path: root/presentation.tex

\documentclass{presentation}

\title{Development of \\ Frequency-Domain Multidimensional Spectroscopy}
\subtitle{---Beyond Two Dimensions---}
\author{Blaise Thompson}

\institute{University of Wisconsin--Madison}
\date{2018-04-23}

\begin{document}
\maketitle

\begin{frame}{Introduction to CMDS}
  \begin{columns}
    \begin{column}{0.6\textwidth}
      \adjincludegraphics[scale=0.2]{presentation/SK_PhDThesis_fsTable-Overview} \\
      \tiny
      Figure courtesy of Schuyler Kain
    \end{column}
    \begin{column}{0.4\textwidth}
      \includegraphics[width=\textwidth]{"literature/BrownEmilyJ1999a_1"}
      \centering
      \\
      \vspace{2\baselineskip}
      $\vec{k_{\text{sig}}} = \vec{k_a} - \vec{k_b} + \vec{k_c}$
      \vspace{2\baselineskip} \\
      \tiny \raggedright
      Figure: \\
      Brown, E., Zhang, Q. and Dantus, M. (1999). \\
      The Journal of Chemical Physics, 110(12), pp.5772-5788.
    \end{column}
  \end{columns}
\end{frame}

\begin{frame}{Introduction to CMDS: microscopic picture}
  \adjincludegraphics[width=\textwidth]{"mixed_domain/simulation overview"}
\end{frame}

\begin{frame}{Diversity}
  Great diversity of experimental strategies under the ``umbrella'' of CMDS:
  \vspace{\baselineskip} \\
  \begin{columns}
    \begin{column}{0.5\textwidth}
      Experimental geometry...
      \begin{itemize}
        \item $\vec{k_a} - \vec{k_b} + \vec{k_c}$
        \item $\vec{k_a} + \vec{k_b} + \vec{k_c}$
        \item $\vec{k_a} - \vec{k_a} + \vec{k_b} + \vec{k_c} + \vec{k_d}$
      \end{itemize}
    \end{column}
    \begin{column}{0.5\textwidth}
      Dimensions explored...
      \begin{itemize}
        \item MIR \& visible: DOVE, TRSF
        \item fully visible: TREE, CARS
        \item frequency-frequency: 2DES/2DIR, ``Resonant-(Raman/IR)''
        \item frequency-delay: TG, TA
        \item delay-delay: 3PE, MUPPETS
      \end{itemize}
    \end{column}
  \end{columns}
  \vphantom{m} \\
  Or 3D.. or 4D: many possibilities not yet popular enough to name  
  % SAY: based on the same basic ability to scan pulses in frequency, delay etc
  %   all possible with one instrument
  %   each has been focus of entire careers...
\end{frame}

\begin{frame}{My focus}
  Focus on the \emph{pipeline} of CMDS:
  \begin{itemize}
    \item throughput
    \item quality
    \item diversity
  \end{itemize}
  \vphantom{M} \\
  Unlock the true potential of these instruments:
  \begin{itemize}
    \item automated calibration
    \item 2D, 3D, 4D...
    \item full diversity of possible hardware combinations, rapid development
    \item powerful and flexible detection strategies
    \item data processing tools
  \end{itemize}
\end{frame}

\section{Tunability}  % ===========================================================================

\begin{frame}{Tunability}
  \centering \huge
  Control and Calibration of \\
  Optical Parametric Amplifiers
\end{frame}

\begin{frame}{Two strategies for CMDS}
  Two strategies for collecting multidimensional spectra:
  \vspace{\baselineskip} \\
  \begin{columns}
    \begin{column}{0.4\textwidth}
      Time Domain
      \begin{itemize}
        \item broadband pulses
        \item resolve spectra interferometrically
        \item fast (even single shot)
        \item robust
      \end{itemize} 
    \end{column}
    \begin{column}{0.4\textwidth}
      Frequency Domain
      \begin{itemize}
        \item narrowband pulses
        \item resolve spectra by tuning OPAs directly
        \item slow (lots of motor motion)
        \item fragile
      \end{itemize}
    \end{column}
  \end{columns}
\end{frame}

\begin{frame}{Two strategies for CMDS}
  \begin{columns}
    \begin{column}{0.5\textwidth}
      Time Domain
      \includegraphics[width=\textwidth]{"literature/SinghAkshay2014a_2"}
      \tiny
      Figure: \\
      Singh, A., Moody, G., Wu, S., Wu, Y., Ghimire, N., Yan, J., Mandrus, D., Xu, X. and Li, X.
      (2014).
      Coherent Electronic Coupling in Atomically Thin MoSe$_2$. Physical Review Letters, 112(21).
    \end{column}
    \begin{column}{0.5\textwidth}
      Frequency Domain
      \adjincludegraphics[width=\textwidth]{presentation/singh_czech}
      More \hl{bandwidth}.
      Crucial for electronic states, band structure.
    \end{column}
  \end{columns}
  % SAY:
  % - time domain: broadband, interferometric, fast
  % - frequency domain: narrowband, direct, slow but broad bandwidth
  % - bandwidth crucial for exploring diverse states
\end{frame}

\begin{frame}{Bandwidth}
  A lot more bandwidth... through the usage of OPAs
  \adjincludegraphics[width=\textwidth]{opa/OPA_ranges}
  \phantom{M} \hfill but how to make this strategy easy and \hl{robust}?
  % SAY: previous spectrum was SHS
  % SAY: crystal angle, temporal and spatial overlap
\end{frame}

\begin{frame}{Automation}
  \begin{columns}
    \begin{column}{0.5\textwidth}
      \adjincludegraphics[width=\textwidth]{opa/autotune_preamp}
    \end{column}
    \begin{column}{0.5\textwidth}
      Fully automated OPA tuning
      \begin{itemize}
        \item less than 1 hour per OPA
        \item can be scheduled during down time
        \item high quality from global analysis 
        \item reproducible
        \item unambiguous representations automatically generated to assess health
      \end{itemize}
      \vspace{\baselineskip}
      Other calibration also needed, automated.
    \end{column}
  \end{columns}
  % SAY: one example for one stage in one model of OPA (out of four in WG)
\end{frame}

\section{Acquisition}  % ==========================================================================

\begin{frame}{Acquisition}
  \centering \huge
  Control of the MR-CMDS \\
  Instrument
\end{frame}

\begin{frame}{The instrument}
  Many kinds of component hardware
  \begin{itemize}
    \item monochromators
    \item delay stages
    \item filters
    \item OPAs
  \end{itemize}
  $\sim10$ settable devices, $\sim25$ motors, multiple detectors. \\
  \vphantom{M} \\
  \hl{Coordination} problem.
\end{frame}

\begin{frame}{Acquisition}
  PyCMDS---unified software for controlling hardware and collecting data.
  \adjincludegraphics[width=\textwidth]{acquisition/screenshots/000}
\end{frame}

\begin{frame}{Acquisition}
  Capabilities of PyCMDS:
  \begin{itemize}
    \item reconfigures itself based on available hardware (modularity)
    \item multithreaded (up to 2x speed enhancement)
    \item queued acquisitions
      \begin{itemize}
        \item long scans, short window of calibration---large duty cycle needed
        \item shortly after implementation, two weeks of data collection yielded as many pixels as
          the previous three years
      \end{itemize}
    \item extensibility
      \begin{itemize}
        \item easy to add new hardware, new sensors, and new acquisition strategies
        \item typical addition $\sim100$ lines of new Python code
      \end{itemize}
  \end{itemize}
\end{frame}

\begin{frame}{Extensibility}
  \begin{columns}
    \begin{column}{0.25\textwidth}
      \adjincludegraphics[width=\textwidth]{presentation/hardware}
    \end{column}
    \begin{column}{0.7\textwidth}
      Easy to add new hardware to PyCMDS
      \begin{itemize}
        \item In 2016, a new OPA was added to the picosecond system in one day.
        \item In 2017, we added multiple delay stages to the femtosecond system. Implementation
          took between one and four hours.
      \end{itemize}
      Once added, new hardware is immediately available for scanning in a multidimensional space
      with other hardware---no additional programming needed!
    \end{column}
  \end{columns}
\end{frame}

\section{Measurement enhancements}  % =============================================================

\begin{frame}{Measurement enhancements}
  \centering \huge
  Measurement enhancements
\end{frame}

\begin{frame}{Digital processing}
  \large
  boxcar averager \hfill digitize immediately
  \begin{columns}
    \begin{column}{0.5\textwidth}
      \adjincludegraphics[width=\textwidth]{presentation/IMG_20180422_145632}
    \end{column}
    \begin{column}{0.5\textwidth}
      \adjincludegraphics[width=\textwidth]{presentation/measure_digitize}
    \end{column}
  \end{columns}
  \vphantom{M} \\
  \begin{itemize}
    \item cheaper, fewer points of failure
    \item more flexibility for different detector configurations
    \item shot-level statistics, processing sequences
    \begin{itemize}
      \item configurable through simple python script
    \end{itemize}
    \item $\sim3\times$ faster
  \end{itemize}
\end{frame}

\begin{frame}{Dual chopping}
  \begin{columns}
    \begin{column}{0.5\textwidth}
      \adjincludegraphics[width=\textwidth]{presentation/IMG_20180422_145757}
    \end{column}
    \begin{column}{0.5\textwidth}
      \begin{tabular}{ r | c | c | c | c }
        & A       & B          & C          & D                       \\
        signal    &            &            & \checkmark &            \\
        scatter 1 &            & \checkmark & \checkmark &            \\
        scatter 2 &            &            & \checkmark & \checkmark \\
        other     & \checkmark & \checkmark & \checkmark & \checkmark
      \end{tabular}
    \end{column}
  \end{columns}
  \centering
  \vphantom{M}
  $\mathsf{I_{signal} = A - B + C - D}$ \\
  \phantom{M} \\
  \raggedright
  Isolate signal that depends on \emph{all} indecent beams.
  \begin{itemize}
    \item{no scatter}
    \item{no competing signals}  % e.g. 2k1+k3 vs k1+k2+k3
    \item{no voltage offset or room lights}
  \end{itemize}
\end{frame}

\begin{frame}{Digital processing}
  \adjincludegraphics[width=\textwidth]{presentation/chopped}
  % SAY: this is a GOOD example
  % - we now regularly extract signals that would be completely invisible on the 'detected' side
\end{frame}

\section{Processing}  % ===========================================================================

\begin{frame}{Processing}
  \centering \huge
  Data processing
\end{frame}

\begin{frame}{Processing}
  \begin{columns}
    \begin{column}{0.6\textwidth}
      Great! We have \emph{lots} of CMDS data. \\
      Now what? \\
      \vphantom{M} \\
      Working with multidimensional data is hard...
      \begin{itemize}
        \item storage
        \item visualization
        \item post-processing
        \item fitting, modeling
      \end{itemize}
      and the dimensions are always changing! \\
      \vphantom{M} \\
      WrightTools---software to process CMDS.
    \end{column}
    \begin{column}{0.4\textwidth}
      \adjincludegraphics[width=\textwidth]{presentation/TOC}
    \end{column}
  \end{columns}
  % SAY: multidimensional includes > 2 dimensions
\end{frame}

\subsection{Universal format}  % ------------------------------------------------------------------

\begin{frame}{Universal format}
  WrightTools defines a \emph{universal file format} for CMDS.
  \begin{itemize}
    \item store multiple multidimensional arrays
    \item metadata
  \end{itemize}
  Import data from a variety of sources.
  \begin{itemize}
    \item previous Wright Group acquisition software
    \item commercial instruments (JASCO, Shimadzu, Ocean Optics)
    \item simulation packages
  \end{itemize}
\end{frame}

\subsection{Flexible data model}  % ---------------------------------------------------------------

\begin{frame}{Flexible data model}
  Flexibility to transform into any desired ``projection'' on component variables.
  \adjincludegraphics[width=\textwidth]{processing/fringes_transform}
  % SAY: 'derived dimensionality' - DJM
  % - including expressions
\end{frame}

\subsection{Integrations}  % ----------------------------------------------------------------------

\begin{frame}{Integrations}
  \begin{columns}
    \begin{column}{0.6\textwidth}
      \adjincludegraphics[width=\textwidth]{PbSe_global_analysis/movies_fitted}
    \end{column}
    \begin{column}{0.4\textwidth}
      \begin{itemize}
        \item WrightTools as a backend
        \item puts models and experiments on the same footing
        \item makes custom modeling work easier
        \item more general-purpose modeling coming soon
      \end{itemize}
    \end{column}
  \end{columns}
\end{frame}

\section{Conclusion}  % ===========================================================================

\begin{frame}{Conclusion}
  \begin{columns}
    \begin{column}{0.5\textwidth}
      \adjincludegraphics[height=1.5in]{opa/autotune_preamp}
      \adjincludegraphics[width=\textwidth]{acquisition/screenshots/000}
    \end{column}
    \begin{column}{0.5\textwidth}
      \adjincludegraphics[width=\textwidth]{presentation/chopped}
      \phantom{M}
      \adjincludegraphics[width=\textwidth]{processing/fringes_transform}
    \end{column}
  \end{columns}
\end{frame}

\begin{frame}{Acknowledgments}
  \begin{columns}
    \begin{column}{0.5\textwidth}
      Wright Group
      \begin{itemize}
        \item Kyle Sunden
        \item Natalia Spitha
        \item Darien Morrow
        \item Jonathan Handali
        \item Nathan Neff-Mallon
        \item Kyle Czech         
        \item Dan Kohler
        \item Erin Boyle
        \item Paul Hebert
        \item Skye Kain
        \item John
        \item (and more...)
      \end{itemize}
    \end{column}
    \begin{column}{0.5\textwidth}
      Committee
      \begin{itemize}
        \item Kyoung-Shin Choi
        \item Randall Goldsmith
        \item Tim Bertram
      \end{itemize}
      \phantom{M}
      UW-Madison Chemistry Department
      \begin{itemize}
        \item Rob McClain
        \item Pam Doolittle
        \item Bill Goebel
        \item Steve Myers
      \end{itemize}
      \phantom{M}
      Friends and family \\
      \phantom{M} \\
      You, the audience!
      \hl{Questions?}
    \end{column}
  \end{columns}
\end{frame}
  
\section{Supplement}  % ===========================================================================

\begin{frame}{TOPAS-C}
  One of four models of OPAs used within the Wright Group.
  \includegraphics[width=\textwidth]{opa/TOPAS-C}
  Two ``stages'', each with two motorized optics.
\end{frame}

\begin{frame}{Tuning}
  % TODO: curve plot?
  Tuning curves---recorded correspondence between motor positions and output color.
  \vspace{\baselineskip} \\
  Exquisite sensitivity to alignment and lab conditions---tuning required roughly once a week.
  \vspace{\baselineskip} \\
  Manual tuning is difficult...
  \begin{itemize}
    \item high dimensional motor space
    \item difficult to asses overall quality
    \item several hours of work per OPA (sometimes, an entire day for one OPA)
  \end{itemize}
\end{frame}

\begin{frame}{Preamp}
  \includegraphics[width=\textwidth]{opa/preamp}
\end{frame}

\begin{frame}{Modular hardware model}
  \adjincludegraphics[scale=0.25]{acquisition/hardware_inheritance}
\end{frame}

\begin{frame}{Modular sensor model}
  Can have as many sensors as needed.
  \vspace{\baselineskip} \\
  Each sensor contributes one or more channels.
  \vspace{\baselineskip} \\
  Sensors with size contribute new variables (dimensions).
\end{frame}

\begin{frame}{Domains of CMDS}
  CMDS can be collected in two domains:
  \begin{itemize}
    \item time domain
    \item frequency domain
  \end{itemize}
\end{frame}

\begin{frame}{Time domain}
  Multiple broadband pulses are scanned in \emph{time} to collect a multidimensional interferogram
  (analogous to FTIR, NMR).
  \vspace{\baselineskip} \\
  A local oscillator must be used to measure the \emph{phase} of the output.
  \vspace{\baselineskip} \\
  This technique is...
  \begin{itemize}
    \item fast (even single shot)
    \item robust
  \end{itemize}
  pulse shapers have made time-domain CMDS (2DIR) almost routine.
\end{frame}

\begin{frame}{Frequency domain}
  In the Wright Group, we focus on \emph{frequency} domain ``Multi-Resonant'' (MR)-CMDS.
  \vspace{\baselineskip} \\
  Automated Optical Parametric Amplifiers (OPAs) are used to produce relatively narrow-band pulses.
  Multidimensional spectra are collected ``directly'' by scanning OPAs against each-other.
  \vspace{\baselineskip} \\
  This strategy is...
  \begin{itemize}
    \item slow (must directly visit each pixel)
    \item fragile (many crucial moving pieces)
  \end{itemize}
  but! It is incredibly flexible.  
\end{frame}

\begin{frame}{Selection rules}
  MR-CMDS can easily collect data without an external local oscillator.
  \vspace{\baselineskip} \\
  This means... [BOYLE]
\end{frame}

\begin{frame}{Central loop}
  At its core, PyCMDS does something very simple...
  \vspace{\baselineskip} \\ 
  Set, wait, read, wait, repeat.  % TODO: better figure
  \vspace{\baselineskip} \\
  Everything is multi-threaded (simultaneous motion, simultaneous read).
  \begin{itemize}
    \item decrease scan time by up to $\sim2\times$, more for complex experiments
  \end{itemize}
\end{frame}

\end{document}