Background

Solid oxide fuel cells (SOFCs) normally operate at considerably high temperatures (>700°C) to facilitate ionic charge transport and electrode kinetics [1, 2]. Encountered by issues such as limited material selection and poor cell durability, many researchers have tried to reduce the operating temperature [3–5]. However, lower operating temperature led to a significant sacrifice in energy conversion efficiency due to the resulting increase in ohmic and activation losses [1].

There are roughly two ways to minimize the ohmic loss surging at lower operating temperatures. One is to reduce the thickness of the electrolyte, and the other is to synthesize materials with higher ionic conductivities. First, the strategy to reduce in electrolyte thickness has been carried out by many research groups [6–10]. Shim et al. demonstrated that a fuel cell employing a 40-nm-thick yttria-stabilized zirconia (YSZ) can generate a power density of 270 mW/cm2 at 350°C [11], while Kerman et al. demonstrated 1,037 mW/cm2 at 500°C from a 100-nm-thick YSZ-based fuel cell [12]. Another approach of minimizing ohmic loss is using electrolytes with higher ionic conductivities. Gadolinium-doped ceria (GDC) has been considered as a promising electrolyte material due to its excellent oxygen ion conductivity at low temperatures [23].

Figure 1
figure 1

XPS spectra of (a) Ce 3 d and (b) Gd 4 d core levels of GDC thin films.

Full size image

We applied the ALD technique, thus enabling excellent step coverage to fabricate the ultrathin conformal YSZ layer using a commercial ALD system (Plus-100, Quros Co., Ltd., Osan, South Korea) [24, 25]. Prior to the deposition of a YSZ thin-film, zirconia and yttria films were separately deposited and characterized for a systematic study. Both films were fabricated by repeating the sequence of precursor pulse (3 s), purge (20 s), oxidant pulse (1 s), and purge (10 s). Tetrakis(dimethylamido)zirconium, Zr(NMe2)4, and Tris(methylcyclopentadienyl)yttrium, Y(MeCp)3, were used as precursors for zirconium and yttrium, respectively. The precursor was delivered using an electropolished stainless steel bubbler fed by Ar gas with 99.99% purity. O2 gas was used as the oxidant, and stage temperature was set to 250°C. The temperatures of canisters with charged precursors were 40°C and 180°C, and the line temperatures were 60°C and 210°C for zirconia and yttria deposition, respectively. The growth rates of both zirconia and yttria films during the initial 1,000 cycles were approximately 1 Å/cycle. Although these growth rates were somewhat lower than the reported values (1.2 to 1.5 Å/cycle) [11], the film thickness increased proportionally with the deposition cycles. XPS analyses were performed to determine the chemical composition of an approximately 100-nm-thick zirconia film and an approximately 100-nm-thick yttria film. The atomic concentrations in the zirconia thin-film were as follows: for Zr 3d, it was 41.6%, and for O 1s, it was 58.4%; they were somewhat different from the expected stoichiometry of ZrO2. It is attributed to the fact that reduced zirconium (e.g., Zr0 3d5/2 or Zr2+ 3d5/2) was partially combined with O2 during the ALD process, as indicated in the curve fitting result of Figure 2a [26]. The atomic concentrations of the yttria thin-film were Y 3d = 40.9% and O 1s = 59.1%, which are well aligned with the stoichiometry of Y2O3. The Y 3d5/2 peak was located at a binding energy of 156.7 eV, as shown in Figure 2b [27].

Figure 2
figure 2

XPS spectra of (a) Zr 3 d and (b) Y 3 d core levels of zirconia/yttria thin films.

Full size image

Subsequently, YSZ thin films were fabricated by co-deposition of zirconia and yttria. Zirconia was deposited prior to yttria deposition. Yttria mole fraction in the ALD YSZ thin-film was controlled by changing the ratio of deposition cycles for zirconia and yttria. The yttria mole fraction is widely known to determine oxygen ion conductivity in the YSZ, and 8% mole yttria was reported to render the maximum oxygen ion conductivity [1]. When the ratio of zirconia and yttria ALD cycle was 7:1, the atomic concentrations of the YSZ thin-film were as follows: Zr 3d = 24.2%, Y 3d = 3.6%, and O 1s = 72.1%, which were also determined by an XPS analysis. The Y2O3 mole fraction, x, in the YSZ chemical formula of (ZrO2)1−x(Y2O3) x was approximately 0.07. In the case of the YSZ thin-film, the XPS spectra corresponding to an under-stoichiometric ZrO2 did not appear, unlike those in the zirconia thin-film.

Design of AAO-supported GDC/YSZ bilayered thin-film fuel cell

A commercial AAO (Synkera Technology Inc., Longmont, CO, USA) template with an 80-nm pore and a 100-μm height was used as the substrate to leverage their high density of nanopores and resulting electrochemical reaction sites [28, 29]. Pt electrode was fabricated by a commercial sputter (A-Tech System Ltd.). Pt with 99.9% purity was used as the Pt target, and the T-S distance was 100 mm. The deposition was conducted at room temperature, and the direct current power was set to 200 W. The Pt anode was deposited on the AAO template in an area of 10 × 10 mm2. Dense Pt anodes were deposited at a 5-mTorr Ar pressure, having the growth rate of approximately 60 nm/min. Subsequently, YSZ and GDC electrolytes with an area of 9 × 9 mm2 were deposited on the Pt anode. The critical thickness ratio of the YSZ layer to the GDC layer to prevent the reduction of ceria, which was determined considering the distribution of oxygen activity through the thickness of a bilayer, was reported to be approximately 10−4 at 800°C and was expected to decrease further at lower temperatures [30]. For this reason, the required minimum thickness of the YSZ layer for electron blockage, if the thickness of GDC layer is 420 nm, is only approximately 0.4 Å. However, a much thicker YSZ film (40 nm) was deposited on the anode side to compensate the rough morphological variations of the Pt-coated AAO surface. The GDC layer, which was 420-nm thick, was then deposited on the YSZ layer. Oxygen reduction reaction happening at the cathode is widely known to cause a significantly greater activation loss compared with the hydrogen oxidation reaction occurring at the anode [1]. In order to facilitate cathode reaction, a porous Pt cathode was prepared by depositing at a much higher Ar pressure of 90 mTorr than that used for anode deposition (5 mTorr Ar). The cathode thickness was approximately 200 nm. The growth rate still remained at approximately 60 nm/min. The Pt cathode, which effectively determines the nominal area of active cell, was deposited using a mask with 1 × 1 mm2 openings.

Electrochemical evaluation of thin-film fuel cells

Thin-film fuel cells with 850-nm-thick GDC and 850-nm-thick Sn0.9In0.1P2O7 (SIPO) electrolytes were fabricated to study further how the ALD YSZ layer have the influence on electrochemical performance [31]. Except for the electrolyte, other cell components were equal to those for GDC/YSZ bilayered thin-film fuel cell. For a comparison with GDC-based cells (cell 1, Pt/GDC/Pt), we fabricated SIPO-based cells (cell 2, Pt/SIPO/Pt). It is postulated that the electrolytes deposited with the same deposition process have identical microstructures [20]. As shown in Figure 3a,b, both the 850-nm-thick dense GDC and SIPO electrolytes did not show any evident pinhole. However, the OCV of approximately 0.3 V for cell 1 was significantly lower than that for cell 2 (approximately 1.0 V). This result indicates that the lower OCV of the GDC-based cells may have originated from oxygen permeation through the GDC electrolyte and/or ceria reduction, not from gas leakage through pinholes. In order to verify the effect of the ALD YSZ layer, we characterized electrochemical performances of GDC/YSZ bilayered thin-film fuel cell (cell 3, Pt/GDC/YSZ/Pt), which has a 40-nm-thick ALD YSZ layer at the anodic interface as shown in Figure 4. As expected, the OCV of cell 3 with the ALD YSZ layer stayed at a decent value of approximately 1.07 V, unlike that of cell 1 (approximately 0.3 V). This discrepancy indicated that the ALD YSZ layer played a successful role as a functional layer to suppress the issues that originated from thin-film GDC electrolyte such as the electronic current leakage and the oxygen permeation [15–17]. The thicknesses of GDC layers in cells 1 and 3 were 850 and 420 nm, respectively. Originally, it was intended for the comparison of the two samples with the same GDC thickness, but a 420-nm-thick GDC-based cell showed highly unstable outputs in the measured quantities. While the peak power density of the cell (cell 3) with an YSZ blocking layer reached approximately 35 mW/cm2, that of the single-layered GDC-based cell (cell 1) showed a much lesser power density below approximately 0.01 mW/cm2, as shown in Figure 5a,b.

Figure 3
figure 3

FE-SEM cross-sectional images of cells 1 and 2. (a) A GDC single-layered thin-film fuel cell (cell 1) and (b) a SIPO single-layered thin-film fuel cell (cell 2).

Full size image
Figure 4
figure 4

FE-SEM cross-sectional image of a GDC/YSZ bilayered thin-film fuel cell (cell 3).

Full size image
Figure 5
figure 5

Electrochemical performances of cells 1 and 3. (a) A 850-nm-thick GDC electrolyte fuel cell (cell 1) and (b) a 460-nm-thick GDC/YSZ electrolyte fuel cell (cell 3) measured at 450°C.

Full size image

To evaluate the stability of GDC/YSZ bilayered thin-film fuel cell (cell 3), the OCV and the peak power density were measured for 4 h at 450°C, as shown in Figure 6. While reduction of the OCV was negligible, the peak power density sharply decreased by approximately 30% after 4 h. This sharp performance degradation in the AAO-supported thin-film fuel cells was previously studied by Kwon et al. [32]. They ascribed the reason to the agglomeration of the Pt thin-film without microstructural supports. In line with the explanation, the agglomeration of Pt particles was clearly visible when comparing the surface morphologies before and after a cell test, and the degradation of power output caused by the Pt cathode agglomeration was also confirmed through AC impedance measurements. Nevertheless, the stability of AAO-supported GDC/YSZ thin-film fuel cells was relatively superior to ‘freestanding’ thin-film fuel cells with silicon-based substrates [33]. Actually, the configuration of the AAO-supported thin-film fuel cells was maintained after 10 h at 450°C. However, it was reported that freestanding thin-film fuel cells were all broken before 1 h in the same operational conditions [29, 33].

Figure 6
figure 6

OCV and peak power density of GDC/YSZ thin-film fuel cell (cell 3) versus dwell time at 450 °C.

Full size image

Conclusions

In this study, we implemented and suggested a promising feasibility of a thin-film low-temperature SOFC using a bilayered electrolyte configuration on the AAO platform. GDC has suffered from its chemical instability and the resulting electronic leakage under a reduction environment. In a thin-film configuration for securing a decent oxygen ion conductivity even at low temperatures (as an LT-SOFC), oxygen permeation through the GDC film became problematic as well. This paper reports that an insertion of a very thin ALD YSZ layer between the anode Pt and the GDC electrolyte significantly improved the electrochemical performance of a cell. At 450°C, a thin-film fuel cell with 850-nm-thick GDC electrolyte showed an OCV of approximately 0.3 V and a power density of approximately 0.01 mW/cm2. On the other hand, a thin-film fuel cell with a bilayered electrolyte consisting of a 40-nm-thick YSZ and a 420-nm-thick GDC reached an OCV of approximately 1.07 V and a power density of approximately 35 mW/cm2. From these results, it was confirmed that the YSZ layer successfully acted as a protective layer. The cell performance is expected to further improve through the microstructural optimization of electrode interfaces and adjustment of chemical compositions of each film.

While the fully functional YSZ layer presented here is already very thin (40 nm), there are good chances of reducing the thickness even further considering that a theoretical approach predicted an YSZ-to-GDC thickness ratio of 0.01% would suffice to guarantee electron blockage [30].

Authors’ information

SJ and IC are students in the Graduate School of Convergence Science and Technology, Seoul National University. YHL, JP, and JYP are graduate students in the School of Mechanical and Aerospace Engineering, Seoul National University. MHL is a professor in the School of Engineering at the University of California, Merced. SWC is a professor in the School of Mechanical and Aerospace Engineering, Seoul National University.