1. Simon Van Gorp 1/x Thesis defense 28 th of February, 2011 Search for physics beyond the standard electroweak model with the WITCH experiment Simon Van Gorp 28 th of February 2011, Leuven Promotor: Prof. Dr. Nathal Severijns

2. Outline Simon Van Gorp Thesis defense28 th of February, 2011 2/x WITCH – Motivation – Overview – Status 2007 Simbuca – Graphics card – Buffer gas routines – An example June 2011 experiment – Data set – Reconstruction of the data – Simulations – results Nonneutral plasmas – Boundary with one-particle regime – Penning trap excitations One species Multiple species Conclusion

3. Physics motivation Simon Van Gorp Thesis defense28 th of February, 2011 3/x EXP [1]: |C S /C V | < 0.07 |C T /C A | < 0.09 =>Search for scalar (or Tensor) Interactions Low energy (couple 100 eV)!  Need for scattering free source [1]: Severijns, N., Beck, M., & Naviliat-Cuncic, O. (2006).Rev. Mod. Phys., 78(3), 991.

4. Overview Simon Van Gorp Thesis defense28 th of February, 2011 4/x mm

5. Experimental setup Simon Van Gorp Thesis defense28 th of February, 2011 5/x Penning traps – Preparation trap Helium buffer gas (10 -3 - 10 -4 mbar) Possible excitations – Decay trap scattering-free source Energy determination with retardation spectrometer – Conversion of radial in axial energy

6. Time situation of the PhD Simon Van Gorp Thesis defense28 th of February, 2011 6/x October 2007 – 35 Cl contamination (ratio 25:1) – Charge exchange in REXTRAP (  1/2 =70 ms) and WITCH (  1/2 =8 ms) – Unwanted ionization effects (sudden discharges) => Upgrade campaign to tackle those issues (WITCH 2.0) November 2009 – Still small ionization that was not noticed before was solved by installation of a wire Not covered in my thesis but in PhD thesis of Michael Tandecki Our goal was in sight – Measure a – Prepare the tools for such the analysis of a

7. Simbuca Simon Van Gorp Thesis defense28 th of February, 2011 7/x 10 4 – 10 6 ions / trap cycle stored up to a few seconds in the decay trap. Simulation time scales with O(N 2 ) – Tree codes O(N log(N)) – Scaled Coulomb approach Novel approach by using the GPU instead of conventional CPU: Simbuca code – Complete simulations package – Different buffer gas routines and integrators – Importing realistic field maps

8. Integrators and buffer gas models Integrators: 4 th and 5 th order Runga Kutta with adaptive step size and error control. 1 st order (predictor corrector) Gear method. Buffer gas models: Langevin or polarizability model (= for all mases) Ion Mobility based model ( ≈ for all mases) HS1 SIMION model 8/18 Simon Van Gorp – MPI Heidelberg – 14.02.2012

9. Why a GPU? GPU -high parallelism -very fast floating point calculations -SIMD structure (pipelining!) Stream processor ≈ CPU = Comparable with a factory assembly line with threads being the workers Geforce 8800 GTX Simon Van Gorp Thesis defense 28 th of February, 2011 9/x

10. Chamomile scheme Calculating gravitational interactions on a Graphics Card via the Chamomile scheme from Hamada and Iitaka (in 2007). Why a GPU? -parallelism! -only 20 float operations -CUDA programming language for GPU’s i-particles piece available for each ‘assembly line’ j-particles piece presents itself sequentially to each line force is the output of each line [7]: T. Hamada and T. Iitaka, arXiv.org:astro-ph/0703100, 2007 Simon Van Gorp Thesis defense 28 th of February, 2011 10/x

11. Chamomile scheme: practical usage Function provided by Hamada and Iitaka: Gravitational force ≈ Coulomb Force Conversion coefficient: Needed: - 64 bit linux - NVIDIA Graphics Card that supports CUDA - CUDA environment v2.3 - 4.0 Not needed: -CUDA knowledge -… Simon Van Gorp Thesis defense 28 th of February, 2011 11/x

12. GPU vs CPU GPU blows the CPU away. The effect becomes more visible with even more particles simulated. Simulated is a quadrupole excitation for 100 ms with buffer gas. This takes 3 days with a GPU compared to 3-4 years with a CPU! GPU improvement factorCPU and GPU simulation time Simon Van Gorp Thesis defense 28 th of February, 2011 12/x

13. Simbuca: outlook and future WITCH Behavior of large ion clouds Mass separation of ions Smiletrap (Stockholm) Highly charged ions Cooling processes ISOLTRAP (CERN) In-trap decay Determine and understand the mass selectivity in a Penning trap ISOLTRAP(Greifswald) isobaric buncher, mass separation and negative mass effect CLIC (CERN) Simulate bunches of the beam Piperade (Orsay and MPI Heidelberg) Simulate mass separation of ion species Simon Van Gorp - Scientific meeting - 16.02.2011 13/21

14. Simon Van Gorp – TCP Saariselkä- 14.04.2010 14/24

15. Data analysis: 3 (or 4) steps Simon Van Gorp – WITCH collaboration meeting – 21 February 2012 15/13 1. reconstruct the experimentally obtained spectrum from the data 2. Simulate the experimentally obtained spectrum, taking into account the experimental conditions (3.) verify your simulations with experimental observations The observed beam spot The energy distribution of the ions in the trap Ratio  `s/ions from the PhD 4. Fit the two spectra to extract the  angular correlation coefficient a

16. Experimental conditions June 2011 Simon Van Gorp – WITCH collaboration meeting – 21 February 2012 16/13 ISOLDE target broke few days before the actual run. Replaced with used target. => low 35 Ar yield (5.10 5 compared to 2.10 7 in yieldbook) HV electrode could not be operated as intended. Not-optimal focus of the electrodes caused a loss off 40% Losses in the decay-trap -> A low statistics experiment (~2600 ions/trapload). losses in the decay-trap due to non-optimized voltages and timings. The red curve (better settings) shows a more constant behavior

17. Proof of recoil ions Simon Van Gorp – WITCH collaboration meeting – 21 February 2012 17/13 - Guassian bell shape indicates the observation of recoil ions - Position distribution shows the presence of recoil ions and missing counts along the Y- axis.

18. measurements Simon Van Gorp – WITCH collaboration meeting – 21 February 2012 18/13 500 ms cooling in the cooler-trap. Afterwards capture in the decay-trap. Measurement with and without retardation voltages. Reconstruction via: - Subtraction - Regression analysis - Overshoot peak - Fitting the data

19. Normalization (1) : subtraction Simon Van Gorp – WITCH collaboration meeting – 21 February 2012 19/13 Difference of measurements with and without retardation voltage applied. (normalized via regression analysis). Correct the data for 35 Ar half-life and losses in the decay-trap. Scale factor f equals 3.540(3)

20. (less good) normalizations (2,3) Simon Van Gorp – WITCH collaboration meeting – 21 February 2012 20/13 Data set 2: normalization on the overshoot peak Data set 3: normalization via a fit function of the data

21. Simulations: Simon Van Gorp – WITCH collaboration meeting – 21 February 2012 21/13 Compare obtained spectra with simulated spectra. Therefore: 1. Simbuca simulates the ion-cloud in the decay-trap. 2. Ion-cloud parameters are fed to a MC simulation program (SimWITCH). Comsol multiphysics program is used to extract electric fieldmaps given the electrode voltages Magnetic fieldmaps from the magnet manufacturer Buffergas collisions and excitations are handled by Simbuca

22. Simulations: Simbuca Simon Van Gorp – WITCH collaboration meeting – 21 February 2012 22/13 Due to limited time the traps were not properly optimized: Transfer time was not set ideally 32.5 us instead of 38.5 us. -mean energy of 4.5 eV (instead of 0.2 eV) -ions positions in the decay-trap is 15 mm lower than the center

23. Simulations: SimWITCH (1) Simon Van Gorp – WITCH collaboration meeting – 21 February 2012 23/13 Simulations for - All retardation voltages (0V, 150V, 250V, 350V, 600V) - All charge states (1 +,2 +,3 +,4 +,5 + ) 1 + : 75(1)% 2 + : 17.3(4)% 3 + : 5.7(2)% 4 + : 1.7(2)% 5 + : < 1 % Including the charge state distribution (as measured with LPC trap) we can extract %ions reaching the MCP depending on the retardation step and a -> Fit the data with a linear combination of a=1 and a=-1 to obtain the final result for the beta-neutrino angular correlation factor a.

24. Simulations: SimWITCH (2) Simon Van Gorp – WITCH collaboration meeting – 21 February 2012 24/13 Ions are not properly focused on the MCP, due to the lower HV settings applied. The applied voltages are not high enough to pull the ions of the magnetic field lines. - Ions are lost on SPDRIF01 electrode. - The higher the charge-state of the daughter ion the better the focus. Input spectra 1+1+ 2+2+

25. Extracting a Simon Van Gorp – WITCH collaboration meeting – 21 February 2012 25/13 The preliminary result from the analysis yields a = 1.12 (33) stat c 2 / n = 0.64 SM value of a =0.09004(16). Not including actual experimental conditions yields a = 2.62 (42) !! => This stresses the importance of simulations!! a=-1 a=1

26. Conclusion and outlook Simon Van Gorp – WITCH collaboration meeting – 21 February 2012 26/13 Conclusion: - Seems to have solved unwanted ionization - Magnetic shield and RFQ allow much more testing time. - First determination of a on the decay of 35 Ar with the WITCH experiment. Outlook: - Experiment in October already increased the available statistics and solved the losses in the decay trap and in the spectrometer. - Count rate can be improved by: 10 (ISOLDE) * 50 (measurement time) * 2 (measurement cycle) * 2 (focussing electrode efficiency) * 4 (tuning in the B-field) = 8000 times more statistics -> sqrt(8000)=90 meaning that it is possible to reduce the statistical error to 0.5 %

27. Simon Van Gorp Thesis defense28 th of February, 2011 27/x

28. Non-neutral plasmas: an outlook for WITCH Simon Van Gorp Thesis defense28 th of February, 2011 28/x When trapping a large amount of ions, the cloud`s own electric field will create an E x B drift force for the ions with Good agreement between calculated and fit value (factor 2). Indications that around 10 4 ions the ion motion behaves like a nonneutral plasma

29. Boundary single particle & nonneutral plasma regime Simon Van Gorp Thesis defense28 th of February, 2011 29/x When storing around 5000 and 20000 ions the ions behave like a nonneutral plasma (in good comparison with [x]) - Energy broadening due to Coulomb repulsion - Resistance to excitations due to electric field of the ion cloud [x]: Nikolaev et al. (2007). RCM, 21(22), 3527–3546 Single ions regime: Nonneutral plasma regime:

30. Single ion species trapped Simon Van Gorp Thesis defense28 th of February, 2011 30/x Plot centered 133 Cs ions vs. duration of the quadrupole excitation Losses due to Coulomb effects Resonant excitation frequency tends to be more positive (as in Ref. [x]) [x]: F. Ames et al. (2005). NIM A, 538, 17–32

31. Multiple ion species trapped Simon Van Gorp Thesis defense28 th of February, 2011 31/x When multiple ion species are trapped a more negative frequency is favored [x] Seems to depend on the N (not on n) There is a large resistance to the applied excitation due to shielding of E cloud No C C ratio 25% to 10% Nx2 [x]: Herlert, A., et al. (2011). Hyperfine Interactions, 199, 211–220. 10.1007/s10751-011-0316-6.

32. Conclusion and Outlook Simon Van Gorp Thesis defense28 th of February, 2011 32/x Conclusion – A versatile Penning trap simulation package is the first application that uses a GPU to calculate the Coulomb interaction between ions in the Penning trap. – First analysis and determination of a on the decay of 35 Ar with the WITCH experiment Outlook – Simbuca will continue to be used by WITCH and other experiments. – Mass purification in Penning traps is a new field that is gaining interest – Investigate the properties of the non-neutral plasma in the WITCH Penning traps – New phase for WITCH, i.e. extensive investigation of systematic effects

33. Conclusion and outlook Simon Van Gorp – WITCH collaboration meeting – 21 February 2012 33/13 Conclusion: - Seems to have solved unwanted ionization - Magnetic shield and RFQ allow much more testing time. - First determination of a on the decay of 35 Ar with the WITCH experiment. Outlook: - Experiment in October already increased the available statistics and solved the losses in the decay trap and in the spectrometer. - Count rate can be improved by: 10 (ISOLDE) * 50 (measurement time) * 2 (measurement cycle) * 2 (focussing electrode efficiency) * 4 (tuning in the B-field) = 8000 times more statistics -> sqrt(8000)=90 meaning that it is possible to reduce the statistical error to 0.5 %