Answer by Joel Jean:
The future of solar?
Let’s start with the future of the world.
To properly frame any discussion about the future of any kind of energy, we need to keep a few facts in mind:
(1) Climate change is a real and present threat to the future of human life and all other life on Earth. Suppose we want to minimize our (children’s) risk of encountering the very worst impacts of climate change. That translates to reducing global greenhouse gas (GHG) emissions ~80% by 2050. Since ~60% of global emissions stem from energy use, we need to deploy low-carbon energy technologies at massive scale, starting yesterday.
***Details: Below is a plot of typical ranges of lifecycle (“cradle to grave”) emissions (or carbon intensity) of different energy technologies (units: grams of CO2-equivalent per kilowatt-hour (kWh) of electric output). The green dashed line is a projection of the average U.S. carbon intensity required to cut emissions by 80% (from 1990 levels) by 2050 and keep global warming below 2ºC. . Wind and concentrating solar power (CSP) are by far the lowest . Geothermal and solar photovoltaics (PV) are comparable. Hydro and nuclear are higher but in some cases still within range of the 2050 target . Natural gas , coal , and even coal with carbon capture and storage (CCS)  are far above the acceptable limit.
(2) Solar is by far the largest energy resource available on Earth—renewable or otherwise. All other energy sources—aside from nuclear, geothermal, and tidal—come from sunlight. Fossil fuels are just solar power integrated over millions of years using dinosaurs (and other carbon-based life forms) as batteries. Wind and wave power is merely solar power absorbed unevenly across the Earth’s surface, leading to thermal gradients and mass flow. Among low-carbon energy sources, only solar, wind, and possibly nuclear can reach the terawatt (TW)-scale deployment needed to satisfy ever-growing global energy demand (currently ~17 TW average).
(3) Solar photovoltaics is growing fast—faster than any other energy technology. Cumulative installed PV capacity worldwide has doubled every two years (43% CAGR) since the year 2000, reaching ~200 gigawatts-peak (GWp) in 2014. This Moore’s Law-like growth shows no sign of slowing, though slow it must, as naive extrapolation leads us to some untenable conclusions: If PV capacity were to keep growing at the current rate, solar panels would satisfy all world electricity demand within a decade, cover the Earth by 2050, and form a Dyson sphere around the sun just after 2100.
Just for fun, here’s the naive extrapolation:
That said, solar PV accounted for only ~1% of our total electricity consumption last year, so there’s clearly a lot of headroom left.
OK. So now we know a few things: Climate change is happening, we need lots of low-carbon energy to stop it, solar is one of our only practical options, and solar PV is growing faster than anyone ever imagined.
But how do we turn sunlight into useful energy? What’s the future of PV? And are there other non-PV solar technologies in the R&D pipeline?
Let’s talk about the technology.
Solar Photovoltaics (PV)
Solar photovoltaics (aka “solar cells”) are by far the leading solar technology in terms of total deployment*. PV is quite nice: It’s truly modular (a single PV module is no less efficient than a huge array), it operates silently and at low temperatures, and it doesn’t require much maintenance over its 25+ year lifetime.
*Aside from solar heaters, which are used widely in China for heating domestic water and in the U.S. for keeping swimming pools warm. Solar heating can’t be compared directly with PV since its output is heat [GW-thermal] rather than electricity [GW-electric].
We typically name PV technologies by the material (or material class) used to absorb light: crystalline silicon (c-Si), gallium arsenide (GaAs), hydrogenated amorphous silicon (a-Si:H), cadmium telluride (CdTe), copper indium gallium diselenide (CIGS), copper zinc tin sulfide (CZTS), organics, perovskites, or colloidal quantum dots (QDs), to name a few.
It’s convenient to think about these technologies in terms of material complexity, which corresponds roughly to the number of atoms in a unit cell, molecule, or other repeating unit of a material [4,5]. Material complexity is related to the degree of disorder at the nanoscale. For current PV technologies, higher material complexity often translates to lower technological maturity, materials use, processing temperatures, and processing complexity. These traits often open up new applications by enabling novel technical attributes, such as visible transparency, flexibility, and new form factors.
With PV technologies, it’s really hard to predict what will be the long-term winner.
Crystalline silicon (c-Si) is king today, with ~90% of the global PV market, and I believe it will continue to dominate for at least the next decade. Silicon PV is abundant, efficient, reliable, and proven, but it absorbs light poorly. That drawback results in thick, heavy, inflexible solar cells and modules with relatively high manufacturing costs. For silicon, there’s not much room to grow in terms of cell efficiency (25% current lab record), although production modules continue to improve: Typical modules are 16-21% efficient, with multicrystalline (mc-Si) technologies at the low end and single-crystalline (sc-Si) technologies at the high end.
Today’s commercial thin-film (TF) PV technologies, including CdTe, a-Si:H, and CIGS, overcome some of the challenges of c-Si—they use much less material and can be made at relatively low cost with high efficiency. CdTe dominates the thin-film market with simple manufacturing and high efficiency (21% cell record, production modules up to ~15%) but has major intrinsic scaling issues: Tellurium is about 4x less abundant than gold in the Earth’s crust, and it’s hard to extract from copper ores. Amorphous silicon is abundant, cheap, and flexible, but its maximum efficiency (13.4% current cell record) is likely too low to compete with crystalline silicon. CIGS is efficient (21.7% cell record, modules up to ~15%) but tough to make reliably, and it also runs into materials scaling issues with indium, gallium, and selenium.
In the PV R&D community, we pursue emerging thin-film PV technologies, such as perovskites and quantum dots, for 2 primary reasons: (1) They may someday be able to reach a lower cost per watt than silicon and current thin films due to simpler manufacturing and reduced materials use, and (2) They offer new functionality, including transparency, flexibility, and extremely light weight, and may open up new applications for PV.
Examples of emerging thin films include CZTS, organic PV, dye-sensitized solar cells (DSSCs), perovskites—which have largely swallowed the organic and dye-sensitized PV R&D communities—and colloidal quantum dot PV (QDPV). Perovskites are extremely promising, with impressive material characteristics and cell efficiencies improving at an unprecedented rate (up to over 20% in ~3 years). But we shouldn’t get too excited yet—there’s still a lot of work to be done, in particular on lifetime, air and water stability, and new cell designs. Although still relatively inefficient (~9% cell record), QD solar cells are also improving fast and can be processed entirely at room temperature from solution, which may someday lead the way to the fabled “solar paint” (which, contrary to popular belief, does not yet exist in any practical form).
See the bottom of this post for several FAQs about solar PV.
Concentrating Solar Power (CSP), aka Solar Thermal
CSP uses mirrors or lenses to concentrate sunlight onto a tank of molten salt or other working fluid, which is then used to boil water and drive a steam turbine. CSP systems have been used for decades, but they only work effectively in places with high direct radiation*—such as the southwestern U.S., southern Europe, northern Africa, and other locations near the equator.
*The MIT Future of Solar study recently analyzed the cost of CSP in Worcester, Massachusetts, and… nope, not a chance.
CSP is not modular like PV—high temperatures require many mirrors over large areas, and turbines are much more efficient at large scale—so unfortunately you won’t have a solar thermal generator on your roof anytime soon. Here’s a picture of the new Ivanpah CSP plant in the Mojave desert (opened Feb. 2014), with over 340,000 mirrors:
Global CSP deployment today is lower than PV deployment by about 2 orders of magnitude. As for the future, CSP will become more and more important as penetration of solar and wind increases, because it can potentially overcome the natural intermittency of those resources (discussed further below) using built-in thermal energy storage (on the time scale of 4-8 hours).
Sunlight can catalyze chemical reactions that use water and CO2 to produce liquid or gaseous fuels (e.g., hydrogen, methane, various alcohols and hydrocarbons). These “solar fuels” have a unique role in a future low-carbon energy economy, since they could help decarbonize transportation—especially by air and sea, where electric-powered transport may be impractical. Solar fuels could also become a key energy storage technology for counteracting solar intermittency.
All that said, solar-to-fuels technology is far from proven—my MIT colleague Bob Jaffe would say that there are many “tooth fairies'” worth of fanciful technological advances that still need to be made to get solar fuels to market at competitive cost.
Technologically, the future of solar energy looks bright.
So what’s stopping solar from taking off? And what might limit it in the future?
Well, there are a few things that might be worth thinking about: cost, intermittency, and scaling issues (i.e., materials and land use). Let’s focus on PV for now.
Solar PV is getting cheaper by the month. Average system costs in the U.S. are now below $2/W for utility scale (>1 MW) systems and just over $3/W for residential (usu. <10 kW) systems . And that doesn’t include subsidies (e.g., 30% federal investment tax credit (ITC), accelerated depreciation, various state renewable portfolio standards (RPSs), and net metering in some places), which in a competitive market can reduce the effective price for consumers. In regions with high direct sunlight, such as southern California, CSP is cost-competitive with PV .
So what’s the problem?
System cost [$/W] isn’t a complete metric. Solar energy technologies will only take off if they can produce and deliver electricity more cheaply than alternatives.
The usual way of comparing the cost of generating electricity with different technologies is the levelized cost of energy (LCOE), in units of $/kWh (usu. “cents per kilowatt-hour”) or $/MWh.
The LCOE of a power plant includes upfront capital costs, operation and maintenance costs (including fuel), subsidies, an assumed discount rate, and the total electricity produced over the plant lifetime. Typical LCOEs are $0.06-0.08/kWh for coal, natural gas, and hydro and ~$0.10/kWh for nuclear.
Most LCOE estimates for unsubsidized solar PV today range from ~$0.10/kWh to $0.40/kWh, depending on the location (i.e., amount of sunlight) and type of system (large utility-scale systems are cheapest, small residential systems most expensive). In places like Hawaii, where fuels are hard to come by and hence expensive*, the LCOE of conventional fossil fuel generation is much higher and more volatile, and solar is much more attractive. But LCOE alone isn’t the whole story either.
*Because it lacks local fossil fuel resources, Hawaii uses imported oil to generate most of its electricity, unlike the rest of the U.S. As a result, electricity prices in Hawaii are highly correlated with global oil prices.
Even in places where solar costs more than coal (and gas and nuclear)—and even when subsidies are included—you can sometimes still save money by putting solar panels on your roof. Why?
Well, your solar electrons aren’t competing directly with the grid’s electrons on cost. Electrons from a coal plant (e.g.) are purchased by a utility and transmitted to your house on the grid. The utility can’t give you the electrons at the same price they were bought for: It has to make a profit, and it has to amortize the costs of building and maintaining the grid. So the actual retail price of grid electricity—what you see on your electricity bill—is substantially higher than the underlying LCOE. Under current regulations (in many states), your rooftop solar electrons can compete with grid electrons at the retail price, and if your PV system makes more electrons than you use, you can sell them back to the utility at the retail price—a practice called net metering. It’s basically a subsidy for solar and a great deal for you, although utilities don’t like it because you’re not paying your share of the grid upkeep. Future policies will likely close that loophole by forcing homeowners with PV systems to pay some fixed cost for grid use.
Fortunately, it seems likely that the cost (LCOE) of solar will continue decreasing steadily, which means that unsubsidized solar will eventually be cheaper than fossil generation in many places—especially if a price is placed on greenhouse gas emissions to account for their negative environmental externalities.
For solar PV in particular, it’s important to note that the total system cost is no longer dominated by the solar panel itself. Everything else that goes into a PV system—inverters, transformers, wiring, racking, installation labor, customer acquisition, permitting, taxes, financing, business overhead—add up to well over half of the total cost of solar PV in the U.S. today. We refer to these non-module costs as balance-of-system (BOS) costs, and in many ways, they’re harder to reduce than module costs. To realize a solar-powered future, we need to innovate and reduce BOS costs.
In most places on Earth, sunlight isn’t always available. Some of the variations in available solar energy are predictable or deterministic (e.g., diurnal and seasonal cycles and local climate), while others are unpredictable or stochastic (e.g., cloud cover and weather).
When solar is deployed at large scale (several percent of total electricity generation) on a given grid, electricity markets will likely change significantly. After a solar PV system is installed, it costs almost nothing to operate. Zero-variable-cost generation means that solar energy will be used whenever it’s available (i.e., when the sun is shining). PV will thus replace fossil-fueled generators (i.e., coal and gas) with the highest variable costs, reducing marginal electricity prices.
But the impact on actual market prices depends strongly on the generation mix—i.e., how much coal, gas, nuclear, wind, etc. is deployed on the grid: Solar tends to stop producing when the sun goes down—and when the clouds come out—so other generators are forced to ramp up (cycle) more rapidly and more often, increasing wear-and-tear and hence their operating costs.
In many places, adding solar PV without energy storage doesn’t substantially reduce the net load that must be supplied by other technologies (total demand – solar generation). It simply shifts the time of peak load slightly later into the evening, when the sun goes down and everyone goes home after work and starts watching TV, cooking, reading Quora, coding, or otherwise consuming electricity (this is the origin of the famous CAISO).
A few complementary technologies would ease the pain of intermittency. In decreasing order of current technological and economic feasibility: Enhanced grid infrastructure (e.g., improved long-distance transmission, demand response, and other smart grid concepts) could adjust demand to meet varying solar supply, or allow geographical averaging to smooth out minute-to-minute variations due to clouds. Grid-scale energy storage (e.g., pumped hydro, compressed air, or big batteries) could store energy during the day and discharge it at night. And solar fuels could someday make solar energy truly dispatchable.
Scaling: Materials and land use
The land use issue is actually not that big a deal. To satisfy all U.S. electricity demand with solar PV at average solar insolation levels, we would need on the order of 50,000 square kilometers —which sounds like a lot, until you find out that we currently dedicate ~100,000 square kilometers to producing corn ethanol satisfying only ~10% of U.S. gasoline demand.
Materials use is a bigger concern: Covering thousands of square kilometers with PV will require huge amounts of raw materials, from the elements used in solar cells to supporting commodity materials, such as steel, glass, and concrete [4,5].
Our analysis suggests that most commodities will not be major obstacles to scaling, except perhaps the flat glass used to cover today’s c-Si PV modules [4,5]. For glass, aluminum, and copper, the amount of material required to satisfy 100% of 2050 world electricity demand exceeds 6 years of current global production, which indicates that PV might eventually become a major driver for those commodity markets.
Critical elements will be limiting for some technologies:
- For c-Si, silicon is not an issue, but silver conductors will need to be replaced by copper.
- For CdTe, Te is a huge obstacle to scaling: At current rates of global Te production, we would need 1500 years to extract all the Te required to satisfy 100% of 2050 world electricity demand with CdTe PV [4,5].
- Emerging thin films are generally more scalable than current commercial technologies, since they mostly use small amounts of abundant, widely-produced elements: For example, to satisfy 100% of 2050 demand with PbS QDPV, we only need the equivalent of 23 days of global lead production and 7 hours of current sulfur production [4,5].
So what is the future of solar?
The main takeaway is that there will be more solar than you think. To spare future generations from the worst impacts of climate change, we’re going to need a lot more solar (and wind and nuclear) generation capacity—10-100 times what’s currently deployed. Today’s technologies (mostly crystalline silicon PV) can—and likely will—scale up to multiple terawatts of capacity worldwide by 2030 without any major technological advances.
The main obstacle is cost: Global PV growth thus far has largely been driven by federal and local subsidies. That said, PV is already cost-competitive with fossil fuels in some places, and system costs (and prices) continue to decline. And even though current technologies will likely plateau at some minimum sustainable cost floor, it’s clear that there are many new and exciting solar technologies in the pipeline, with many new and exciting applications to come. I can’t wait.
 J. E. Trancik and D. Cross-Call, Environ. Sci. Technol., 2013, 47, 6673-6680.
 M. Z. Jacobson, Energy Environ. Sci., 2009, 2, 148-173.
 U.S. Energy Information Agency (EIA), 2014.
 MIT Future of Solar Energy Study, 2015, in preparation.
 J. Jean et al., 2015, submitted.
*Just to be clear, the views expressed here are my own. My opinions are informed by my involvement in the MIT Future of Solar Energy Study but don’t necessarily reflect the final conclusions of that study. I encourage you to read the report when it’s released later this year (and others) and come to your own conclusions about the future of solar.
**Feel free to use the figures in this answer for educational purposes. Figures without citations are my own, and proper attribution is appreciated.
Solar PV FAQs
Does a solar panel produce more energy than it takes to manufacture it? In other words, is the energy payback time (EPBT) shorter than the lifetime of the panel?
YES! A typical silicon PV module today produces as much energy as it took to manufacture it in less than 2 years (<1 year for CdTe), and continues to operate with minimal efficiency loss for at least 25 years.
I hear solar cells can only convert 15% (or 20%) of incoming sunlight into electricity. Why are solar cells so inefficient? I mean, my body can convert a Big Mac into useful energy at 25% efficiency. Why can’t you scientists do better?
Physics is tough to beat. Thermodynamic limits (see Shockley-Queisser limit) cap the ultimate efficiency of typical solar cells based on a single material at ~31%. Using multiple materials—as in multijunction or tandem cells—boosts the theoretical maximum efficiency (e.g., to ~49% for 3 junctions). Some “exotic” approaches (e.g., carrier multiplication, hot-carrier collection, and intermediate band cells) can theoretically bypass the Shockley-Queisser limit, but none has yet achieved practical efficiency gains.
But in the end, we don’t really care about efficiency anyway; we care about the cost of energy [$/kWh]. Photosynthesis is on the order of 1% efficient at converting sunlight into chemical energy, yet the U.S. still tries to grow corn to make ethanol. That said, it’s worthwhile to work on improving the efficiency of solar cells because higher efficiencies can decrease module and system costs.
Why are production PV modules so much less efficient than record cells in the lab?
Intrinsic scaling losses: Scaling from small cells (~1 square centimeter) to large modules with multiple interconnected cells (~100 square centimeters) incurs physical scaling losses. Electrons must travel farther, increasing resistive losses. Shadowing from electrodes reduces the available light. Longer wires in modules dissipate more power, while spacing between cells reduces the module active area. The output of a module is often limited by its worst-performing cell.
Extrinsic manufacturing losses: While researchers typically target high efficiencies without much regard to cost, manufacturers may sacrifice efficiency to reduce cost, improve yield, and increase throughput. Fabrication techniques that produce high efficiencies in the lab may not scale to large areas. High-quality materials used in research labs may be too expensive for high-volume manufacturing.
Do solar cells work when it’s cloudy?
Yep, although somewhat less efficiently. Clouds and atmospheric particulates turn direct sunlight (what you see when you look right at the sun) into diffuse light (what you see when you look anywhere else in the sky, or at the ground). A solar cell doesn’t really care where photons come from—whether scattered from a cloud or transmitted straight through the atmosphere. Once a photon of a given color is inside a solar cell, it’s converted into an excited electron with the same probability as any other photon of that color.
But a solar cell DOES care about how many photons are incident on it—that is, the light intensity. Higher intensities generally give higher efficiencies. Because some direct sunlight is scattered back into space by clouds—and because more light is lost to reflection at the front surface of the panel at wide angles of incidence—the light intensity that a solar cell sees is lower on overcast days than on clear days, leading to lower efficiency and lower power output.
On a related note, concentrating solar technologies (CSP and CPV) don’t work at all when it’s cloudy: You can’t concentrate light that’s coming from all directions.
Are solar cells toxic?
Yep—but only if you eat too many of them. In all seriousness, silicon solar cells are about as benign as any other piece of technology that you probably wouldn’t eat. Cadmium telluride is a bit worse—cadmium is toxic—but the amount of Cd inside a typical solar cell is quite small. The CdTe layer used in cells today is about 1/50th the thickness of a sheet of paper, or roughly 2 µm, and it’s sealed up tight between two thick glass sheets. As for emerging technologies, you’d have to eat a lot (!) of solar cells to get lead poisoning from lead sulfide quantum dot PV or methylammonium lead halide perovskite PV.
When solar electricity becomes cheaper than current grid electricity (“grid parity”), will it completely displace all other generation technologies?
Nope. We’ll probably still need other types of generation—certainly at night, and likely during northern winters as well. Like any other free market, electricity markets generally follow the laws of supply and demand. Increasing supply of zero-marginal-cost electricity from solar PV depresses electricity prices at exactly the hours when solar is available (i.e., when the sun is shining). For any given electricity grid and cost of PV, there will likely be a natural break-even point for solar penetration—above that level of deployment, additional solar generation will no longer be profitable, and further investment in solar is unlikely. The story might change if grid-scale electricity storage gets way, way cheaper or if CSP (with built-in thermal storage) is deployed widely.