Concentrated Solar Power

How concentrated solar power works

Start with a magnifying glass and a sunny afternoon. The lens gathers sunlight falling across its whole face and crushes it into a bright point hot enough to scorch wood. Concentrated solar power is that trick built at the scale of a power station: replace the glass with a field of thousands of mirrors, and replace the scorched wood with a receiver full of fluid you genuinely want to boil.

The plant has four jobs, and they run in a chain. Collect: mirrors track the sun across the sky and reflect its direct beam onto a small target. Absorb: a receiver — a black-coated tube or a tower-top boiler — turns the concentrated beam into heat in a working fluid. Store (optional but pivotal): the hot fluid, or heat transferred from it, charges a tank of molten salt that can sit for hours. Convert: the heat boils water to steam (or heats a gas), and a turbine-generator turns that into electricity — the exact same Rankine or Brayton power block you'd find in a coal or gas plant. The only difference from fossil generation is the heat source: a mirror field instead of a furnace.

The crucial physical fact is that only direct sunlight can be focused. A mirror images the sun's disc; it can redirect the collimated beam (direct normal irradiance, DNI) to a focal point, but diffuse skylight scattered by clouds and haze arrives from every direction and cannot be concentrated. That single constraint is why CSP lives in cloudless deserts and why a hazy temperate sky that's fine for flat PV panels is worthless to a concentrator.

The governing physics: concentration, cosine, and Carnot

Three numbers decide whether a CSP design works: how much you concentrate, how much sunlight your tilted mirrors actually catch, and how efficiently the turbine converts the resulting heat.

Concentration ratio is the aperture area gathering light divided by the receiver area absorbing it. It governs the achievable temperature, because a hotter receiver also radiates more heat away (Stefan–Boltzmann, ∝T⁴); a higher concentration ratio lets you reach a higher equilibrium temperature before re-radiation losses dominate.

Geometric concentration ratio:
  C = A_aperture / A_receiver

  Parabolic trough / linear Fresnel (line focus):  C ≈ 30 to 100 suns  → ~400 °C
  Power tower / parabolic dish    (point focus):    C ≈ 500 to 1000+    → 565 to 1000+ °C

Thermodynamic ceiling on C (for a perfect concentrator):
  C_max,3D = 1 / sin²(θ_sun) ≈ 46,000   (point focus)
  C_max,2D = 1 / sin(θ_sun)  ≈ 215      (line focus)
  where θ_sun ≈ 0.267° is the sun's angular half-width.

The cosine effect is the dominant optical loss in a tower field. A heliostat must bisect the angle between the incoming sun and the tower, so its mirror face is rarely pointed straight at the sun. The usable reflected energy scales with the cosine of half the sun-to-tower angle. Mirrors on the far side of the tower from the sun lose a third or more of their light to cosine alone — which is exactly why northern-hemisphere tower fields bulge to the polar (north) side, where cosine losses are smallest.

Carnot sets the prize for going hotter. A heat engine's ideal efficiency is η = 1 − T_cold/T_hot (absolute temperatures). Push the receiver from a trough's 390°C (663 K) to a tower's 565°C (838 K) against a 40°C (313 K) condenser, and the Carnot ceiling climbs from 53% to 63%. Real steam cycles capture a fraction of that — about 38% for a trough, 42% for a tower — but the gain is real money, and it's the whole reason towers exist. Push receivers toward 700°C+ with supercritical CO₂ Brayton cycles and you chase 50% power-block efficiency.

Overall solar-to-electric efficiency is a product of stages:

  η_solar→elec = η_optical × η_receiver × η_cycle × η_parasitic

  Tower example:
    η_optical   ≈ 0.55   (cosine, shadowing, blocking, reflectance, spillage, attenuation)
    η_receiver  ≈ 0.88   (convection + re-radiation losses from the hot receiver)
    η_cycle     ≈ 0.41   (565 °C steam Rankine)
    η_parasitic ≈ 0.90   (pumps, tracking drives, salt heating)

  η_net ≈ 0.55 × 0.88 × 0.41 × 0.90 ≈ 0.18  → ~18% annual net

Worked example: sizing a 100 MW power tower

Take a 100 MW (electric) molten-salt tower in a Mojave-class site and run the numbers from sunlight to grid.

Design-point DNI:        950 W/m²  (clear noon)
Net solar-to-elec eff:   18%
Target net output:       100 MWe

Required intercepted solar power at the field:
  P_solar = P_elec / η_net = 100 MWe / 0.18 = 556 MW(thermal-equivalent at the mirrors)

Required mirror aperture (at design DNI):
  A_mirror = P_solar / DNI = 556e6 W / 950 W/m² ≈ 585,000 m²  (~0.59 km² of glass)

Heliostat count (115 m² mirrors, typical):
  N = 585,000 / 115 ≈ 5,100 heliostats

Receiver flux:
  A receiver of ~1000 m² absorbing ~600 MWth runs at an average
  flux near 0.5 to 1.0 MW/m² — among the highest steady heat fluxes
  in any industrial process, which is why receiver tubes are Inconel.

Now add storage. To run 100 MWe for 10 hours after sunset you must bank 1000 MWh of electrical output, which at 41% cycle efficiency is about 2440 MWh of thermal energy in salt:

Thermal energy to store:
  Q = 1000 MWh_e / 0.41 ≈ 2440 MWh_th = 8.78e12 J

Salt mass (solar salt, c_p ≈ 1.53 kJ/kg·K, ΔT = 565 − 290 = 275 K):
  m = Q / (c_p · ΔT) = 8.78e12 / (1530 · 275) ≈ 20.9 million kg ≈ 21,000 tonnes

That is the real reason these tanks are enormous: Crescent Dunes
stored ~32,000 tonnes of salt for ~10 hours at 110 MWe.

This is also why CSP storage is cheap per kWh: a tank of cheap nitrate salt and steel holds energy at roughly 20 to 30 $/kWh of thermal capacity, an order of magnitude below lithium batteries — but only if you've already paid for the turbine to convert it.

Collector types compared

	Parabolic trough	Power tower	Linear Fresnel	Parabolic dish
Focus geometry	Line	Point (central)	Line	Point
Tracking	1-axis	2-axis (each heliostat)	1-axis	2-axis (whole dish)
Concentration	30 to 100	500 to 1000+	10 to 40	1000 to 3000
Operating temp	~390 °C	565 °C (to 1000+ °C)	~300 to 400 °C	650 to 750 °C
Power block	Steam Rankine	Steam Rankine / sCO₂	Steam Rankine	Stirling engine (per dish)
Storage friendly	Yes (oil→salt or direct salt)	Yes (direct molten salt)	Marginal	No (each dish is tiny)
Land use	Moderate	High (big spread field)	Lowest (mirrors near ground)	Modular
Maturity	Most installed GW	Highest-efficiency frontier	Cheapest mirrors, niche	Demo / off-grid only
Flagship plant	SEGS, Solana, Noor I–II	Ivanpah, Crescent Dunes, Noor III	Puerto Errado 2	Maricopa, Tooele

Molten-salt storage and the two-tank loop

The feature that distinguishes CSP from every other solar technology is that its energy is already thermal, and thermal energy is cheap to bank. The dominant scheme is the two-tank molten-salt loop:

• Solar salt. A eutectic of 60% sodium nitrate and 40% potassium nitrate (NaNO₃/KNO₃). It melts at ~220°C, is liquid and pumpable from ~290°C, and is stable to ~600°C before it begins to decompose. Above that it releases NOₓ and corrodes steel — the hard upper temperature wall that caps single-tank tower temperatures near 565°C.
• Cold tank. Holds salt at ~290°C, kept liquid by electric trace heating. Pumps draw from here and send salt up to the receiver.
• Receiver charging. Concentrated flux heats the salt to ~565°C as it passes through the receiver tubes.
• Hot tank. An insulated tank stores the 565°C salt with losses of only ~1°C per day — exceptional retention for a tank holding tens of thousands of tonnes.
• Discharge. On demand, hot salt is pumped through a steam generator, gives up its heat to make turbine steam, and returns to the cold tank — closing the loop.

The freezing risk is the operational nightmare: let solar salt drop below ~220°C anywhere in the piping and it solidifies into a plug that can crack pipes and take a plant offline for months. The 2016 hot-tank salt leak and subsequent freezing problems at Crescent Dunes idled the plant for the better part of a year, a textbook lesson in how thermal storage's cheapness comes with brittle operational margins.

Real-world systems and figures

Plant	Type	Capacity	Storage	Notes
SEGS (California, 1984–1990)	Trough	354 MW total (9 plants)	None (gas backup)	The pioneer fleet; proved trough CSP at utility scale
Ivanpah (California, 2014)	Tower ×3	392 MW	None	Direct steam towers; uses gas for morning warm-up
Crescent Dunes (Nevada, 2015)	Tower	110 MW	~10 h molten salt	First large salt tower in US; plagued by salt-freeze outage
Noor Ouarzazate (Morocco, 2016–2018)	Trough + Tower	~510 MW	3 to 7+ h salt	Among the largest CSP complexes in the world
Gemasolar (Spain, 2011)	Tower	20 MW	15 h salt	First tower to run 24 h continuously on stored salt
Cerro Dominador (Chile, 2021)	Tower	110 MW	17.5 h salt	Extreme-DNI Atacama site; near-baseload solar
Mohammed bin Rashid (Dubai, 2023)	Tower + Trough	700 MW CSP	~15 h salt	World's tallest solar tower (~262 m) at commissioning

Across the fleet, levelized cost of electricity for new CSP-with-storage sits in the rough range of $0.07 to $0.12/kWh in the best sites — competitive only where long-duration dispatchability is the requirement. For pure daytime energy, utility PV at $0.02 to $0.04/kWh wins decisively.

When CSP makes sense (and when it doesn't)

• Very high DNI is non-negotiable. Build only where direct normal irradiance exceeds ~2000 kWh/m²/year — the Mojave, Atacama, Sahara fringe, Australian interior, Rajasthan. Below ~1800 the economics collapse.
• You need long-duration, dispatchable clean power. CSP's edge is 8 to 15+ hours of cheap thermal storage to deliver evening-peak or near-baseload output. If you only need daytime energy, choose PV.
• Scale is large. The turbine power block has fixed overhead; CSP economics only work at the tens-to-hundreds-of-MW scale, never rooftop.
• High-temperature process heat is the goal. An emerging niche: feeding 400 to 1000°C heat directly to industrial processes (desalination, mineral processing, hydrogen, even cement) where electricity-then-resistance-heating would waste exergy.

Pick something else when the sky is hazy or cloudy (use PV, which still harvests diffuse light), when you need a small or modular installation (PV), or when you need only daytime energy and storage isn't required (PV undercuts CSP roughly 3-to-1 on daytime kWh).

Common misconceptions and failure modes

• "CSP is just bigger solar panels." No — there are no semiconductors anywhere. CSP makes heat and runs a turbine; it is thermodynamically closer to a coal plant (minus the coal) than to a PV array. This is also why CSP, like any thermal plant, needs cooling water or dry cooling at the condenser.
• "More mirrors always means more power." Far-field heliostats suffer steep cosine, atmospheric-attenuation, and spillage losses; past a certain field radius each added mirror returns less than it costs. Field layout is an optimization, not just "add glass."
• Salt freezing. The dominant operational failure. Trace heating, freeze-protection recirculation, and careful drain-down procedures are mandatory; a single solidified plug can idle a plant for months (Crescent Dunes, 2016–2017).
• Receiver thermal fatigue. Receiver tubes cycle from ambient to ~565°C every single day and endure flux gradients across their face. Thermal-fatigue cracking of Inconel receiver panels is a known life-limiter; flux is actively "aimed" by steering heliostats off-point to flatten hot spots.
• Avian flux hazard. Real for direct-steam towers: birds flying through the concentrated flux above Ivanpah were singed ("streamers"). Modern designs mitigate with standby-aim-point dispersal so the off-tower flux never forms a lethal hot zone in mid-air.
• Mirror soiling. Desert dust on mirrors can cut reflectance several percent per week; large robotic and waterless wash fleets are a continuous operating cost that PV (with far less surface per watt) largely avoids.