In this project, I designed a high temperature thermophotovoltaic (TPV) system and assessed its system-level performance in the context of concentrated solar power (CSP) with thermal energy storage (TES). Until now, TPV has largely been discarded as an option for commercial energy production, due to its seemingly low efficiency. However, in this work, I showed that there exists a pathway to engineering TPV that has the potential to compete with the most efficient heat engines on the planet, namely combined cycles. Nonetheless, TPV could offer other advantages that would give CSP an advantage over fossil-based alternatives, with the most significant advantage being the potential to achieve an almost order of magnitude lower cost (e.g., $0.1/W-e vs. $1/W-e). This potential could be a game changer for CSP and utility-scale solar. The system concept introduced and designed combines the great economic advantages of TES with the potential for low cost and high performance derived from TPV cells fabricated on reusable substrates, with a high reflectivity back reflector for photon recycling. The fact that the results suggest that another technology other than turbomachinery has the potential to reach 60%, and could set a new record for heat engine efficiency is a finding that could change the trajectory of the power industry.

Using TPV as a Power Cycle for Concentrated Solar Power (CSP) with Thermal Energy Storage (TES)

One potential application for TPV has been in the context of solar energy conversion and specifically as a power cycle for CSP, instead of a turbine based heat engine. State of the art CSP plants with TES are capital intensive, but presently have a levelized cost of electricity (LCOE) in the range of ~ 13.5-20 cents/kWh. Since LCOE scales with the inverse of the overall system efficiency (solar-to-electric), one potential route to decreasing the LCOE is to use a more efficient power cycle, since it is the largest source of inefficiency in the full system. Current CSP power cycles are steam based Rankine cycles with an efficiency in the range of 35-40% and one potential way to improve the LCOE of CSP would be to consider using a solid-state heat engine such as TPV. Here, the question is not only can one achieve a major boost in efficiency, but also whether it is possible to achieve a lower cost, faster response/ramp times/rates, longer life and lower maintenance costs than a turbine. It should be noted that the cost of the turbine in a CSP plant is not negligible as it is typically on the order of $1/W-e, while to total plant costs are typically ~ $4-6/W-e. However, by comparison, if the electrical power output of a TPV based power cycle was on the order of 50kW-e/m2 of cell area and the cost of the cells could be reduced to the order of $10,000/m2 via the use of reusable substrates, one could potentially achieve costs on the order of $0.1/W-e as shown in figure 1. Such an advancement would first require high efficiencies e.g., > 50% but would also require that such efficient cells be fabricated on reusable substrates. However, initial demonstrations by Morral et al. suggest the possibility of such advancement is not unfounded. Nonetheless, in the following, I outline how such a system might be designed.

Fig 1. Cost to power ratio of TPV power block as a function of TPV cell cost.

Considering the limits on TPV cell performance due to recombination, one must operate the emitter at high temperatures so that (1) a material with a reasonably high band gap can be used, and (2) so that the heat flow from photons above the bandgap greatly exceeds the rate of heat leakage to the environment. Generally, this corresponds to temperatures above 1000°C, but one must also then think about what medium could be used to store the heat at such high temperatures. In this respect, we envision that one could use a TPV based power cycle in the context of CSP with TES, by using a receiver made of graphite, with liquid metal Sn serving as a primary heat transfer fluid (HTF). Initial models and ongoing experiments suggest that this approach can reach the same range of receiver efficiencies as existing plants 80-90%, but would require higher sunlight concentrations > 5000 kW/m2. Furthermore, Sn and graphite exhibit no chemical interaction at any temperature, despite the fact that they reside in the same column of the periodic table. Thus, Sn(l) can be used as a HTF in a graphite piping network, since it melts at 232°C and does not boil until 2602°C, and will not exhibit any corrosion. Furthermore, commercial molten metal pumps made of graphite exist, and with proper retrofitting to keep the motor thermally isolated, they could be used at the requisite temperatures. Sn, however, is exceedingly expensive to use as a TES fluid, and therefore one must use another material to store the heat at high temperature.

Here, it should be appreciated that TPV, unlike a turbine, is best suited for a constant temperature heat input, similar to a Carnot engine. Therefore, the most exergetically efficient embodiment would be to store all of the heat at the highest possible temperature via a phase change material (PCM). It should be noted that a heat engine that takes in heat at a constant temperature has the potential to reach higher efficiencies than a heat engine that takes in heat over a range of temperatures because there is less entropy transfer into the system. Thus, a solid-state heat engine such as TPV has an intrinsic thermodynamic advantage over a turbine. There are very few classes of materials that are stable above 1000°C, but there is one choice that can not only act as a PCM in the target temperature range but also has a somewhat anomalously large heat of fusion and happens to be very low cost, namely silicon (Si). Si is the most abundant solid element on earth, with a cost for bulk metallurgical grade Si ~ $1-2/lb, a melting point of 1414°C and a large heat of fusion of 1.92 MJ/kg. One could encapsulate Si in tubes made from inexpensive refractories, such as mullite or alumina stored in large, sealed closed end tubes, stacked in a larger tank. Here, the liquid metal Sn can flow in between the spaces between parallel tubes to either deliver or extract heat from the Si melting/solidification occurring inside the tubes. With such an approach, due to the extremely high convective heat transfer coefficients of liquid metals, the system could be operated near isothermally, with temperature differences on the order of 50°C.

For example, as illustrated in Fig. 2, to charge the Si PCM TES, the Sn could be pumped to the receiver and heated from the melting point of the storage medium (Si – 1414°C) to ≈ Tmelt+ΔT1 , via sensible heating. A high flow rate and small temperature rise ΔT ≈ 50°C is needed to maximize the system level exergetic efficiency. The liquid Sn at Tmelt+ΔT1  could then be routed to the TES tank where it melts the storage medium and is cooled back to Tmelt, whereby it is recirculated back to the tower. To discharge the thermal battery, the Sn is then circulated from the TES tank to the power cycle. Here I envision a TPV based power cycle, which consists of a large array of graphite tubes that serve as near blackbody emitters. The TPV modules are located in between successive columns of graphite tubes and are irradiated with blackbody emission between  Tmelt and Tmelt-ΔT2 , where again a small  ΔT2 is used to preserve high exergetic efficiency and the heat is transferred to the emitter via sensible cooling in the Sn (see Fig. 2). A major benefit of using TES, aside from its low cost, is that dispatchability is enabled through the rate of discharge, which can be controlled by the flow rate of the Sn through the graphite emitter pipes (see Fig. 2).

Fig 2. Schematic conceptual layout of a utility-scale TPV system. Red and white arrows on pipes indicate the flow path for charging (red) and discharging (white) the thermal storage.

In the aforementioned system concept, the TPV modules would need to be engineered with the spectral control strategy integrated into the cell/module itself. As shown in Figure 3, high-efficiency InGaAs cells grown on reusable InP substrates somehow backed with a highly reflective (e.g., Ag) layer to serve as an inexpensive omnidirectional high-efficiency IR reflector is one potential option. This layer, along with the rest of the cell would then need to be engineered to minimize below band gap absorption and efficiently convert the upper 15-25% of the spectrum, which is peaked between 1.72-1.97 micron (1200-1414°C). The cells would then need to be actively cooled behind the substrates with a water or oil cooled heat sink keeping them at ambient temperature to maximize their performance. The entire power cycle unit would need to be heavily insulated from the environment and held in a vacuum to minimize convective and conductive heat leakage from the emitter (graphite pipes) to the TPV cells. Furthermore, the entire power cycle must be sufficiently large (i.e. ≈ 1-10 MW) to minimize edge effects and so the power generated inside greatly exceeds the heat leakage to the environment.

Fig 3. Schematic of multijunction TPV power block with silver back surface reflector.

It is worth mentioning that when one integrates a BSR, the priorities for further optimization change, and the path to further increase efficiency involves maximizing the collection (EQE) and voltage associated with the photons with energies significantly above the band gap. This might then suggest that further optical filtering or multiple junctions could provide some additional benefits. Figure 4 shows the efficiency of power block with a dual junction TPV cell (InGaAs bottom cell) as a function of the bandgap of the top cell. It can be seen that there is an optimum bandgap (Eg2  ~ 0.94 eV) for the top material that yields the maximum efficiency. This optimum occurs because of the relative trade-off between thermalization losses in the bottom cell (InGaAs cell) vs. the fictitious topping cell. The addition of this second TPV cell increases the overall efficiency above the critical barrier of 60%  and conceptually, additional junctions could increase this even more.

Fig. 4. System efficiency for a dual junction TPV cell. The inset shows the peak in efficiency at 0.95 eV.

The significance of the 60% barrier is important because it represents the highest heat engine efficiency that has ever been achieved commercially, which is accomplished through the usage of a tandem/combined (Brayton + Rankine) turbine based cycle. Turbomachinery based heat engines are the most efficient and cost-effective heat engines at present and are therefore the most widely used devices in the utility-scale power generation industry. Their costs are well-established and unlikely to see significant decreases in the future, and their performance is limited by thermodynamics – which is fundamental. Turbines currently operate very close to their fundamental thermodynamic limits and thus there is little room for significant improvement, other than increasing the operating temperatures. From this perspective, it is then quite remarkable to note that TPV may have the potential to reach the same or possibly higher performance than a combined cycle. Unlike, turbine engineering, which has been highly successful over the last few decades and has already capitalized on most opportunities for further advancement, TPV has much more room for improvement. Thus, it is remarkable to consider TPV as an alternative power cycle for CSP or grid-scale storage applications, particularly given its potential to cost 2-10X less than a turbine. Furthermore, it is interesting to consider the additional advantages associated with using a TPV power cycle at the utility scale. For example, one could imagine arranging the TPV cells in such a way that they can be mechanically moved into and out of view of the emitter. In this way, one could control the output of the system and shift from zero to full load within seconds, which could allow for load following of other renewables, providing the grid with great flexibility.

Fig 5. Asegun Henry’s team at Georgia Tech are developing a thermal storage system that uses the liquid tin as a transport fluid. The storage medium itself would be metallurgical grade silicon or an aluminum-silicon alloy, which is much more cost effective. The technology could help increase the efficiency of energy systems based on concentrated solar thermal energy