gpt pro 5.2 attempt on Math-863

Determine which countable ordinals $\beta$ have the property that, if $\alpha=\omega^{^\beta}$, then in any red/blue colouring of the edges of $K_\alpha$ there is either a red $K_\alpha$ or a blue $K_3$. This property holds for $\beta=2$ and not for $3\leq \beta <\omega$ (Specker [Sp57]) and for $\beta=\omega$ (Chang [Ch72]). The first open case is $\beta=\omega^2$ (see [591] ). Galvin and Larson [GaLa74] have shown that if $\beta\geq 3$ has this property then $\beta$ must be 'additively indecomposable', so that in particular $\beta=\omega^\gamma$ for some $\gamma<\omega_1$. Galvin and Larson conjecture that every $\beta\geq 3$ of this form has this property. See also [590] . References [Ch72] Chang, C. C., A partition theorem for the complete graph on {$\omega\sp{\omega }$} . J. Combinatorial Theory Ser. A $1972$, 396-452. [GaLa74] Galvin, Fred and Larson, Jean, Pinning countable ordinals . Fund. Math. $1974/75$, 357-361. [Sp57] Specker, Ernst, Teilmengen von Mengen mit Relationen . Comment. Math. Helv. $1957$, 302-314.

FORMAL RESTATEMENT

Let $\beta$ be a countable ordinal and set $\alpha:=\omega^\beta$ (interpreting the problem's $\omega^{^\beta}$ as $\omega^\beta$). For a red/blue colouring $c:[\alpha]^2\to\{\text{red},\text{blue}\}$, ask whether there must exist either

• a red $K_\alpha$, i.e. $X\subseteq\alpha$ with order type $\alpha$ and $c(\{x,y\})=\text{red}$ for all distinct $x,y\in X$, or
• a blue $K_3$, i.e. $x<y<z<\alpha$ with all three edges blue.

Problem: determine exactly which countable ordinals $\beta$ satisfy $\omega^\beta\to(\omega^\beta,3)^2$.

QUICK LITERATURE/CONTEXT CHECK

I will not import external results beyond what the problem text explicitly states. The text states: the property holds for $\beta=2$ and fails for all finite $\beta\ge 3$ (Specker), and holds for $\beta=\omega$ (Chang). It also states a necessary condition for $\beta\ge 3$ proved by Galvin--Larson: $\beta$ must be additively indecomposable (hence of the form $\omega^\gamma$). The first open case is stated to be $\beta=\omega^2$.

ATTACK PLAN

Proof-track ideas.

• Attempt to generalize Chang's $\beta=\omega$ argument to $\beta=\omega^2$ by iterating a suitable "pinning" or canonical decomposition along the tower $\omega^{\omega\cdot n}$.
• Use the additively indecomposable structure of $\beta$ (i.e. $\beta=\omega^\gamma$) to get strong tail-isomorphism properties of $\alpha=\omega^\beta$ and try to build a red copy by transfinite recursion under no-blue-triangle.

Disproof-track ideas.

• Try to build explicit triangle-free-blue colourings on $\omega^\beta$ when $\beta$ is not additively indecomposable (e.g. $\beta=\gamma+\delta$ with $\delta>0$), by splitting into blocks and coloring cross-block edges in a controlled way.
• Search for the smallest additively indecomposable open case ($\beta=\omega^2$) and attempt a Specker-type construction adapted to the normal form of ordinals $<\omega^{\omega^2}$.

I did not resolve the classification. I record two self-contained lemmas about the ordinal notions mentioned in the statement.

WORK

FAST REALITY CHECK (stated base cases)

• $\beta=1$ gives $\alpha=\omega$; then $\omega\to(\omega,3)^2$ holds by the constructive argument of Lemma lem:omega_case above.
• $\beta=2$ ($\alpha=\omega^2$) is stated in the problem text to satisfy the property (Specker).
• Finite $\beta\ge 3$ are stated in the text to fail (Specker).
• $\beta=\omega$ ($\alpha=\omega^\omega$) is stated in the text to satisfy the property (Chang).
• The first open case is $\beta=\omega^2$.

Lemma 1 (additively indecomposable ordinals are exactly powers of $\omega$). Let $\beta$ be a nonzero countable ordinal. Call $\beta$ additively indecomposable if for all ordinals $\gamma,\delta<\beta$ one has $\gamma+\delta<\beta$. Then $\beta$ is additively indecomposable if and only if $\beta=\omega^\eta$ for some ordinal $\eta$.

Proof. ($\Rightarrow$) Write $\beta$ in Cantor normal form: \[ \beta=\omega^{\eta_0}c_0+\omega^{\eta_1}c_1+\cdots+\omega^{\eta_m}c_m, \] where $m\ge 0$, $\eta_0>\eta_1>\cdots>\eta_m$ are ordinals and $c_i\in\mathbb{N}$ with $c_i\ge 1$. If either $m\ge 1$ or $c_0\ge 2$, we will produce $\gamma,\delta<\beta$ with $\gamma+\delta\ge\beta$, contradicting additively indecomposable.

If $m\ge 1$ (so there is a nonzero "tail"), set \[ \gamma:=\omega^{\eta_0}c_0 \quad\text{and}\quad \delta:=\omega^{\eta_1}c_1+\cdots+\omega^{\eta_m}c_m. \] Then $\gamma<\beta$ (because $\delta>0$) and also $\delta<\beta$ (because $\delta<\omega^{\eta_0}$ while $\beta\ge\omega^{\eta_0}$). But by definition of ordinal addition for Cantor normal form with strictly decreasing exponents, \[ \gamma+\delta=\omega^{\eta_0}c_0+\omega^{\eta_1}c_1+\cdots+\omega^{\eta_m}c_m=\beta, \] which contradicts $\gamma+\delta<\beta$.

If instead $m=0$ but $c_0\ge 2$, set \[ \gamma:=\omega^{\eta_0}(c_0-1)\quad\text{and}\quad\delta:=\omega^{\eta_0}. \] Then $\gamma<\beta$ and $\delta<\beta$. Moreover ordinal addition here agrees with multiplication by a natural number: \[ \gamma+\delta=\omega^{\eta_0}(c_0-1)+\omega^{\eta_0}=\omega^{\eta_0}c_0=\beta, \] again contradicting additively indecomposable.

Therefore we must have $m=0$ and $c_0=1$, i.e. $\beta=\omega^{\eta_0}$.

($\Leftarrow$) Now assume $\beta=\omega^\eta$. Let $\gamma,\delta<\omega^\eta$. Write $\gamma$ and $\delta$ in Cantor normal form. In particular, the largest exponent appearing in either is some ordinal $<\eta$. Let $\theta<\eta$ be the maximum of those leading exponents. Then $\gamma<\omega^{\theta+1}$ and $\delta<\omega^{\theta+1}$. We claim that if $a,b<\omega^{\theta+1}$ then $a+b<\omega^{\theta+1}$. Indeed $\omega^{\theta+1}=\omega^{\theta}\cdot\omega$, so for each $a<\omega^{\theta+1}$ there exist $m<\omega$ and $r<\omega^{\theta}$ with $a=\omega^{\theta}m+r$; similarly $b=\omega^{\theta}n+s$ with $n<\omega$ and $s<\omega^{\theta}$. Then, since $r<\omega^{\theta}$ and $n>0$ implies $r+\omega^{\theta}n=\omega^{\theta}n$ (and if $n=0$ the statement is trivial), we compute \[ a+b=(\omega^{\theta}m+r)+(\omega^{\theta}n+s)=\omega^{\theta}m+(r+\omega^{\theta}n)+s=\omega^{\theta}(m+n)+s. \] Because $m+n<\omega$ and $s<\omega^{\theta}$, we have $\omega^{\theta}(m+n)+s<\omega^{\theta}(m+n+1)\le\omega^{\theta}\cdot\omega=\omega^{\theta+1}$. Applying this with $a=\gamma$ and $b=\delta$ gives $\gamma+\delta<\omega^{\theta+1}$. Finally, because $\theta+1\le\eta$ we have $\omega^{\theta+1}\le\omega^\eta$ and hence $\gamma+\delta<\omega^{\theta+1}\le\omega^\eta=\beta$. Thus $\beta$ is additively indecomposable.$\square$

Lemma 2 (tail self-similarity of $\omega^\beta$ for $\beta>0$). Let $\beta>0$ be an ordinal and set $\alpha:=\omega^\beta$. Then for every $\xi<\alpha$, the tail interval $[\xi,\alpha):=\{\eta:\xi\le \eta<\alpha\}$ is order-isomorphic to $\alpha$. Equivalently, $\xi+\alpha=\alpha$ for all $\xi<\alpha$.

Proof. Since $\beta>0$, $\alpha=\omega^\beta$ is a (nonzero) power of $\omega$, hence by Lemma lem:add_indec it is additively indecomposable as an ordinal. In particular for any $\xi<\alpha$ we have $\xi+\alpha=\alpha$. Consider the map $f:\alpha\to [\xi,\alpha)$ given by $f(\eta)=\xi+\eta$. Ordinal addition is strictly increasing in the second argument, so $f$ is strictly increasing. Its image is $[\xi,\xi+\alpha)=[\xi,\alpha)$ because $\xi+\alpha=\alpha$. Thus $f$ is an order-isomorphism from $\alpha$ onto the tail $[\xi,\alpha)$.$\square$

VERIFICATION

• Lemma lem:add_indec: the forward direction uses only Cantor normal form and explicit choices of $\gamma,\delta$ witnessing decomposability. The backward direction reduces the sum to a smaller power $\omega^{\theta+1}$ using the leading exponent. All steps are ordinal-arithmetic checks.
• Lemma lem:tail_iso: verifies that the tail has the same order type via the explicit order-isomorphism $\eta\mapsto \xi+\eta$.

FINAL

UNRESOLVED

(i) Strongest proved partial result. The ordinal constraint "$\beta$ additively indecomposable" appearing in the problem text is equivalent to "$\beta=\omega^\eta$" (Lemma lem:add_indec), and for $\alpha=\omega^\beta$ (with $\beta>0$) every tail $[\xi,\alpha)$ is order-isomorphic to $\alpha$ (Lemma lem:tail_iso). These structural facts are consistent with the conjectural "self-similarity" needed to build red copies of $K_\alpha$.

(ii) First gap (crisp). Decide whether $\omega^{\omega^2}\to(\omega^{\omega^2},3)^2$ holds (the first open case $\beta=\omega^2$), i.e. either construct a triangle-free-blue colouring of $[\omega^{\omega^2}]^2$ with no red $K_{\omega^{\omega^2}}$, or give a proof that none exists.

(iii) Top 3 next moves.

1. Using Lemma lem:tail_iso, attempt a transfinite recursion that builds an increasing red set of type $\omega^{\omega^2}$ from red sets of type $\omega^{\omega\cdot m}$, under the hypothesis "no blue triangle".
2. Isolate a "one-step" strengthening: prove that for some $m$, any triangle-free-blue colouring on $\omega^{\omega\cdot(m+1)}$ contains a red $\omega^{\omega\cdot(m+1)}$ provided it contains cofinally many red $\omega^{\omega\cdot m}$ blocks.
3. Try to explicitly generalize Specker-type counterexamples from $\omega^n$ ($n\ge 3$) to exponents with Cantor normal form having at least two terms (i.e. $\beta$ not additively indecomposable), to re-prove the Galvin--Larson necessity direction by direct construction.

(iv) Minimal counterexample structure. A minimal counterexample at $\beta=\omega^2$ would be a colouring on $\omega^{\omega^2}$ with no blue triangle and no red subset of order type $\omega^{\omega^2}$. By restriction to smaller initial segments, such a colouring would still need to contain red copies of the largest known true cases (e.g. a red $\omega^\omega$), while carefully arranging blue edges between blocks to block extension to order type $\omega^{\omega^2}$.