The following post has been submitted by Michael Lammert
Engineering Challenges for Developing the SunCell’s Concentrator Photovoltaic Cells and Geodesic Dome Array
Michael Lammert (AKA Dr. Mike)
Brilliant Light Power (BLP) has made significant progress in the development of their SunCell over the last several months, including the ability to operate at a sustained energy output without having to supply ignition energy to the system, at an output level sufficient to heat a blackbody radiator to 3000ºK. Although there certainly is still a lot of work left to do to perfect the light generation portion of the SunCell system, there remains the large task of developing the concentrator photovoltaic (CPV) cells and the geodesic dome array to capture the energy from the blackbody radiator. The primary challenges that I see in the development of the SunCell’s CPV cells and the geodesic dome array include:
- increasing the metal thickness on the illuminated side of the cell sufficiently to handle the high currents in the large area triangular shaped cells needed to form a geodesic dome array,
- improving the design of the semiconductor layers within the CPV cell to accommodate higher current densities,
- achieving extremely low resistance and low stress connections between cells that form the geodesic array, and
- verifying that the blackbody radiator is emitting energy uniformly at the distance the CPV cells will be from the radiator.
Considerable information about the SunCell can be found on Brilliant Light Power’s website. In particular, Brilliant Light Power’s 12-6-2016 video presentation on the SunCell provides information on the SunCell’s design and future development plans. Since it is doubtful that many took the time to listen to the entire two hour presentation, below is some of the important information from that presentation as I understand it:
- Min 41-44: A peak pulse power of 5MW can be achieved with the average pulse power being 1MW with a pulse width of about 1msec. 1000 pulses per sec are needed to achieve 1MW output power.
- Min 44: The initial target temperature for the blackbody radiator will be 3000ºK. (The temperature is to be increased to 3500ºK after the triple junction cells become available.)
- Min 46: The triple junction CPV cells will be built by Masimo. (Data on one commercial concentrator solar cell built by Masimo- Part No. BFG3-2BB-1010-A10 can be found on-line. Also, technical papers discussing their bi-facial growth, triple junction CPV cells can be found on-line.)
- Min 52: The CPV cells will be arrayed in a geodesic dome design.
- Min 54-55: Low energy light (IR) will be reflected back to the blackbody source to improve efficiency.
- Min 57: The CPV cells will be grown by MOCVD (metal-organic chemical vapor deposition).
- Min 64: The ignition can be turned off when blackbody radiator reaches ~3000ºK. (My assumption is that once an ignition source is no longer needed, the rate at which energy is delivered to the blackbody radiator is controlled by rate that fuel is supplied to the system, although this was not stated in the presentation.)
- Min 65-67: The plan for adding CPV cells to the system includes starting with an operating temperature of 3000ºK, use single junction cells with about10% efficiency to achieve an output 33KW. Next about 20% efficiency double junction cells will be used to achieve 66KW output power, and then 40% efficiency triple junction cells will be used to achieve ~130KW output power. Finally, blackbody radiator temperature will be increased to 3500ºK to get output of about 250KW. (Another statement was made that the target output power for the initial “pilot” application of the triple junction CPV cells will be 100KW.)
- Min 65: The blackbody graphite dome is 6”diameter and will output 330KW at 3000ºK
- Min 123: The CPV cells will be cooled by water at a 40ºC output temperature from the cooling radiator and at a flow rate to return water at 80ºC from the CPV cells back to the radiator.
Although there is a lot of information about the SunCell design on Brilliant Light Power’s website and in their videos, the details of the CPV array design are not given. For the purposes of my analysis of the engineering challenges for the array design, I am making these assumptions:
- The design of the triple layer CPV cells will be based on a modified and scaled version of an existing commercial concentrator solar cell made by Masimo (Part No. BFG3-2BB-1010-A10).
- The CPV cells will be fabricated on 100mm diameter MOCVD wafers, and the cells can be fabricated within 1mm of the edge of the wafer.
- From the diagrams of the SunCell it appears that the geodesic dome layout is a basic “2V” design. The “2V” design of a geodesic dome for a hemisphere would consist of 5 triangles on top, 15 triangles in a second row and 20 triangles in the third row. The diagram for the SunCell indicates that the CPV cell array would contain another row of 20 triangles to form a ¾ sphere for a total of 60 triangular cells. Of the 60 triangular cells, 15 cells would be equilateral triangles (A cells) of length A, and 45 cells would be isosceles triangles (B cells) with one side of length A and two sides of length B where A/B = 1.131.
- The 60 triangular CPV cells will be connected in series to achieve the highest output voltage at the minimum current, which will minimize losses in electrical connections and losses in wires delivering the power to its load.
- The metal grid pattern on the illuminated side of the CPV cell will have a single bus bar along the edge that will be connected to the adjacent cell in the geodesic dome array.
- The target initial “pilot” output power for the triple junction array will be 100KW as stated in the 12-6-2016 presentation.
Requirements for the CPV Cell Design
The key issues in designing the CPV cells to be used in the geodesic dome array include the optimization of the cell sizes and the modifications that are required to the existing Masimo concentrator solar cell to make it work in the SunCell. The optimum cell size for the geodesic array is the largest equilateral triangular cell (the A cell) that can be fitted on a 100mm diameter wafer. Assuming that the wafer can be patterned to within about 1mm of its edge, the largest A cell that can fit on the 100mm wafer is about 8.5cm (3.35”) on a side. The smaller B cell will easily fit on the 100mm wafer. The areas of the A and B cells are 31.3cm2 and 26.3 cm2, respectively, although this active area will be reduced by the area of a bus bar used to make the connection between the cells. The cell sizes also may be reduced by the need to have a space on the bus bar edge of the cells to make a connection from the cell to the backside of an adjacent cell.
The first issue with scaling a cell size from a little more than 1cm2 to 31cm2 is the defect density in the MOCVD wafer growth. About 50 1cm2 cells can be fabricated on a 100mm wafer. If there is a defect in the MOCVD growth (or other defect introduced in fabrication) these defects may cause the cell performance to be so poor that the cell is rejected as a defective cell. Similar defects in the large area CPV cells fabricated for the SnCell will degrade their performance; however, it would not be economical to reject many of the large area cells as defective. Therefore the expected spread in the efficiencies of acceptable large area CPV cells will be much larger than the spread shown on the data sheet for the existing 1cm2 cell. The average efficiency of the large area CPV cells is expected to be lower than that achieved on the existing 1cm2 cells, but it would be difficult to predict how much lower without knowing the defect density associated with fabricating the smaller cells.
The single biggest factor in scaling the existing 1cm2 Masimo concentrator solar cells up to a much larger area CPV cell is the need to scale up the metal conductor thickness to handle the higher currents. A triangular cell that is to be connected to an adjacent triangular cell is not the ideal shape for a CPV cell because the metal grid line length will vary from close to zero at the ends of the bus bar to the height of the triangular cell at the center of the bus bar. The metal thickness should be scaled to the length of the longest conductor length (distance to a connected bus bar) which is 0.5cm for the existing cell. (This assumes that electrical connections are made to both bus bars in an application of the existing device.) It’s hard to say what the metal configuration will be on the triangular cells, but the minimum longest conductor length in the metal grid would be the height of the triangular cell, about 7cm, so the metal thickness should be scaled by 7cm / 0.5cm = 14. Furthermore, the existing device is rated at a maximum current density of 16A/cm2 at a 1200X illumination concentration. Assuming the scaled CPV cells have a similar 3.0 volt output/cell, the cell current density will have to be more than 21A/cm2 to achieve 100KW of output power. The metal thickness should be increased by another 30% to accommodate the higher current densities. Therefore, the metal thickness should be increased by a factor of 18 times the thickness of the metal used on the existing devices. My guess is that it will take a considerable development effort to increase the metal thicknesses by this factor on the triangular CPV cells both from difficulties in patterning the thick conductor lines and in maintaining a low stress in the thick metal lines so the cells are not bowed by the stress in the metal lines. The backside metal thickness would need to be increased by a similar factor. Increasing the back metal thickness should not be much of a challenge other than to make sure the thicker metal has low stress and does not bow the CPV cell. (Note: If the thickness of the front side metallization was 200µm, about 1/3 the cell thickness, the voltage drop along the center grid fingers would be about 0.3V for 10% metal coverage of the front surface. Ideally, the front metallization voltage drops would be less than 0.1V.)
One other issue with triangular shaped CPV cells is that if the cell is laid out with minimum length metal lines (straight lines perpendicular to the bus bar) the current will not flow into the bus bar uniformly (near zero at the ends, and very high in the center). Non-uniform current flow into the bus bar will increase the difficulty of achieving low resistance connections between the cells.
Another issue with the very thick metal grid lines on the illuminated side of the CPV cells is that with the cell’s close proximity to the blackbody radiator the thick metal lines may shadow a significant portion of the illumination coming from the blackbody radiator. This issue can be resolved by making the sides of the metal lines highly reflective and maintaining that reflectivity through the life of the CPV cell.
The geodesic dome layout for the CPV array is probably the best way to form a nearly spherical structure out of similar geometric shapes (triangles) that can be connected at adjacent edges to form a series connected array of cells. However, forming a geodesic dome array with series connected CPV cells requires the fabrication of 4 different cells, 1 A cell and 3 B cells. The A cell has all sides the same length so the only A cell design is required to have a bus bar along just one side of the cell. However 3 layouts are needed for the B cells. One layout has a bus bar connection on the “a” side; one layout has a bus bar connection on the “b” side to the right of the “a” side; and the final layout has a bus bar connection on the “b” side to the left of the “a” side.
Other modifications to the design of the existing concentrator solar cell to optimize it for use in the SunCell system include: 1) optimize semiconductor thicknesses for operation in 3000-3500ºK blackbody radiation spectrum rather than the solar spectrum, 2) modify semiconductor layers as necessary to increase operating current densities by a minimum of 30%, and 3) scale up the size of the bus bar to accommodate a very low resistance connection to the adjacent cell. It should be a simple engineering modeling problem to make the adjustments to optimize the layers in triple junction cell for 3000-3500ºK blackbody radiation. Since the original Masimo concentrator solar cell was no doubt designed for operation at the maximum possible solar concentration (current density) and minimum internal series resistanc, it is really hard to see how additional optimization can increase the current density by at least 30%. The plan the increase the blackbody radiator to 3500ºK and operate the CPV cells at a 130% higher current density than the existing Masimo concentrator solar cells may be a much harder task than is anticipated. The width of the bus bar on the CPV cells should be determined experimentally by the resistance requirements for the connections between the cells. (See the next section.)
Other Issues in Building the CPV Cell Array
Electrical Connections between the Cells
Probably the next biggest issue for the CPV cell array design, after designing the actual cells, is how to make reliable and very low resistance electrical connections between the cells. If each cell has an output voltage of about 3.0 volts and the cells are series connected, the array of cells has to operate with an output voltage of 180V and a current of 556A to produce 100KW of power. Assuming a loss of 1% of the system operating voltage is tolerable in connecting series resistance drops, the maximum resistance allowable per cell connection is .01 * 180V / 556A / 60 connections = 54 micro-ohms per connection. Achieving a low resistance connection is further complicated by the non-uniform current flow into the front side bus bar.
Another issue with the cell to cell connections is how to connect the bus bar of metal grid on the illuminated side of the cell to the backside metallization of the adjacent cell. Space for the connection between the cells will have to be built into the design of the cell. Another issue with connections between the cells is that the metal used to connect the cells will be exposed to the illumination and may absorb a significant amount of heat from the blackbody radiator. This in turn will cause thermal expansion or the metal potentially putting a significant stress on the electrical connections and the CPV cells. A thermal analysis should be done to determine the expected temperature rise in the electrical connection. If possible the design of the electrical connection should accommodate some thermal expansion of the connecting metal without stressing the CPV cells.
One final issue with the electrical connection is that edge of the cell having the backside contact can not be directly connected to the heat sink. The active cell under this contact will get hotter than the portion of the cell connected directly to the heat sink. A thermal analysis should be done to verify the non-heat-sinked edge of the cell only gets a few degrees hotter than the heat-sinked portion of the cell. A significant heat rise in this portion of the cell could lead to an early degradation of the cell’s output power.
Available Power Calculation
In the 12-6-2016 video the available light output power from the 6 inch diameter sphere at a temperature of 3000ºK was given as 330KW. Indeed, the calculation for the, emitted radiation of a 6 inch diameter spherical black body radiator having an emissivity of 0.99 is 331KW. However, the radiator used in the SunCell is not a full sphere, but rather more like a ¾ sphere. Also, it might be more reasonable to assume an emissivity value of some what less than 0.99, perhaps 0.95. The actual available light power would be about 238KW at 3000ºK and about 442KW at 3500ºK assuming the blackbody radiator is a ¾ sphere with a .95 emissivity. If a CPV array could be built with an efficiency of 0.40, the maximum available output power from the system would be about 100KW at 3000ºK and about 175KW at 3500ºK.
Nonuniformity of Light Output Due to Flange on the Graphite Black Body
The actual blackbody radiator is not a sphere, but rather ¾ of a sphere with a large flange connecting the top hemisphere to the bottom base. Assuming that the cells in the geodesic array are connected in series to maximize the output voltage, each cell needs to see the same illumination intensity to equalize the current flowing through each cell. However cells near the flange (40 of the 60 array cells) will see a different intensity than the top 20 cells which should see fairly uniform light from the spherical region of the black body radiator. It is expected that the flange will be cooler than top portion of the black body sphere and therefore emit less light. However, the cells near the flange will be much closer to flange surface (~1.3”) than the top cells are from the spherical surface (~2.3”) and therefore could potentially see a much higher light intensity even if the flange is a little cooler. Cells near the flange potentially could have a large non-uniformity of light intensity over the cell area, which could severely degrade the cell performance. If the light intensity varies too much within cells that are near the flange to achieve a reasonable cell performance, the black body radiator will have to be redesigned to eliminate the effect of the flange. This issue can be evaluated before the CPV cells are added to the system just by measuring the light intensity at the positions where the cells will be located in the final design. It would be desirable to have the light uniformity to be within 10% over the entire CPV cell array, including the cells near the flange. (Note that because of the way the triangular cells are laid out, some cells will only have a tip of the cell near the flange, while the adjacent cell will have a large area near the flange.) The primary concern with localized regions of high illumination intensity is shortening the life of the CPV cell.
Loss of Usable Power in A Cells
The geodesic dome is assumed to consist of 15 A cells and 45 B cells with the A cells being about 20% larger than the B cells. Assuming the CPV array is wired with all cells connected in series (and all connections are between adjacent cells), the same current must flow through all cells. This means that the 15 A cells will not be operating at their maximum output efficiency. Their 20% lower output current will be partially offset by an increase of about 8% in the A cell output voltage, resulting in their power output being about 12% less than their maximum output. The overall loss in system efficiency will be about 3%.
Cooling the CPV Cell Array
According to the 12-6-2016 video presentation, the CPV cells are to be water cooled by a system radiator that pumps 40ºC water to the cells and receives water at 80ºC back from the cells. If it is assumed that the array is outputting 100KW at 40% efficiency, then the cooling system needs to dissipate 150KW. If my calculations are correct, a water flow of 14.2gal/min (or .24gal/min to each cell) will be needed to remove 150KW from the CPV cell array with a 40ºC rise in the water temperature. It would seem to be favorable to have all cells operating at the temperature, but this would require each cell to be supplied with its own cooling line from the radiator outlet manifold and then another outlet line from each cell back to the radiator input. It will be interesting to see if some of the cooling lines to the CPV cells are series connected to reduce cooling plumbing at the expense of not maintaining all cells at the same temperature.
Another issue with the CPV cell cooling is that the cooling system needs to be able to handle the full 250KW output from the blackbody radiator for a short time in case the load is accidently removed from the CPV cell array output. Hopefully, the SunCell will be designed with sensors that will shut down the system if the load is lost. Also, the system needs to be designed to shut down immediately if the cooling water system fails. It would be desirable to detect a cooling water failure quickly enough that the CPV cell array would not be damaged from overheating.
One additional comment on the water cooling system is that it might have to provide 25-50% more cooling than required just by the CPV cell array. A lot of wasted heat will have to be dissipated from the lower portion of the blackbody radiator. Other portions of the system also may need some active cooling. One concern is that size of the radiator in the current SunCell design may be too small to dissipate the >150KW of excess heat for an electrical output of 100KW.
Improving the SunCell Efficiency By Reflecting the IR Spectrum Back to the Blackbody
The 12-6-2016 video states that the system efficiency can be improved by reflecting the IR portion of the spectrum back to the blackbody radiator. While this is a true statement, the efficiency of the CPV cells will not be improved and the power output of the SunCell will not be increased by reflecting the IR spectrum back to the blackbody since the maximum allowable current density in the CPV cells will be the limiting factor in the available output power. The real benefit of reflecting the IR spectrum back to the blackbody radiator will be a small reduction in the cooling requirements for the CPV cells.
Brilliant Light Power’s CPV Cell Development Plan and Target Output Power Goals
Brilliant Light Power seems to have a reasonable CPV cell development plan of first starting with single junction CPV cells and eventually moving to high efficiency triple junction CPV cells. My assumption is that BLP will work with Masimo to develop the much thicker metallization required for large area CPV triangular cells on the lower cost single junction cells. These same low cost single junction CPV cells can be used to develop the technology required make the connections between the cells. It makes a lot of sense to finalize the development of thick CPV cell grid metal (and thick backside metal) and the development of the electrical connection technology on lower cost cells before applying that technology to the more expensive triple junction CPV cells. It is not clear what benefit is derived by having an interim step of using double junction CPV cells between the initial phase of single junction CPV cells and the pilot phase that uses the triple junction CPV cells.
As discussed earlier the ¾ sphere blackbody radiator used in the SunCell will have a maximum light output of only about ¾ of what was claimed in the 12-6-2016 video presentation, and therefore the maximum electrical output for 0.40 efficiency CPV cells is really only about 100KW for a blackbody radiator temperature of 3000ºK and about 175KW for a blackbody radiator temperature of 3500ºK. Even to output 100KW the new CPV cells must be designed to run at a 31% higher current density than Masimo’s existing concentrator solar cell (and a130% higher current density to output 175KW). It’s not certain that the initial triple junction CPV cell can be designed to operate reliably with even a 31% increase in the allowable current density. If the CPV cells can only operate reliably at the maximum rated current density as Masimo’s concentrator solar cell, the SunCell would have a maximum output power of no more than 76KW and the blackbody operating temperature would have to be lowered to about 2840ºK.
The primary reliability concerns with the CPV cell array are degradation of the cells over time (including catastrophic failure of the cells) and degradation of the electrical connections between the cells. It would probably be a good idea to verify the reliability of the existing Masimo concentrator solar cells by operating them at their maximum rated current density, or perhaps even force failures by operating them at about 30% above their maximum rated current to determine the failure mechanisms for stressed operation. Even if Masimo concentrator solar cells are very reliable, it’s possible that changes are made to some of the semiconductor layers in the CPV cell in an attempt to increase to maximum operational current density will introduce some new reliability issues. The electrical connections are a potential reliability issue because of the high current flowing between cells. It would be good to use a technology for making connections to the cell bus bar that has already been proven reliable for high currents.
Another issue with the SunCell’s CPV array is that localized areas on individual cells may be subjected to illumination intensities above the spec limit of the cells. The final design of the CPV cells should be checked to verify that localized high intensities will not cause rapid degradation of the entire cell.
One probable method of operating of the SunCell system is to adjust blackbody temperature to maintain consistent output power. If the output power decreases due to CPV cell array degrading or the connecting resistance increasing, the operating method of increasing the blackbody temperature to maintain output power may just increase the rate at which the CPV cells or the electrical connections degrade. An independent method of monitoring the CPV array is needed.
It is going to be quite an engineering challenge for Brilliant Light Power’s SunCell to achieve a “pilot” goal of an output of 100KW with the blackbody radiator operating at a temperature of 3000ºK due to the difficulty of developing a thick metal technology and increasing the allowable CPV cell current density (by about 30%). If higher current density cells can not be developed, it still should be possible to achieve an output power of about 76KW, assuming that the existing Masimo concentrator solar cells are reliable at their maximum rated concentration (1200X). However, even operating the SunCell system at 76KW will require the successful development of a thick metal technology. It might be desirable to initially target an output power of about 50KW until reliability can be demonstrated a higher power levels. There are numerous other engineering challenges for the development of a reliable SunCell CPV cell array including developing a method to achieve low resistance cell to cell connections, verifying the blackbody illumination intensity is uniform, developing a cooling system for the CPV cells and other hardware, verifying system reliability, and the other issues noted in this post. It will be interesting to follow the engineering solutions to these challenges.