Illustration of difference between FET and SET transistors, in the context of charge offset drift. Upper panel: this shows the standard “S-shaped” control curve for a FET; note that a small change in the threshold voltage (different colors) does not affect the basic on/off behavior outside of the threshold region. Lower panel: in contrast, the basic control curve for a SET transistor is periodic, with a period corresponding to putting one more electron on the gate capacitor. Because of this periodicity, a small change in the charge offset can substantially alter the behavior of the transistor, turning an “off” state into an “on” state, and vice versa.
Example of time dependence of for a limited segment of run 2.10 , using method I. We note that there is a TLF present, with an amplitude of about . Over the course of this three-day measurement, except for one very short time near 19.1, the charge offset appears to be stable — neither drifting nor jumping. These data correspond to 2.10F in Table II. We note that, in this and all other data plots, the statistical uncertainty error bar is too small to be seen.
Time dependence for all of run 2.10, using method I . The data in Fig. 2 are encompassed here between days 19 and 22 in the middle panel. The vertical dashed lines refer to deliberate temperature excursions or other events, as noted in the text above each panel. The vertical dotted lines mostly refer to mechanical events in the dilution refrigerator (specifically, transfer of liquid helium); in the upper panel, the first three dotted lines refer to accidental losses of data. Since the control curve for a SETT, as illustrated in Fig. 1, is periodic, we can only measure modulo ; for the particular set of data shown here, the monotonic drift in the first suggests that in fact changes by many .
Measurement of for a device fabricated at PTB in Germany, using method I . This data are denoted in the tables as 2.42B, C. Between 14.5 and , the device was warmed to room temperature and opened to the air. The measurement temperature was between 0.03 and . The vertical dotted lines refer to transfer of liquid helium.
Measurement of for a device fabricated at NIST in Boulder, CO, USA, using method I . This data are denoted in the tables as 2.45. The measurement temperature was between 0.15 and . The vertical dotted lines refer to transfer of liquid helium.
An example of for a Si device, using method II; note the scale change (previous figures had a range from 0 to , while here the range is ) . As in previous plots, dotted lines denote transfers of liquid He, and dashed lines refer to specific temperature excursions. Over the entire period, the difference between minimum and maximum values of is . We note that much of this change was caused by the rise in temperature to . Data correspond to run 2.26A in Table III.
Dependence of drain current on gate voltage for device 2C-1Y . This device likely had an unintentional tunnel junction which reduced the total size of the device, thus making it possible to see at least one Coulomb blockade peak up to room temperature. At the highest temperature, the downward spikes are due to a single dominant TLF.
Power spectral density of charge fluctuations vs frequency at the base temperature , for the same device as in Fig. 7; various colors refer to four different measurement dates . At these temperatures, there was a single dominant TLF at , which was absent at . The stability of this TLF is demonstrated by both the extended period of time during which it was present, as well as the fact that there were multiple thermal cycles up to room temperature between the various measurements.
A single Coulomb blockade peak for temperatures between 37 and , for the same device as in Fig. 7. In this range, there was a single dominant TLF (indicated by the large additional amount of noise on the middle pair of curves) which moved through the bandwidth as a function of temperature. Upper pair: 89 and . Middle pair: 69 and . Lower pair: 41 and . As in Fig. 8, the stability of this TLF is demonstrated by both the long period of time between the first set of measurements and the second, as well as by the multiple thermal cycles up to room temperature between the two sets of measurements.
Temperature dependence of the amplitude of power spectral density fluctuations measured at and in the range of , over various periods of time, for the same device as in Fig. 7. Main: The peaks at about 60 and correspond to dominant TLFs, with a Lorentzian power spectral density; the rest of the data corresponds to typical noise. Note that these two TLFs were stable over long periods of time as indicated by the repeated measurements of the same peaks. Inset: At low temperatures, there were two TLFs dominant, below 0.1 and between about 1 and . In contrast to the two TLFs at higher temperatures, these two were much less temperature dependent, although they were highly gate-voltage-dependent.
Compendium of a large number of measurements of , mostly on devices fabricated at NIST, Gaithersburg, MD, USA, sorted by grade [amount of drift]. Rows with gray background correspond to measurements where a transient relaxation was observed. Columns: “ grade” represents a rough ranking of the amount of ; “A” corresponds to very little drift. “Notes:” “Inline” and “angled” correspond to source and drain leads parallel to or perpendicular to island; “caged” refers to a device in an electrostatic cage (Ref. 9). “Fabrication details:” Dates refer to dates of fabrication; specific times refer to time of deposition of ; for device WH72-2 in run 2.20A, the tunnel junctions were fabricated by oxidation in ozone; “blanket” devices refer to transistors with a layer of deposited on top of either the tunnel junction or the center of the island. “Precondition:” Amongst other treatments were annealing at in either or forming gas (mixture of and ). “Cooldown details:” Mostly these give the date and time when we reached a particular temperature; “illuminated” refers to an experiment where we exposed the device to light from a light emitting diode (LED) at base temperature. “Norm/sc” refers to normal or superconducting state. “” refers to a measurement of the power spectral density of the short-term (typically ) noise at .
Similar to the previous compendium, except that this table is sorted by date of measurement, amongst the devices fabricated at NIST, Gaithersburg. This table is included for two reasons: (1) in order to follow the behavior of a single device when multiple treatments were performed (e.g., Run 2.10) and (2) in order to assess the possiblity that the quality of the drift is correlated with date of fabrication, in case some fabrication parameter (e.g., deposition system base pressure) was drifting over the months or years.
Compendium of measurements on Si-based devices, fabricated at NTT, Tokyo, Japan and at Cornell University/NIST Gaithersburg, both of USA. In contrast to the metal-based devices (Table II), the Si-based devices show no significant charge offset drift. Notes: PADOX, V-PADOX, and “tunable” refer to various device architectures (Ref. 2 and 27).
Article metrics loading...
Full text loading...