X-ray absorber structures are integral to the design and performance of the microcalorimeter sensor. Their photolithographic integration enables exquisite determination of such details as thermal time constants and spectral resolution performance but can influence fabrication process flow. While many groups [2–7] have developed processes to fabricate electroplated bismuth absorbers for microcalorimeters, the requirements on x-ray absorbers for the ATHENA X-IFU instrument pose unique fabrication challenges. Specifically for X-IFU, we need to fabricate free standing absorbers with six narrow 4–5 µm diameter stems supporting a large 313 µm2 area absorber composed of 1.05 µm Au and 5.51 µm Bi and capped by a thin Ti/Au coating[1]. The stem diameter is chosen to achieve structural and thermal needs of the absorber while being feasible to fabricate. However, small stem diameter can lead to a mechanically weak stem which may not be suitable for extended geometries and certain film thickness limits. To enhance the structural integrity of the stem, a funnel-shape photoresist mold using a proximity exposure has been developed to control the shape of the stem. Backfilling the stems with electroplated Au was also incorporated to increase the strength of thin stems in supporting the mechanical stresses applied to them from the large absorbers. Additionally, to optimize for post-patterning substrate cleanliness for absorber yield, we have developed a method of plating the thick bismuth absorbers using a photoresist mold while controlling the bismuth grain growth with a leveling process while electroplating. Finally, to separate the pixels from each other, we use an ion mill etch to remove the Ti/Au capping layer as well as the Ti/Au seed layer. Although others have backfilled stems [8] and electroplated bismuth through a photoresist mold [4, 5, 7], we are combining all of these patterning steps, requiring numerous repatterns over the sacrificial resist layer supporting the electroplated mushroom absorber overhanging the TES. Figure 1 has a cartoon of the pixel architecture (Left) and a table listing our process fabrication steps (Right).
Historically, our calorimeter group first developed a procedure to yield overhanging mushroom shaped x-ray absorbers with electroplated gold and bismuth films suspended above a transition edge sensor on a silicon nitride membrane [9, 10]. Our NASA group has demonstrated an average 2.25 eV energy resolution in prototype arrays of spectrometer pixels with 1.5 µm Au + 3 µm Bi absorbers on 275 µm pitch [11]. Recently revised ATHENA requirements for lower heat capacity and higher quantum efficiency imposed a fabrication change to thinner Au (1.0 µm) and thicker Bi (5.5 µm) as well as an increase in absorber pixel size to 313 µm2 [1]. Our initial attempts to yield these new absorbers resulted in absorber touches to the substrate degrading pixel performance. Further, some pixels collapse completely when the stems break near the base. SEM imaging revealed that the thinner Au stems filled with bismuth grains were not mechanically strong enough to support the larger area thicker bismuth absorbers. Figure 2 shows examples of absorber collapse (top and bottom left) and hollow stems that can break during release (top and bottom right). We measured the film stresses of the individual layers and found the titanium adhesion layer to be tensile, the gold to be mostly stress free and the bismuth to be compressive. With hollower stems from thinner electroplated Au and the additional stress from the thicker bismuth film, our traditional stem design was not mechanically strong enough to support the larger area absorbers required by ATHENA.
Our solution to making mechanically stable absorbers was threefold. First, we smoothed the shape of stems using a double exposure of the resist and a reflow bake. Next, we increased the tensile titanium in the seed layer deposition to counteract the compressive stress of the bismuth. Third, we pattern atop the seed layer and backfill the stems with electroplated Au to make them mechanically more robust.
While developing our new process to mechanically strengthen the stems supporting the overhanging absorbers, we also worked out a procedure to electroplate the absorbers through a photoresist mold. Previously, we have defined our absorbers by a long ion mill through the entire Au/Bi stack. Because the Bi grain size increases with thickness [2], we have found that after a long ion mill, a fraction of absorbers have small metal bridges thermally shorting them together, degrading badly the energy resolution in the affected pixels. Electroplating the bismuth through a photoresist mold and then following with a much shorter ion mill to clear the remaining Ti/Au from both the capping and seed layers appears to solve the problem of thermally shorted pixels. Figure 1 (Right) details our new fabrication flow with the additional steps highlighted in red. As will be shown later in this paper, despite the complications from the additional fabrication steps, our first delivered devices showed no degradation in energy resolution, no additional broadening in the low energy tails of the x-ray spectrum and we were still able to achieve the desired heat capacity. Figure 3 shows SEM images of an array that yielded after electroplating Bi through a photoresist mold. The absorbers are free standing and flat with 5.0 µm gaps between pixels.