling redundancy:

$$
(x,y)
\sim
(cx,c^{-1}y),
\qquad
c\neq0.
$$

Thus the rank-one determinantal variety has dimension:

$$
d+\ell-1.
$$

and codimension:

$$
d\ell-(d+\ell-1)
=
(d-1)(\ell-1).
$$

The condition of Six Exponentials:

$$
d\ell>d+\ell
$$

is equivalent to:

$$
(d-1)(\ell-1)>1.
$$

Thus:

$$
\boxed{
\text{Six Exponentials works once rank-one codimension }\ge2.
}
$$

---

# 5. The Only Nondegenerate Boundary: $2\times2$

If:

$$
d,\ell\ge2,
$$

but:

$$
d\ell\le d+\ell,
$$

then:

$$
(d-1)(\ell-1)\le1.
$$

Thus the only possibility is:

$$
\boxed{
d=\ell=2.
}
$$

In this case the codimension of the rank-one locus is:

$$
(2-1)(2-1)=1.
$$

that is, it is a hypersurface.

Its single basic determinant equation is:

$$
\boxed{
z_{11}z_{22}
-
z_{12}z_{21}
=
0.
}
$$

---

# 6. Four Exponentials Conjecture

The Four Exponentials Conjecture asserts:

If:

$$
x_1,x_2
$$

are linearly independent over $\mathbb Q$, and:

$$
y_1,y_2
$$

are also linearly independent over $\mathbb Q$, then among the four numbers:

$$
e^{x_1y_1},
\quad
e^{x_1y_2},
\quad
e^{x_2y_1},
\quad
e^{x_2y_2}
$$

at least one is transcendental.

That is:

$$
\boxed{
\text{There is no nondegenerate }
2\times2
\text{ rank-one matrix entirely inside }
\mathcal L_{\mathrm{alg}}.
}
$$

---

# 7. From Round 2 to Round 3: A Linear Obstruction Becomes a Quadratic Obstruction

Round 2 uses Baker rigidity:

> $\mathbb Q$-independent algebraic logarithms cannot produce new $\overline{\mathbb Q}$-linear relations.

That is, it primarily controls:

$$
\boxed{
\text{degree-1 relations}.
}
$$

But for a $2\times2$ rank-one matrix:

$$
Z=
\begin{pmatrix}
z_{11}&z_{12}\\
z_{21}&z_{22}
\end{pmatrix},
$$

the rank-one relation is:

$$
\boxed{
z_{11}z_{22}
-
z_{12}z_{21}
=
0.
}
$$

This is a degree-$2$ relation.

Thus:

$$
\boxed{
\text{Baker linear rigidity}
\not\Rightarrow
\text{Four Exponentials}.
}
$$

ACTM calls this gap the:

**Linear-to-Quadratic Barrier**.

---

# 8. Why $2\times3$ Has an Additional Layer of Constraints

For a:

$$
2\times3
$$

rank-one matrix:

$$
Z=
\begin{pmatrix}
z_{11}&z_{12}&z_{13}\\
z_{21}&z_{22}&z_{23}
\end{pmatrix},
$$

all $2\times2$ minors vanish:

$$
z_{11}z_{22}-z_{12}z_{21}=0,
$$

$$
z_{11}z_{23}-z_{13}z_{21}=0,
$$

$$
z_{12}z_{23}-z_{13}z_{22}=0.
$$

These relations are not completely independent, but the rank-one variety has codimension:

$$
(2-1)(3-1)=2.
$$

Thus, compared with $2\times2$:

$$
\boxed{
\text{one determinantal constraint}
\longrightarrow
\text{two-dimensional determinantal constraint structure}.
}
$$

The interpolation / auxiliary-function machinery of Six Exponentials can precisely cross this overdetermined threshold.

---

# 9. Determinantal Surplus

ACTM defines a working quantity:

$$
\operatorname{DS}(d,\ell)
=
d\ell-d-\ell.
$$

Equivalently:

$$
\operatorname{DS}(d,\ell)
=
(d-1)(\ell-1)-1.
$$

Thus:

### $2\times2$

$$
\operatorname{DS}(2,2)=0.
$$

### $2\times3$

$$
\operatorname{DS}(2,3)=1.
$$

### $3\times3$

$$
\operatorname{DS}(3,3)=3.
$$

The known region of Six Exponentials is precisely:

$$
\boxed{
\operatorname{DS}>0.
}
$$

Four Exponentials lies at:

$$
\boxed{
\operatorname{DS}=0.
}
$$

---

# 10. Normalising a $2\times2$ Counterexample

Suppose Four Exponentials is false.

Then there exist:

$$
x_1,x_2,
\qquad
y_1,y_2
$$

each pair $\mathbb Q$-linearly independent, with all four exponentials algebraic.

Since all $x_i,y_j$ are nonzero, we may put:

$$
\lambda=x_1y_1.
$$

Define:

$$
a=\frac{y_2}{y_1},
\qquad
b=\frac{x_2}{x_1}.
$$

Then:

$$
Z
=
\lambda
\begin{pmatrix}
1&a\\
b&ab
\end{pmatrix}.
$$

Since the two pairs are each $\mathbb Q$-linearly independent:

$$
a\notin\mathbb Q,
\qquad
b\notin\mathbb Q.
$$

If all exponentials are algebraic, then:

$$
\boxed{
\lambda,
\quad
a\lambda,
\quad
b\lambda,
\quad
ab\lambda
\in
\mathcal L_{\mathrm{alg}}.
}
$$

---

# 11. Multiplier Spaces

Fix:

$$
0\neq
\lambda
\in
\mathcal L_{\mathrm{alg}}.
$$

Define:

$$
\boxed{
\mathfrak M_\lambda
=
\left\{
c\in\mathbb C:
c\lambda
\in
\mathcal L_{\mathrm{alg}}
\right\}.
}
$$

---

# 12. Proposition R3.1: $\mathfrak M_\lambda$ Is a $\mathbb Q$-Vector Space

If:

$$
c,d\in\mathfrak M_\lambda,
$$

then:

$$
c\lambda,
\quad
d\lambda
\in
\mathcal L_{\mathrm{alg}}.
$$

Since:

$$
\mathcal L_{\mathrm{alg}}
$$

is a $\mathbb Q$-vector space:

$$
(c+d)\lambda
\in
\mathcal L_{\mathrm{alg}},
$$

and:

$$
(qc)\lambda
\in
\mathcal L_{\mathrm{alg}}
$$

for all:

$$
q\in\mathbb Q.
$$

Therefore:

$$
\boxed{
\mathfrak M_\lambda
\text{ is a }
\mathbb Q\text{-vector space}.
}
$$

Moreover:

$$
1\in\mathfrak M_\lambda.
$$

so:

$$
\mathbb Q
\subseteq
\mathfrak M_\lambda.
$$

---

# 13. Reformulating Round 2 in Multiplier-Space Language

Round 2 has proved:

If:

$$
c\in\overline{\mathbb Q}
$$

and:

$$
c\lambda
\in
\mathcal L_{\mathrm{alg}},
$$

then:

$$
c\in\mathbb Q.
$$

Thus:

$$
\boxed{
\mathfrak M_\lambda
\cap
\overline{\mathbb Q}
=
\mathbb Q.
}
$$

This means:

> Any non-rational element of the multiplier space must be transcendental.

That is:

$$
\boxed{
c\in
\mathfrak M_\lambda
\setminus
\mathbb Q
\Rightarrow
c
\text{ transcendental}.
}
$$

---

# 14. The Multiplier Form of Four Exponentials

The normalisation above shows:

A counterexample to Four Exponentials is equivalent to the existence of:

$$
0\neq\lambda
\in
\mathcal L_{\mathrm{alg}},
$$

and:

$$
a,b
\in
\mathfrak M_\lambda
\setminus\mathbb Q
$$

such that:

$$
ab
\in
\mathfrak M_\lambda.
$$

Thus the Four Exponentials Conjecture is equivalent to:

$$
\boxed{
a,b\in
\mathfrak M_\lambda\setminus\mathbb Q
\Rightarrow
ab\notin\mathfrak M_\lambda.
}
$$

For every nonzero:

$$
\lambda\in\mathcal L_{\mathrm{alg}}
$$

the displayed implication holds.

---

# 15. Multiplicative Rectangle Barrier

Thus Four Exponentials can be rewritten as:

> There is no nontrivial multiplicative rectangle in the multiplier space $\mathfrak M_\lambda$
>
> $$
> \{1,a,b,ab\}
> $$
>
> with:
>
> $$
> a,b\notin\mathbb Q.
> $$

ACTM calls this the:

$$
\boxed{
\text{Multiplicative Rectangle Barrier}.
}
$$

Note that this is an equivalent structural reformulation of Four Exponentials, not a newly proved theorem.

---

# 16. Multiplier Intersection

Fix:

$$
b
\in
\mathfrak M_\lambda
\setminus
\mathbb Q.
$$

Define:

$$
\boxed{
W_\lambda(b)
=
\mathfrak M_\lambda
\cap
b^{-1}\mathfrak M_\lambda.
}
$$

That is:

$$
W_\lambda(b)
=
\left\{
c\in\mathbb C:
c\lambda
\in\mathcal L_{\mathrm{alg}},
\quad
bc\lambda
\in\mathcal L_{\mathrm{alg}}
\right\}.
$$

Since:

$$
\mathfrak M_\lambda
$$

and:

$$
b^{-1}\mathfrak M_\lambda
$$

are both $\mathbb Q$-vector spaces:

$$
W_\lambda(b)
$$

is also a $\mathbb Q$-vector space.

---

# 17. $W_\lambda(b)$ Contains at Least $\mathbb Q$

Since:

$$
b
\in
\mathfrak M_\lambda,
$$

for any:

$$
q\in\mathbb Q,
$$

we have:

$$
q\lambda
\in
\mathcal L_{\mathrm{alg}},
$$

and:

$$
bq\lambda
=
q(b\lambda)
\in
\mathcal L_{\mathrm{alg}}.
$$

so:

$$
\boxed{
\mathbb Q
\subseteq
W_\lambda(b).
}
$$

Thus:

$$
\boxed{
\dim_{\mathbb Q}W_\lambda(b)
\ge1.
}
$$

---

# 18. Proposition R3.2: The Exact Upper Bound Supplied by Six Exponentials

## Proposition

For:

$$
0\neq\lambda
\in
\mathcal L_{\mathrm{alg}}
$$

and:

$$
b
\in
\mathfrak M_\lambda
\setminus\mathbb Q,
$$

the Six Exponentials Theorem implies:

$$
\boxed{
\dim_{\mathbb Q}
W_\lambda(b)
\le2.
}
$$

---

# 19. Proof

Suppose, to the contrary:

$$
\dim_{\mathbb Q}
W_\lambda(b)
\ge3.
$$

Then we may choose:

$$
c_1,c_2,c_3
\in
W_\lambda(b)
$$

linearly independent over $\mathbb Q$.

Define:

$$
x_1=1,
\qquad
x_2=b.
$$

Since:

$$
b\notin\mathbb Q,
$$

we have:

$$
x_1,x_2
$$

linearly independent over $\mathbb Q$.

Next define:

$$
y_j=c_j\lambda,
\qquad
j=1,2,3.
$$

Since:

$$
\lambda\neq0
$$

and the $c_j$ are $\mathbb Q$-independent:

$$
y_1,y_2,y_3
$$

are linearly independent over $\mathbb Q$.

Also, since:

$$
c_j\in W_\lambda(b),
$$

we have:

$$
c_j\lambda
\in
\mathcal L_{\mathrm{alg}},
$$

and:

$$
bc_j\lambda
\in
\mathcal L_{\mathrm{alg}}.
$$

Thus all six:

$$
e^{x_i y_j}
$$

are algebraic numbers.

This contradicts the Six Exponentials Theorem.

Hence:

$$
\dim_{\mathbb Q}
W_\lambda(b)
\le2.
$$

This completes the proof.

---

# 20. The One-Dimensional Version of Four Exponentials

If:

$$
\dim_{\mathbb Q}
W_\lambda(b)
\ge2,
$$

since:

$$
1\in W_\lambda(b),
$$

we may choose:

$$
a\in
W_\lambda(b)
\setminus
\mathbb Q.
$$

Then:

$$
\lambda,
\quad
a\lambda,
\quad
b\lambda,
\quad
ab\lambda
\in
\mathcal L_{\mathrm{alg}}.
$$

and:

$$
a,b\notin\mathbb Q.
$$

This gives precisely a counterexample to Four Exponentials.

Conversely, every counterexample to Four Exponentials produces such an $a$ after normalisation.

Thus:

$$
\boxed{
\text{Four Exponentials}
\iff
\dim_{\mathbb Q}
W_\lambda(b)=1
}
$$

For all admissible:

$$
(\lambda,b)
$$

the displayed equivalence holds.

Since:

$$
\mathbb Q\subseteq W_\lambda(b),
$$

this is also equivalent to:

$$
\boxed{
W_\lambda(b)=\mathbb Q.
}
$$

---

# 21. Compressing Six to Four into One Dimension

Known Six Exponentials:

$$
\boxed{
1
\le
\dim_{\mathbb Q}
W_\lambda(b)
\le2.
}
$$

The Four Exponentials conjecture:

$$
\boxed{
\dim_{\mathbb Q}
W_\lambda(b)=1.
}
$$

Thus ACTM compresses the core unknown into:

$$
\boxed{
2
\stackrel{?}{\longrightarrow}
1.
}
$$

This is not a proof, but it supplies a very clearly defined research target.

---

# 22. A More Symmetric Two-Subspace Version

Let:

$$
U,V
\subseteq
\mathbb C
$$

be finite-dimensional $\mathbb Q$-vector spaces.

Define the product set:

$$
UV
=
\{uv:u\in U,\ v\in V\}.
$$

Since:

$$
\mathcal L_{\mathrm{alg}}
$$

is a $\mathbb Q$-vector space, it suffices to check products of basis elements to decide:

$$
UV
\subseteq
\mathcal L_{\mathrm{alg}}.
$$

If:

$$
\dim_{\mathbb Q}U=d,
\qquad
\dim_{\mathbb Q}V=\ell,
$$

and:

$$
UV
\subseteq
\mathcal L_{\mathrm{alg}},
$$

then from the bases:

$$
x_1,\ldots,x_d
$$

and:

$$
y_1,\ldots,y_\ell
$$

we obtain:

$$
x_i y_j
\in
\mathcal L_{\mathrm{alg}}.
$$

---

# 23. The Product-Space Version of Six Exponentials

Six Exponentials implies that when:

$$
UV
\subseteq
\mathcal L_{\mathrm{alg}}
$$

holds, we cannot also have:

$$
d\ell>d+\ell.
$$

Thus:

$$
\boxed{
d\ell\le d+\ell.
}
$$

If:

$$
d,\ell\ge2,
$$

then the only size not yet excluded by Six Exponentials is:

$$
\boxed{
(d,\ell)=(2,2).
}
$$

---

# 24. The Product-Space Version of Four Exponentials

The Four Exponentials Conjecture is equivalent to:

> There do not exist two $\mathbb Q$-vector spaces
>
> $$
> U,V\subseteq\mathbb C
> $$
>
> satisfying:
>
> $$
> \dim_{\mathbb Q}U\ge2,
> \qquad
> \dim_{\mathbb Q}V\ge2,
> $$
>
> and:
>
> $$
> UV
> \subseteq
> \mathcal L_{\mathrm{alg}}.
> $$

In other words:

$$
\boxed{
UV
\subseteq
\mathcal L_{\mathrm{alg}}
\Rightarrow
\min
\left(
\dim_{\mathbb Q}U,
\dim_{\mathbb Q}V
\right)
=1.
}
$$

This is the most symmetric reformulation of Four Exponentials in ACTM Round 3.

---

# 25. Bilinear Capacity

ACTM defines the candidate concept:

$$
\operatorname{BC}
(\mathcal L_{\mathrm{alg}})
$$

as the capacity of the algebraic-log locus to accommodate bilinear product spaces.

Rather than compressing it into a single scalar, this round first records the feasible dimension pairs:

$$
\mathcal B
=
\left\{
(d,\ell):
\exists U,V,\ 
\dim U=d,\ 
\dim V=\ell,\ 
UV\subseteq\mathcal L_{\mathrm{alg}}
\right\}.
$$

Six Exponentials tells us:

$$
(d,\ell)\in\mathcal B
\Rightarrow
d\ell\le d+\ell.
$$

Four Exponentials predicts:

$$
(d,\ell)\in\mathcal B
\Rightarrow
d=1
\text{ or }
\ell=1.
$$

---

# 26. The $\pi$-Normalised Window

A particularly natural base generator is:

$$
\lambda=2\pi i.
$$

Since:

$$
e^{2\pi i}=1,
$$

we have:

$$
2\pi i
\in
\mathcal L_{\mathrm{alg}}.
$$

In this case:

$$
\mathfrak M_{2\pi i}
=
\left\{
c\in\mathbb C:
e^{2\pi i c}
\in
\overline{\mathbb Q}^{\times}
\right\}.
$$

Round 2 gives:

$$
\boxed{
\mathfrak M_{2\pi i}
\cap
\overline{\mathbb Q}
=
\mathbb Q.
}
$$

That is:

If:

$$
c\in\overline{\mathbb Q}\setminus\mathbb Q,
$$

then:

$$
e^{2\pi i c}
$$

is transcendental.

---

# 27. The Form of Four Exponentials in the $\pi$ Window

If there exist:

$$
a,b
\in
\mathfrak M_{2\pi i}
\setminus
\mathbb Q
$$

such that:

$$
ab
\in
\mathfrak M_{2\pi i},
$$

then:

$$
1,
\quad
e^{2\pi i a},
\quad
e^{2\pi i b},
\quad
e^{2\pi i ab}
$$

are all algebraic numbers.

This would directly give a counterexample to Four Exponentials.

Thus FEC predicts:

$$
\boxed{
a,b
\in
\mathfrak M_{2\pi i}\setminus\mathbb Q
\Rightarrow
ab
\notin
\mathfrak M_{2\pi i}.
}
$$

This supplies a natural interface between ACTM and the earlier research on "$\pi$ generation / compression".

---

# 28. Ratio Interpretation

If:

$$
\mu
\in
\mathcal L_{\mathrm{alg}},
$$

and:

$$
\lambda\neq0,
$$

then:

$$
b=\frac{\mu}{\lambda}
$$

satisfies:

$$
b\in\mathfrak M_\lambda.
$$

Thus:

$$
\mathfrak M_\lambda
$$

can be understood as:

> algebraic-log ratios with $\lambda$ as the denominator.

Round 2 tells us:

Any non-rational ratio:

$$
\frac{\mu}{\lambda}
$$

must be transcendental if $\lambda,\mu$ are independent over $\mathbb Q$.

Round 3 instead studies:

> Can the product of two such transcendental log-ratios still be a log-ratio for the same $\lambda$?

The answer predicted by the Four Exponentials conjecture is:

$$
\boxed{
\text{No}.
}
$$

---

# 29. Rational Visibility

Consider a singular square matrix:

$$
M
\in
M_n(\mathcal L_{\mathrm{alg}}).
$$

The Matrix Coefficient Conjecture of Dasgupta–Kakde proposes:

If:

$$
\det M=0,
$$

then there exist:

$$
0\neq w,v\in\mathbb Q^n
$$

such that:

$$
\boxed{
w^T M v=0.
}
$$

ACTM interprets this as:

> A matrix singularity within the algebraic-log locus should have a rationally visible witness.

---

# 30. Rational Invisibility of a $2\times2$ Rank-One Matrix

Let:

$$
M=xy^T.
$$

Then:

$$
w^TMv
=
(w^Tx)(y^Tv).
$$

If:

$$
x_1,x_2
$$

are linearly independent over $\mathbb Q$, then for:

$$
0\neq w\in\mathbb Q^2,
$$

we have:

$$
w^Tx\neq0.
$$

Likewise, if:

$$
y_1,y_2
$$

are linearly independent over $\mathbb Q$, then:

$$
y^Tv\neq0
$$

for any:

$$
0\neq v\in\mathbb Q^2.
$$

Thus:

$$
\boxed{
w^TMv\neq0
}
$$

holds for all nonzero rational $w,v$.

Thus a counterexample to Four Exponentials is precisely a:

$$
\boxed{
\text{singular but rationally invisible}
}
$$

$2\times2$ algebraic-log matrix of the kind displayed above.

---

# 31. The Position of the 2026 Matrix Coefficient Conjecture

Dasgupta and Kakde formally published the Matrix Coefficient Conjecture framework in 2026.

Its $n=2$ case is equivalent to the Four Exponentials Conjecture.

Thus the connection between ACTM Round 3 and contemporary research is very direct:

$$
\boxed{
\text{Four Exponentials}
}
$$

can be viewed simultaneously as:

1. a transcendence problem;
2. rank-one matrix exclusion;
3. a multiplicative rectangle barrier;
4. a multiplier-intersection dimension problem;
5. rational visibility of singular logarithmic matrices.

---

# 32. The Structural Rank Direction

The stronger Structural Rank Conjecture studies:

$$
M
=
\sum_{k}
M_k\lambda_k,
$$

where:

$$
M_k
\in
M_{m\times n}(\mathbb Q),
$$

and:

$$
\lambda_k
\in
\mathcal L_{\mathrm{alg}}
$$

are linearly independent over $\mathbb Q$.

It predicts that the actual matrix rank should equal the generic rank obtained by replacing:

$$
\lambda_k
$$

with independent indeterminates.

This direction has known equivalences with special cases of Schanuel.

Thus ACTM's:

$$
\text{relation visibility}
$$

and:

$$
\text{generic-vs-special rank collapse}
$$

are not isolated concepts.

---

# 33. The Genuine Unknown of Round 3

Round 2 has already handled:

$$
\boxed{
\mathfrak M_\lambda
\cap
\overline{\mathbb Q}
=
\mathbb Q.
}
$$

Thus:

$$
a,b
$$

in a counterexample to Four Exponentials must be transcendental.

The difficulty now is therefore not:

> Can an algebraic irrational multiplier preserve algebraic descent?

This has already been excluded by Gelfond–Schneider / Baker.

The genuine unknown is:

> Can certain transcendental multipliers generated by algebraic logarithm ratios form a two-dimensional multiplicatively closed fragment within the same multiplier space?

That is:

$$
\boxed{
a,b,ab
\in
\mathfrak M_\lambda
}
$$

where:

$$
a,b\notin\mathbb Q.
$$

---

# 34. Multiplier Intersection Dimension

Formally define:

$$
\boxed{
\operatorname{MID}_\lambda(b)
=
\dim_{\mathbb Q}
\left(
\mathfrak M_\lambda
\cap
b^{-1}\mathfrak M_\lambda
\right).
}
$$

This applies to:

$$
b
\in
\mathfrak M_\lambda
\setminus
\mathbb Q.
$$

It is known that:

$$
\boxed{
1
\le
\operatorname{MID}_\lambda(b)
\le
2.
}
$$

The Four Exponentials Conjecture is equivalent to:

$$
\boxed{
\operatorname{MID}_\lambda(b)=1.
}
$$

---

# 35. The Research Value of This Quantity

MID rewrites the problem from:

$$
\text{four transcendence decisions}
$$

into:

$$
\boxed{
\text{the intersection dimension of two }\mathbb Q\text{-vector spaces}.
}
$$

That is, numerical decisions become:

$$
\text{linear-algebraic geometry of multiplier spaces}.
$$

This fits precisely the methodological objective of ACTM:

$$
\boxed{
\text{output classification}
\rightarrow
\text{generator-space structure}.
}
$$

---

# 36. New Research Targets for Round 3

Instead of directly proving that:

$$
e^{x_i y_j}
$$

at least one is transcendental, one can instead try to prove:

$$
\boxed{
\mathfrak M_\lambda
\cap
b^{-1}\mathfrak M_\lambda
=
\mathbb Q.
}
$$

This is an extremely concrete intersection-rigidity problem.

---

# 37. An Initial Set of Research Subproblems

## R3-Q1

For which special:

$$
\lambda
\in
\mathcal L_{\mathrm{alg}}
$$

can one directly prove:

$$
\operatorname{MID}_\lambda(b)=1?
$$

---

## R3-Q2

If:

$$
\operatorname{MID}_\lambda(b)=2,
$$

then there exists:

$$
a\notin\mathbb Q
$$

such that:

$$
\lambda,
a\lambda,
b\lambda,
ab\lambda
\in
\mathcal L_{\mathrm{alg}}.
$$

What additional arithmetic / period constraints must these four logarithms satisfy?

---

## R3-Q3

Can one construct invariants for:

$$
W_\lambda(b)
$$

of the following types:

- height;
- denominator growth;
- approximation exponent;
- monodromy;
- period;
- definability;

as listed above?

---

## R3-Q4

Six Exponentials gives:

$$
\dim W\le2.
$$

Exactly which analytic ingredient fails to compress:

$$
2
$$

down to:

$$
1?
$$

This should be directly related to:

$$
\operatorname{DS}=0
$$

and the criticality of a single determinant relation.

---

# 38. Round 3 Assessment

**PASS — Simultaneous Transcendence Frontier Compressed**

This round completes:

1. Reformulation of Six Exponentials as rank-one algebraic-log matrix exclusion.
2. Interpretation of the condition:
   $$
   d\ell>d+\ell
   $$
   as determinantal codimension $>1$.
3. Identification of the only nondegenerate boundary:
   $$
   2\times2.
   $$
4. Positioning Four Exponentials as the first quadratic hidden-relation barrier.
5. Definition of the multiplier space:
   $$
   \mathfrak M_\lambda.
   $$
6. Reformulation of Round 2 as:
   $$
   \mathfrak M_\lambda
   \cap
   \overline{\mathbb Q}
   =
   \mathbb Q.
   $$
7. Definition of the multiplier intersection:
   $$
   W_\lambda(b).
   $$
8. A proof from Six Exponentials that:
   $$
   \dim_{\mathbb Q}W_\lambda(b)\le2.
   $$
9. A proof that Four Exponentials is equivalent to:
   $$
   W_\lambda(b)=\mathbb Q.
   $$
10. Definition of:
    $$
    \operatorname{MID}_\lambda(b).
    $$
11. Compression of the Six-to-Four gap into:
    $$
    2\rightarrow1.
    $$
12. Construction of the product-space formulation:
    $$
    UV\subseteq\mathcal L_{\mathrm{alg}}.
    $$
13. Construction of the $\pi$-normalised multiplier window.
14. Connection to the rational-visibility framework of the 2026 Matrix Coefficient Conjecture.

---

# 39. The Formal Entry Point for Round 4

The next round need not rush to reproduce the proof of Six Exponentials.

A more valuable entry point is to study directly:

$$
\boxed{
\operatorname{MID}_\lambda(b)=2
}
$$

exactly which additional structures this hypothetical counterexample state forces.

That is, suppose for contradiction that there exist:

$$
\lambda,
\quad
a\lambda,
\quad
b\lambda,
\quad
ab\lambda
\in
\mathcal L_{\mathrm{alg}},
$$

where:

$$
a,b\notin\mathbb Q.
$$

Then study the layers in turn:

### Layer 1

By Baker / Round 2:

$$
a,b
$$

must be transcendental.

### Layer 2

Study:

$$
ab
$$

whether it can be algebraic or transcendental, and the structural differences between these cases.

### Layer 3

Denote the four algebraic exponentials by:

$$
\alpha=e^\lambda,
\quad
\beta=e^{a\lambda},
\quad
\gamma=e^{b\lambda},
\quad
\delta=e^{ab\lambda}.
$$

Study the determinant relation:

$$
\log\alpha\log\delta
-
\log\beta\log\gamma
=
0
$$

and its branch-independent / branch-dependent structure.

### Layer 4

Compare the Matrix Coefficient Conjecture:

$$
w^TMv=0
$$

with the ACTM multiplier intersection:

$$
W_\lambda(b)
$$

to see whether a more direct bridge can be established.

### Layer 5

Try to find an intermediate obstruction weaker than "all four numbers are algebraic" but still sufficient to cause a contradiction.

The objective of Round 4 is not to claim a direct proof of Four Exponentials, but:

$$
\boxed{
\text{classify the hypothetical }
\operatorname{MID}=2
\text{ state}.
}
$$

---

# 40. One-Sentence Conclusion

> **Six Exponentials has compressed algebraic-log multiplier overlap to at most two dimensions; what Four Exponentials lacks is precisely a proof that every non-rational multiplier can retain only the necessarily existing one-dimensional rational intersection with the original multiplier space.**

Core formula:

$$
\boxed{
\operatorname{MID}_\lambda(b)
=
\dim_{\mathbb Q}
\left(
\mathfrak M_\lambda
\cap
b^{-1}\mathfrak M_\lambda
\right)
}
$$

It is known that:

$$
\boxed{
1\le
\operatorname{MID}_\lambda(b)
\le2.
}
$$

Four Exponentials predicts:

$$
\boxed{
\operatorname{MID}_\lambda(b)=1.
}
$$

This is the formal frontier of ACTM Round 3.

---

# References

1. **Schneider method / Six Exponentials Theorem**, Encyclopedia of Mathematics.  
   https://encyclopediaofmath.org/wiki/Schneider_method

2. Michel Waldschmidt, expository material on interpolation determinants and the Six Exponentials Theorem.  
   https://webusers.imj-prg.fr/~michel.waldschmidt/articles/pdf/EID.pdf

3. Samit Dasgupta, **Ranks of matrices of logarithms of algebraic numbers, I: The theorems of Baker and Waldschmidt–Masser**, *Essential Number Theory*, 2023.  
   https://msp.org/ent/2023/2-1/ent-v2-n1-p05-p.pdf

4. Samit Dasgupta and Mahesh Kakde, **Ranks of Matrices of Logarithms of Algebraic Numbers II: The Matrix Coefficient Conjecture**, *Advances in Mathematics* 487 (2026), 110753.  
   DOI: 10.1016/j.aim.2025.110753  
   arXiv: 2408.08178

5. Four Exponentials Conjecture status reference, TheoremDB, reviewed 2026.  
   https://theoremdb.org/statements/P4/

---

**End of ACTM Round 3**