Pinchoff
The PinchoffFit fits a one dimensional I-V curve to a hyperbolic tangent to determine the parameters of the tangent describing the measurements the best. The fit function used is
where \(a\) is the amplitude, \(b\) the slope and \(c\) the shift. \(d\) is the sign of tanh, which may be different for different types of readout methods, e.g. rf. Note that there is a more sophisticated fit function in the sim package. These parameters as well as the residuals of the fit are retained as features. Based on the first derivative of either the fit or normalized data, the active range as well as transition voltage of the gate swept is determined. The active/valid voltage range is generally indicated by \([L, H]\), while the transition voltage is indicated by a \(T\). These values, together with the signal strength at each of these voltages is saved to metadata of the (QCoDeS) dataset under the nt.meta_tag key.
Figures Fig. 14 and Fig. 15 show the fit together with extracted features. They have been plotted using the plot_fit and plot_features methods.
Fig. 14 Example of a pinchoff fit.
Fig. 15 Some of the pinchoff features explained.
- Pinchoff labels
To allow for supervised machine learning to determine the quality, pinchoff curves need to be labelled. nanotune uses to labels, good (1, True) and poor (0, False). In general, a good curve is one showing a clear transition between open and closed regime and a poor doesn’t. However, there are many cases in-between, such as curves that don’t start at max/open regime but slightly below or gates that pinch off in stages, with some noise, or with a small slope. It is up to the labeller to decide which types of imperfection belong to which category. Ideally, this decision is made beforehand, to ensure consistent labelling.