Origin and Reduction of $1/f$ Magnetic Flux Noise in Superconducting Devices

Magnetic flux noise is a dominant source of dephasing and energy relaxation in superconducting qubits. The noise power spectral density varies with frequency as $1/f^{\alpha }$, with $\alpha \lesssim 1$, and spans 13 orders of magnitude. Recent work indicates that the noise is from unpaired magnetic defects on the surfaces of the superconducting devices. Here, we demonstrate that adsorbed molecular ${\mathrm{O}}_2$ is the dominant contributor to magnetism in superconducting thin films. We show that this magnetism can be reduced by appropriate surface treatment or improvement in the sample vacuum environment. We observe a suppression of static spin susceptibility by more than an order of magnitude and a suppression of $1/f$ magnetic flux noise power spectral density of up to a factor of 5. These advances open the door to the realization of superconducting qubits with improved quantum coherence.

Figure 1

(a) X-ray magnetic circular dichroism (XMCD) at the oxygen $K$ edge for a native Al film and an Al film exposed to air. The native film (top) shows no XMCD signal, while the air-exposed film (bottom) shows a clear XMCD signal at 531 eV. (Traces are offset for clarity.) A similar XMCD signal at the oxygen $K$ edge is seen for Nb films exposed to air (not shown). (b) Oxygen $K$-edge x-ray absorption spectroscopy (XAS) of an Al thin film cooled in the presence of $5\times {10}^{-8}\mathrm{Torr}$ ${\mathrm{O}}_2$. Beginning at around 45 K, we observe a sharp peak at 531 eV and a broad spectral feature from 535–550 eV which we ascribe to adsorbed molecular ${\mathrm{O}}_2$. (Traces are offset for clarity.) Dashed lines are from DFT simulations for ${\mathrm{Al}}_2{\mathrm{O}}_3$ (XAS at 50 K) and for ${\mathrm{O}}_2/{\mathrm{Al}}_2{\mathrm{O}}_3$ (XMCD and XAS at 10 K); see the Supplemental Material [18] for details.

Figure 2

(a) Schematic of hermetic grade 5 titanium enclosure for susceptibility and flux noise measurements. The enclosure incorporates weld-in SMA feedthroughs and a single $2.75^{\prime \prime }$ ConFlat gasket. (b) Schematic of the sample prep chamber. The chamber incorporates a turbo pump, an ion pump, and a transfer arm used to install the NEG in the sample cell.

Figure 3

Temperature-dependent flux threading a square-washer Nb SQUID (350 pH; see inset) cooled in a conventional vacuum (closed red symbols) and cooled following vacuum bake and ${\mathrm{NH}}_3$ passivation (open blue symbols). The upper (lower) branches correspond to cooling fields of $+128\mu \mathrm{T}$ ($-128\mu \mathrm{T}$). The magnitude of the flux change is proportional to the density of magnetically active surface spins [11].

Figure 4

(a) Flux noise spectra of SQUID device ${\mathrm{SiN}}_x-4$ before (upper trace) and after (lower trace) vacuum bakeout and ${\mathrm{NH}}_3$ passivation. (Inset) Device layout. (b) Flux noise spectra of SQUID device ${\mathrm{SiN}}_x-6$ before (upper trace) and after (lower trace) vacuum bakeout and UV illumination.