Effect of hydrogen on the integrity of aluminium–oxide interface at elevated temperatures

Hydrogen can facilitate the detachment of protective oxide layer off metals and alloys. The degradation is usually exacerbated at elevated temperatures in many industrial applications; however, its origin remains poorly understood. Here by heating hydrogenated aluminium inside an environmental transmission electron microscope, we show that hydrogen exposure of just a few minutes can greatly degrade the high temperature integrity of metal–oxide interface. Moreover, there exists a critical temperature of ∼150 °C, above which the growth of cavities at the metal–oxide interface reverses to shrinkage, followed by the formation of a few giant cavities. Vacancy supersaturation, activation of a long-range diffusion pathway along the detached interface and the dissociation of hydrogen-vacancy complexes are critical factors affecting this behaviour. These results enrich the understanding of hydrogen-induced interfacial failure at elevated temperatures.


Supplementary Note 1:
The MEMS heating chip. We designed and produced a MEMS heating chip specifically for samples prepared from bulk materials using FIB (Fig. 1). Unlike conventional MEMS heaters, which use a membrane-shaped heating area fixed in the center of the bulk handle 1,2 , this heating chip has a free-standing thick heating block with a Si hotplate (Fig. 1 b). The hotplate is connected to the handle part of the chip with springs and has mounting bars to attach the sample at the free end. Serpentine metal filaments were deposited on the hotplate as the heater and the temperature sensor.
Compared with traditional MEMS heaters, this design provides more convenience for transferring lift-out samples cut from bulk material. During heating, the springs can adaptively counterbalance the thermal expansion of the hotplate, preventing the bulging of the heating membrane that occurs in conventional MEMS heaters as well as minimizing sample drift. These springs can also be used to isolate thermal dissipation, helping to maintain an even temperature distribution in the hotplate with a low power supply. Finally, the small heating volume lends to a quick response with thermal equilibrium reached within a blink. The temperature is measured from the resistance of the metal coil using a calibrated temperature-resistivity relationship.

Supplementary Note 2:
Volume estimation. After normal hydrogenation, multiple giant cavities formed on both the pillar surface and the neighboring substrate area (see Fig. 4b), making reasonable estimation of the cavity volume nearly impossible. To simplify this problem, we placed only the extruding part of the pillar into the irradiation zone (see Supplementary Fig.   2a). As can be seen from Supplementary Fig. 2b&c, after hydrogenation, small protocavities were evident only at the top and along the middle part of the pillar. The illuminated area can be seen as a cylinder with height (H=940 nm) and diameter (D=315 nm). The average size of the proto-cavities (d=30 nm, h=8.5 nm, l=50 nm, defined in the inset in Supplementary Fig.2b) after hydrogenation were measured from the TEM image. Assuming the proto-cavities are spherical in shape, the radius of the protocavities were calculated as r = . As shown in Supplementary Fig. 3, the diameter at the marked point increased from 314.0 nm up to 320.8 nm, which meets good with the expectation.

Supplementary Note 3:
Diffusivity estimation. We used the same calculation methods used in our previous The point defect clusters are characterized to be interstitial Frank loops induced by irradiation 6 . Both these FIB-induced defects and implanted Ga can be cleaned out by thermal annealing 7 . Supplementary Fig. 1a-b shows the results from thermal annealing at 200 ℃. It can be seen that after annealing, the initially 'dirty' pillar became clean, as evidenced by the smooth thickness contour. Besides, we observed from SEM image that some spherical particles formed on the lamella surface after annealing, which was proved to be Ga by energy dispersive X-ray analysis ( Supplementary Fig. 6).
To verify the FIB effect on the giant cavity formation process, annealed samples