sPHENIX tracking discussion

[Martin Purschke <purschke AT bnl.gov>]

All, especially Tom, Mike, and Tony,

there are numerous complaints that files are easily getting overwritten
in overleaf, which seems to be some race condition between the web
interface and the overleaf git export.

Luckily, whenever something gets committed to the git repo, it leaves a
trace. Tom was looking for some text about MAPS that he had written that
disappeared in a subsequent commit. It dos NOT take a tag or so to make
those past changes accessible later.

If you issue the command "git log" to see what has been going on, you get

> commit cb314d8c288d323e4f8c3af8456bf13567b97a48
> Author: John Lajoie <lajoie AT iastate.edu>
> Date:   Thu Oct 15 21:02:59 2015 +0000
> 
>     Update on Overleaf.
> 
> commit 5695f8e278643261fe4a001e27bf57898861257d
> Author: Thomas Hemmick <tkhemmick AT gmail.com>
> Date:   Thu Oct 15 20:40:15 2015 +0000
> 
>     Update on Overleaf.
> 
> commit de69855348560a869455de1eefc510e6933783bf
> Author: Thomas Hemmick <tkhemmick AT gmail.com>
> Date:   Thu Oct 15 20:38:25 2015 +0000
> 
>     Update on Overleaf.
> 
> commit daf86ed149bd2aa5315c6c45d269999428ba196a
> Author: Jin Huang <blackcathj AT gmail.com>
> Date:   Thu Oct 15 18:14:46 2015 +0000
> 
>     Update on Overleaf.
...

and so on.

Tom advised that the text in question, now gone, contained the string
"latch up". I made a list of all commits with those long hex codes

> git log | grep commit | awk '{print $NF}'  > commits

and then went through them one by one to see when the string would appear.

> $ cat search_latchup.sh
> #! /bin/sh
> 
> git checkout master > /dev/null 2>&1
> 
> for c in $(cat commits) ; do
> 
>     echo $c
>     git checkout $c  > /dev/null 2>&1
>     if grep -iq 'latch up' tracker/tracker.tex ; then
>       echo -n "$c  "
>       grep -i 'latch up' tracker/tracker.tex
>     fi
>     git checkout -f master > /dev/null 2>&1
> done

Lo and behold, in this case, cb314d8c288d323e4f8c3af8456bf13567b97a48 is
the one. For Tom's benefit, I attach the corresponding tracker.tex
version, but now you have the recipe.

Best,
        Martin

-- 
Martin L. Purschke, Ph.D.        ;   purschke AT bnl.gov
                                 ;   http://www.phenix.bnl.gov/~purschke
                                 ;
Brookhaven National Laboratory   ;   phone: +1-631-344-5244
Physics Department Bldg 510 C    ;   fax:   +1-631-344-3253
Upton, NY 11973-5000             ;   skype: mpurschke
-----------------------------------------------------------------------
%% tracker.tex  Tracker  Ed, Yasuyuki, Itaru 20p    
 
% Ed's file copy-pasted by a.m.

%  Hi guys
%  I just made a pass over the document using to restructure the "section",
%  "subsection", and "subsubsection" markings to better match the DOC
%  file sent around by Ed.
%                                   TKH 9-10-2015
%
% I edited both of those sections to make our plan more clear to the reader
% and fill out the discussion of heavy-flavor jet neededs (I think our
% design requirements on b-jets are still incomplete and need additional 
% numbers put to them...)
% - MPM 10/14/2015

\section{Physics requirements}
Three elements of the sPHENIX physics program drive the performance 
parameters of sPHENIX tracking. The measurement of the upsilon family of 
quarkonia states , fragmentation functions at high and at low $z$, and heavy 
flavor tagged jets together set the momentum resolution specification for the 
Tracker . To fully utilize the available RHIC luminosity, the tracking 
systems should have large, uniform acceptance and be capable of fast readout. 
 
Measuring fragmentation functions at low $z$ requires looking for possibly 
wide angle correlations between a trigger jet and a charged hadron. This 
places only moderate requirements on the momentum resolution ($\Delta p/p 
\simeq 1\%\cdot p$), but reinforces the requirement of a large acceptance and 
tracking purity at low momentum. Fragmentation functions at high $z$ also 
place more stringent requirements on momentum resolution and can be a design 
constraint for momenta well above 10~GeV/c. In order to unfold the full 
fragmentation        function, $f(z)$, the smearing due to momentum 
uncertainty should be very small compared to the corresponding smearing due 
to the calorimetric jet measurement for a cleanly identified jet. For a 
40~GeV jet this condition is satisfied by a tracking momentum resolution of 
$\Delta p/p \simeq 0.2\%\cdot p$ or better up to momenta near 40~GeV/c.
The measurement of the $\Upsilon$ family places the most stringent 
requirement on momentum resolution at lower momenta. The large mass of the 
upsilon means that one can primarily focus on electrons with momenta of 
$\sim4-10$~GeV/c. The $\Upsilon(3S)$ has about 3\% higher mass than the 
$\Upsilon(2S)$ state and to distinguish them clearly one needs invariant mass 
resolution of $\sim$100 MeV, or $\sim1$\%.  This translates into a momentum 
resolution for the daughter $e^\pm$ of $\sim$ 1.2\% in the range 
$4-10$~GeV/c.  
The $\Upsilon$ measurement also generates requirements on the purity and 
efficiency of electron identification.  The identification needs to be 
efficient because of the low cross section for $\Upsilon$ production at RHIC, 
and it needs to have high purity against the charged hadron background to 
maintain a good signal to background ratio. This requires minimizing track 
ambiguities. For a multi-layer device such as a silicon tracker one must 
optimize the number of tracking layers, their spacing, and the segmentation  
of the strip layers. For a continuous tracking device such as a TPC one must 
optimize the two-track separation through the appropriate choice of 
granularity of the readout plane, and control of space charge and pile-up 
effects.  
Tagging heavy-flavor jets introduces the additional tracking requirements. At 
minimum this demands the ablity to measure the displaced track originating 
from a D or B meson decay, as described in Section ?. The $c\tau$ for D and B 
decays is 123~$\mu$m and 457~$\mu$m, respectively, and the displaced track 
would need to be identified with a resolution sufficient to distinguish these 
decays against backgrounds. Furthermore, heavy-flavor jet identification 
algorithms such as DCA-counting methods require multiple large DCA tracks to 
be found simultaneously within a jet and will require a large single track 
efficiency to keep the overall identification suitably efficient. Other heavy 
flavor jet identification methods such as those based on fully reconstructing 
individual secondary vertexes can place additional demands on the individual 
track position resolution and impact the inner pixel segmentation.

Translating these needs into a detector requirement can be done only by 
performing detailed simulations with a specific tracking configuration, 
followed by evaluation of the tracking performance.

%{\blue overview written by Ed.}\\
{\blue Information form Mike and Dennis to be put together by Tony}\\
for upsilons, jets, hadrons, jets+X Au+Au, \pp, \pAu.

\section{Overview of Tracking Options}

Here we present brief summaries of design options that are currently under 
consideration for central
tracking. The physics program requires that the tracking system will provide 
measurements of displaced
vertices for the identification of heavy flavor decays and sufficient 
momentum resolution to separate the
three upsilon states and measure jet fragmentation functions. The summaries 
that follow include technical options for inner pixels (reconfigured PHENIX 
pixels and MAPS sensors) and outer tracking (Si strips and TPC) and the steps 
required to realize each project. Detailed design considerations are 
discussed later in the chapter.

\subsection{Reuse of the PHENIX Pixel Detector}
% {\blue to be written by Yasuyuki and Itaru -- baseline, resue of the VTX}\\
\subsubsection{Detector Description} % Yasuyuki and Itaru:  Describe the 
reused pixels (grammar edits by MPM)
In this option, we re-use the pixel detector ladders of the PHENIX VTX 
detector.
The PHENIX VTX is composed of 4 layers of silicon detectors, and its inner 
two layers are
made of 30 ladders of hybrid pixel detectors. The first layer, made of 10 
ladders,
is at the radius of approximately 2.5 cm from the beam pipe and the second
layer, made of 20 ladders, is at 5 cm. Currently they are positioned to 
measure charged particles also sampled by the PHENIX central arms and so 
cover about 80\% of 2$\pi$. 

One pixel ladder is made of 4 sensor modules on a stave made of CFRP, which 
provides
mechanical support and cooling. A sensor module is made of a pixel sensor
(5.672cm (L)$\times$ 1.50 cm (w) $\times$ 200 $\mu$m (t)) and 4 ALICE
pixel read-out chips bump bonded on the sensor. Each of ALICE chip read-out 
8192
pixels, arranged as $32 \times 256$. The size of an individual pixel is 
$\Delta \phi \times \Delta z =$ 
50$\mu$m $\times$ 425 $\mu$m. A pixel ladder is read-out from both ends by 
read-out boards
called SPIRO through high density buses. The total radiation length of a 
pixel ladder is
approximately 1.28\%. These pixel detectors were built by RIKEN. In addition 
to the 30 ladders 
in PHENIX VTX, RIKEN also has 6 unused spare ladders.

After the 2016 run, PHENIX detector and VTX detector will be decomissioned, 
and
the pixel ladders will be saved along with their read-out electronics. We 
plan to
reconfigure the 30 pixel ladders and the 6 spare pixel ladders to make two 
inner
pixel layers for the sPHENIX tracker. The radial position of the two layers 
and the number of
ladders in each layers will be adjusted so that the detector will cover 
2$\pi$ in azimuth.
A tentative configuration of the pixel layers are shown in 
Table~\ref{tbl:SiStripSummary}
% geometry, channel count, support concept, sensor options, material, readout 
scheme

%\begin{itemize}
%\item{Pixel reconfiguration and support}
%\item{Pixel readout bus and electronics}
%\item{Pixel cooling and cabling}
%\end{itemize}

\subsubsection{Justification of Design Choices} % Yasuyuki and Itaru:  
Justify the reused pixels

Precision vertex position measurement is needed to measure heavy quark 
production.
Measurement of $D$ mesons in high $p_T$ and $b$-tagged jet is one of the main 
physics goals of sPHENIX.
Two layers of pixel detector will provides this good DCA measurement 
capability. DCA resolution of 70 $\mu$m
for $p_T>1$ GeV/$c$ was achieved for the existing pixel detectors. The DCA 
resolution at high $p_T$ was about
50 $\mu$m, limited by the beam profile width.

The main advantage of this option that we have 30 pixel ladders and 6 spares 
along with read-out
electronics. The detector has been used in the experiment and it is 
demostrated that they can
achieve good DCA resoluiton that is sufficient to separate heavy flavor 
electrons from charm and bottom
decays from light hadron sources and to separate charm decay electrons and 
bottom decay electrons.
Read-out speed of approximately 10 kHz has been demonstrated in real physics 
experiment.
This option provides the most cost effective implemenation of two inner most 
tracking layer of sPHENIX
for heavy quark measurement in terms of both of the budget and the manpower.

% A drawback of this option is that the detector has been used many years and 
some of the pixels are dead.


% Why the momentum resolution, channel count for pixel/strip/TPC, material 
thickness or rad length,  choice of geometry, etc.
\subsubsection{Steps to Project Completion} % Yasuyuki and Itaru: Steps to 
realize reused pixels
{\bf Tabular or text description of steps in the project between now and 
installation.}
Step for reconfiguration.
\begin{itemize}
\item Define configuration of the two pixel layers. The best configuration 
from the available number
of pixel ladders will be defined. (FY2016)
\item Chose best pixel ladders to be used (FY2016)
\item Design mechanical support of two layers of pixel ladders and SPRIO 
read-out boards, assembly fixtures of pixels, and the power/cooling 
interface. (FY2018)
\item Produce mechanical support structure (FY2018)
\item Install pixel ladders into the mechanical support structure (FY2019)
\item Install SPIRO read-out boards into the support structure and connect to 
the pixel ladders (FY2019)
\item Install the reconfigured pixel into sPHENIX tracker (FY2019)
\end{itemize}
% For instance: Production and testing of prototype strip sensors between now 
and Oct 2016. Design studies for the reconfiguration of the pixel now-Oct 
2016. Evaluation of alternate pixel options Fall 2015-fall 2016. Production 
of prototype strip ladders with prototype readout Oct 2016-Oct 2017. 
Conceptual design of the pixel/strip/TPC throughout FY17.  TPC prototype 
production and testing FY17. Choice of final tracking technology Jan 2018. 
Large scale prototypes of the strip or TPC Jan 2018-Oct 2018.  Final 
engineering design of the strip/TPC starting in early FY18 and concluding in 
4QY18. Start procurement process for major Tracking production components 
4QFY18. Production of Tracking components beginning in spring 2019 …..  

\subsection{MAPS Pixel}
%{\blue new pixels Mike McCumber and Ming Liu, LANL people}
\subsubsection{Detector Description} % Mike and Ming: Describe the MAPS pixels
In this option, we construct a new set of inner tracking sensors to improve 
the precision of the tracking system for an expanded physics reach with the 
tracker and to both mitigate known issues with the reconfiguration of the 
PHENIX pixel sensor. The technology choice for the new set of silicon sensors 
would be the same monolithic active pixel sensor (MAPS) based vertex detector 
technology developed by the ALICE experiment, see 
Figure~\ref{fig:ALICE_MAPS_Stave}.

%(geometry, channel count, support concept, sensor options, material, readout 
scheme)
It is highly desired to have a low mass high efficiency precision vertex 
tracking detector for the heavy flavor physics program in sPHENIX. An 
exploration of silicon pixel sensor technology with low power consumption, 
thus less supporting materials, is the only way to go beyond the current 
pixel systems. The latest MAPS based pixel vertex detector being developed 
for the ALICE Inner Tracking System Upgrade has made significant reduction of 
material budget from $\sim 1\%$ down to $\sim 0.3\% X_{0}$, and much reduced 
pixel size of $\sim O(30\mu$m $\times 30\mu$m). The same technology albeit 
with a different readout configuration has been adopted in the latest STAR 
inner vertex detector upgrade at RHIC.~\cite{STAR:MAPS-Upgrade} For sPHENIX a 
minimum of two layers are required to provide the coverage with high tracking 
efficiency and precision. The exact dimension and configuration is under 
development but we foresee the inner most radius could be as close to IP as 
possible, yet still outside the beam pipe, at a radius $R \sim 2.4$cm. With 
very low power front-end electronics ($40$ nW per channel), the whole 
detector can be cooled by air with minimal cooling materials. From those 
currently underdevelopment, the ALICE ITS ALPIDE architecture provides a 4 
$\mu$sec integration time and smaller than 50 $mW/cm^{2}$ power consumption 
appears best suited to our needs.

We expect the characteristics of the inner MAPS tracker to closely resemble 
those of the ALICE inner ITS. The inner ladders of the ALICE ITS extend 27 cm 
in length. These will be suitable for the sPHENIX coverage for radii within 6 
cm. Should layers outside of 6 cm be required, longer staves will need to be 
produced to preserve the pseudo-rapidity coverage for all vertexes within 
$\pm$10cm. A summary of the geometry for the inner tracker is provided in 
Table~\ref{tbl:MapsSummary}.

\begin{table}[h]
\begin{center}
\caption{Summary of geometries for an inner MAPS tracker with 3 layers.}
\begin{tabular}{|cc|ccc|ccc|}
\hline\hline
       &        &          & sensor   &           &                   & &     
           \\
 Layer & radius & pitch    & length   & depth     & total thickness   & 
length & area    \\
       &  (cm)  & ($\mu$m) & ($\mu$m) & ($\mu$m) &     $X_0\%$ &  (cm)  &  
(m$^2$)\\
\hline
 1 & 2.4        & 28 & 28 & 50 & 0.3 & 27          &     0.041\\
 2 & $\sim$4    & 28 & 28 & 50 & 0.3 & 27          & $\sim$0.068\\
 3 & $\sim$6-15 & 28 & 28 & 50 & 0.3 & $\sim$27-39 & $\sim$0.102-0.368\\
\hline

\end{tabular}
\end{center}
\label{tbl:MapsSummary}
\end{table}

\begin{itemize}
\item{Pixel reconfiguration and support}
\item{Pixel readout bus and electronics}
\item{Pixel cooling and cabling}
\end{itemize}

\subsubsection{Justification of Design Choices} % Mike and Ming: Justify the 
MAPS pixels

A new set of tracking layers utilizing a developed yet modern technology, 
such as the MAPS sensors discussed above, can alleviate some concerns with 
the reconfiguration of the existing PHENIX sensors. We discuss the following 
issues in some detail:
\subitem[Tracking inefficiencies] The existing pixel layers in PHENIX have a 
90\% and 77\% tracking efficiency for the first and second layers 
respectively. For track-counting methods this will represent an immediate 
efficiency drop to $(90\%\times77\%)^{3} = 33\%$, degrading and potentially 
endangering our ability to measure b-jets. Mitigation strategies such as 6 
hit tracking, wherein only one of the first two layers are utilized, will 
suffer a higher rate of fake reconstructions and remove the precision 
placement at the vertex needed by the heavy flavor program. MAPS sensors 
being developed for the ALICE upgrade have shown much larger tracking 
efficiencies (>99\%), would recover these lost signals, and add new 
capabilities by better defining the track position.  It is important to note 
that the difficulties encountered by the PHENIX silicon detectors (loss of 
contact via flaws in bump/wire bonds) are intrinsically eliminated in the 
MAPS design since these contacts are part of the silicon chip and not applied 
as a post-production step.

\subitem[Material thickness] The existing pixel ladders have a 1.28\% 
radiation thickness per layer. Modern MAPS based ladder designs can improve 
dramatically on those values with a thickness per layer of only 0.3\% 
$X_{0}$. The integrated material of the Tracker impacts the momentum 
resolution of electrons and this constraint is driving the placement of the 
outer tracking layers in the silicon strip option to 80 cm. A new inner 
tracking subsystem composed of 2 or 3 MAPS based silicon sensor layers would 
integrate to a smaller total material budget which could be utilized to 
either improve the Upsilon mass resolution, reduce the outer tracking layer 
cost by moving the outer layers to a smaller radius, or some combined 
optimization of both.

\subitem[Tracking precision] The existing pixel ladders have large pixel 
dimensions by modern standards. The pitch of the pixels ($50um\times425um$) 
is optimized in one direction to facilitate DCA based measurements in the 
transverse plane. The ALICE MAPS design can achieve much narrower pitch 
($30um\times30um$) opening the possibility of reconstructing individual 
secondary vertexes with two-track comparisons and an additional approach to 
b-jet identification.

\subitem[DAQ rate limit] A system test of the existing pixels ran at 14 kHz 
with 100\% dead time. This is somewhat under the need of sPHENIX for 15 kHz 
with a small dead time for rare trigger collection and impacts slightly the 
entire physics program. A new set of tracking layers would free sPHENIX to 
establish a larger event collection rate and ensure the 15 kHz design goal is 
fully achieved.

\subitem[TPC integration] The integration of a compact TPC would offer 
additional benefits that will be detailed in a later subsection. The Tracker 
is a unified system and must be designed as such. The final design of the 
inner silicon will thus be influenced, particularly in the placement of the 
large radii layers by design choices of the outer tracking. A newly 
constructed set of materially-thin tracking layers has the flexibility to 
meet the design needs should the TPC option need a third layer of pixels to 
complete the matching between the TPC and pixel layers.

\subsubsection{Steps to Project Completion} 
% Mike and Ming: Steps to realize MAPS pixels
% all of this is heavily dependent upon our ability to secure funding... Ming 
can you fill this in? -MPM

{\bf Tabular or text description of steps in the project between now and 
installation.}
For instance: Production and testing of prototype strip sensors between now 
and Oct 2016. 
Design studies for the reconfiguration of the pixel now-Oct 2016. 
Evaluation of alternate pixel options Fall 2015-fall 2016. 
Production of prototype strip ladders with prototype readout Oct 2016-Oct 
2017. 
Conceptual design of the pixel/strip/TPC throughout FY17.  
TPC prototype production and testing FY17. 
Choice of final tracking technology Jan 2018. 
Large scale prototypes of the strip or TPC Jan 2018-Oct 2018.  
Final engineering design of the strip/TPC starting in early FY18 and 
concluding in 4QY18. 
Start procurement process for major Tracking production components 4QFY18. 
Production of Tracking components beginning in spring 2019 …..

\subsection{Strip option} 
%{\blue to be written by Itaru}\\
\subsubsection{Detector Description} % Itaru:  Describe the strip detector

%geometry, channel count, support concept, sensor options, material, readout 
scheme

The silicon strip tracker covers the acceptance $|\eta|<1$ and full azimuthal 
coverage, $\Delta\phi=2\pi$.
The silicon strip tracker consist of 3 stations (from inner to outer, called 
S0, S1 and S2) as illustrated in Figure~\ref{fig:3D_Overview}. The S0 and S1 
stations consisted of two layers which are staggered each other between 
adjacent sensor modules for hermeticity. The radius and dimensions of each 
station and layers
are summarized in the table~\ref{tbl:SiStripSummary}.

\begin{figure}[htb!]
\centering
\includegraphics[width = 0.4\textwidth]{figs/SiTracker_3D_Overview.png}
\caption{CAD drawing of the silicon strip tracker.}
\label{fig:3D_Overview}
\end{figure} 


\begin{table}[h]
\begin{center}
\caption{Summary of geometries for silicon strip tracker.}
\begin{tabular}{|ccc|ccc|cc|}
\hline\hline
        &       &        &       & sensor &       &                  &\\
Station & Layer & radius & pitch & length & depth & total thickness  & area\\
             &             &  (cm)   &   ($\mu$m)    &  (cm)                 
&      ($\mu$m)   &     $X_0\%$       &  (m$^2$)\\
\hline
 Pixel  & 1          &      2.4  & 50                   & 0.425               
& 200                  &          1.3              & 0.034\\
 Pixel  & 2          &      4.4  & 50                   & 0.425               
& 200                  &          1.3              & 0.059\\
\hline
  S0a   & 3          &      9.5  & 60                   & 8                   
 & 240                  &          1.35            & 0.152\\
  S0b   & 4          &    10.5  &  240               & 2                      
 & 240                   &         1.35             & 0.185\\
  S1a   & 5          &     44.5  & 60                  & 8                    
 & 240                  &          1                  & 3.3\\
  S1b   & 6          &     45.5  &  240               & 2                     
 & 240                   &         1                  & 3.5\\
  S2     & 7          &     80.0  &  60                  & 8                  
   & 320                   &          2                  & 10.8\\
\hline

\end{tabular}
\end{center}
\label{tbl:SiStripSummary}
\end{table}

%Y. Akiba (To Nakagawa-san, I slightly changed the text here. I want to make 
S0 and S1 240um thick.)
As discussed in a later section, we have developed a prototype sensor for S2. 
The sensor was manufactured by HPK.
This prototype sensor is 320 $\mu$m thick, the standard thickness of HPK. 
HPK has technology to manufacture 240 $\mu$m thickness sensors
as well. We are now prototyping the sensor for S1 layer. HPK is manufacturing 
the prototpye sensor with both
of 320 $\mu$m and 240 $\mu$m. Our aim is to make both S0a/b and S1a/b to be 
240 $\mu$m thick or even thinner to
achieve minimize the multiple scattering within the sensor material. The 
details of the technology is to be discussed
in the later section.

The FPHX chip which was developed for the FVTX detector is employed as the 
readout chip of the silicon tracker 
for the sPHENIX. The readout scheme is therefore very similar to that of 
FVTX. The primary reasons for this choice are
(1) we can use adapt the read-out chain of FVTX with minium change and (2) 
the power consumption of the chip is only 64mw
per chip (128 channel). Another option we considered is to use SVX4, which 
was used for PHENIX VTX stripixel detector and
MPC-EX detector. However, the power consumption of SVX4 is more than 5 times 
hihger than that of FPHX chip.
It is unavoidable to have liquid cooling system for SVX4 readout though, the 
FPHX chip opens up an option 
to operate the silicon module with an air cooling system. This is the major 
advantage to make the cooling system 
simpler with less materials.  

The number of channels for silicon module, ladder, and sensors are summarized 
in table~\ref{tbl:SiStripChannelSummary}.

\begin{table}[h]
\begin{center}
\caption{Number of channel summary for the silicon strip tracker.}
\begin{tabular}{ccccc}
\hline\hline
 station  & sub-layer & silicon module & $\#$ of ladder &\# of sensor\\
          &           & per ladder     &                & \\
\hline 
   s0     &   2       &     3          &   36        &    216\\
   s1     &   2       &     7          &   48        &    672\\
   s2     &   1       &    14          &   48        &    672\\
\hline 
\end{tabular}
\end{center}
\label{tbl:SiStripChannelSummary}
\end{table}          

Shown in Figure~\ref{fig:StripReadout} is the conceptual layout design of 
strip and readout lines for the S2 sensor. The dimension of the active area 
is 96mm$\times$92.16mm, which is divided into 12$\times$12 blocks. Strips are 
running parallel to the beam line. 
Silicon sensors readout lines are aligned perpendicularly (azimuthal 
direction) with respect to the strip direction as illustrated in 
Figure~\ref{fig:StripReadout}. Upper and lower 6 blocks are connected upwards 
and downwards, respectively. This way the number of readout channels are 
reduced to save the cost. Although signals in 6 blocks are combined and 
cannot be distinguished in the readout electronics, the origin of hit block 
is to be identified in offline analysis by requiring hit matching between 
other station and/or layers. An expected occupancy is sufficiently low which 
allows us to combine the signal in readout and let an offline tracking 
algorithm to identify the actual fired block. Number of combined blocks are 
less for s0 and s1 because the number of blocks are fewer due to smaller size 
of their sensors.

\begin{figure}[htb!]
\centering
\includegraphics[width = 0.8\textwidth]{figs/SiTracker_StripReadout.pdf}
\caption{Schematic layout strip and readout lines of the sensor.}
\label{fig:StripReadout}
\end{figure}

%geometry, channel count, support concept, sensor options, material, readout 
scheme

\begin{itemize}
\item{Pixel reconfiguration and support}
\item{Pixel readout bus and electronics}
\item{Pixel cooling and cabling}
\end{itemize}

\subsubsection{Justification of Design Choices} % Itaru:  Justify the strip 
detector
The momentum resolution is limited by the multiple scattering in the middle 
(s1) layer which is proportional to the total radiation length of materials. 
Following efforts in the design are under consideration/development to 
minimize the material budget as low as possible. 

\begin{itemize}
 \item{Air cooling system instead of liquid one}
 \item{240$\mu$m thick silicon sensor instead of 320$\mu$m.}
\end{itemize}
As discussed in the previous section, the choice of the FPHX chip help to 
reduce the material budget due to smaller heat generation which can be 
handled by the air cooling system. Another efforts to reduce the radiation 
length is the thickness of silicon sensors. If 240$\mu$m thick silicon 
sensors demonstrates satisfactory performance, then it will reduce the 
material budget for the silicon sensor part by 25\% from the standard 
320$\mu$m. The 240$\mu$m thick sensor is known to draw larger dark current 
and it is still under developing stage. 

The cost of readout is greatly reduced by combining signal across multiple 
blocks in azimuthal direction without widening strip pitch. This way the 
momentum resolution is not sacrificed by reducing number of readout channels. 


%Why the momentum resolution, channel count for pixel/strip/TPC, material 
thickness or rad length,  choice of geometry, etc.

\subsubsection{Steps to Project Completion} % Itaru: Steps to realize strip 
detector
{\bf Tabular or text description of steps in the project between now and 
installation.}
For instance: Production and testing of prototype strip sensors between now 
and Oct 2016. Design studies for the reconfiguration of the pixel now-Oct 
2016. Evaluation of alternate pixel options Fall 2015-fall 2016. Production 
of prototype strip ladders with prototype readout Oct 2016-Oct 2017. 
Conceptual design of the pixel/strip/TPC throughout FY17.  TPC prototype 
production and testing FY17. Choice of final tracking technology Jan 2018. 
Large scale prototypes of the strip or TPC Jan 2018-Oct 2018.  Final 
engineering design of the strip/TPC starting in early FY18 and concluding in 
4QY18. Start procurement process for major Tracking production components 
4QFY18. Production of Tracking components beginning in spring 2019 …..


\subsection{TPC option}
\subsubsection{Detector Description}

The TPC design follows the classical cylindrical double-sided TPC layout used 
in several other experiments {\red [reuse citations]} with a central membrane 
electrode located at the middle of the interaction region, which divides the 
TPC into two mirror-symmetric volumes shown in fig.~\ref{fig:tpc_global}.

\begin{figure}[htb!]
\centering
\includegraphics[width = 0.8\textwidth]{figs/tpc_scheme.png}
\caption{Schematic layout of TPC main elements.}
\label{fig:tpc_global}
\end{figure}

In each such volume the readout plane is located on the endcap inner surface, 
facing the gas volume. The electric field, transporting primary ionization to 
the readout plane is formed by the membrane electrode set to the highest 
voltage bias on one side and by the the readout plane at ground potential on 
the other. The electrical drift field is constrained by the field cage along 
the inner and the outer cylindrical surfaces of the TPC.

The two mirror-symmetrical parts of the TPC form a common gas volume filled 
with the gas mixture, which transports primary ionization to the readout 
plane on each TPC endcap surface. The same gas that transport primary 
ionization also works as the medium for the amplification elements located in 
front of the readout planes. These amplification elements are built based on 
several layers of micropattern gaseous detectors.

Other TPC subsystems directly related to the main volume are the channel 
readout system; high voltage distribution systems for the drift field and for 
the amplification elements; gas circulation, control and purification system; 
TPC calibration systems. Operation and readout of different service 
subsystems is handled by the TPC slow control system.

Clearly the exact specifications of any of the tracking detectors will 
require additional R\&D work to finalize.  Nevertheless, it is appropriate to 
demonstrate an ``existence proof'' in the form of a design that would match 
and exceed all necessary performance parameters.  Because of the extensive 
R\&D used to establish, in great detail, the operational principles for the 
ALICE TPC Upgrade project, this example makes for a very effective initial 
model.  In the coming discussion, we use the ALICE upgrade plans and R\&D 
results to show that an implementation along these lines would work for 
sPHENIX.  In the later discussion, we give more detail on how these 
parameters are likely to be even further optimized for our application.  As 
such, even though we present exact figures here, these should be considered 
as exact figures of a ``straw design'' and will certainly evolve with time.

\begin{center}
\begin{tabular}{|r|c|c|l|}
\hline
Parameter & Value & Unit & Comment\\
\hline
Gas enclosure & $<$30 $\rightarrow$ 80 & cm &  \\
Active volume & 30 $\rightarrow$ $<$80 & cm & \\
Length & $\pm$ 80 & cm & $\left| \eta \right|~<~1$\\
Pad Size & 1.2 $\times$ 10 &$mm^2$ & 45 radial segments\\
Channel Counter & 123700 $\times$ 2 & & \\
Avalanche Technology & Quadruple GEM & & Same as ALICE upgrade design\\
Gas gain & 2000 & & Same as ALICE\\
Field Cage Central Potential & 32,000 & V & $400\frac{V}{cm}$, conservatively 
high\\
\hline
%\caption{Summary of the basic TPC design parameters.  Here are have assumed 
$Ne-CO_2$ gas since this places the highest voltage burden on the field cage.}
\end{tabular}
\label{Table:TPCDesign}
\end{center}

\subsubsection{Justification of Design Choices} % Tom and Sasha: Justify the 
TPC 
The basic geometry of the sPHENIX TPC is driven by several factors.  As is 
typical for cylindrical collider detectors, the division between the "barrel" 
acceptance and future endcap upgrades is located at $\eta~=~\pm~1$ as 
enforced by the existing BaBar coil.  The radial extent of the TPC is driven 
differently on the outside and inside surfaces.  Both the inner and outer 
walls will have two layers: 
\begin{itemize}
\item the "field cage" defines the active volume of the detector
\item gas enclosure defines the physical size of the detector and its 
coupling to other parts of the spectrometer.  
\end{itemize}
At the outer radius, the location of the gas enclosure is defined by the 
surrounding systems. Currently the EMCAL inner radius is 90~cm.  Furthermore, 
we envision adding some hadron ID capability as we upgrade the spectrometer 
for use at the future electron-ion collider.  Some devices (e.g. the BaBar 
DIRC or ALICE-like HMPID) would fit within 10~cm and so we currently 
constrain the outer TPC gas enclosure to $r=80cm$. The outer radial extent of 
the tracking volume itself will depend upon the exact design of the field 
cage and would then be less than 80cm.  

Contrarily, the inner radius is defined by the inner tracking volume. 
Possible interference between the TPC inner gas enclosure and a 2-3 layer 
vertex pixel detector is a loose constraint.  The driving factor defining in 
locating the TPC inner field cage is the radius at which the TPC begins to 
track with high efficiency.  Currently we are taking the inner field cage 
radius to be 30 cm.  This figure is a compromise between being safely outside 
of the vertex detector, maximizing the length of the measurement, and 
experiencing particle densities similar to those at inner portion of the 
ALICE TPC.

The pad size, gas choice, avalanche technology, gas gain, and central 
potential of the TPC are facets of a single optimization problem necessary to 
define the best design.  Furthermore, these parameters must be correctly 
matched to the electronics by having the charge collection time less than the 
amplifier shaping time.  All of this must be optimized not only for momentum 
resolution but also excellent two-particle response for closely spaced 
tracks.  Despite this complexity, we can constrain the design well enough to 
demonstrate its feasibility in at least one incarnation.

The driving characteristic of the TPC performance is that it should have 
momentum resolution sufficient to cleanly separate the $\Upsilon$ 1s, 2s, and 
3s states.  As shown in the discussion on simulations, mass resolution 
performance exceeding that of the baseline solution is achieved through forty 
radial measurements, each of $\sigma \sim 120~\mu m$.  Pixels of full-width 
$a$, measure position with a resolution of $\sigma~=~\frac{a}{\sqrt{12}}$ 
(less than $\frac{a}{3}$).  Charge division between neighboring cells of 
pitch a, typically measures position to better than $\frac{a}{10}$.  Thus, an 
appropriate pixel size is $1.2~mm$ around the azimuth and $1~cm$ in radius.  
This leads to a total channel count summed across both readout ends below 
250,000.

Traditionally, TPC devices were described as "low rate" detector elements, as 
defined by the long single event collection time and the need for "gating" to 
prevent positive ion buildup within the TPC gas volume.  Both of these 
arguments have become obsolete in recent years.  Traditional calculations 
assumed that an event would be lost if a second event occurred within the 
readout time.  If we were to use the "slow" gas of the ALICE detector 
($Ne-CO_2$ 90-10; $v_d~=~30\frac{\mu m}{ns}$), an 80 cm drift would 
correspond to a full drift time of $26~\mu sec$.  At a 15kHz collision rate, 
this would put a second event in the readout time window with a probability 
of $\sim\frac{1}{3}$.  In recent years, it has been realized that because the 
TPC measures trajectory by time, it is naturally able to distinguish overlaid 
events via their apparent displaced event vertex.  As an example, the ALICE 
upgrade is designed to have their TPC, with more than 80 $\mu s$ drift time, 
measuring collisions at rates up to 50 kHz!  The key to this revolution is 
the development of electronics that dismisses the concept of a single event 
and instead simply reads out the TPC continuously over time.  The net result 
for sPHENIX is that by using this same readout principle, for all reasonable 
gas choices, the "speed" of the gas will never be an event rate limiting 
factor.   Gas speed requirements are bracketed by the competing goals of 
minimizing particle pileup within an event (wherein slower is best) and 
minimizing event pileup (wherein faster is best).  The advent of continuous 
readout electronics shifts the balance away from faster gasses and toward 
slower gasses.

The second consideration is "gating" the TPC.  Avalanche of the primary 
charge necessarily produces a large quantity of positive ions.  Introduction 
of all these ions into the TPC volume would distort the electric field to the 
point that the TPC could no longer function.  Because drift speeds of ions 
are dramatically slower than electron drift speeds, the time to neutralize 
these ions generates a "dead interval" for the TPC.  Although schemes for 
fast flush of the positive ions have been developed, most of the R\&D on TPC 
readout currently involves using Micro Pattern Gas Detectors (MPGD) as these 
can be configured to reject $\sim$99\% of the positive ions allowing the TPC 
to remain live at all times.

\begin{figure}[htb!]
\centering
\includegraphics[width = 0.8\textwidth]{figs/Ne-CO2_90-10_log.png}
\caption{Calculated drift characteristics of $Ne-CO_2$ gas (same as used by 
ALICE).}
\label{fig:gasdiffusion}
\end{figure} 

Figure~\ref{fig:gasdiffusion} shows the characteristics of the $Ne-CO_2$ gas 
mixture used by ALICE.  The best single point resolution is achieved by using 
a drift field that simultaneously minimizes the diffusion and also resides at 
a plateau in drift velocity.  No such plateau exists for $Ne-CO_2$ and so the 
operating conditions for this gas would be defined merely by the diffusion, 
implying a drift field of $400~\frac{V}{cm}$, with a transverse diffusion of 
$150\frac{\mu m}{\sqrt{cm}}$, longitudinal diffusion of $220\frac{\mu 
m}{\sqrt{cm}}$, and drift velocity of $30\frac{\mu m}{nsec}$.  Note that 
these values are not read off the graph, but taken from ALICE wherein they 
have also added 5\% $N_2$ to the gas mixture to improve voltage stability.  
Note also that the lack of plateau in drift velocity requires the TPC to have 
a highly uniform electric field, thereby placing tight constraints on the 
field cage design.

Traditional wire chamber TPC designs image the positive ion cloud around the 
wire onto a segmented pad plane.  Thus, the spread of the charge as compared 
to the pad segmentation is controlled by the wire-to-pad gap.  Using MPGD 
readout, the electron cloud itself must be spread wide enough to allow for 
charge sharing among multiple pads for every avalanche.  This is an extreme 
challenge for ALICE because of their large pad size ($\sim 4 mm$ in azimuth). 
 Their saving grace is the large amount of diffusion due to the long drift 
distance and purposeful "hole misalignment" among the GEM-based avalanche 
scheme.  For sPHENIX, the job is simpler.  Accounting for a hole-induced 
charge spread of $\sim 400 \mu m$, we can add this in quadrature to the 
diffusion induced charge spread to find that charge clouds will have width in 
the range $400~\mu m~<~\sigma~<~1400 \mu m$, which is ideally matched to the 
proposed segmentation.  Furthermore the longitudinal diffusion will induce a 
time spread in the farthest signal ($\frac{\sigma_{long}}{v_d}$) of less than 
70 nsec.  The SAMPA chip is designed with a minimal shaping time of 190~nsec, 
meaning that our signals will have the same characteristic shape independent 
of drift length (an ideal feature).  Finally, as shown in the later sections 
we are considering a chevron or ``zig-zag'' shape for our pads which should 
not only improve the resolution beyond $\frac{a}{10}$, but also ensure good 
charge sharing for all avalanches, including the narrowest we envision.

Although we have shown that "ALICE gas" would have performance that would 
satisfy all design criteria, it is likely that further study will reveal even 
better choices.  A more detailed discussion of other gasses is included below.

\subsubsection{Steps to Project Completion} % Tom and Sasha: Steps to realize 
the TPC
As compared to previous TPC devices used in heavy ion experiments, the 
sPHENIX TPC will be quite small.  The STAR TPC has a radius of 2 meters, 
ALICE has a radius of 2.5 meters and sPHENIX will have only an 80 cm radius.  
Because the device is shorter than an average person, a variety of 
simplifications in the construction will result.  Although, many of the 
typical issues are simpler to address than they were for the larger devices, 
there are nonetheless challenges.  The principle design challenges involve 
maximizing the active volume in such a small device and developing a robust 
gain stage.

Both STAR and ALICE use an ``air'' gap ($N_2$ for STAR and $CO_2$ for ALICE) 
as the principle voltage gap between the high voltage center of the field 
cage and the protective ground shield.  A principle advantage of an air gap 
is that the overall material thickness is minimal.  In sPHENIX, we propose to 
use an air gap on the inner surface where the material budget is critical, 
but use a well designed solid gap (polyimide or $FR_4$) on the outer gap 
where material budget is less of an issue, but active area is at a premium.

Although a significant R\&D program (including research from ALICE and 
sPHENIX institutions) has shown in bench tests and beam tests the viability 
of the quad-GEM approach, it has yet to be realized in a working device.  
Furthermore, experience shows that scaling a small area GEM detector to full 
size can be fraught with difficulty.

For these reasons we have chosen a conservative plan to lead us through the 
prototyping stages to full detector development with minimal risk.  We 
propose that our first prototype ($v1$) include a full sized field cage that 
could/would be used as the final device for sPHENIX.  In this way, we will 
begin the gain stage prototyping with single full scale modules.  This is a 
natural approach since the authors have already completed an intensive 
small-scale investigation of several different gain technologies at the time 
of this writing.  By addressing the scale issue from the outset, we will 
minimize risk of delays to the project itself.

Here we formulate a series of tables outlining the project milestones for 
each subset of the project.  The first table involves the construction of the 
field cage.  As mentioned previously, the prototype v1 field cage is intended 
to be used for the final TPC in order to minimize risk associated with the 
development of the gain stage.  The field cage will be designed with a 
central potential assuming ALICE gas since this is the highest field gradient 
gas currently under consideration.  The design of the field cage should 
implement gas connections, laser ports, mechanical outlines for the modules 
(6 or 8 modules populating each end), and an aluminum stripe pattern on the 
central membrane to act as the electron source under laser illumination.  One 
should note that design/construction of the field cage includes not only the 
field cage itself, but also the design/construction for the mandrels and 
other fixtures necessary to construct the the field cage.

\begin{center}
\begin{tabular}{|l|c|c|c|}
\hline
Item & Start Date & End Date\\
\hline
Design of Field cage &  now & April 2016\\
Procurement of Field Cage Components&May 2016& September 2016\\
Construction of Field Cage&October 2016& March 2017\\
Testing of Field Cage&April 2017&June 2017\\
\hline
\end{tabular}
\end{center}

The full detector will use 6(8) gain stage modules at each end for a total of 
12(16) modules.  Prototyping involves making a single full-scale module of 
some design and coupling this module to the final field cage.  The 
prototyping will take place in three rounds:  v1, v2, and pre-production.  
Because the interface to the field cage was already specified by the field 
cage design, the various gain stage designs focus only on the details of the 
MPGD implementation, HV distribution, electronic readout, and cooling.  To 
maximally decouple these issues we are choosing to define our prototype 
stages as follows:
\begin{itemize}
\item The v1 prototype will use off-the-shelf electronics and not implement 
cooling.  This device only tests the MPGD performance.
\item The v2 prototype will apply lessons learned for the MPGD and add a 
connector pattern compatible with the readout card development.  It will be 
tested using off-the-shelf electronics adapted to the connector pattern.
\item The pre-production prototype will further apply lessons learned for the 
MPGD and additionally implement cooling and electronic card mechanical 
support.  This device is intended as the final design, pending validation 
under stringent tests.
\end{itemize}

The start of the v1 gain stage design immediately follows the end of the 
field cage design.  The schedule below assumes that procurement of the 
pre-production prototype should wait until project funds could be available 
and we have assumed June 2018 as that date.  This leaves 6 months of schedule 
contingency in the overall prototyping phase.  The readiness review following 
the complete suite of tests on the pro-production module would be held in 
December 2018.

\begin{center}
\begin{tabular}{|l|c|c|c|}
\hline
Item & Start Date & End Date\\
\hline
Design of v1 gain stage&April 2016&June 2016\\
Procurement of v1 gain stage&July 2016&September 2016\\
Construction of v1 gain stage&October 2016&October 2016\\
Bench Tests of v1 gain stage&November 2016&November 2016\\
Beam Tests of v1 gain stage&December 2016&December 2016\\
Design of v2 gain stage&January 2017&March 2017\\
Procurement of v2 gain stage&April 2017&June 2017\\
Construction of v2 gain stage&July 2017&July 2017\\
Bench Tests of v2 gain stage&August 2017&August 2017\\
Beam Tests of v2 gain stage&September 2017&September 2017\\
Design of pre-production gain stage&October 2017& December 2017\\
Schedule Contingency&January 2018&June 2018\\
Procurement of pre-production gain stage&June 2018&August 2018\\
Construction of pre-production gain stage&September 2018&September 2018\\
Tests of pre-production gain stage&October 2018&November 2018\\
Readiness Review&December 2018&December 2018\\
\hline
\end{tabular}
\end{center}

To complete the project, after a successful readiness review will require 
that we procure, assemble, test, and install all remaining modules and 
spares.  We will produce 14(18) modules of the 12(16) required so that we 
have two complete modules as spares.  Because of the modular nature of the 
system, procurement, construction, testing, and assembly will all happen in 
parallel, with each step having a delayed start as compared to the rest.  The 
construction task is the most time consuming and is solely responsible for 
the periods required for testing and assembly.  Final testing of the full 
system happens only after the assembly is fully complete.

\begin{center}
\begin{tabular}{|l|c|c|c|}
\hline
Item & Description &Start Date & End Date\\
\hline
Procurement of module parts&&January 2019&June 2019\\
Construction of 18 modules&1 weeks per module&April 2019&August 2019\\
Testing of 18 modules&&April 2019&September 2019\\
Assembly of 16 modules&&April 2019&September 2019\\
Schedule Contingency&&October 2019&November 2019\\
Final Test&&December 2019&December 2019\\
\hline
\end{tabular}
\end{center}

The cooling design of the electronics for the final gain stage must be 
completed as part of the design of the pre-production gain stage listed 
above.  Given that modern liquid-based cooling systems run at pressures below 
atmospheric, it is possible that details of the cooling source could affect 
the implementation of the cooling on the cards themselves.  As such, we have 
scheduled the design for the overall cooling system following the design of 
the v2 gain stage and preceding the design of the pre-production.  We have 
assumed schedule contingency lasts until the readiness review, thereby 
indicating 10 months of schedule contingency.  Alternatively, this schedule 
could be further relaxed to target detector delivery as the end date for an 
additional 12 month of schedule contingency.

\begin{center}
\begin{tabular}{|l|c|c|c|}
\hline
Item & Start Date & End Date\\
\hline
Design of Cooling System&April 2017&June 2017\\
Procurement of Cooling System Parts&July 2017& September 2017\\
Construction of Cooling System&October 2017& December 2017\\
Testing of Cooling System&January 2018&February 2018\\
Schedule Contingency&March2018&December 2018\\
\hline
\end{tabular}
\end{center}

Because the prototyping uses a full sized field cage, it will be necessary to 
have a high capacity recirculating gas system available from the very 
beginning of testing (summer 2017).  Fortunately, such a gas system already 
exists at Stony Brook and will be used for all bench and beam tests.  Thus, 
the final gas system need only be available after the TPC arrives in 
interaction hall.  It should, nonetheless be available soon thereafter to 
allow in situ measurements of cosmic ray tracks prior to first beam.

\begin{center}
\begin{tabular}{|l|c|c|c|}
\hline
Item & Start Date & End Date\\
\hline
Design of Gas System&November 2019&January 2019\\
Procurement of Gas System Parts&February 2019&May 2019\\
Construction of Gas System&June 2019&September 2019\\
Closed Loop Testing of Gas System&October 2019&October 2019\\
Schedule Contingency&November 2019&December 2019\\
Connection to the TPC&January 2020&January 2020\\
\hline
\end{tabular}
\end{center}

The laser system is not, in principle, necessary until the detector arrives 
in the hall.  It would, however, dramatically improve the testing capability 
after assembly is complete and before delivery of the device to the hall.  
For this reason, we put the target date of the initial completion of the 
laser system just prior to the final test of the assembled detector.  In this 
scenario, the system will initially be constructed in the TPC assembly lab 
and then later moved to it's home in the 8 O'Clock hall.

\begin{center}
\begin{tabular}{|l|c|c|c|}
\hline
Item & Start Date & End Date\\
\hline
Design of Laser System&September 2018&December 2018\\
Procurement of Laser System Parts&January 2019&March 2019\\
Construction of Laser System&April 2019&July 2019\\
Testing of Laser System&August 2019&August 2019\\
Schedule Contingency&September 2019&November 2019\\
Use of the Laser System with the Assembled TPC&December 2019&December 2019\\
Movement of the laser to the 8 O'Clock Hall&February 2020&February 2020\\
\hline
\end{tabular}
\end{center}

The TPC is scheduled, with contingency, to arrive in the detector hall in 
January 2020.  The next few months will be a connection and commissioning 
period.  The full system will be installed and operating in summer 2020.

\section{Detector performance: Momentum and Mass Resolution}
%summary plots (momentum resolution, efficiency, etc...) {\blue Ron, Tony}\\
%Tables of specifications\\
\subsection{Magnetic Field}
Tracking in PHENIX was complicated by the fact that the magnetic field in the 
central arms was of a complex shape, dropping off with radius via a 
$\sim$Gaussian profile, and having significant radial component, particularly 
near the pole tips resulting an a mild charge-independent focusing character 
to the field.  The Babar magnet produces a significantly simpler field as is 
indicated in Figure~\ref{fig:FieldTracker}.  The sPHENIX application of this 
coil is rather close to the original design (EMCAL inside the coil and 
tracking extending to $\sim$80~cm and thereby befits from the design choices 
made by BaBar.  A standard solenoid with length equal to diameter has 
significant radial magnetic field components at each end and thereby does not 
produce an idealized field shape.  A return yoke with a small opening ({\it 
e.g.} STAR) will compensate for this shortcoming while severely limiting 
possibilities for upgrades in the forward direction.  Fortunately, the BaBar 
magnet attacks this classic problem using an increased winding density at 
each end, thereby sacrificing uniformity of the field at large radius, for an 
extended ``sweet spot'' of field in the middle.  It is no accident, 
therefore, that the region in which sPHENIX plans to install tracking 
features a close-to-ideal magnetic field shape.  It should further be noted 
that the calculations of Figure~\ref{fig:FieldTracker} are done with a 
"future upgrade" return yoke that allows for a full capability tracker in the 
forward direction.

\begin{figure}[htb!]
\centering
\includegraphics[width = 0.4\textwidth]{figs/FieldAndTracker.PNG}
\caption{The BaBar magnet field superimposed with the dimensions of the 
tracker volume.  This calculation includes the effect of the field return as 
envisioned for future upgrades (forward arm spectrometer).  The dashed line 
indicates the inner radius of the TPC tracking volume.}
\label{fig:FieldTracker}
\end{figure}

Because of the near-ideal field, the momentum resolution effects due to 
multiple scattering and position resolution, can be calculated analytically 
and in closed form.  These results can be further extended to include 
Bremsstrahlung losses (significant only for the electrons) using a 
semi-analytical technique.  In this section, we will highlight the analytic 
and semi-analytic results and show that they are in excellent agreement with 
full GEANT simulations (as they must be).  Further detailed work on the GEANT 
simulations is discussed later in the Chapter.

\subsection{Gl{\"u}ckstern Formula}

In 1963, R.L. Gl{\"u}cktern~\cite{Gluckstern:1963NIM} analytically solved the 
problem of charge particles traversing a uniform magnetic field while 
experiencing the effects of multiple Coulomb scattering.  
Figure~\ref{fig:PosRes} shows the contribution to the transverse momentum 
resolution from the detector's position resolution.  For a uniform magnetic 
field, the sagitta is proportional to the inverse of the particle's 
transverse momentum, thereby making the fractional error in the transverse 
momentum rise linearly with the transverse momentum of the particle itself.  
Assuming $N$ equally spaced measurements each carrying error $\sigma_x$, we 
get the final result:
\begin{equation}
\frac{\sigma_{p_T}}{p_T}=\sqrt{\frac{720}{N+4}}\frac{\sigma_x}{0.3L^2B}p_T
\end{equation}
\begin{figure}[htb!]
\centering
\includegraphics[width = 0.8\textwidth]{figs/PosRes.PNG}
\caption{This figure demonstrates the contribution of position resolution to 
momentum resolution.}
\label{fig:PosRes}
\end{figure}

Multiple Coulomb scattering is addressed in Figure~\ref{fig:MulScat} wherein 
we see that since the material-dependent angular deflection varies as the 
inverse transverse momentum the contribution this contribution is effectively 
a constant yielding the final result:
\begin{equation}
\frac{\sigma^{MS}_{p_T}}{p_T}=\frac{0.052}{\beta 
BL}\sqrt{\frac{L}{\chi_0}}\left[ 1 + 0.038 ln\left( 
\frac{L}{\chi_0}\right)\right]
\end{equation}
\begin{figure}[htb!]
\centering
\includegraphics[width = 0.8\textwidth]{figs/MulScat.PNG}
\caption{This figure demonstrates the contribution of multiple Coulomb 
scattering to momentum resolution.}
\label{fig:MulScat}
\end{figure} 

The Gl{\"u}ckstern result is simply the sum of these two terms in quadrature 
and yields the familiar expression:
\begin{equation}
\frac{\sigma_{p_T}}{p_T}=\sqrt{a^2+\left( b p_T \right)^2}
\end{equation}
wherein the constant $a$ is driven primarily by the material thickness 
($\chi_0$), and the constant $b$ is driven primarily by the intrinsic 
detector resolution ($\sigma_x$).

\subsection{Beyond Gl{\"u}ckstern}
The Gl{\"u}ckstern formula correctly indicates the transverse momentum 
resolution of a spectrometer with a uniform distribution of material and 
uniformly-spaced measurements along the radius with equal position 
resolution.  Two extensions of this result are important to the sPHENIX 
spectrometer design.  First, since out $\eta$ bite is significantly larger 
than the original PHENIX design and most detector designs result in different 
transverse and longitudinal position resolutions, we must account for the 
differing behavior of the zed component.  Second, since our primary design 
goal involves the measurement of upsilon states via the dielectron decay 
channel, we must explicitly account for Bremsstrahlung losses as the 
electrons traverse our detector volumes.

The extension to three dimensions is conceptually simple, and adds both a 
multiple scattering and position-error term for the $p_Z$ resolution. 
Bringing all the terms together in quadrature we determine the fractional 
error in the full momentum to be:
\begin{equation}
\frac{\sigma_p}{p}=\sqrt{\left( 
\frac{\sigma_{MS}}{\sqrt{\sin{\theta}}}\right)^2  +  \left( 
\sigma_{det}p\sin{\theta}\right)^2  +  \left( 
\sigma_{\theta}^{det}\cos{\theta}\right)^2  +  \left( 
\frac{\sigma_{\theta}^{MS}}{\sqrt{\sin{\theta}}} \frac{\cot{\theta}}{p}  
\right)^2}
\end{equation}
where the first and third terms are the familiar Gl{\"u}ckstern result and 
the second and third terms are the contribution of the z-component.

Finally, we must consider the effects of Bremsstrahlung.  The passage of an 
electron through material will result in the occasional loss of energy via 
the emission of single photons.  As a consequence, the electron's subsequent 
trajectory will be determined by the reduced momentum.  For this reason, some 
electrons will be entirely unaffected (having never emitted a photon) while 
others will measure a momentum below that at the production point, thereby 
creating low-side tails in both the momentum and mass spectra.  The 
asymmetric nature of the effect precludes the description of Bremsstrahlung 
as a simple resolution term.  We have handled this by a trivial Monte Carlo 
calculation.

Figure~\ref{fig:Brem} summarizes the necessary input to the calculation.  
Here $k$ is the fractional energy loss due to the Bremsstrahlung photon, 
whose probability (cross section) follows the parametric function shown in 
the figure.  This analytic approximation yields an infinite total cross 
section, and must be integrated across the limits $k_{min}$ to $k_{max}$ to 
determine the photon yield, $N_{\gamma}$, as a function of the material 
thickness in radiation lengths $\frac{L}{\chi_0}$.  In our semi-analytical 
calculation, Bremsstrahlung is folded into the analytical resolution function 
using $N_{\gamma}$ as the mean to a Poisson probability distribution for the 
number of photons emitted and $\frac{d\sigma}{dk}$ for their energy spectrum.

It is worth noting that all the detailed considerations here concern only the 
change in the momentum magnitude.  It is shown elsewhere that shifts in the 
apparent initial direction, generate mass errors orders of magnitude smaller 
than those already considered and and hence ignored with no loss in validity 
of the generated results.

\begin{figure}[htb!]
\centering
\includegraphics[width = 0.8\textwidth]{figs/Brem.PNG}
\caption{This figure indicates how Bremsstrahlung is calculated via single 
photon emission as a contributor to momentum resolution.}
\label{fig:Brem}
\end{figure} 

\subsection{Results}
Our analytic results and full GEANT simulation results are in excellent 
agreement and it is up to the taste of the reader to decide which of them is 
the validation for the other.  The analytic results, however, uniquely lend 
insight into which terms in any design limit the momentum resolution.  At the 
lowest momentum, the material term will dominate favoring the TPC-based 
design.  At the highest momentum, the position resolution term will dominate, 
favoring the Silicon-based design.  If we could afford to deploy MAPS-based 
sensors at all layers of tracking, this would produce momentum and mass 
resolution superior to any known option, since these devices are both thin 
and precise.  Such an choice would, however, eliminate hadron-ID capability 
in the range necessary for the future Electron-Ion Collider.

The analytical and semi-analytical calculations use the summary tables below 
to characterize the detector geometry and material.

\begin{table}[H]
\begin{center}
\begin{tabular}{cccccccc}
\hline
layer & radius & sensor length & $\theta$ resolution & Thickness & 
$\frac{\Delta L}{L}$ & $c_{ms}$ & $\sigma_{ms}$ \\
 & (cm) & (mm) & (mrad) & \% $\chi_0$ & & (mrad) & (mrad) \\
\hline
1 & 2.7 & 0.425 & 4.5  & 1.3 & 0.94 & 1.8 & 1.7 \\
2 & 4.6 & 0.425 & 2.7  & 1.3 & 0.90 & 1.8 & 1.6 \\
3 & 9.5 & 8     & 2.5  & 1.35& 0.79 & 1.8 & 1.4 \\
4 & 10.5 & 2    & 0.55 & 1.35& 0.77 & 1.8 & 1.4 \\
5 & 44.5 & 8    & 0.52 & 1.0 & 0.02 & 1.6 & 0.03 \\
6 & 45.5 & 2    & 0.13 & 1.0 & 0    & 1.6 & 0 \\
\hline
\end{tabular}
\caption{Table showing the input parameters used to characterize the baseline 
silicon tracker for the analytic and semi-analytic momentum and mass 
resolution calculations.}
\end{center}
\end{table}

\begin{table}[H]
\begin{center}
\begin{tabular}{cccccc}
\hline
layer & radius & Thickness & $\frac{\Delta L}{L}$ & $c_{ms}$ & $\sigma_{ms}$ 
\\
 & (cm) & \% $\chi_0$ & & (mrad) & (mrad) \\
\hline
VTX 1 & 2.7 & 1.3 & 0.95 & 1.8 & 1.7\\
VTX 2 & 4.6 & 1.3 & 0.92 & 1.8 & 1.7\\
air   & 15  & 0.1 & 0.73 & 0.03 & 0.02\\
Field cage & 30 & 1.0 & 0.55 & 1.12 & 0.5\\
\hline
\end{tabular}
\caption{Table showing the input parameters used to characterize the hybrid 
TPC tracker for the analytic and semi-analytic momentum and mass resolution 
calculations.}
\end{center}
\end{table}

\begin{table}[H]
\begin{center}
\begin{tabular}{ccccc}
\hline
Option & $\sigma_{ms}$ & $\sigma_{det}$ & $\sigma_{det}^{\theta}$ & 
$\sigma_{\theta}^{ms}$ \\
\hline
Baseline & 0.011 & 0.00057 & 0.0012 & 0.0031\\
Hybrid & 0.0066 & 0.00105 & 0.0011 & 0.0025 \\
\hline
\end{tabular}
\caption{Extended Gl{\"u}ckstern parameters for each tracking option.}
\end{center}
\end{table}

Figure~\ref{fig:SiliconPt} shows an overlay of multiple results for the 
baseline tracking design (reuse of PHENIX silicon pixels plus new silicon 
strips).  Note that this figure plots only the transverse momentum component. 
 The black dots are a full GEANT simulation  and the blue lines are analytic 
calculations.  The solid blue line is a first-principles calculation of the 
result using the Gl{\"u}ckstern formula.  Despite the fact that the formula 
is derived assuming that the detector material is uniformly distributed 
throughout the tracking volume, the agreement is excellent.  The dashed blue 
line is a modest improvement to the fit wherein the functional form was held 
fixed, but the material thickness and position resolution were allowed to 
vary as free parameters.  The light blue line in this figure indicates the 
contribution of the position resolution term alone.  Clearly the resolution 
in this detector design is nearly entirely dominated by the multiple 
scattering term.  Improved performance (in momentum resolution, mass 
resolution, and pattern recognition) could only result from lowering the 
material budget (TPC or MAPS) and not from improving the position resolution.

\begin{figure}[htb!]
\centering
\includegraphics[width = 0.4\textwidth]{figs/SiliconPt.PNG}
\caption{Transverse momentum resolution of the baseline silicon tracking 
detector design.  Black dots are from a full GEANT4 simulation.  The dashed 
blue line is a free fit to the Gl{\"u}ckstern formula, whereas the solid blue 
line is a first-principles prediction using the material list.  The light 
blue line shows the contribution from position resolution alone demonstrating 
that the baseline design is entirely multiple-scattering limited.}
\label{fig:SiliconPt}
\end{figure} 

Figure ~\ref{fig:SiliconTPCComp} shows a comparison of the analytical results 
for the momentum resolution for the baseline silicon tracker version and the 
TPC-based option.  Note that this plot shows the full momentum resolution 
(not merely the transverse component), different from the result in 
Figure~\ref{fig:SiliconPt}.  Solid lines indicate the result at $\eta$=0 and 
are thereby the same as the result from the prior figure ({\it i.e.} where 
$p_t=p$).  The dashed lines indicate the result averaged over tracks across 
the full aperture $\left| \eta \right| = \pm 1$.  The TPC is considered in 
two different ways:  standalone as indicated in pink and with a vertex 
constraint as indicated in red.  At the very lowest momentum, that TPC 
standalone results are better than those including a vertex constraint since 
these do not suffer from the multiple scattering limits of the entrance 
window, beam pipe, and vertex detector.  At higher momentum, using the vertex 
constraint improves the TPC-alone result due to the increased level arm of 
the measurement.  Clearly, during a physics analysis, one would user a 
properly weighted average of these two calculations thereby indicating that 
the lower envelope of these results indicates the detector system's ultimate 
performance.

Considering first the $\eta$=0 results (left panel, solid lines), we see that 
the hybrid TPC design provides superior performance until 
$p_T=10~\frac{GeV}{c}$ beyond which the baseline solution is better.  
However, when averaged over the full aperture (dashed lines), the hybrid 
solution has superior performance over the entire plotted range (out to 
14$\frac{GeV}{c}$) pushing the momentum at which the silicon performance 
becomes superior out to $\sim 17\frac{GeV}{c}$.  The relevant range for decay 
electrons from the Upsilon is roughly $4-10\frac{GeV}{c}$

\begin{figure}[htb!]
\centering
\includegraphics[width = 0.8\textwidth]{figs/SiliconTPCComp.PNG}
\caption{Momentum resolution of the baseline sPHENIX tracker and the hybrid 
TPC option.  Solid lines in the left panel are for $\eta$=0 ($p_T=p$), dashed 
lines are averaged over the entire aperture.}
\label{fig:SiliconPTPCComp}
\end{figure} 

The right hand panel of Figure~\ref{fig:SiliconTPCComp} shows the full 
momentum resolution of the various designs as a function of $\eta$ for 
discreet momenta 1, 3, 5, 7, and 9 $\frac{GeV}{c}$.

\begin{figure}[htb!]
\centering
\includegraphics[width = 0.8\textwidth]{figs/MassResolution.PNG}
\caption{Simulated Mass Resolution for the Upsilon states.  Black dots are 
the results from GEANT simulations.  The black histogram is the result of the 
semi-analytic calculation.  Blue curves are free fits to the $\Upsilon(1s)$.}
\label{fig:MassResolution}
\end{figure} 

Figure~\ref{fig:MassResolution} multiple calculations for both the baseline 
silicon option and for the TPC-hybrid option on reconstruction of the 
$\Upsilon(1s, 2s, 3s)$ states.  In each panel, the black dots are the result 
of a full GEANT calculation (described in detail later in the chapter).  
These results can be compared to those shown in the black solid histogram, 
which is the result of the semi-analytical calculation that uses the 
Gl{\"u}ckstern formula generalized to 3D, with the Bremsstrahlung 
after-burner applied.  Finally, the blue curves show the result of a free fit 
to the so-called ``Crystal Ball'' function that is used to extract the 
gaussian width of the $\Upsilon$(1s) state.  As one would expect based upon 
the previous results, both detector solutions provide adequate separation of 
the states with the TPC-hybrid solution producing a peak width 
$\sim\frac{2}{3}$ that of the baseline option.

\begin{figure}[htb!]
\centering
\includegraphics[width = 0.8\textwidth]{figs/BremLoss.PNG}
\caption{Statistics loss of Upsilons due to Bremsstrahlung.}
\label{fig:BremLoss}
\end{figure} 

A close examination of Figure~\ref{fig:MassResolution} shows that not only 
are the shapes of the mass resolution plots different in the two options, but 
for the same running period the yield is better for the thinner option.  This 
effect is quantified in Figure~\ref{fig:BremLoss}.  Here we quantify the loss 
in statistics due to Bremsstrahlung driving counts outside the $3\sigma$ 
momentum window.  The loss for $\frac{x}{\chi_0}$=7.9\% is 24\% of single 
tracks meaning a pair loss of 42\%.  As indicated in the right panel, these 
losses will be smaller as the detector gets thinner with pair losses of 23\% 
for a hybrid solution with TPC and reuse of PHENIX pixels and 12\% for a 
hybrid solution with a TPC and MAPS detectors of thickness equal to that 
being constructed for the ALICE ITR upgrade.


\section{Design Details}
The sections below give a detailed discussion of the designs of each detector 
option.
\subsection{Inner Tracker (Pixel)}

{\blue Yasuyuki for baseline}
\subsubsection{PHENIX VTX pixel detector }
\subsubsection{Pixel reconfiguration and support}
% \subsubsection{Pixel readout bus and electronics}
\subsubsection{Pixel cooling and cabling}
liquid cooling is required.
external connection are control/data fibers to SPIRO, LV power, and sensor 
bias.
{\blue LANL for new option}


\subsection{Outer Silicon Tracker}


The silicon strip tracker consisted of 3 stations as shown in 
Fig.\ref{fig:3D_Overview}. The S0 and S1 stations consisted of two layers 
which are separated each other by 1cm in radial direction.
Front layer (a) has fine pitch in azimuthal direction for better momentum 
resolution whereas back layer (b) has finer pitch in beam direction than the 
layer (a) for the collision vertex position determination.  Sensor modules 
are staggered each other between
adjacent sensor modules for hermeticity.  Shown in Figure~\ref{fig:3D_S1} is 
the 3D CAD design of S1 station. 


\begin{figure}[htb!]
\centering
\includegraphics[width = 0.4\textwidth]{figs/SiTracker_Silicon_3D_S1.pdf}
\caption{CAD drawing of the silicon strip tracker S1 layer. In order to 
minimize the dead area, every 
other sensor modules are staggered.}
\label{fig:3D_S1}
\end{figure} 

\subsubsection{Strip sensors {\blue Itaru}}

Each sensor is divided in cells of 9.6mm in beam direction $\times$ 7.68mm 
active area.
Each cell consists of 128 strips of 58 $\mu$m $\times$ 9.6mm. As illustrated 
in 
Figure~\ref{fig:3D_S0S1S2},  the active area of S2, S1, and S0 silicon strip 
sensors are 
segmented into 12$\times$10, 6$\times$10, 2$\times$10 cells, respectively. 
Readout lines run perpendicularly with respect to longitudinal strip 
direction. For S1 and S2
sensors, the readout line connects multiple strips in different cells to save 
channel counts.
1 channel in S2 sensor read 6 strips whereas it reads 3 strips in S1 sensor. 
Thus the readout
hit information do not distinguish the origin of the hit among multiple 
cells. However the 
offline reconstruction algorithm will be able to distinguish the cell by 
requiring matching
hits in different stations and/or layers. The chances to misidentify the 
wrong cell is very
low due to expected low occupancy in S1 and S2 sensors.  The expected channel 
occupancy is $\sim$ 0.2$\%$ in S1 and 0.1$\%$ in S2 even in central Au+Au 
collision.  

The top and bottom half cells are readout upwards and downwards, respectively.
At the edge of the sensor, the readout lines are connected to readout pads 
which are
aligned to match the FPHX chip pattern as shown in 
Figure~\ref{fig:Silicon_S1_MechanicalDrawing}.  


\begin{figure}[htb!]
\centering
\includegraphics[width = 0.7\textwidth]{figs/SiTracker_Silicon_S0S1S2.pdf}
\caption{Illustration of S2, S1, and S0 silicon strip sensors. The active 
areas are segmented into 
12$\times$10, 6$\times$10, 2$\times$10 cells, respectively. }
\label{fig:3D_S0S1S2}
\end{figure} 

This mechanical drawing of S1 sensor is designed by HPK for the standard 
thickness
320 $\mu$m. The breakdown voltage is to be $>$250V, and fully depleted 
voltage is 
to be less than 100V. The bias resistance is 5 to 15M$\Omega$ per strip. 


\begin{figure}[htb!]
\centering
\includegraphics[width = 
0.8\textwidth]{figs/SiTracker_Silicon_S1_MechanicalDrawing.pdf}
\caption{The mechanical drawing of S1 sensor designed by HPK.}
\label{fig:Silicon_S1_MechanicalDrawing}
\end{figure} 


\subsubsection{Strip HDI {\blue Itaru}}

\subsubsection{Strip sensor module {\blue Itaru}}
\subsubsection{Stave (cooling) {\blue Itaru}}
\subsubsection{Ladder {\blue Itaru}}
\subsubsection{Barrel {\blue Itaru}}
\subsubsection{Space Frame {\blue Itaru}}
\subsubsection{Readout Electronics and Cabling {\blue LANL}}
\subsubsection{SiTracker installation and integration {\blue Don Lynch}}

\subsection{TPC}

\subsubsection{TPC readout plane} % Sasha and Tom
The TPC readout plane is the most composite and multifunctional TPC element. 
On the inner side of the TPC volume the readout plane is subdivided into 
conductive readout elements (pads) connected to the electronic readout 
channels sitting on the opposite side of the plane. The amplification element 
that receives primary ionization and multiplies it to the level suitable for 
the electronics is directly coupled to the readout plane on the inner side of 
the TPC volume. The readout plane also handles some additional communications 
like high voltage, calibration and other services. The electronic and 
mechanical design of the readout plane is a multiparametric task that 
involves matching the performance of almost all TPC subsystems.

The pad structure of the readout plane is considered to be formed by 
concentric circular rings of pads spanning between the innermost and outmost 
TPC radii as schematically shown in fig.~\ref{fig:chevron_pads}. 
\begin{figure}[htb!]
\centering
\includegraphics[width = 0.6\textwidth]{figs/chevron_pads}
\caption{Schematic layout of the TPC padrows and chevron pads. {\red add some 
dimensions}}
\label{fig:chevron_pads}
\end{figure}
The radial pad size is $\sim$1~cm. The pads will have a so-called "chevron" 
shape that enforces charge sharing between the neighboring pads for the most 
precise determination of the hit azimuthal position. The transverse of the 
dimension of the pads is compatible with the size of the avalanche determined 
by electron difusion during both the drift and avalanche stage. The best 
position resolution will result from designing the diffusion to be primarily 
from the gas stage, with little diffusion from the drift.

Although it has been shown that the "straw model" resulting in $\sim$250,000 
pads will result in excellent performance, the actual number of padrows, size 
and shape of the pads, are subject of further optimization during the R\&D 
period.

The TPC amplification element is based on several layers of Gas Electron 
Multiplier (GEM) detectors. Traditional Muti-Wire Proportional Chamber (MWPC) 
technology is not considered because it a) cannot provide desired $r\phi$ 
resolution of 100~$\mu$m {\red need connection to MC section} and b) the MWPC 
require gating to stop ion backflow, and that significantly limits the data 
taking rate.

Four GEM layers are considered in the current scheme of the amplification 
element. Each GEM will provide the gain in the range of typically a few 
thousand suitable for the readout electronics considered for the TPC. The 
gain range is driven by competing factors.  Higher gains will improve the 
signal:noise and improve $\frac{dE}{dx}$ results, but will also increase the 
Ion Back Flow (IBF).  ALICE intends to run at a gain of 2000 with SAMPA chip 
readout.  ALICE results also {\red [HBD results?]} demonstrate high stability 
of GEM operation in the environment of high energy heavy heavy ion collisions.

The amplification element is shown in fig.~\ref{fig:gem_layers}. 
\begin{figure}[htb!]
\centering
\includegraphics[width = 0.8\textwidth]{figs/quattro_gem}
\caption{Schematic view not to scale of the readout element built with four 
layers of GEMs. yellow lines show electron paths, brown lines show the ion 
paths for one single hole (simulation) from {\red add Ref.}}
\label{fig:gem_layers}
\end{figure} 
Besides providing the gas amplification suitable for the electronics the 
element also has two other functions. 

The transverse size of the ionization cloud arriving to the readout plain 
from the drift volume is defined by the gas diffusion and by the longitudinal 
(w.r.t. to the drift direction) magnetic field. As shown in later sections, a 
strong magnetic field limits transverse diffusion significantly.  Some 
diffusion will, however, be necessary to spread the signal charge across 
multiple electrodes. By using several layers of GEM, an effective 
"hole-misalignment" diffusion term is introduced.  Often the hole patterns on 
the individual GEMS (holes make equilateral triangles) are rotated by 90 
degrees to avoid large regions of accidental hole alignment (Moire pattern).  
ALICE plans to not only charge the orientation of the holes layer-by-layer, 
but also their spacing.

The most important task for operating TPC at high rates is to suppress the 
Ion Back Flow (IBF) that leads to the built up of the space charge in the 
drift volume of the TPC and, as a result, to the distortion of the of the 
primary ionization arrival point onto readout plane. The ions are coming from 
the collision particles ionizing gas in the TPC volumes but mainly from the 
avalanches in GEM layers. The second contribution is by several orders of 
magnitude (the gas gain factor) is larger than the first contribution. The 
number of ions arriving from lower (closer to readout plane) GEM layers is 
higher than from the upper layers. Properly configuring electric fields 
inside and between the GEM layers one can significantly suppress the second 
contribution by redirecting ions to drift towards the back planes of upper 
GEMs as shown in fig.~\ref{fig:gem_layers}. Recent studies made by ALICE 
collaboration {\red[need citations]} and shown in fig.~\ref{fig:ibf} 
demonstrates dependence of the IBF on the charge measurement resolution at 
the constant gas gain. 
\begin{figure}[htb!]
\centering
\includegraphics[width = 0.6\textwidth]{figs/ibf_alice.png}
\caption{{\red needs ref}}
\label{fig:ibf}
\end{figure} 
This results show that the IBF can be reduced to $\approx2$\% for the charge 
determination resolution of approximately 12\%. These is considered 
acceptable level for operating ALICE detector.

\subsubsection{TPC field cage} % Tom and Sasha
The basic function of the TPC field cage is to provide a uniform drift field 
from the central membrane to the detector modules at each end.  This field 
cage is traditionally defined by a series of conducting rings held at 
uniformly decreasing potential by a precision-matched chain of resistors.  
The field cage is then surrounded by a gas enclosure.  Both for safety 
considerations and to avoid stray electric fields in neighboring detectors, 
the gas enclosure is usually grounded.  Figure~\ref{fig:FieldCage} shows the 
configuration found on the outer shell of the STAR TPC.  Both the field cage 
and the gas enclosure are made structurally rigid using a hexcell honeycomb 
sandwich structure.

\begin{figure}[htb!]
\centering
\includegraphics[width = 0.4\textwidth]{figs/FieldCage-STAR.PNG}
\caption{Scale drawing of the outer field cage and gas enclosure for the STAR 
TPC.}
\label{fig:FieldCage}
\end{figure} 

The field cage electrodes are made as a double-layer of staggered rings, one 
facing the operating gas and the other embedded in the field cage wall.  The 
latter ring serves to shape the field and minimize nonuniformities in the 
drift volume.  Dry nitrogen gas flows through the 5.7~cm gap, exceeding by 
slightly more than a factor of two the "rule~of~thumb" gap dielectric 
strength of $1\frac{kV}{mm}$ when operating at a central potential of 27~kV.  
Although in STAR the inner gas enclosure is skipped (exposing the field cage 
strips to outside air and stressing inner detectors with electric field) in 
the sPHENIX application we have more than enough room between the inner 
silicon pixels for an inner gas enclosure.  Scaling to an identical safety 
factor as used by STAR, we would require a $5.7cm\frac{34kV}{27kV}=7.2cm$ gap.

An ``air'' gap of this size would be undesirable for the outer TPC wall since 
it would limit the active volume and degrade the momentum resolution.  
Because the TPC is followed by the EMCAL, we can safely afford to solve the 
field issue using a solid of high dielectric strength.  The concern over this 
solution is two-fold.  First, the dielectric field strength of common 
materials is found to reduce with time in a variety of materials as shown in 
Figure~\ref{fig:Dielectrics}.  Much of this variation ({\it e.g.} FR4) is 
dominated by micro-gas bubbles within the material which can carbonize over 
time.  Secondly, dependent upon material, solid material high voltage gaps, 
can be subject to permanent failure during a discharge event or over-time 
corona current.  
\begin{figure}[htb!]
\centering
\includegraphics[width = 0.6\textwidth]{figs/HVPF.PNG}
\caption{Dielectric strengths of various common circuit card materials, 
reproduced from figures by Sierra Proto Express, a Palo Alto-based circuitry 
company specializing in high voltage circuit card for both terrestrial and 
satellite applications.}
\label{fig:Dielectrics}
\end{figure} 

sPHENIX is working with the Sierra Proto Express company to develop a robust 
solid core solution for the outer field cage that would maximize the 
reliability and longevity of the device.  Although a multi-material, layered 
ultimate design is likely, the table below shows the required thicknesses for 
safety factors of 3X and 5X in the design assuming a single material type and 
neglecting contributions other than the insulator itself.  Calculations here 
use the worst-case aging estimates from Sierra for each material type.  These 
initial calculations seem promising, meaning that the "air gap" solution is 
presently considered only as a fallback option.  If the solid option 
realization has a sifficiently small radiation length, it can also be 
considered for the entrance window, thereby simplifying the design.

\begin{center}
\begin{tabular}{|c|c|c|c|c|}
\hline
Material & $\chi_0$ (cm) & Volt/mil & 3X Safety & 5X Safety\\
\hline
FR4 & 16.76 & 150 & 1.72~cm (10.3\%$\chi_0$) & 2.88~cm(17.2\%$\chi_0$)\\
Kapton & 28.58 & 500 & 0.52~cm (1.8\%$\chi_0$) & 0.86~cm(3.0\%$\chi_0$)\\
HVPF & 28.57 & 2000 & 0.13~cm (0.45\%$\chi_0$) & 0.22~cm(0.75\%$\chi_0$)\\
\hline
\end{tabular}
\end{center}

The endplates of the TPC will likely be shaped in a "spoke" pattern and house 
either 6 gain stage modules on each end or 8 gain stage modules.  The 
6-module solution provides the largest live area and least material that 
could disturb future upgrade plans.  The 8-module design would use smaller 
GEMS and thereby involve a moderately reduced risk.

\subsubsection{TPC gas} %Tom and Sasha
The choice of gas is a critical design criterion for the TPC as this choice 
affects the potential of the central membrane, single point resolution, 
multiple scattering, and positive ion feedback.  Currently, the largest 
effort has been put into considerations of the so-called ``T2K'' gas and the 
``ALICE'' gas.  The characteristics of these two gasses are summarized in 
Figure~\ref{fig:GasCompare}.

\begin{figure}[htb!]
\centering
\includegraphics[width = 0.6\textwidth]{figs/GasCompare.PNG}
\caption{Comparison of the electron transport properties of T2K gas 
($Ar$:$CF_4$:$iC_4H_{10}$ 95:3:2) and ALICE gas ($Ar$:$CO_2$ 90:10).}
\label{fig:GasCompare}
\end{figure} 

The T2K gas has exceptionally low diffusion ($~50\frac{\mu m}{\sqrt{cm}}$) at 
a rather low drift field of ($\sim 120\frac{V}{cm}$).  At 85 cm of drift, the 
central membrane would only require a voltage of $\sim 10~kV$.  It is also a 
"fast gas" with a drift velocity at minimum diffusion of $\sim 60 \frac{\mu 
m}{nsec}$ and thereby a full TPC drift time of $14~\mu sec$.  Furthermore, 
T2K gas is near a velocity plateau, thereby relaxing the constraints on field 
uniformity.  Conversely, the ALICE gas has a significantly higher minimum 
diffusion ($\sim 150 \frac{\mu m}{\sqrt{cm}}$) and lower drift velocity 
($~22\frac{\mu m}{nsec} \rightarrow 34~\mu s$).  Recently ALICE added $N_2$ 
to the mix which slightly increases the drift speed as noted earlier.

The naive assumption favors low diffusion, high speed, and a velocity plateau 
making the T2K gas an apparent winner.  For our application, further 
consideration is required.  Since we will likely be using MPGD-style 
avalanche, we {\bf rely} upon the diffusion of the charge to spread the 
avalanche across the pads.  To achieve a sigma of 
$\frac{1}{2}$pad-width~=~$500\mu m$, would require at least 1 meter of drift 
length.  In the ALICE gas, this level of diffusion is achieved after only 10 
cm of drift.  Furthermore, since the slow shaping time of the SAMPA chip (190 
nsec) is significantly longer than the charge collection time in either gas, 
the effective voxel occupancy of T2K gas is more than twice that of ALICE 
gas.  For heavy ion collisions with embedded high momentum Jets, this might 
be a significant limiting factor to the detector performance.  Furthermore a 
significant limiting factor for TPC operation at high rate is ion feedback.  
Selecting a light dominant noble gas ({\it e.g.} Ne instead of Ar) can 
increase the ion mobility by factors approaching 10X.  STAR (Ar-based) is 
dominated by space charge from the primary ionization.  After upgrade ALICE 
will have positive ion feedback from the ungated avalanche stage, but not 
from the primary beam.  To first approximation, if we were to use an Ar-based 
(low ion mobility) gas we would encounter both the primary ionization load 
and the ion backflow load.

\begin{figure}[htb!]
\centering
\includegraphics[width = 0.6\textwidth]{figs/GasCompare2.PNG}
\caption{Alternative gas choices that may be better compromises for sPHENIX.}
\label{fig:GasCompare2}
\end{figure} 

As stated earlier, we have concentrated this far on T2K gas and ALICE gas as 
two extreme choices.  Figure~\ref{fig:GasCompare2} shows two different 
Neon-based gas choices with difference choices of quench ($CF_4$ vs. $CH_4$). 
 Both of these combinations have a reasonably flat velocity curve near the 
diffusion minimum and would require a rather low central membrane potential 
($\sim 10~kV$), reasonable drift velocity, and a diffusion value that is 
enough to ensure good charge sharing for all drift lengths.  Among the 
parameters plotted here, both would appear to be superior to the ALICE gas, 
with the edge going to the $CH_4$ mixture.  This last one is effectively STAR 
gas with the argon swapped for Neon.

An additional, but critical parameter regards the ion mobility in the TPC 
gas.  Positive ion transport through a gas follows the form:
\begin{eqnarray}
v_{ion~drift}=KE
\end{eqnarray}
where K is the ion mobility (typically in units of $\frac{cm^2}{Volt \cdot 
sec}$) and E is the applied electric field.  Since the prior expression is 
dependent upon the number density of the gas, it is frequently rewritten in 
terms of a reduced mobility $K_0$ and a field/number density $\frac{E}{N}$ in 
units of the Townsend (Td).  For reference, room temperature gasses have 1~Td 
equating to $E~=~250\frac{V}{cm}$.  Although the ion mobility does indeed 
vary with applied field, it is typically flat until $\sim$10~Td, which is far 
in excess of any drift field we would consider for the TPC.  Therefore, the 
ion velocity will be proportional to the applied electric field, which should 
be as high as possible to minimize space charge effects.

Ion velocity in mixed gasses is a well understood phenomenon and follows 
Blanc's Law for which:
\begin{eqnarray}
\frac{1}{K_{TOT}}=f_1\frac{1}{K_{11}}+f_2\frac{1}{K_{22}}
\end{eqnarray}
in exact analogy to resistors in parallel.  This was shown for Argon mixtures 
in 1977 by Charpak and Sauli and measured more recently for neon mixtures as 
illustrated in Figure~\ref{fig:Blanc}.
\begin{figure}[htb!]
\centering
\includegraphics[width = 0.6\textwidth]{figs/Blanc.PNG}
\caption{Inverse ion mobility as a function of gas fraction demonstrating 
Blanc's Law.}
\label{fig:Blanc}
\end{figure} 
Since the TPC gas is likely to be dominated by the noble component, it is 
interesting to compare the ion mobilities for three pure noble gasses as 
shown in the table below.
\begin{center}
\begin{tabular}{|c|c|c|c|}
\hline
Gas & K ($\frac{cm^2}{Volt \cdot sec}$) & $v_{D}\left( E=130\frac{V}{cm} 
\right)$ &$v_{D}\left( E=400\frac{V}{cm} \right)$\\
\hline
Ar & 1.51 & 196 & 604 \\
Ar-$CH_4$ 90:10 & 1.56 & 203(STAR) & 624 \\
Ar-$CO_2$ 90:10 & 1.45 & 189 & 582\\
Ne & 4.2 & 546 & 1680\\
Ne-$CH_4$ 90:10 & 3.87 & 503 & 1547 \\
Ne-$CO_2$ 90:10 & 3.27 & 425 & 1307(ALICE)\\
He & 10.2 & 1326& 4080\\
He-$CH_4$ 90:10 & 7.55 & 981 & 3019\\
He-$CO_2$ 90:10 & 5.56 & 722 & 2222\\
T2K & 1.46 & 190(ILC) & 584\\
\hline
\end{tabular}
\end{center}
It is clear that the space charge issues in STAR and ALICE are of an entirely 
different nature.  in STAR, the ion mobility is low enough that the positive 
argon ions from the primary charge generate track distortions.  In ALICE, 
both the noble gas choice (Ne instead of Ar) and the high drift field, 
dramatically reduce the distortions due to the space charge from the primary 
ionization.  After upgrade, ALICE will struggle primarily with the ion back 
flow from the amplification stage.  For sPHENIX, we should choose to avoid 
corrections from both these contributions and therefore we are driven toward 
Ne or He as the noble gas and the use of a strong drift field.

\subsubsection{TPC fab} % Tom and Sasha

Because of the size of the TPC, the fabrication of all parts could, in 
principle, be accomplished at any of our collaborating institutions 
worldwide.  That said, it would nonetheless be simplest of the field cage 
assembly was accomplished locally with smaller parts made around the world.  
This model proved quite effective in building the PHENIX Hadron Blind 
Detector, wherein the individual parts were of manufacture at the Weizmann 
Institute of Science in Israel and the assembly was accomplished at Stony 
Brook University.

Because of the need to maintain active area to the largest radius, our 
designs for the TPC field cage and gas enclosure will be biased toward the 
thinnest of robust designs.  Thus, the STAR and ILC field cage designs are 
the most appropriate as models for our work.  Those devices were manufactured 
using large mandrels upon which layers of flexible circuit card and honeycomb 
were applied.  Each mandrel is designed to release by "collapsing" to smaller 
radius after the TPC shell is cured, thereby releasing the shell.  The 
completed shells are then outfitted with aluminum spoke-like endcaps and a 
central membrane to form the completed field cage.  We intend to design the 
field cage to safely hold the highest potential currently under investigation 
(ALICE gas $\sim37~kV$).

The open ports between the spokes of the endcaps will be filled with 
"mechanical blank" modules to allow the field cage to become gas tight during 
the prototyping stage.  This will allow full testing of the high voltage 
stability of the field cage without any of the gain stage modules in place.

During the prototyping stage, single items of prototype gain stage module 
will be built.  Because of the finite size of these units, a list of possible 
institutions are capable of prototype construction including Weizmann, Stony 
Brook, BNL, Yale, Wayne State, and UT.  All of these institutions have past 
experience in the the PHENIX HBD or in the ongoing construction of the inner 
TPC layers for the ALICE upgrade.  We envision two full sized prototypes 
whose design is driven by results from our ongoing TPC gain stage R\&D which 
has been funded by the EIC R\&D program.  As described below, we have already 
garnered extensive experience in multiple gain stage technologies, as well as 
a number of clever readout scheme applications.

The so-called "pre-production prototype" will be the third and final stage of 
full sized prototype construction.  Barring any discovered deficiencies, 
"production" would involve the manufacture of the remaining gain stage 
modules as well as spare units.  As with the prior work, it is likely that 
much of this effort will take place "off site" from the location of the field 
cage itself, with working modules shipped via clean dust-free packaging.

\subsubsection{TPC Electronics}
The geometry of the TPC is assumed to have 30\,cm inner radius and 80\,cm
outer radius, resulting in the cross-section in radius direction as
5500\,cm$^{2}$ ($\equiv$550000\,mm$^{2}$). {\red this is per one side} With 
the pad size of
6\,mm$^{2}$, {\red shall 6 sq.mm, or any other number be motivated in 
previous sections?} We will readout $\sim$90k channels for each side, and 
$\sim$180K
channels in total. We have to wait for the results from the detail simulation,
but we assume that the number of readout channels to be 200K. We need the
readout electronics that can accommodate 200k channels with 10\,\%
spares.

The interaction rate in sPHENIX for Au+Au collisions is $\sim$25\,kHz, which
is very similar spec or a bit more relaxing as compared to the ALICE TPC
upgrade specification. The ALICE TPC at LHC is required to readout the
signals continuously at 50\,kHz in Pb+Pb collisions. Therefore, we will
base our electronics on the ALICE TPC upgrade electronics.
Figure~\ref{fig:ALICEDAQ} shows the block diagram of signal processing
for ALICE TPC upgrade electronics.
\begin{figure}[htb]
  \centering
  \includegraphics[width=0.8\linewidth]{figs/ALICE_TPC_Readout_Scheme.pdf}
  \caption{Block diagram of signal processing for ALICE TPC upgrade}
  \label{fig:ALICEDAQ}
 \end{figure}
We explain the diagram from the end part of the signal processing. The DCS
stands for Data Control System, and online farm is the computer system
where the data are stored and processed for analysis. The LTU provides
the timing and trigger signal to CRU (common readout unit), which is the
post-processing system where some online calibrations and event reconstruction
is performed.

The FEC stands for front end card, consists of SAMPA chips which
amplify and shape the analog signals and digitize. The DSP (data
processing unit) is also on the chip which formats the digital
data into a data packet (this also performs baseline suppression,
i.e., zero-suppression of the raw data). The packet is then sent to
GBTx followed by VTTx. They convert the data packet into optical signals.
In the PHENIX case, DCM-II is defined as the destination of the data to
be sent optically. Since, we don't perform calibration or reconstruction
online, we wouldn't need CRU as long as the data format for DCM-II is
simple enough that the DSP can fully handle.

The block diagram of the SAMPA chips is shown in Figure~\ref{fig:SAMPA}.
\begin{figure}[htb]
  \centering
  \includegraphics[width=0.8\linewidth]{figs/ALICE_SAMPA.pdf}
  \caption{Block diagram of ALICE SAMPA chip}
  \label{fig:SAMPA}
 \end{figure}
In the ALICE design, there will be 5 SAMPA chips multiplexed by
2GBTx ASICs. One SAMPA chip accept 32 inputs, therefore one FEC
can process 160 inputs. The ALICE TPC will install 121 FECs per
readout segment module. The TPC will be equipped with 18 segments
in each side, 36 segments in total.
It should be noted that the simple math using the number of FECs,
segments and channels per FEC doesn't give the 550K channels 
of electronics that the ALICE claims to need. We don't discuss
here about it, since this is not relevant for estimation of
our needs for electronics. We show the FEC block diagram
designed by the ORNL group in Figure~\ref{fig:TPCROC}, and the
actual connection scheme to the TPC in Figure~\ref{fig:FEC_Config}.
\begin{figure}[htb]
  \centering
  \begin{minipage}{0.48\linewidth}
  \includegraphics[width=0.8\linewidth]{figs/FEC_Diag.pdf}
  \caption{Conceptual Board Schematics of ALICE TPC FEC}
  \label{fig:TPCROC}
 \end{minipage}
\begin{minipage}{0.48\linewidth}
  \includegraphics[width=1.0\linewidth]{figs/ALICE_TPC_ROC_Config.pdf}
  \caption{Installation image of ALICE TPC FEC to TPC}
  \label{fig:FEC_Config}
  \end{minipage}
 \end{figure}
The cables from TPC to the FECs are part of the connector on the
FECs. So, we don't need to prepare additional signal cables for
this purpose. Therefore, once the FEC is designed and fabricated
with the optimal form factor that fits to the TPC, the only
remaining off-board parts are power supplies and their cables, and
optical fibers to DCM-II.
The power supply that ALICE chose is Wiener PL 500 F12. One power
supply can take care of two readout segments, which is 242 FECs
as shown in the diagram in the Figure~\ref{fig:TPCPowerdiagram}.
\begin{figure}[htb]
  \centering
  \includegraphics[width=0.4\linewidth]{figs/power_distr_PL500_F12_2.pdf}
  \caption{Diagram of power supply for ALICE TPC FEC}
  \label{fig:TPCPowerdiagram}
 \end{figure}
ALICE needs 18 of them + spare, but we need only 6 of them + spare
(we accounted for totally 7 modules for now).
The picture of power supply itself is shown in the
Figure~\ref{fig:TPCPowerSupply}.
\begin{figure}[htb]
  \centering
\begin{minipage}{0.48\linewidth}
  \includegraphics[width=1.0\linewidth]{figs/Maraton_distri_back.JPG}
\end{minipage}
\begin{minipage}{0.48\linewidth}
  \includegraphics[width=1.0\linewidth]{figs/PL512.JPG}
\end{minipage}
\caption{Power Supply module, Wiener PL 500 F21, employed for ALICE TPC 
electronics.}
\label{fig:TPCPowerSupply}
\end{figure}
We learned that the cables for connecting FECs and power supplies
are provided by CERN and the cost is not known. We have to estimate
by ourselves for our TPC case, but we suppose the cost is negligibly
small compared to FECs or power supplies themselves.

Based on the difference between the number of channels between the
ALICE case and our case, without taking the contingency into
account, the cost for 200k channel electronics for our TPC will
be $\sim$\$2-2.2M.
There is still uncertainty on the time scale on the SAMPA chip
development, and this will affect to our FEC development. If the
SAMPA chip development will be significantly delayed, we may
consider taking the old ALICE TPC electronics scheme; design the
FECs with PASA and ALTRO chips, which is basically the split of
the function to be integrated into the SAMPA chip. Another concern
is the optical readout module. The ALICE uses rad-hard optical
module which is a common device throughout the LHC experiments.
The sPHENIX can use the commercially available optical module
if it meets the radiation condition of the RHIC. We will also
investigate this option that will likely reduces the cost.

\subsubsection{TPC cooling and cabling} % Tom and Sasha

Our cooling requirements on the TPC electronics will be significant.  
Although we are only cooling $\frac{1}{2}$ as many channels, these channels 
are distributed over only $\frac{1}{10}$ as much surface area.  Therefore the 
power required from our cooling plant will be less overall, but we will need 
to design for very effective heat transfer to the cooling lines.

\begin{figure}[htb]
  \centering
  \includegraphics[width=0.35\linewidth]{figs/Cooling.png}
  \caption{Diagram of the cooling plant in use the the ALICE TPC.  The 
cooling plant is an underpressure system so that any leak results in gas 
bubbling into the coolant rather than coolant dripping into the detector.}
  \label{fig:Cooling}
 \end{figure}

Figure~\ref{fig:Cooling} shows the configuration of the cooling plant 
currently in use by the ALICE experiment.  The key feature of this cooling 
plant is that the coolant is delivered at pressures below one atmosphere so 
that in the event of a leak, gas is introduced into the coolant rather than 
coolant introduced into the gas.  The ALICE resistor chains dissipate a 
significant high power (8W in each of 4 resistor bars).  Higher power in the 
resistor chain is driven by the need for robust performance in the face of 
stray currents due to nearby ionization.  Although the track density on 
sPHENIX and ALICE are very similar, the change load onto the ALICE TPC frame 
is much higher.  Among STAR, ALICE, and ILC, only ALICE water cools their 
resistor chain.  Since our power dissipation will be the least of these three 
applications, we are safest to not water cool the resistor chain and thereby 
retire the risk of water leaking into the chamber from the outset.

The cable plant for the TPC includes a pair of shielded coaxial high voltage 
lead whose diameter will be under $\frac{1}{2}"$ ({\it e.g.} Dielectric 
Sciences 2125: $100~kV; \o$ 0.4").  Each sector will receive bias for the 
GEMstack as 8 independent voltages.  The readout cards, will receive DC power 
input, optical connections for slow control and optical connections for data 
output.  The whatever extent possible this significant cable plant will be 
localized so as to align with the end cap spokes to minimize radiation depth 
for endcap detectpr systems.

\subsubsection{Pixel/TPC installation and calibration} %  Don Lynch Tom and 
Sasha

The assembly order for sPHENIX specifies that the "Tracker Barrel" assembly 
will be inserted from the end after the calorimeters have already been 
installed onto the magnet.  Although the utility of all the sPHENIX 
calorimeters leading into the Electron-Ion Collider era is clear, the 
tracking situation is less clear.  This distinction is one of the reasons why 
multiple tracking solutions are currently under consideration.  The TPC/MAPS 
solution is currently considered as the optimal choice for the EIC due both 
to minimal radiation length and excellent hadron-ID capability at low 
momentum.  However, to retain future flexibility and simple upgrade 
capability, the tracking systems, regardless of inner details, will be 
mounted to an external space frame.  This space frame will carry the load of 
the tracking detectors, and define the so-called Tracker Barrel.

TPC calibration will be achieved using a laser system, similar in philosophy 
to that used by STAR and prototypes for the ILC.  Because the work function 
of aluminum is low, a UV flash will release electrons.  Both the STAR TPC and 
the ILC TPC prototype used a pattern of aluminum applied to the central 
membrane to produce these reference tracks.  The pattern used by STAR 
consists of lines shown in Figure~\ref{fig:CentralMembrane}, whereas that of 
the ILC was a pattern of dots.  The laser system will not only provide an 
initial reference calibration, but can be fired at regular intervals (PHENIX 
fires their EMCAL laser at 1 Hz) during data collection to provide a 
continuously calibration of the drift velocity and space charge distortions.  
Gain calibrations can be roughly estimated using cosmic rays, but final 
calibration will use collision data.

\begin{figure}[htb!]
\centering
\includegraphics[width = 0.4\textwidth]{figs/CentralMembrane.jpg}
\caption{Photograph of the central membrane of the STAR TPC.  The pattern of 
Aluminum strips is used to release electrons via laser flash as a calibration 
signal.}
\label{fig:CentralMembrane}
\end{figure} 

\section{Simulations}  
{\blue Tony coordinates this section, including Alan and Ron, also some LANL 
people} \\

\subsection{TPC Simulations}
The TPC simulations performed target a realistic representation of the 
cluster size and two-hit resolution  based on design parameters \textit{which 
are consistent with those described in the previous section}.  Thus far the  
effects of space charge have not been included, but \textit{ the effects of 
space charge should be within the tolerances of already working TPC 
experiments }.

GEANT4 is used to record energy deposits in a cylindrical volume of gas.  In 
the results shown below, the volume was filled with Argon ("G4\_Ar") in order 
to simulate $dE/dx$ of the T2K gas used in the ILC TPC prototype.  The energy 
deposits are recorded in discrete radial regions of the cylindrical volume.  
For each region, a Poisonnian random number of ionization electrons are 
produced along the track trajectory according to measured values of the 
average ionization per energy deposit for the simulated gas.  Each electron 
is then randomly diffused in 3 dimensions according to measured values in the 
gas with a desired electric field \textit{ see table below }.  The average 
diffusion is then added in quadruture with a 300 $\mu$m diffusion to emulate 
diffusion during the amplification stage of readout.

The $r - \phi$ readout is simulated using a plane of rectangular pads.  Each 
electron adds to a 12-bit ADC for each pad directly in proportion to the 
number of diffused electrons reaching the pad (gain fluctuations are not 
currently simulated).  For the $z$ direction, the analogue timing response is 
not simulated but rather each electron is placed into a time "bin" with width 
corresponding to $\sqrt{12}$ times the expected width of the shaper pulse.  
If anything the two-hit resolution in the $z$ direction is pessimistic in the 
current simulation.

\begin{figure}[h]
\includegraphics[page=1,clip,trim= 0px 260px 0px 
0px,width=\textwidth]{figs/drift.pdf}
\caption{ schematic illustration of electrons drifting to the pad plane in 
the TPC sim
ulation (pad size not drawn to scale) }
\label{fig:tpc_pad_drift}
\end{figure}

After the pad ADC has been recorded in each time bin, clustering is performed 
to group pad,time-bin pairs into 3-dimensional detector hits to be passed to 
the track-finding algorithm.  In the current simulation, clustering is 
performed independently for each pad,time pair corresponding to a given 
radial pad index.  This is done for simplicity, but the pad length in the 
direction corresponding to radial measurement is much larger than the typical 
electron diffusion so not much generality is lost in this simplification.  
Figure \ref{fig:tpc_pad_drift} shows the readout and clustering process 
schematically.

\begin{figure}[h]
\includegraphics[width=\textwidth]{figs/tpc_mom_resolution.pdf}
\caption{ comparison of the momentum resolution of the simulated TPC with an 
analytic calculation in one unit of pseudorapidity }
\label{fig:tpc_mom_res}
\end{figure}

\begin{figure}[h]
\includegraphics[width=\textwidth]{figs/TPC_Ups_res.pdf}
\caption{ Upsilon 1S, 2S, and 3S resolution for the simulated TPC along with 
an analytic calculation }
\label{fig:TPC_Ups_res}
\end{figure}

In addition to the TPC, the two inner pixel layers from the baseline sPHENIX 
proposal are included in the tracking setup.  The clustering is performed on 
the silicon hits as in the baseline silicon simulation.

Finally, the clusters are passed to a hough-transform-based algorithm for 
pattern recognition and fit with a Kalman filter.  The momentum resolution 
achieved is compared to an analytic calculation for the particular TPC 
simulated in Figure \ref{fig:tpc_mom_res}.  The mass resolution for single 
Upsilons of the 1S, 2S, and 3S states along with an analytic calculation 
given the TPC parameters is shown in Figure \ref{fig:TPC_Ups_res}.

\begin{itemize}
\item[] Single particle momentum resolution {\blue Tony, Alan, Ron}
\item[] Upsilon mass resolution {\blue Tony, Alan, Ron}
\item[] Pattern recognition {\blue Tony, Alan, Ron}
\item[] \AuAu HIJING{\blue Tony, Alan, Ron}
\item[] Contribution of rejection vs efficiency for electrons {\blue Tony, 
Alan, Ron}
\end{itemize}

\section{Prototyping and R\&D}

{\blue Itaru for strips;\\ Mike McC, Ming for new Pixels;\\ Tom, Sasha, Ron 
for TPC;\\ Mike McC for strip electronics;\\ Takao, Ron for TPC electronic}

\subsection{Silicon Tracker R\&D {\blue Gaku}}
\subsubsection{Overview}

The silicon tracker R\&D has been performed since 2014. In particular in 
2015, we are focusing on the design of the silicon sensors and HDIs and on 
the production of the prototype module. The prototype module testing will 
start at RIKEN in early 2016.

\subsubsection{Sensor Prototype}

As shown in Sec.~\ref{sec:OuterSiTracker}, the outer silicon tracker consists 
of the three sub-detectors, i.e., S0, S1, and S2 from the inner. We have 
totally five prototype silicon sensors for the S2 detector (see 
Fig.~\ref{fig:S2sensor}). The prototype AC coupled sensors, fabricated by 
Hamamatsu Photonics K.K. in 2014--2015, have the thickness of 320\,$\mu$m, 
the active area of $96 \times 92.16$\,mm$^2$, and $128 \times 24$ readout 
channels. According to the quality assurance test at RIKEN in March 2015 
(\textcolor{red}{correct?}), we found neither bad channel nor bad strip for 
all sensors, and obtained the bias voltage $V_\text{fd} \approx 50$\,V and 
the breakdown voltage $V_\text{breakdown}$ larger than 250\,V. This 
performance was satisfactory for the requirements for the S2 detector 
(\textcolor{red}{OK?}).

\begin{figure}
\centering
\includegraphics[width=8.5cm, keepaspectratio]{./figs/S2sensor.jpg}
\caption{The prototype silicon sensors for the S2 layer.}
\label{fig:S2sensor}
\end{figure}

For the S1 detector, we have a more strict requirement for the total material 
budget to achieve a good momentum resolution (\textcolor{red}{refer to 
simulation section}).
Thus we consider the usage of the silicon sensors with the thickness of 
either 240\,$\mu$m or 320\,$\mu$m depending on their performances.
Compared with the latter one, the sensor with the thickness of 240\,$\mu$m is 
25\,\% smaller contributory to the material budget, however is expected to be 
accompanied with 2--3 times larger dark current.
Totally eight prototype silicon sensors for the S1 detector will be produced 
at Hamamatsu Photonics in December 2015, i.e., four sensors with the 
thickness of 240\,$\mu$m and the other four sensors with the thickness of 
320\,$\mu$m.
They will be inspected at RIKEN in early 2016. Figure~\ref{fig:S1sensor} 
shows the layout of the prototype silicon sensor for the S1 detector.

\begin{figure}
\centering
\includegraphics[width=8.5cm, keepaspectratio]{./figs/S1sensor.pdf}
\caption{Layout for the silicon sensor for the S1 detector.}
\label{fig:S1sensor}
\end{figure}

\subsubsection{HDI prototype}

HDI for the outer tracker has been designed by taking into account 1) the 
effects of both common mode and differential mode noises, 2) the geometrical 
size, and 3) the material budget.
The second point is carefully treated for the S0 and S1 detectors, since 
these detectors must be compact for installations into tiny space (see 
Sec.~\ref{sec:OuterSiTracker}).
The third point is important especially for the S1 detector where effects of 
a multiple scattering would be significant.
Thus we started the HDI R\&D with putting a higher priority to the S0 and S1 
detectors compared with the S2 detector.

The current design of the HDI for the S1 detector has seven layers of 
flexible printed circuits and has the size of 30\,mm (W) $\times$ 500\,$\mu$m 
(H). Depth of the HDI has the variation from 10\,cm to 40\,cm depending on 
its location in the ladder, which is explained in 
Sec.~\ref{sec:OuterSiTracker}. The size of the HDI is well below the 
geometrical requirement. The prototype HDI has an approximately 0.5\,\% 
(\textcolor{red}{to be updated}) radiation length in average.
The seven layers of flexible printed circuits have been designed with a 
special emphasis on a reduction of an unwanted microstrip antenna, and on 
good impedance matching even at $\sim 40$\,cm distant from the FPHX readout 
chip.
Production of the prototype HDIs for the S1 detector will be performed by 
Yamashita Materials Corporation and will be ready for further assembly at the 
end of 2015.

\subsubsection{Silicon Module Assembly}

One silicon sensor, two HDIs with ten FPHX chips for each, and one support 
frame are assembled into a silicon module as shown in 
Fig.~\ref{fig:SiliconModule}.
In the assembly, the bonding of the FPHX chips to the silicon sensor and 
their attachment to the two HDIs are performed as well. Hayashi watch-works 
co., ltd, having the experience to produce the PHENIX VTX detectors, will be 
employed for the assembly of the prototype module.

\begin{figure}
\centering
\includegraphics[width=8.5cm, keepaspectratio]{./figs/SiliconModule.jpg}
\caption{Layout for the silicon sensor module for the S1 detector.}
\label{fig:SiliconModule}
\end{figure}

\subsubsection{Silicon Module Test Plan}

The first prototype module for the S1 detector will be tested at RIKEN. Since 
the sPHENIX outer tracker uses the  essentially same electronics as the 
PHENIX FVTX detector, for example, the FPHX readout chip and front-end 
circuits, the test bench at RIKEN that was originally setup for the high 
multiplicity trigger development for the FVTX detectors. This test bench 
setup can be used again to test prototype sensor modules with minor 
modification in readout configuration. The test is planned in 2016 spring and 
summer.

\subsubsection{Ladder Prototype}

Schematic view of the prototype ladder for the S1 detector is shown in 
Fig.~\ref{fig:SiliconLadder}. The ladder consists of the hollow frame made by 
carbon and silicon modules to be attached to the top and bottom of the hollow 
frame in series.
Currently the design of the prototype ladder is ongoing. Here we take account 
of the capability of air-cooling the FPHX chips and the mechanical vibration 
of the ladder due to an air-blowing inside the frame itself.

\begin{figure}
\centering
\includegraphics[width=8.5cm, keepaspectratio]{./figs/SiliconLadder.jpg}
\caption{Layout for the silicon sensor ladder for the S1 detector.}
\label{fig:SiliconLadder}
\end{figure}

\subsection{MAPS R\&D}
Among the vulnerabilities of the MAPS technology is the tendency for single 
pixels to "latch up" in a state of large current draw that can, if not 
promptly reset, lead to burn-out of single pixels.  The initial operation of 
the STAR HFT device (the first large-scale MAPS deployment in a major 
experiment) not only discovered this phenomenon, but also found the 
appropriate mediating course of action.  Due to the intrinsically low power 
of the MAPS sensors, the lock up of even a single pixel produces a measurable 
change in Dc current draw and can thereby be used to shut down and 
subsequently reset the chip.  Following the establishment of this protocol, 
STAR experienced a very low rate of failures.

In summer of 2015, two students from SBU working together with scientists 
from LBNL performed measurements on latch up rates in the ALPIDE (ALICE PIxel 
DEtector) chip by exposing the chip to beams of various heavy ions produced 
by the Berkeley 88" cyclotron.  Figure~\ref{fig:CycloTest} shows the 
experimental setup used.

\begin{figure}
\centering
\includegraphics[width=0.7\textwidth]{figs/CycloTest.PNG}
\caption{The right hand panel shows the experimental setup used to expose 
ALPIDE MAPS chips to highly ionizing radiation to induce "Latch Up" events 
and study their rate.  The table indicates the various ion species used 
during the test.}
\label{fig:CycloTest}
\end{figure}

Identification of latch up events was achieved by monitoring the analog 
current draw of the ALPIDE chip and reacting soon thereafter with a prompt 
shutdown of the chip.  Four such latch up events are shown in 
Figure~\ref{fig:LatchUp}.  No pixels were lost with the established protocol 
and the latch up rate was well within the specs required for ALICE operation. 
 Because the MAPS sensors in sPHENIX would be placed at similar distance to 
the beam pipe, our latch up criteria would be looser than those required by 
ALICE.  Based partly on the successful completion of these tests, it is 
likely that the ALPIDE variant of MAPS sensor will be selected for production 
by ALICE.

\begin{figure}
\centering
\includegraphics[width=0.6\textwidth]{figs/LatchUp.PNG}
\caption{The detection of Latch Up events in the ALPIDE MAPS chip is 
identified by steps in the DC current draw.}
\label{fig:LatchUp}
\end{figure}



\subsection{TPC Tracker R\&D}
\subsubsection{Overview}

R\&D for the development of new gas-based detectors is very active around the 
world.  In addition to the aforementioned efforts on upgrades to ALICE and 
design of the TPC for the ILC, both ATLAS and CMS have taken new gas detector 
designs to an advanced stage.  The principle development has been the 
invention of the Micro Pattern Gas Detector or MPGD.  Both of the two basic 
MPGD technologies (GEM from Sauli and $\mu$MEGA from Giomataris) had 
successful applications in the COMPASS experiment.  The PHENIX HBD pioneered 
a very unique GEM application using evaporated coatings of CsI to create a 
large area gaseous UV photon detector.  ATLAS plans include large ($1 \times 
2~m^2$) $\mu$MEGA chambers and CMS plans to use large GEM chambers as part of 
their upgrade to higher luminosity running.

Several years ago, BNL began hosting the EIC R\&D program, which has had 
active participation from many institutions with diverse historical 
backgrounds.  Much of the EMCAL R\&D work discussed in later chapters was 
performed under the guise of the so-called "Calorimeter Consortium" from 
within the EIC R\&D program.  The calorimeter consortium features leading 
contributions from both STAR (UCLA) and PHENIX (BNL) institutions.  Even 
prior to the formation of the calorimeter consortium, the "Tracking and PID 
Consortium" (aka RD6) formed and also featured broad support from 
institutions with varied historical backgrounds:  Stony Brook (PHENIX), BNL 
(PHENIX), Yale (STAR), University of Virginia (JLab:~Hall~A), and Florida 
Institute of Technology (CMS).  All these institutions are also members of 
the CERN RD51 Collaboration which is dedicated to the development and 
advancement of MPGD technology.  Furthermore, our Yale collaborators have 
been central figures in the R\&D leading to the designs utilized for the 
ongoing ALICE TPC upgrade.

In this section, we will give an overview of the relevant R\&D that we have 
conducted to date and an outlook for how this research will continue toward 
the development of the TPC detector.  This discussion will necessarily be an 
incomplete summary of RD6, since that research has broader goals beyond TPC 
development (forward planar GEM trackers, GEM-based RICH, ...).

\subsubsection{Recent R\&D Efforts}

BNL and Yale maintain joint custody of an apparatus designed characterize gas 
properties as shown in Figure~\ref{fig:DriftApparatus}.  Electrons liberated 
by either a laser source (measured time $\rightarrow$ measured velocity) or 
$^{55}Fe$ photon conversion (known energy) follow the vertical drift tube and 
generate signals at the base (mostly via GEM-induced avalanche).  Some of the 
results from these measurements are summarized in 
Figure~\ref{fig:GasParameters}.  Most of these measurements focus on Ne-based 
mixtures as is appropriate for a TPC in heavy ion collisions where space 
charge is of a concern.  Among the many results, the $Ne-CH_4-CO_2$ mixture 
is especially intriguing since it not only exhibits a long plateau in drift 
velocity as a function of field, but it is among the gasses with the lowest 
probability of charge attachment.

\begin{figure}[htb!]
\centering
\includegraphics[width = 0.65\textwidth]{figs/DriftApparatus.PNG}
\caption{Apparatus used to measure electron drift velocities and electron 
attachment.}
\label{fig:DriftApparatus}
\end{figure} 

\begin{figure}[htb!]
\centering
\includegraphics[width = 0.8\textwidth]{figs/GasParameters.PNG}
\caption{Results for drift velocity and electron attachment in a variety of 
gasses.}
\label{fig:GasParameters}
\end{figure} 

These gas results and {\bf extensive} ANSIS field calculations lead to the 
design of the small prototype TPC shown in Figure~\ref{fig:SmallTPC}.  This 
TPC has a unique feature in that it combines the typical drift volume of all 
TPCs with a wire-mesh wall on the particle exit face of the cube.  Behind the 
wire mesh is a CsI GEMstack that will allow this "TPC/HBD" hybrid device to 
measure not only the trajectory (momentum), $\frac{dE}{dx}$ (hadron ID), but 
even the generation of cherenkov light (electron ID).  The result on the 
right shows that this device has excellent energy resolution 
($\frac{dE}{dx}$) since the width of the $^{55}Fe$ peak is not significantly 
broader than the statistical fluctuations intrinsic to the primary charge.  
This prototype is not only used to test possibilities for cherenkov light 
detection, but it is additionally used to evaluate the performance of various 
options for the readout plane segmentation.

\begin{figure}[htb!]
\centering
\includegraphics[width = 0.8\textwidth]{figs/SmallTPCSpectrum.PNG}
\caption{Photograph of the TPC/HBD prototype along with a measurement of the 
$^{55}Fe$ energy deposit spectrum.}
\label{fig:SmallTPC}
\end{figure} 

The TPC prototype is presently equipped with a readout pattern similar to 
that described in Figure~\ref{fig:chevron_pads}.  This shape of pads 
significantly improves the position resolution of the device at the cost of a 
small, but correctable differential non-linearity.  The result shown in the 
upper left corner of Figure~\ref{fig:TPCPositionResolution} shows the 
comparison between the reconstructed position of a precise ionization source 
and that reconstructed by charge sharing.  The regular structure of this 
correlation, with a period equal to the pad pitch is an indication of the 
level of differential non-linearity induced by the chevron design.  By 
quantifying the difference between the true position and that from the 
reconstruction as a function of the reconstructed position, we determine the 
correction that must be applied to data.  The result of applying this 
correction improves the measurement to a precision of $\sigma~=~99~\mu m$, 
exactly as needed for the final TPC device.

\begin{figure}[htb!]
\centering
\includegraphics[width = 0.8\textwidth]{figs/TPCPositionResolution.PNG}
\caption{Measurements of the differential non-linearity of chevron pad 
response and demonstration of the resolution achieved after correction.}
\label{fig:TPCPositionResolution}
\end{figure} 

The item of central and critical importance in designing the TPC will be 
handling the Ion Back Flow (IBF).  MPGD-based gain stages (along with 
streaming electronics), are the principle advance that make a TPC into a high 
collision rate device.  These techniques are new and require extensive R\&D.  
Although ALICE considers the R\&D phase of their TPC upgrade (up to the stage 
of technology choice) to be complete, our R\&D plan allows us time to improve 
further upon these plans and benefit from their experience.  
Figure~\ref{fig:IonBackFlowSetup} shows a schematic of the Yale apparatus 
used to perform the ion back flow measurements.  This particular figure shows 
the configuration used to measure the so-called "hybrid gain stage" solution 
that features a microMEGA plane preceded by a two stage GEMstack.  

\begin{figure}[htb!]
\centering
\includegraphics[width = 0.8\textwidth]{figs/IonBackFlowSetup.PNG}
\caption{Ion back flow measurements for the ALICE upgrade were made by the 
Yale group (collaborators through EIC R\&D) not only for the quad-GEM 
solution, but also for other alternative readout schemes.  This figure shows 
their apparatus as setup for measurements of a hybrid gain stage utilizing 
two GEMs followed by a microMEGA.}
\label{fig:IonBackFlowSetup}
\end{figure} 

\begin{figure}[htb!]
\centering
\includegraphics[width = 0.8\textwidth]{figs/IBF.jpg}
\caption{Resolution of the hybrid gain vs ion back flow percentage.}
\label{fig:IBF}
\end{figure} 

Figure~\ref{fig:IBF} shows the results for IBF as measured for the hybrid 
gain stage.  The vertical axis denotes energy resolution, by plotting the 
ratio of the sigma for the main $^{55}Fe$ peak over its mean.  This 
measurement is lower-limited to the natural width of just below 10\% driven 
by the statistics of the primary ionization.  The horizontal axis shows the 
level of ion back flow measured at a gain of 2000.  Nearly all gasses behave 
in the same manner tracing out a banana curve.  The reason for this shape is 
simple. In any multi-stage avalanche process, statistical fluctuations in the 
first stage of amplification generate the limiting resolution.  Therefore 
making the initial GEM gain as high as possible yields the best energy 
resolution.  That said, ions resulting from the first stage of avalanche are 
perfectly coupled back into the TPC volume.  The band is formed by smoothly 
changing the fraction of the gain carrier by the first GEM; high gain in the 
first GEM optimizes resolution, while low gain in the first GEM optimizes 
IBF.  It is worth noting that these results are, in fact, significantly 
better than the final design chosen by ALICE.  Because the running conditions 
might drive the balance between energy resolution and ion feedback to 
different optimal points ({\it e.g.} optimize $\frac{dE}{dx}$ resolution for 
EIC and optimize IBF for Au+AU collisions), it might be worthwhile to adopt a 
HV solution similar to ALICE wherein the voltage is supplied separately to 
each side of each GEM, allowing run-time decisions about the operating point 
along the banana curve.

\begin{figure}[htb!]
\centering
\includegraphics[width = 0.8\textwidth]{figs/HowardGrid.jpg}
\caption{Ion clearing times in the stacked grid scheme.}
\label{fig:HowardGrid}
\end{figure} 

Another intriguing option for dealing with IBF involves re-inventing the 
gating grid rather than eliminating it.  A typical gating grid involves a 
single wire or mesh plane whose voltage decides whether it it transparent to 
both electrons and ions or opaque to both.  Because of the very slow drift 
speed of the ions, the minimum time for the grid to be off after recording an 
event is often multiple hundreds of microseconds!  Howard Weiman has proposed 
a new grid style in which the grid has many layers and the ion clearing time 
is not set by the full depth of the grid, but rather by the distance to the 
nearest layer.  Using such a grid, one would leave it open for a significant 
time (until it filled with ions) and then briefly dump the ions prior to 
opening again.  Initial estimates of such a structure indicate that it could 
nearly entirely eliminate ion back flow at the cost of a "duty cycle" of 
80-90\%.  Using Garfield and ANSYS calculations, we have simulated designs 
for such a grid using both multiple wire planes and multiple mesh planes.  
These results are shown in Figure~\ref{fig:HowardGrid}.  Indeed, the ion 
clearing time is reduces to 10's of microseconds, but there is a long tail 
due to ions starting in low field areas within the grid structure.  The net 
result with the currently calculated grid design exhibits a long tail meaning 
tht more clever designs would be necessary to achieve an IBF of exactly zero, 
characteristic of old-style gating grids.

Although PHENIX collaborators who worked on the PHENIX HBD, have gained 
extensive experience with GEM avalanche devices, these GEMs were roughly the 
same size as a standard sheet of paper.  Challenges are certain as GEMstacks 
become larger.  Shown in Figure~\ref{fig:SoLID} is a picture of the GEM 
chamber that we took to test beam in October 2013.  At the time, this device 
(designed and built by our U. Va. colleagues) was among the largest GEM 
chambers yet built.  At that same test beam an equal sized chamber designed 
and built by our Fla. Tech. colleagues was tested as well.  Both chambers 
worked flawlessly with the U. Va. chamber producing a resolution below 80 
$\mu m$.  Note below, the new ideas from WIS, will be investigated for 
"tiling" a large area with self-supporting GEMstack structures.

\begin{figure}[htb!]
\centering
\includegraphics[width = 0.8\textwidth]{figs/SoLID.png}
\caption{The RD6 GEM detector shown here is among the largest GEM chambers 
yet built.  It was tested at Fermilab test beam and yielded a position 
resolution of better than 80 $\mu m$}
\label{fig:SoLID}
\end{figure} 


\subsubsection{Bench test results} 
\subsubsection{Test beam and future plans}
\subsubsection{TPC electronics R\&D plan}
Currently, we are following the R\&D plan for ALICE TPC electronics,
namely, prototype version 1, pre-production prototype and the
final production. The first version prototype should meet the schedule
for second prototype of TPC development with which we should do the
beam test with high multiplicity events.
\\


\section{Alternative technologies}
{\blue Ed does the introduction}\\
{\red for the reviewers, to show that we considered other choices as well}

\subsection{Other tracking alternatives that could be applicable} 
\subsection{Alternate pixel options} % Mike McC and Ming
\subsection{Alternate TPC readout plane options} %Sasha \& Tom
As discussed previously, we are currently investigating a list of possible 
alternate technologies for the readout plane.  These alternatives include 
both the possibility of changing a classic gating grid to implement a prompt 
flush for positive ions (possibly resulting in a TPC with zero ion back flow, 
and the cost of ``duty cycle'') and variations of the scheme for the 
MPGD-based gas amplification stage.  Already discussed is the ongoing work to 
implement a hybrid $\mu$MEGA/GEM detector that would benefit from the 
superior ion back flow characteristics of the $\mu$MEGA and achieve 
remarkable stability by lowering the $\mu$MEGA gain requirements via the 
assistance of the GEMstack.

A unique suggestion has been tested at WIS.  In this case, small 
self-supporting hexagonal GEMstacks were developed that could be used to 
populate any large surface.  These devices would feature the robust 
performance of smaller GEMS while still maintaining a nearly hermetic 
acceptance.

\subsection{Alternate electronics solutions} %Eric


Text, tables, and figure calls go here.
 
%Once you cite references, delete this text being used
%to avoid a BibTeX error~\cite{Adare:2010ux}.