diff --git a/.gitignore b/.gitignore index 5a396e8..25669e1 100644 --- a/.gitignore +++ b/.gitignore @@ -306,4 +306,5 @@ TSWLatexianTemp* F4D1-Analysis-and-Geometry-on-Manifolds/Analysis_on_Manifolds_Notes.pdf V4A1-Algebraic-Geometry-I/Algebraic_Geometry_I_Notes.pdf V5A2-Habiro-Rings/Habiro_Rings_Notes.pdf -V5B1-Advanced-Topics-in-Complex-Analysis/Advanced_Complex_Analysis_Notes.pdf \ No newline at end of file +V5B1-Advanced-Topics-in-Complex-Analysis/Advanced_Complex_Analysis_Notes.pdf +V5A2-Habiro-Rings/._wordcount_selection.tex diff --git a/V5A2-Habiro-Rings/Lecture 4.tex b/V5A2-Habiro-Rings/Lecture 4.tex index 9c98abe..0104500 100644 --- a/V5A2-Habiro-Rings/Lecture 4.tex +++ b/V5A2-Habiro-Rings/Lecture 4.tex @@ -89,12 +89,12 @@ \section{Lecture 4 -- 15th November 2024}\label{sec: lecture 4} \end{align*} as desired. \end{proof} -We can modify the Nahm sum to rid ourselves of the $q^{a/2}$ factor in (\ref{eqn: q-difference equation of 1x1 Nahm sum}).\marginpar{The exposition that follows is drawn from the erratum discussed at the begining of lecture 5. The discussion below is likely extremely error-prone and the reader is encouraged to consult \cite{NahmAsymptotics} for full details.} +We can modify the Nahm sum to rid ourselves of the $q^{a/2}$ factor in (\ref{eqn: q-difference equation of 1x1 Nahm sum}).\marginpar{The exposition that follows is drawn from the erratum discussed at the begining of lecture 5. The reader is encouraged to consult \cite{NahmAsymptotics} for full details.} \begin{corollary}\label{corr: q-difference equation of modified 1x1 Nahm sum} Let $$f_{a}^{\mathrm{mod}}(t,q)=f_{a}(q^{-a/2}t,q)=\sum_{n\geq0}(-1)^{an}\frac{q^{\frac{1}{2}an^{2}-\frac{1}{2}an}}{(q;q)_{n}}t^{n}$$ Then the modified Nahm sum satisfies the $t$-deformed $q$-difference equation - \begin{equation}\label{eqn: q-difference equation} + \begin{equation}\label{eqn: q-difference equation of modified 1x1 Nahm sum} f_{a}^{\mathrm{mod}}(t,q) - f_{a}^{\mathrm{mod}}(qt,q) = (-1)^{a}t\cdot f_{a}^{\mathrm{mod}}(q^{a}t,q). \end{equation} \end{corollary} @@ -111,7 +111,7 @@ \section{Lecture 4 -- 15th November 2024}\label{sec: lecture 4} f_{a}(t,q)=\exp\left(\frac{V(t)}{h}\right)g_{a}(t,q) \end{equation} \begin{equation}\label{eqn: qt 1x1 Nahm sum ansatz} - f_{a}(t,q)=\exp\left(\frac{V(qt)}{h}\right)g_{a}(qt,q) + f_{a}(qt,q)=\exp\left(\frac{V(qt)}{h}\right)g_{a}(qt,q) \end{equation} in what follows, with $V(t)\in\QQ[[t]]$ satisfying $V(0)=0$. \Cref{eqn: 1x1 Nahm sum ansatz,eqn: qt 1x1 Nahm sum ansatz} takes are related by the formula we now describe. \begin{proposition}\label{prop: V functional equation for 1x1 Nahm ansatz} @@ -134,18 +134,22 @@ \section{Lecture 4 -- 15th November 2024}\label{sec: lecture 4} \end{proposition} \begin{corollary}\label{corr: exponent ratio for 1x1 Nahm sum antsatz} Let $V(t)$ be as in the Nahm equation ansatz. The ratios of the exponential factors satisfy - $$\exp\left(\frac{V(qt)}{h}\right)/\exp\left(\frac{V(t)}{h}\right)\sim\exp((\partial^{\log}V)(t))(1+O(h)).$$ + $$\exp\left(\frac{V(qt)}{h}\right)/\exp\left(\frac{V(t)}{h}\right)\sim\exp((\partial^{\log}V)(t))(1+O(h))$$ + for $q=e^{h}$. \end{corollary} \begin{proof} - We have + We have by \Cref{prop: V functional equation for 1x1 Nahm ansatz} \begin{align*} \exp\left(\frac{V(qt)}{h}\right) &= \exp\left(\frac{V(t)}{h}+(\partial^{\log}V)(t)+\frac{h}{2}((\partial^{\log})^{2}V)(t)\right)\\ &=\exp\left(\frac{V(t)}{h}\right)\exp\left((\partial^{\log}V)(t)\right)\exp\left(\frac{h}{2}((\partial^{\log})^{2}V)(t)\right) \end{align*} - where the first term of the product above cancels in the ratio and the final term of the product expands to a power series in $h$. + where the first term of the product above cancels in the ratio and the final term of the product expands to a power series in $h$ with constant coefficient 1. \end{proof} -Of special interest to us will be the final terms of the product above, where we denote $Z(t)=\exp((\partial^{\log}V)(t))$ and $\widetilde{Z}(t,q)=\exp\left((\partial^{\log}V)(t)\right)\exp\left(\frac{h}{2}((\partial^{\log})^{2}V)(t)\right)$. -This allows us to produce a functional equation of the $g_{a}$ appearing in \Cref{eqn: 1x1 Nahm sum ansatz,eqn: qt 1x1 Nahm sum ansatz} which we will soon revisit. Note specializing at $q-1$ produces +Denoting $Z(t)=\exp((\partial^{\log}V)(t))$ and $\widetilde{Z}(t,q)=\exp\left((\partial^{\log}V)(t)\right)\exp\left(\frac{h}{2}((\partial^{\log})^{2}V)(t)\right)$, we divide (\ref{eqn: q-difference equation of modified 1x1 Nahm sum}) by the exponential prefactor $\exp\left(\frac{V(t)}{h}\right)$ and the ansatz \Cref{eqn: 1x1 Nahm sum ansatz,eqn: qt 1x1 Nahm sum ansatz} to observe +$$g_{a}(t,q)-\widetilde{Z}(t,q)g_{a}(qt,q)=(-1)^{a}\cdot t\cdot\prod_{i=0}^{a-1}\widetilde{Z}(q^{i}t,q)\cdot g_{a}(q^{a},t)$$ +which we seek to show lies in $\QQ[[t,q-1]]$. + +Specializing at $q=1$ (ie. $h=0$) produces \begin{equation}\label{eqn: specialized Z functional equation} 1-Z(t)=(-1)^{a}t\cdot Z(t)^{a} \end{equation} diff --git a/V5A2-Habiro-Rings/Lecture 5.tex b/V5A2-Habiro-Rings/Lecture 5.tex index 431e330..e981e2f 100644 --- a/V5A2-Habiro-Rings/Lecture 5.tex +++ b/V5A2-Habiro-Rings/Lecture 5.tex @@ -1,26 +1,39 @@ \section{Lecture 5 -- 22nd November 2024}\label{sec: lecture 5} -We continue our discussion of $q$-series and in particular a property of the modified Nahm sum considered in \Cref{corr: q-difference equation of modified 1x1 Nahm sum}.\marginpar{As it stands, the proofs and structure for this lecture are more rough than usual, and will be updated in due course.} +We continue our discussion of $q$-series and in particular a property of the modified Nahm sum considered in \Cref{corr: q-difference equation of modified 1x1 Nahm sum}. The following definition is due to Konsevich-Soibelman \cite{DTInvariants}. \begin{definition}[Admissable Series]\label{def: admissable series} - A series $f\in\ZZ((q))[[t]]$ is admissable if it can be written as + A series $f\in\ZZ((q))[[t]]$ such that $f\equiv 1\pmod{(t)}$ is admissable if it can be written as $$f=\prod_{n\geq1}\prod_{i\in\ZZ}(q^{i}t^{n};q)_{\infty}^{a_{n,i}}$$ such that for each $n$ only finitely many $a_{n,i}$ are nonzero. \end{definition} \begin{remark} - These $a_{n,i}$'s are precisely Donaldson-Thomas invariants. + These $a_{n,i}$'s are precisely Donaldson-Thomas invariants that arise in Gromov-Witten theory and enumerative geometry. \end{remark} Admissable series force an algebraicity condition on the $q$, allowing $f$ to be written as an element of $\ZZ[q][[t]]$. Up to a condition on the residue of the series $f$ mod $(t)$, series in $\ZZ((q))[[t]]$ admit such an expansion. \begin{proposition} - Let $f\in\ZZ((q))[[t]]$. If $f\equiv1\pmod{(t)}$ then $f$ admits a unique expansion as an admissable series. + Let $f\in\ZZ((q))[[t]]$. If $f\equiv1\pmod{(t)}$ then $f$ admits a unique expansion as a series of the form + $$f=\prod_{n\geq1}\prod_{i\in\ZZ}(q^{i}t^{n};q)_{\infty}^{a_{n,i}}.$$ \end{proposition} +\begin{proof}[Proof Outline] + This can be solved for truncated polynomials so for $f\equiv 1\pmod{(t)}$, it suffices to consider $f\equiv 1\pmod{(t^{m})}$ and solve inductively to give an expression algebraic in $t$. +\end{proof} This result is in fact much more general and it can be shown that the modified Nahm sum $$f_{a}(t,q)=\sum_{n\geq0}(-1)^{an}\frac{q^{\frac{1}{2}an^{2}-\frac{1}{2}an}}{(q;q)_{n}}t^{n}$$ as previously defined is admissable. The original proof is highly involved, and we will instead offer a simpler exposition of the same result. Recall from \Cref{prop: logarithm at worst simple poles at roots of unity}, we have \begin{equation}\label{eqn: modified Nahm sum as exponent of sum} - (q^{i}t;q)_{\infty}=\exp\left(-\sum_{\ell\geq 1}\frac{1}{\ell}\cdot\frac{q^{i\ell}t^{n\ell}}{1-q^{\ell}}\right) + (q^{i}t;q)_{\infty}=\exp\left(-\sum_{\ell\geq 1}\frac{1}{\ell}\cdot\frac{q^{i\ell}t^{n\ell}}{1-q^{\ell}}\right). \end{equation} -and further recall that $\ZZ((q))[[t]]$ is a $\lambda$-ring -- admits an action with $\NN$ as a multiplicative monoid -- and admits Adams operations $\psi_{n}:\ZZ((q))[[t]]\to\ZZ((q))[[t]]$ by $t\mapsto t^{n},q\mapsto q^{n}$. The Adams operations allow us to rewrite (\ref{eqn: modified Nahm sum as exponent of sum}) as +Furthermore, $\ZZ((q))[[t]]$ is a $\lambda$-ring as we now define. +\begin{definition}[$\lambda$-Ring]\label{def: lambda-ring} + A $\lambda$-ring $A$ is a ring equipped with set maps $\lambda^{k}:A\to A$ for $0\leq k\leq\infty$ such that + \begin{enumerate}[label=(\roman*)] + \item $\lambda^{0}(a)=1$ for all $a\in A$. + \item $\lambda^{1}(a)=a$ for all $a\in A$. + \item $\lambda^{n}(a+b)=\sum_{i+j=n}\lambda^{i}(a)\lambda^{i}(b)$ for all $a,b\in A$. + \end{enumerate} +\end{definition} +Such a ring admits Adams operations $\psi_{n}:\ZZ((q))[[t]]\to\ZZ((q))[[t]]$ by $t\mapsto t^{n},q\mapsto q^{n}$. The Adams operations allow us to rewrite (\ref{eqn: modified Nahm sum as exponent of sum}) as \begin{equation}\label{eqn: modified Nahm sum as exponent of Adams sum} \exp\left(-\sum_{\ell\geq1}\psi_{\ell}\left(\frac{q^{i}t^{n}}{1-q}\right)\right). \end{equation} @@ -31,13 +44,22 @@ \section{Lecture 5 -- 22nd November 2024}\label{sec: lecture 5} \end{definition} Now taking $$\phi(t,q)=-\sum_{n\geq1}\sum_{i\in\ZZ}a_{n,i}q^{i}t^{n}\in\ZZ((q))[[t]]$$ -writing $\frac{\phi(t,q)}{1-q}$ as the plethystic logarithm of $f_{a}(t,q)$, inverse to the plethystic exponential, the coefficients of the expansion as an admissable series will be those coefficients of the plethystic logarithm since the plethystic exponential gives an isomorphism $t\QQ[q^{\pm}][[t]]\to 1+t\QQ[q^{\pm}][[t]]$. It thus suffices to show that the plethystic logarithm of $f_{a}$ is a function that is a sum $\frac{\phi_{0}(t)}{1-q}$ with an element of $\QQ[q^{\pm}][[t]]$. +we have +\begin{equation*}\label{eqn: adams admissable Nahm sum} + \prod_{n\geq1}\prod_{i\in\ZZ}(q^{i}t;q)_{\infty}=\exp\left(\sum_{\ell\geq1}\frac{1}{\ell}\psi_{\ell}\left(\frac{\phi(t,q)}{1-q}\right)\right). +\end{equation*} +We can define the plethystic logarithm as the inverse of the plethystic exponential and observe $\frac{\phi(t,q)}{1-q}$ is the plethystic logarithm of the modified Nahm sum $f_{a}(t,q)$. We then seek to show that this plethystic logarithm $\frac{\phi(t,q)}{1-q}\in\frac{1}{1-q}\ZZ[q^{\pm}][[t]]$ which has a single simple pole at $q=1$. Indeed, this suffices as the behavior at other roots of unity are determined by the Adams operations. + + +Observe the plethystic exponential gives an isomorphism $t\QQ[q^{\pm}][[t]]\to 1+t\QQ[q^{\pm}][[t]]$. It thus suffices to show that the plethystic logarithm of $f_{a}$ is a function that is a sum $\frac{\phi_{0}(t)}{1-q}$ with an element of $\QQ[q^{\pm}][[t]]$. Now using the ansatz \begin{equation}\label{eqn: plethystic exponential ansatz} f_{a}(t,q)=\exp\left(\sum_{\ell\geq 1}\frac{1}{\ell}\frac{\phi_{0}(t^{\ell})}{1-q^{\ell}}\right)g_{a}(t,q) \end{equation} -we show that there exists a chioce of function $\psi_{0}(t)$ so that the remainder $g_{a}(t,q)\in 1+t\QQ[q^{\pm}][[t]]$ and from which the result would follow by application of the Plethystic exponential. But a choice of $\phi_{0}\in t\cdot\QQ[[t]]$ can be made such that $\sum_{\ell\geq 1}\frac{\phi_{0}(t^{\ell})}{\ell^{2}}=-V(t)$. +we show that there exists a chioce of function $\phi_{0}(t)$ satisfying the ansatz above would imply the factor $g_{a}(t,q)\in 1+t\QQ[q^{\pm}][[t]]$ and from which the result of showing $g_{a}(t,q)$ lying in $1+t\QQ[t^{\pm}][[t]]$ would follow by application of the Plethystic exponential. + +But a choice of $\phi_{0}\in t\cdot\QQ[[t]]$ can be made such that $\sum_{\ell\geq 1}\frac{\phi_{0}(t^{\ell})}{\ell^{2}}=-V(t)$ whose plethystic exponential has leading term asymptotics agreeing with $f_{a}(t,q)$ at all roots of unity. This is gives the desired result as stated below. \begin{theorem}[Kontsevich-Soibelman, Efimov]\label{thm: modified 1x1 Nahm sum is admissable} @@ -45,4 +67,14 @@ \section{Lecture 5 -- 22nd November 2024}\label{sec: lecture 5} $$f_{a}(t,q)=\sum_{n\geq0}(-1)^{an}\frac{q^{\frac{1}{2}an^{2}-\frac{1}{2}an}}{(q;q)_{n}}t^{n}$$ is admissable. \end{theorem} +Let us unfurl some of the consequences of \Cref{thm: modified 1x1 Nahm sum is admissable}. +To understand the $\phi_{0}(t)$ function better, we use compute its logarithmic derivative so in conjunction with \Cref{prop: V functional equation for 1x1 Nahm ansatz} we have +$$\sum_{\ell\geq1}\frac{(\partial^{\log}\phi_{0})(t^{\ell})}{\ell}=-\log Z(t)$$ +with $Z(t)$ is the logarithmic derivative of $V(t)$. Note that $Z(t)\in 1+t\ZZ[[t]]$ so $(\partial^{\log}\phi_{0})(t)=-\sum_{n\geq1}c_{n}t^{n}$ for $Z(t)=\prod_{n\geq1}(1-t^{n})^{c_{n}}$ for $c_{n}\in\ZZ$. This shows $n|c_{n}$. +\begin{example} + For $a=0$, $f_{1}(t,q)=(t;q)^{-1}_{\infty}$ so $Z(t)=1-t$ and $\phi_{0}(t)=\pm t$ which agrees with $V(t)$ being the dilogarithm (up to a sign). +\end{example} +\begin{example} + In the case $a=2$ which was discussed in \Cref{sec: lecture 1}, we recover $Z(t)$ as an alternating sum of the Catalan numbers. +\end{example} \ No newline at end of file