Design and Performance of the DUNE 35-ton Prototype Time Projection - - PDF document

SLIDE 1

Design and Performance of the DUNE 35-ton Prototype Time Projection Chamber

The 35-ton Author List

Addresses for the 35-ton Author List

Abstract The DUNE 35-ton prototype time-projection chamber was designed to test the functionality of the components foreseen to be used in the far detector for the DUNE experiment. The Phase I run, completed in early 2014, demon- strated that liquid argon could be maintained at sufficient purity in a mem- brane cryostat. A time projection chamber was installed for the Phase II run, which collected data in February and March of 2016. The Phase II run was a test of the modular anode plane assemblies with wrapped wires, cold readout electronics, and integrated photon detection systems. While the details of the design for the DUNE far detector has since evolved, the 35-ton prototype is a demonstration of the functionality of the basic com-

• ponents. Measurements are performed using the Phase II data to extract

signal and noise characteristics, detector alignment and performance in the gaps between modules, as well as measurements of the electron lifetime and diffusion characteristics. Keywords: Prototype, Liquid Argon, Time Projection Chamber

• 1. Introduction

1

The 35-ton prototype was designed to test the performance of the con-

2

cepts and components to be used in the DUNE far detector. The DUNE

3

far detector is proposed to consist of 40 kTons (fiducial) of liquid argon in

4

four 10 kTon modules located at the 4850’ level of the Sanford Underground

5

Research Facility (SURF) in Lead, South Dakota. The start of installation

6

• f the first 10 kton module is scheduled to begin in 2021. The DUNE far

7

detector modules will be much larger than any previous liquid argon time

8

projection chamber (TPC), and the components must be shipped to the site,

9

Preprint submitted to Nuclear Instruments and Methods in Physics Research ADecember 7, 2016

SLIDE 2

lowered down the shaft, assembled in place, tested, and operated, all in a

10

cost-effective and time-efficient manner. These steps place constraints on

11

the design of the far detector, and compromises must be made in order to

12

satisfy these constraints. To meet the physics goals of DUNE, the perfor-

13

mance of the detector must satisfy basic requirements of spatial, time, and

14

energy resolution, signal-to-noise performance, detection efficiency and up-

15

• time. The design choices must be tested in a prototype before the far detector

16

design is finalized for the far detector and resources are committed. Section 2

17

describes the design of the 35-ton prototype and which design choices for the

18

far detector are tested. Because of the rapid evolution of the far detector de-

19

sign, the choices considered when the 35-ton prototype design was finalized

20

are no longer exactly those considered for the DUNE far detector, although

21

the broad features are the same. Section 2 describes these issues in detail.

22

The data acquisition system is described in Section 3, and the running

23

conditions are summarized in Section 4. Several analyses of the data from

24

the Phase II run of the 35-ton prototype are listed in Sections 6 through 12.

25

These comprise studies of the signal and noise performance of the system,

26

the relative alignment of the external counters and the TPC using cosmic-ray

27

tracks, the measurement of the relative time between the external counters

28

and the TPC using tracks that cross the anode-plane assembly (APA) vol-

29

umes, alignment and charge characteristic measurements using tracks that

30

cross between one APA’s drift volume to another’s, a measurement of the

31

electron lifetime, and studies of diffusion of drifting electrons. A summary

32

and outlook is given in Section 13.

33

• 2. Detector Design

34

The 35-ton prototype was designed by the LBNE collaboration and project

35

effort in order to test the design of the LBNE far detector, a massive liquid-

36

argon TPC to be located at SURF [? ]. In order to ship the APA’s from

37

their manufacturing site to SURF in standard high-cube shipping contain-

38

ers, lower them down the shaft at SURF and assemble them in place, they

39

are limited in size to 6m×2.5m. Amplifiers and digitizers are placed in the

40

cryostat in order to reduce thermal noise and simplify the cabling. It is more

41

cost-effective to place the readout electronics on the ends of the APA’s, and

42

thus the far detector has two layers of APA’s: one with electronics read out

43

• n the top and one on the bottom. As the total volume of liquid argon is

44

very large and the drift length should be limited to 3.6 m in order to combat

45

2

SLIDE 3

the effects of electron lifetime and diffusion, the APA’s are installed inside

46

the active volume and read out on both sides.

47

It is predicted that a steep angle of 45◦ between the wires of separate wire

48

planes optimizes the physics reach [? ], by providing a high degree of spatial

49

resolution for measuring the displaced vertex positions in π0 → γγ decay,

50

with subsequent showering of the photons some distance away from the π0

51

• vertex. This steep angle, coupled with the aspect ratio of the APA frames

52

and the need to read out the wires only on one edge of the APA in order to

53

reduce cable runs and make the electronics accessible, results in the design

54

choice that the angled wires are to be wrapped around from one side of the

55

APA frame to the other and back. The angled wires are the U- and V -plane

56

induction wires. There are two planes of collection wires, one on each side of

57

each APA, that do not wrap around from one side to the other.

58

The wrapped induction-plane wires induce an ambiguity in the inter-

59

pretation of the charge read out on them, as the charge could have been

60

deposited on any of the wire segments, and also at any position along the

61

wire segments. If the angles of the U and V planes were chosen to be ±45◦,

62

this ambiguity would not be possible to break in the data, even for single,

63

isolated hits, as the combinations of U, V , and Z wires cross in multiple

64

• places. In order to break this degeneracy, the angles were chosen to be 45.6◦

65

for the U plane and −44.3◦ for the V plane, and the collection wires were

66

chosen to be vertical.

67

An unistrumented grid wire plane is situated between the U plane and the

68

drift volume, and a grounded mesh is installed between the collection plane

69

and the argon volume inside the APA frame where the photon detectors lie.

70

The 35-ton prototype is designed to test the performance of a detector

71

with these choices. In order to fit inside the membrane cryostat of the Phase I

72

prototype, however, the APA’s and the drift volumes were shortened relative

73

to the far detector design. A drift region as long as possible to fit in the

74

35-ton cryostat was designed, while still having a shorter drift region on the

75

• ther side of the APA in order to test the double-sided readout functionality

76

• f the APA’s. The long drift length of the 35-ton prototype is 2.258 m from

77

the center of the APA to the cathode, while the short drift length is 0.302 m.

78

The APA’s are also smaller versions of the ones to be used in the DUNE

79

far detector. Two tall APA’s are mounted on the ends of the plane of APA’s.

80

They measure 2 m vertically by 0.5 m horizontally, and extend from the

81

bottom of the detector to the top. Their electronics are mounted on the top.

82

Two shorter APA’s are mounted between the two long ones, both 0.5 m wide.

83

3

SLIDE 4

The short APA on top is 1.2 m tall while the short APA on the bottom is

84

0.91 m tall. The electronics for the short APA on the bottom are mounted

85

• n its bottom edge. The layout of the APA’s is designed so that there are

86

horizontal and vertical seams between the APA’s as there are in the DUNE

87

far detector. The aspect ratio of the APA frames in the 35-ton prototype is

88

narrower than the 2.5 m×6.0 m DUNE far detector APA design; the 35-ton

89

APA frame dimensions were chosen so that they would fit in an acess hatch

90

• n the top of the cryostat. The narrower aspect ratio of the 35-ton APA

91

design requires the wrapped induction-plane wires to wrap more times than

92

they do in the DUNE far detector, for the same angle.

93

Preliminary Monte Carlo studies of the DUNE far detector showed a

94

degradation in pattern-recognition performance with wrapping ambiguities.

95

After the 35-ton prototype design was fixed, the DUNE far detector induction-

96

plane wire angle was reduced to 35.71◦ in order to limit the number of cross-

97

ings of any collection-plane wire with any induction-plane wire to one. The

98

small offsets in the U and V angles needed to break the ambiguities in the

99

±45◦ design are no longer needed. One impact of this choice is that all three

100

planes’ data are needed in order to break the ambiguity of any induction-

101

plane hit in the 35-ton data, while matched hits in the collection plane and

102

• ne induction plane are sufficient in the far detector.

103

One of the APA’s in the 35-ton prototype was built without the grounded

104

mesh in order to test its impact on the operations and measurements. In-

105

stalled in the vertical gap between the short middle APA and one of the

106

long APA’s is an electrostatic deflector, which is designed to control the elec-

107

tric field in this difficult-to-model region and make the charge collection on

108

the neighboring wires better understood. The bias on the deflector was not

109

studied in a systematic way however.

110

The 35-ton prototype detector is not in a test beam; cosmic rays provide

111

the particles required to understand its performance. In order to trigger on

112

cosmic rays that provide the most information about the detector, scintilla-

113

tor paddles were intstalled on the four vertical walls of the concrete structure

114

supporting the cryostat, as well as on the top, and a muon telescope consist-

115

ing of further scintillator paddles was installed xx m above the top of the

116

cryostat.

117

Four purity monitors were installed on a vertical support in the liquid

118

argon, outside of the TPC volume. They functioned by laser light illumi-

119

nating a cathode, which emitted electrons, which drift through a short drift

120

volume and were collected by an anode. Comparison of the integrated charge

121

4

SLIDE 5

collected in short pulses between that emitted by the cathode and collected

122

by the anode provided four measurements of the electron lifetime. Electrons

123

that attached to impurities drifted with much smaller velocities and are not

124

included in the short pulse charge integration.

125

• 3. Data Acquisition

126

The currents on the wires were amplified by cold preamplifiers and digi-

127

tized by 12-bit ADC’s, also in the cold volume. The preamplifiers had four

128

selectable shaping-time settings, ?? µs, ?? µs, ?? µs, and ??µs. The data

129

were collected with a shaping time of 3 µs. The rate at which the signals

130

was digitized was 2 MHz, in a continuous stream. The ADC pedestals were

131

configurable – approximately 500 ADC counts for collection-plane wires and

132

900 ADC counts for induction-plane wires. The digitized signals were sent to

133

Reconfigurable Computing Elements (RCE’s) which triggered, buffered and

134

formatted the data for analysis and storage. The RCE’s transferred their

135

data via Ethernet to commodity computers running artdaq [? ], a flexible

136

data-aquisition framework which provides hardware interfaces, event build-

137

ing, logging, and online monitoring functionality. Because the disk-writing

138

speed was limited to approximately 60 MB/s, the data were triggered at ap-

139

proximately 1 Hz from the 35-ton detector. A design including continuous,

140

untriggered readout was considered, but it required zero suppression and a

141

high degree of compression in order to be accommodated by the data rate

142

• limitations. Electronic noise in the detector and the small signals precluded

143

the use of zero suppression, and thus all ADC samples were recorded for

144

all triggered readouts. Data were written in a single output stream by an

145

artdaq aggregator process in ROOT format. No compression was applied, in

146

• rder for CPU not to be a bottleneck in the output data stream. The large

147

electronics noise reduced the effectiveness of compression to a factor of ≈ 2,

148

with a large CPU penalty.

149

Each triggered readout, called an “event”, consisted of 15000 ADC sam-

150

ples on each wire. The trigger decisions were formed by a configurable hard-

151

ware module [? ], which used coincidences of scintillator paddle hits as inputs

152

to the trigger logic. The trigger coincidences chosen for reading out events

153

were East-West coincidences, North-South coincidences, and muon telescope

154

• events. There were FIXME nnnn ADC samples saved before each trigger.

155

5

SLIDE 6

• 4. Running Conditions

156

The nominal drift field in the DUNE far detector design is 500 V/cm.

157

The data collected by the 35-ton Phase II prototype was taken at a field of

158

250 V/cm, however. This lower field magnifies the effects of the lifetime and

159

diffusion on the collected charge as a function of drift distance, and increases

160

the amount of charge that recombines with the argon ions in order to make

161

scintillation light while decreasing the signals on the TPC wires.

162

The electron lifetime measured by the purity monitors was stable at

163

around 3 ms for the duration of the data-taking period. Several short-lived

164

• perational issues, such as power outages and an exhausted supply of liq-

165

uid nitrogen, caused the electron lifetime to drop temporarily. Data used in

166

the analyses presented here are selected from only the high-electron-lifetime

167

running periods.

168

The electronic noise was higher than anticipated in the 35-ton data. In

169

the worst case, a very high amplitude oscillatory noise signal of size 200 ADC

170

counts per channel was seen throughout the detector, and corresponded to

171

a self-sustaining ”high-noise” state. The detector entered this state spon-

172

taneously, though only when the drift field was turned on and the anode

173

wire planes were biased. The high-noise state could be cleared by removing

174

power from the front-end boards, restoring power to them, and re-initializing

175

• them. It was found in the course of the run that switching off the front-end

176

boards of the shortest APA helped to prevent spontaneous triggers of the

177

high-noise state. Sec. 5 describes the characteristics of the data when not in

178

the high-noise state.

179

A number of wires were not read out for part or all of the run. FIXME-NN

180

wires were broken during APA fabrication and testing. The wires remained

181

mechanically secure but their electrical connections were severed during a

182

thermal test. They were jumpered to their neighbors in order to preserve the

183

electrostatic configuration of the APA’s. Some FIXME-number front-end

184

ASICs did not function after the detector cooled down and these channels

185

were not read out for the duration of the run. FIXME-number ASICs failed

186

after the repeated power-cycling needed to exit the high-noise state. A total

187

• f nn% of the TPC channels were not functioning at the end of the run.

188

Nonetheless, enough data were present in order to test the design choices

189

and meet the goals of the prototype.

190

6

SLIDE 7

Figure 1: The Fourier transform of the ADC values read out for all 2048 channels in the 35-ton prototype, for a low-noise run.

7

SLIDE 8

Figure 2: The Fourier transform of the ADC values read out for a single channel in the 35-ton prototype, for a low-noise run. Figure 3: Correlations in the ADC values between pairs of TPC channels in the 35-ton prototype, for a low-noise run.

8

SLIDE 9

• 5. Raw Data Characteristics

191

When not in the high-noise state, the RMS of the ADC’s was approx-

192

imately 30 counts on each channel. A frequency spectrum of this noise is

193

shown in Figure 1. TODO: Noise vs. wire length plot? The noise consists of

194

correlated and uncorrelated components which changed over time. Figure 3

195

shows the correlation of the raw ADC values from one channel to all others

196

in the 35-ton detector readout. Clear correlations are seen in blocks of NN

197

(64? 32?) consecutive channels. The noise has been ascribed to a voltage

198

regulator on the front-end board, which is shared by MM front-end ASICs.

199

This characteristic of the noise is used in the coherent nosie subtraction step,

200

described in Section 6.

201

Data are also affected by bit-level corruption. In a fraction of ADC sam-

202

ples, which depends on the temperature, the channel, and the input current,

203

the low six bits of the ADC can be erroneously reported as 0x0 or 0x3f. If

204

the low six bits are erroneously 0x0, then number represented by the upper

205

six bits is one greater than it would be if the bits had not stuck, and when

206

the low six bits are erroneously 0x3F, then the number represented by the

207

upper six bits is one less than if the bits had not stuck. The probability

208

that the low six bits will stick depends strongly on the proximity of the true

209

input value value to the boundary in which the result would be 0x0 if the

210

bits had not stuck. These fractions of ADC samples vary from 20% to 80%,

211

depending on the factors mentioned above.

212

Furthermore, a fraction of channels exhibit more classic stuck-bit issues,

213

in which a particular bit is never set (or never clear). When the least-

214

siginficant six bits are stuck, it is referred to as a ”stuck code”, and when a

215

single bit is stuck, it is a ”stuck bit”.

216

• 6. Data Processing

217

Before results can be obtained from the data, the noise must be filtered,

218

the stuck codes mitigated, and the electronics and detector response decon-

219

• voluted. The first step is to flag samples as underflow, overflow, stuck (0x0

220

• r 0x3f), or valid, so that subsequent processing steps have access to the raw

221

bit pattern. The pedestal, measured separately and stored in a database, is

222

first subtracted.

223

The first stage of noise filtering addresses the correlated noise induced

224

by the front-end regulators. On each time sample (”tick”), the ADC values

225

9

SLIDE 10

• f neighboring groups of 64 channels (including the channel in question),

226

• mitting those that have stuck-code candidates, are collected, the pedestals

227

subtracted, and the median is found. This median value is then subtracted

228

from the channel’s ADC value.

229

TODO – performance of noise filtering.

230

The next step is the mitigation of stuck codes. If a channel’s raw bit

231

pattern on a tick indicates that the low six bits could be stuck, then the

232

neighboring time samples are examined. If there are samples before and after

233

the sample investigated that are not stuck, and these samples are no more

234

than five ticks away from the sample investigated, then a linear interpolation

235

in ADC vs. tick is performed to fill in the stuck ADC values. If the closest

236

tick without a stuck code candidate is more than five ticks away, the sample is

237

not corrected, and the sample is flagged as bad. If the end of the waveform is

238

encountered before the five-tick condition is met, then a linear extrapolation

239

is performed.

240

The next step in signal processing is filtering and deconvolution, which are

241

collected in one step. For each channel, an FFT is performed on the raw ADC

242

values as a function of time to obtain a frequency-domain respresentation

243

• f the data, which is then multiplied by a the product of a deconvolution

244

kernel and noise filter. The deconvolution kernel is defined separately for

245

induction-plane signals and collection-plane signals, and is the reciprocal

246

• f the FFT of the simulated response of the detector and electronics to a

247

single impulse of charge arriving in a very short time. Poles in the kernel

248

are set to a maximum value so as not to emphasize noise that coincides

249

with a zero in the detector response. The noise filters, one for induction-

250

plane channels and one for collection-plane channels, were constructed from

251

representative waveforms containing visually identifiable signals from tracks

252

traveling roughly perpendicular to the drift field. Portions of the waveforms

253

corresponding to identifiable hits were removed and the spectrum of the

254

remaining waveform was calculated to estimate the noise-only spectrum, and

255

the regions in time near the hits were used to calculate the spectrum of the

256

• signal. The noise filter is then a Wiener function s/(s + b) as a function

257

• f frequency. Generally, frequencies between 20 and 120 KHz were retained

258

while others were filtered out.

259

10

SLIDE 11

Table 1: Signal, noise, and signal-to-noise ratio in the 35-ton prototype for minimum- ionizing particles traveling perpendicular to the electric field near the anode planes, sepa- ratley listed for the collection (C), and the U and V induction planes.

Before Noise Filter After Noise Filter TPC Plane Signal Noise S/N Signal Noise S/N 48.4 28.9 1.6 54.6 14.2 3.9 TPC0 1 46.7 22.0 2.1 36.4 10.3 3.5 2 117.6 19.3 6.1 88.4 7.5 11.9 29.1 29.1 1.0 29.0 12.7 2.3 TPC1 1 58.7 26.1 2.3 55.6 10.7 5.2 2 98.0 19.2 5.1 84.0 9.6 8.8 42.0 19.8 2.1 38.1 9.5 4.0 TPC4 1 2 117.3 13.2 8.9 97.9 4.1 23.7 25.4 20.3 1.3 24.2 8.9 2.7 TPC5 1 44.7 18.0 2.5 46.8 7.9 5.9 2 112.4 14.5 7.8 96.9 5.3 18.4 51.9 29.3 1.8 49.6 11.6 4.3 TPC6 1 35.8 26.1 1.4 37.3 10.2 3.6 2 111.8 20.9 5.4 78.1 9.5 8.2 26.0 32.3 0.8 23.9 13.2 1.8 TPC7 1 49.0 28.9 1.7 50.9 9.6 5.3 2 106.4 31.5 3.4 91.5 13.2 7.0 11

SLIDE 12

• 7. Hit and Track Finding

260

• 8. Alignment of counters and TPC

261

8.1. North-South Muon Triggers

262

8.2. East-West Muon Triggers

263

• Why align?

264

• Describe two methods

265

• Describe differences

266 267

• 9. T0 Measurement from Tracks Crossing the Anode Planes

268

A collected dataset unique to the 35-ton consists of tracks which pass

269

from one drift region to the other, thereby passing through the APA planes.

270

The 35-ton is the only planned experiment in the LAr program with APAs

271

which read out TPC information for multiple drift regions simultaneously

272

before the DUNE far detector, so analysis of these events is important.

273

Since these tracks cross the planes, it is possible to measure the interaction

274

T0 by ensuring the two track segments are aligned across the anode planes.

275

An incorrect T0 would introduce a common timing offset and have the effect

276

• f moving the tracks in each drift region closer to or further from the APAs.

277

This can then be compared to the T0 measured by the external counters and

278

provide a calibration for inter-detector components.

279

TODO: Am I allowed a figure here showing, for example, an EVD with

280

an incorrect T0?

281

Tracks with a shallow APA-crossing angle are selected, to ensure sufficient

282

hits in each drift region, and the crossing time, T counter , is provided by the

283

• counters. Only collection plane hits are used in the analysis and are selected

284

by considering the ‘counter shadow’, hits whose position implies they lie

285

in the section of the detector between the two counters through which the

286

triggering particle passed, and disregarding hits on noisy wires. These hit

287

positions are determined, correcting for T counter , and a linear least square

288

residual fit is applied to the track segment in each drift region separately.

289

The measurement of the interaction time implied by this TPC information,

290

T TPC , is determined by aligning these two tracks.

291

12

SLIDE 13

Two complimentary approaches were considered in order to quantify the

292

alignment between the track segments in each TPC drift volume. Each in-

293

volved a different metric:

294

• the chi-squared value determined when considering all hits across both

295

regions with the best fit line through all hits;

296

• the separation between the lines from each region at x = 0, the centre

297

• f the APA.

298

T TPC is provided by separately minimising either of these metrics. The agree-

299

ment between each was found to be good with the latter method providing

300

slightly more consistent results.

301

The method was validated on simulation and used on data to determine

302

the relation between T counter and T TPC . This is shown in Fig. 4.

303

(a) 35-ton simulation (b) 35-ton data

Figure 4: Difference between interaction time T0 measured by the external counter system and by the TPC data. See the text for the described method. [NB Data figure is a placeholder – I had more stats at one point, I’ll work out where they went...]

From Fig. 4b a systematic offset can be seen between T counter and T TPC .

304

This appears to be due to a miscalibration of 64 ticks (= 32µs) between

305

the detector subcomponents and was unknown before observed in the TPC

306

data. This represents the utility of the method demonstrated to provide

307

precise calibration of the integrated system.

308

Further studies involving the APA crossing tracks involved studying the

309

distributions of the readout time of each hit associated with the crossing

310

• track. It was noted there is a sharp peak in this distribution corresponding

311

13

SLIDE 14

to the interaction time of the event; furthermore this peak was not present

312

for hit on tracks which cross the short centre APA. This APA is the only one

313

without a grounded mesh at its centre so it appears these hits populating

314

the peak at the interaction time partly come from charge drifting ‘backwards’

315

(towards the collection plane) after the particle had passed through due to

316

the field created by the grounded mesh and positively biased collection plane.

317

See Fig. ?? [fig not here yet– want a picture of an EVD with these check

318

marks present]. It is noted that there is no difference between the distribution

319

• f hit times between hits on wires next to the frame of the APA and those

320

towards the middle of its face in the case of mesh; this is the purpose of such

321

a design choice and so has been validated as working as intended.

322

• 10. Z-Gap Crossing Tracks

323

• 11. Electron Lifetime Measurement

324

While the electron lifetime was measured using dedicated purity monitors [REF], the lifetime was not measured inside the TPC field cage using this

• method. So, a measurement of lifetime was made directly from the drifted

ionized charge from cosmic muons in the detector. Such a measurement is im- portant in quantifying the overall concentration of electronegative impurities in the argon, which can cause a significant reduction in the charge collected, especially at long (1m) drift distances. FIXME-possiblywrongnumbers For example, a reduction of 20% in charge collected, versus total charge ion- ized in an interaction, for a drift of 2m is observed with a lifetime of 3ms. These values, in actuality, motivate the DUNE far-detector signal-to-noise re-

• quirements. Quantitatively, the electron lifetime is defined by the following

relationship Qcoll = Q0e−t/τ, where Qcoll is the total collected charge, Q0 is the original charge before drift, t is the drift time of the charge packet, and τ is the electron lifetime. It is related to the concentration of electronegative impurities by the empirical relationship τ =

• i

kini −1 where the ni are the number concentrations of the different species, and ki

325

are constants.

326

14

SLIDE 15

In practice, the electron lifetime is determined by plotting dQ

dx at the anode

327

versus drift distance. The charge, Q, is found by integrating the digitized

328

waveform over some time interval. A correction must be made to the charge

329

to account for the fact that the particle tracks may traverse multiple wires

330

at varying angles. So, dividing the collected charge on a particular wire by

331

the length of the track which lies in the shadow of each wire. The factor

332

γ = p/ cos θ, where p is the track pitch and θ is the angle between the track

333

and the vector perpendicular to the wire and in the plane of the wires, is

334

used to normalize the charge for varying track angle.

335

Due to the specific electronic noise conditions of the 35-ton prototype,

336

it was decided that a better measurement of the charge would be possible

337

by integrating the raw data waveforms, rather than the noise-subtracted and

338

deconvolved signal. Further pre-processing of data has the unwanted effect of

339

introducing more extensive systematic uncertainties. So, charge “hits” were

340

found

341

• 12. Electron Diffusion Studies

342

• 13. Summary

343

The 35-ton prototype successfully demonstrated in Phase I that liquid

344

argon of sufficient purity could be maintained in a membrane cryostat with

345

adequate filtering and circulation. Phase II confirmed that this is also the

346

case when a time-projection chamber and associated electronics and cabling

347

were installed. While the far detector design evolved after the 35-ton design

348

was finalized, and while the noise characteristics of the 35-ton prototype

349

made analyses challenging, a number of analyses were possible that enabled

350

testing of the design ideas of the far detector unique to DUNE.

351

15