MANUAL DE PROCEDIMIENTOS DE SEGURIDAD 1 LEVANTAMIENTO DE PROCESOS.

For strong wall relaxation one expects, that the difference between the pure and the mixed states gets less pronounced. This is visible in fig. 6.13, where the difference almost vanishes. In this case, the determination of τse is of low precision and the uncertainty

reaches values of 20 % and more (see tab. 6.1). A strong difference between mixed and pure states in case of low wall relaxation is shown in fig. 6.12, which was performed with a new storage cell at high temperature (measurement No. 6). The measured magnetic field dependence of the hyperfine populations in this case is close to the theoretical behavior of pure spin exchange interaction as shown in fig. D.4 and the spin exchange relaxation time τse is obtained by the relaxation model with high precision. If the spin exchange

6.8 The Magnetic Field Dependence of Hydrogen Spin Relaxation 93 No. εw13 [%] εs24 [%] σse ηse n0 φT GA ρBRPa 1 98.1_±0.2 97.8_±0.35 23.4 3.09_±0.05 3.71_±0.06 54.0_±0.9 0.9_±0.02 2 98.1_±0.13 98.0_±0.17 23.4 3.07_±0.04 3.70_±0.048 49.9_±0.2 0.97_±0.014 3 97.9_±0.12 97.8_±0.15 23.2 3.14_±0.04 3.81_±0.048 50.4_±0.12 0.99_±0.014 4 98.9_±0.2 99.3_±0.28 22.95 2.63_±0.1 3.22_±0.123 49.8_±0.25 0.85_±0.034 5 97.6±0.13 97.5±0.16 23.75 1.5±0.03 1.77±0.036 45.1±0.35 0.52±0.012 6 98.3±0.12 98.1±0.15 23.75 2.82±0.04 3.34±0.048 51.8±0.31 0.85±0.013 7 97.7±0.13 97.6±0.16 23.4 3.24±0.06 3.90±0.072 50.6±0.17 1.01±0.018 Tab. 6.2: Results of the hydrogen relaxation measurements: The ABS transition efficiencies and spin exchange parameters. More results of the same measurements are listed in tab. 6.1. The value of σse is taken from Allison [All 72] in units 10−16cm2 and ηse is defined as τd/τse, normalized to a storage cell temperature of 100K:ηse=τd/τse

p

T /100K . The atomic density in the storage cell centern0 (normalized to a storage cell temperature of 100K) is given in units

of 1012atoms cm−3. The error bars are of statistical nature.φT GAis the total flux, measured with the TGA in kHz/mA, which is proportional to the ABS intensity. ρBRP_a was obtained by the atomic densityn0, normalized to a flux ofφT GA= 50kHz/mAand scaled such, thatρBRPa →1 forαT GA→1. The scaling factor is 4.13·1012atoms cm−3at a TGA flux ofφT GA= 50kHz/mA corresponding to 8.25·1011atoms at φT GA = 1kHz/mA. The details of the calculation are explained in sec. 6.8.2.

In order to extract the central atomic density by the spin exchange effect, one may use eq. 6.9: The correlation factor ρc was introduced in sec. 6.3 and is defined by eq. C.105

in app. C.6.2. For the atomic sample of the BRP, the molecular flow simulation resulted

ρc = 1.093±0.015. One obtains for the average density of hydrogen atoms:

hn_i= τd τse 1 √ 2ρcLdσse . (6.68)

In case of low recombination (αT GA≥0.97), one can assume, that for the average atomic

density_hn_i= 1₂n0 holds and the integrated areal proton density129Dtortarget thickness

is then given by:

Dt= 2L hni , (6.69)

where L is the length of one wing of the storage cell (20cm). The results for τd/τse and n0 including the statistical uncertainty are listed in tab. 6.1 and tab. 6.2.

Even though there is no strong reason for this relations, the measurement of fig. 6.3 shows, that the polarization loss by spin exchange is proportional to αT GA

r . The mea-

surement of fig. 4.3 shows under the same conditions, that αT GA

r ∝ ρBRPa , so that it is

reasonable to assume, that the normalized atomic density ρBRP

a is proportional to the

atomic density in the storage cell center n0 as measured by the spin exchange effect. A measured value of n0, that was obtained from a measurement, for which αT GAr is

close to one (αT GA

r →1 ⇒ρBRPa → 1), was used to estimate the normalization of ρBRPa .

The values of ρBRP

a obtained by the estimated normalization are listed in tab. 6.2 are

compared to the values ofαT GA

r of tab. 6.1 in fig. 6.15. The new storage cell (measurements

no. 6 and 7) is in agreement with the Monte Carlo (MC) predictions for a homogeneous 129_{More precisely: The areal density of protons in atoms.}

0.4 0.5 0.6 0.7 0.8 0.9 1 0.4 0.5 0.6 0.7 0.8 0.9 1 α_rTGA ρa BRP 1 23 4 5 6 7

Fig. 6.15:Comparison between the density results of the relaxation measurements and the measured ato- mic fraction for the measurements No. 1 to 7 from tab. 6.1 and tab. 6.2. The number of the measure- ment is printed close to the corresponding data point. The solid line is the diagonal, whereαT GAr =ρBRPa .

The dotted lines are the mathematical limits and the dashed line is the a prediction calculated with the re- sults of a molecular flow simulation for a completely homogeneous storage cell surface. The circles repre- sent measurements with the first storage cell in 1997, the squares two measurements shortly after the in- stallation of a new storage cell.

storage cell, while measurement 5, taken after a several weeks of target operation in the HERA ring disagrees with this prediction. The fact, that this data point is located on the diagonal, indicates that most recombination occurs in the beam tube, while recombination in sample- and extension tube - which would have a stronger influence on the TGA sample than on the BRP sample - plays a minor role. As already reasoned above, measurement no. 4 can be ignored. The results are in agreement with the ones shown in fig. 4.4. They indicate again, that - now measured at high temperatures and with a completely different technique - the recombination probability of the sample tube surfaceγST

r is much smaller

than for the beam tubeγBT

r (γrST γrBT) in case of the first storage cell of the 1997 data

taking period, and γ_rST _'γ_rBT for the second.

Besides the analysis of the sampling properties, the absolute value of the derived atomic density can be compared to the target density measured by the HERMES luminosity monitor. The expected totaltarget density at a fluxφT GA= 50.4kHz (3rd measurement)

is _hn_i = 1.855 _·1012cm−3 (see sec. 3.7). The spin relaxation measurements yields an averageatomicdensity of_hn_ia= 1.905_·1012cm−3as listed in tab. 6.2. As the measurement of the luminosity monitor includes also molecular contributions of ballistic flow and rest gas (see sec. 3.5.4), this value has to be multiplied byα0 '0.96, if one wants to compare it to theatomicdensity obtained from spin exchange relaxation. The comparison yields with this correction a relative difference of about 7 %. As the calibration uncertainty of the Lumi is about 6 %, both results are in agreement. However, it can not be used for a significantly improved measurement of σse. But it should be mentioned, that the statistical precision

of the ratio τd/τse is 1.5 % only. One can assume, that the systematical uncertainty of the

BRP measurement is - in case of low recombination - of about the same size. The value of σse could therefore be measured to a high precision with the setup of the HERMES

target - provided, an independent density measurement of high precision is available. It is worth mentioning, that measured atomic target density corresponds to a target thickness of Dt= 7.42·1013nucl cm−2. As the conductance of the storage cell for atomic

6.9 Measurement and Analysis of the Transition Spectra 95

hydrogen is 17380cm3_s−1 _{at a storage cell temperature of 100K (see tab. F.2), the} calculation yields an injected flux of

ΦABS = 2 _hn_i Ctot = 6.62·1016s−1, (6.70)

which is in good agreement to a previous compression tube measurement of 6.4_·1016_s−1 [Stw 97].

6.9 Measurement and Analysis of the Transition Spectra

It will be shown in this section, that the master equation (eq. 6.1) can be used to cal- culate the individual transition probabilitiesWab of each relaxational transition from the

measured hyperfine population numbers - if the following assumptions are correct: • Possible effect of super-paramagnetic clusters can be neglected at higher tempera-

tures. The surface is unpolarized.

• The relaxation is not dominated by the interaction between adsorbed atoms, but by the interaction of the adsorbed atoms with the wall. This can assumed to be true, if the surface coverage is low.

• Wb→a is independent from the injected hyperfine population.

• The transition probabilities are symmetric: Wb→a =Wa→b =Wab.

• Correlated multi photon transitions can be neglected.

If these assumptions are - under the typical HERMES conditions - true, then it is possible to use the master equations (eq. 6.1) for several injection modes and magnetic holding field values as a linear (and over-determined) system of equations for the transition probabilities

Wab - if the injected hyperfine population is known. At any given value of the magnetic

holding field and using the hydrogen injection modes Pe, Pz+ and Pz− (see tab. 3.1),

one has 12 equations and 5 unknown transition probabilities. For a simpler analysis, we define the spectrum Jab(ωab) of a relaxational transition. The spectrum is the ratio of the

transition probabilities, that are required to match the measured hyperfine populations and the squared reduced matrix elements:

Jab(ωab) = Wab |Cab|2 . (6.71) 0 2 4 6 8 10 10-2 10-1 1 10 ω₂₄ ω₂₃ ω₁₄ ω₃₄ ω₁₂ ω/ ω_HFS

Fig. 6.16:Frequency range of hydrogen relaxation transitions for holding field values within 0.7. . .350mT. The size of the reduced matrix elements Cab is in-

dicated by vertical lines. The 2−3 tran- sition has the widest range and a rea- sonably large reduced matrix element over the whole range.

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 10-2 10-1 1 ω/ ωHFS J(B, ω )/a.u. J(B, ω12) J(B, ω34) J(B, ω14) J(B, ω23) 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 10-2 10-1 1 B/B_C J(B, ω )/a.u. J(B, ω12) J(B, ω34) J(B, ω14) J(B, ω23)

Fig. 6.17: Measured spectra of the hydrogen π-transitions at a storage cell temperature of about 100K. In the left graph, the spectra are plotted vs. the transition frequency in the right graph vs. the target holding field. The solid line is the prediction by the saturation prefactor for J(B, ω12) and the dashed line for J(B, ω23).

The spectrum is not identical to the spectral density, but it has the advantage, that it can be determined directly from the measured data, as the squared reduced matrix elements |Cab|2 depend on the magnetic holding field only. The spin exchange relaxation time τse

is the same for all magnetic field values, as the magnetic field dependence of the spin exchange reaction is expressed by the (well known) tensor Mc

ab in eq. 6.1. Even if one

takes the normalization of the hyperfine population into account (P

a Na = 1), one has

still 9 equations and 5 unknowns per field value. This relation is even better in case of deuterium, as there are 11 different standard injection modes. With the definition

W12=W21 =w1 W34=W43=w2

W14=W41 =w3 W23=W23=w4

W24=W42 =w5

(6.72)

the equations for hydrogen in one injection mode are:

0 = N₁inj−N1+_ττ_sed P b,c M1 bcNbNc+_ττd_f{w1(N2−N1) +w3(N4−N1)} 0 = N₂inj−N2+ τd τse P b,c M2 bcNbNc+_ττd f{w1(N1−N2) +w4(N3−N2) +w5(N4−N2)} 0 = N₃inj−N3+ τd τse P b,c M_bc3NbNc+_ττd f{w2(N4−N3) +w4(N2−N3)} 0 = N₄inj₋N4+_ττ_sed P b,c M4 bcNbNc+_ττ_fd{w2(N3−N4) +w3(N1−N4) +w5(N2−N4)} (6.73)

All values - exceptwi andτse are either known or measured directly, so that the equation

system can be solved in case of hydrogen. In case of deuterium, the injected hyperfine populations are less well known, so that an iteration process had to be performed, which will be described in sec. 6.10.2.