width and the center of the resonance is obtained by fitting the function
a(I
FB
) = a
/jointfilesconvert/317304/bg
·
1 −
∆I
FB
I
FB
− I
FB,0
!
, (3.3)
to the data, where the width and the center of the FR are expressed in terms of the
Feshbach current. The goal of this two-step fitting procedure is to extract not only
the fitting parameters ∆
I
FB
and
I
FB,0
(and their uncertainties), but also the correlation
between them. If we find e.g. an anti-correlation, the value of interest (in our case the
scattering length) is better known than the single uncertainties on the fitting parameters
would suggest. We discuss the determination of the uncertainty ∆
a
on the scattering
length in details in appendix A.7.
The results for the calibration of the scattering length are shown in Fig. 3.5. Using only
the data above the FR (
B > B
0
) for the calibration, we obtain a small uncertainty ∆
a
close to the datapoints, i.e. for
a &
20
a
0
. However, in the region
a <
0 the uncertainty
grows significantly. This indicates a rather large uncertainty on the fitting parameters
∆
I
FB
and
I
FB,0
, combined with a strong anti-correlation between them. If we also use
the data for the magnetic field strengths below the resonance, the uncertainties on ∆
I
FB
and
I
FB,0
typically decrease by a factor of around four, while the anti-correlation becomes
smaller. As shown in Fig. 3.5(b), we thus obtain a more precise calibration in the region
a < 0, which is an interesting region for the experiments presented in this thesis.
-10
100
300
30
-5
0
5
10 15
15
10
5
0
0
40
80
-40
scatt. length a (a )
0
scatt. length a (a )
0
B-B (G)
0
calibration a
uncertainty
(a) (b)
Fig. 3.5, Calibration of the scattering length:
(a) Scattering length values computed
from the measured values
R
5
y
/N
above the Feshbach resonance (filled black
dots) and below the FR (open black dots). The black line is a fit with the
theoretically expected behaviour
a
(
B
), given by Eq.
(3.1)
. (b) Uncertainty ∆
a
on the scattering length
a
. Using only the datapoints above the FR in the
fitting procedure yields a large uncertainty ∆
a
in the experimentally interesting
region, i.e. when
a .
0 (dashed line). When including datapoints below the FR
(solid line), ∆a decreases in this region, owing to reduced uncertainties on the
fitting parameters.
57
Comments to this Manuals