LEDs & Diode Sources

1.1What Is an LED?

A light-emitting diode (LED) is a semiconductor device that emits incoherent light when forward-biased current flows across a p-n junction. Electrons injected into the conduction band recombine with holes in the valence band, and in a direct-bandgap semiconductor a significant fraction of these recombinations produce photons with energy approximately equal to the bandgap energy. The result is spontaneous electroluminescence — narrow-spectrum, spatially diffuse, and temporally incoherent — fundamentally distinct from the stimulated emission of lasers and the broadband thermal radiation of incandescent lamps [1, 2].

LEDs convert electrical energy directly into optical energy without an intermediate thermal stage. This gives them inherent advantages in efficiency, lifetime, and controllability compared to incandescent sources. A modern high-power LED can convert over 50% of input electrical power into light (wall-plug efficiency), operate for 50,000 hours or more before significant degradation, switch on and off in nanoseconds, and be manufactured to emit at any wavelength from the deep ultraviolet (210 nm) through the visible spectrum to the mid-infrared (beyond 5 µm) by selecting the appropriate semiconductor material system [1, 3].

1.2Historical Development

The electroluminescence effect in semiconductors was first observed by Henry Joseph Round in 1907 using silicon carbide (SiC) crystals and a cat-whisker detector, and independently by Oleg Losev in the 1920s. Practical visible LEDs did not emerge until Nick Holonyak Jr. demonstrated the first red GaAsP LED at General Electric in 1962. Through the 1970s and 1980s, improvements in AlGaAs and AlGaInP material systems extended efficient LED emission from the infrared through red, orange, and yellow wavelengths, but the blue and green regions remained inaccessible with adequate efficiency [1, 4].

The breakthrough came in the early 1990s when Shuji Nakamura, Isamu Akasaki, and Hiroshi Amano developed high-quality InGaN films on sapphire substrates, enabling efficient blue and green LEDs. This achievement — recognized with the 2014 Nobel Prize in Physics — completed the visible spectrum and made white LED illumination possible through blue LED excitation of yellow phosphor coatings. The resulting solid-state lighting revolution has displaced incandescent and fluorescent lamps across residential, commercial, automotive, and display applications worldwide [1, 4].

1.3LEDs vs Other Sources

LEDs occupy a distinct position in the photonics source landscape. Compared to lasers, LEDs produce spatially incoherent, wider-bandwidth emission from a larger emitting area, but at lower cost, with simpler driving requirements, no threshold behavior, and no speckle. Compared to lamps, LEDs offer dramatically higher efficiency (luminous efficacies exceeding 200 lm/W versus 15–25 lm/W for incandescent), longer lifetime (50,000+ hours versus 1000–3000 hours for halogen), faster modulation (MHz versus DC-only for thermal sources), and wavelength selectivity without filters. Compared to superluminescent diodes (SLEDs), standard LEDs have broader spectral width and lower spatial coherence but are available at much higher optical power and lower cost [1, 2, 5].

In photonics applications, LEDs serve as excitation sources for fluorescence microscopy (replacing mercury arc lamps), pump sources for fiber amplifiers and certain solid-state gain media, illumination sources for machine vision and spectroscopy, and emitters for optical communication at moderate data rates. The spectral width of a typical LED (15–60 nm FWHM) is too broad for high-resolution spectroscopy but well-matched to applications where wavelength-band excitation or illumination is sufficient [2, 5].

2.1Structural Classification

LEDs are classified structurally by their chip architecture and light-extraction geometry. The three principal structures are surface-emitting, edge-emitting, and flip-chip designs [1, 3].

Surface-emitting LEDs emit light from the top face of the semiconductor die perpendicular to the junction plane. The emitting area is typically 200 µm to 2 mm on a side. Light generated in the active region radiates into a Lambertian angular distribution (intensity proportional to cos θ) modified by the chip geometry and encapsulant optics. Surface emitters are the dominant architecture for illumination, signaling, and most photonics applications because they are easy to manufacture, package, and optically couple [1, 3].

Edge-emitting LEDs (EELEDs) emit from a narrow stripe along the cleaved facet of the semiconductor, similar to a laser diode but without a resonant cavity. The emitting aperture is small (typically 1–5 µm × 50–200 µm), producing a highly asymmetric emission pattern. EELEDs are used primarily for fiber-coupled applications where the narrow emitting region can be efficiently coupled into single-mode or multimode optical fibers [1, 3].

Flip-chip LEDs mount the die junction-side-down onto the submount, with light extracted through the transparent substrate (typically sapphire for InGaN devices). This geometry provides a direct thermal path from the junction to the heatsink through the solder bond, dramatically improving thermal management compared to conventional top-emitting designs where heat must conduct through the substrate. Flip-chip architectures are standard for high-power LEDs operating at drive currents above 350 mA [3, 6].

2.2Spectral Classification

LEDs are also classified by their emission wavelength band. Ultraviolet LEDs (UV-A: 315–400 nm, UV-B: 280–315 nm, UV-C: 200–280 nm) use AlGaN material systems with progressively higher aluminum content for shorter wavelengths. Visible LEDs span blue (450–490 nm), green (520–560 nm), yellow (570–590 nm), orange (590–620 nm), and red (620–700 nm) using InGaN (blue/green) and AlGaInP (yellow/orange/red) material systems. Near-infrared LEDs (700–1700 nm) use GaAs, AlGaAs, and InGaAsP material systems. Mid-infrared LEDs (2–5+ µm) use InAsSb and other narrow-gap III-V alloys or type-II superlattice structures [1, 2, 3].

2.3White LED Approaches

White light can be generated from LEDs using three principal approaches [1, 4]:

Phosphor conversion: A blue LED (typically InGaN emitting at 450–460 nm) is coated with a yellow-emitting phosphor (most commonly cerium-doped yttrium aluminum garnet, Ce:YAG). The blue LED emission excites the phosphor, which emits a broad yellow band centered near 560 nm. The combination of transmitted blue light and phosphor-generated yellow light is perceived as white by the human eye. This is the dominant approach for general illumination and is used in virtually all commercial white LEDs. Warm-white variants add a red phosphor to improve color rendering at the expense of slightly reduced efficacy [1, 4].

RGB color mixing: Separate red, green, and blue LEDs are combined optically to produce white light. This approach offers tunable color temperature and full color-rendering capability but requires independent current control of three (or more) channels and suffers from differential aging and thermal drift among the color channels. RGB mixing is used primarily in display backlighting, architectural lighting, and stage lighting where color tunability is essential [1, 4].

UV LED plus RGB phosphors: An ultraviolet LED (typically 365–405 nm) excites a blend of red, green, and blue phosphors. This approach can produce excellent color rendering because the entire visible spectrum is phosphor-generated, but it is less efficient than blue-plus-yellow conversion because of the larger Stokes shift from UV to visible wavelengths. This method finds niche use in high-CRI specialty lighting [1, 4].

3.1Band Structure and Direct vs Indirect Gaps

The electronic band structure of a semiconductor determines whether it can function as an efficient light emitter. In a direct-bandgap semiconductor (GaAs, InGaN, AlGaInP), the minimum of the conduction band and the maximum of the valence band occur at the same crystal momentum (k-vector) in the Brillouin zone. An electron at the conduction band minimum can recombine with a hole at the valence band maximum by emitting a photon, conserving both energy and momentum simultaneously. This radiative recombination process is efficient because it requires no change in crystal momentum [1, 2].

In an indirect-bandgap semiconductor (Si, Ge, GaP), the conduction band minimum and valence band maximum occur at different k-vectors. Radiative recombination requires the simultaneous emission of a photon and absorption or emission of a phonon to conserve momentum. This three-body process has a much lower probability than the direct two-body process, making indirect-gap materials extremely poor light emitters. Silicon, despite being the dominant semiconductor for electronics, emits light with an internal quantum efficiency below 10⁻⁶ — six orders of magnitude worse than a good direct-gap LED [1, 2].

All efficient LED material systems are based on direct-bandgap III-V compound semiconductors (GaAs, GaN, InP and their alloys) or, for certain visible wavelengths, direct-gap II-VI compounds. The bandgap energy determines the photon energy and therefore the emission wavelength [1, 2].

3.2The p-n Junction and Electroluminescence

An LED is fundamentally a forward-biased p-n junction diode fabricated from a direct-bandgap semiconductor. Under zero bias, a depletion region forms at the junction where mobile carriers have diffused away, leaving behind ionized donor and acceptor atoms that create a built-in electric field opposing further diffusion. When a forward voltage exceeding the built-in potential is applied, the depletion region narrows and carriers are injected across the junction: electrons from the n-side into the p-side, and holes from the p-side into the n-side [1, 2].

In the active region near the junction, the injected minority carriers recombine with majority carriers. In a direct-bandgap material, a significant fraction of these recombinations are radiative — the electron drops from the conduction band to the valence band and emits a photon with energy approximately equal to the bandgap energy E_g. The remaining recombinations are non-radiative, producing heat through Shockley-Read-Hall recombination at crystal defects, Auger recombination (energy transferred to a third carrier), and surface recombination at unpassivated interfaces [1, 2].

Modern high-brightness LEDs use a double heterostructure or quantum well active region rather than a simple p-n homojunction. In a double heterostructure, the thin active layer is sandwiched between wider-bandgap cladding layers that confine both electrons and holes to the active region, dramatically increasing the carrier density and radiative recombination rate. Quantum well structures (active layer thickness 2–10 nm) further increase confinement and allow the emission wavelength to be tuned by adjusting the well width in addition to the material composition [1, 2, 3].

3.3Wavelength-Band Gap Relationship

The peak emission wavelength of an LED is determined by the bandgap energy of the active region semiconductor. The photon energy equals the bandgap energy (to first approximation, neglecting small thermal and quantum confinement corrections), and the relationship between photon wavelength and energy is [1, 2]:

Where: λ = peak emission wavelength (nm), E_g = bandgap energy (eV). The constant 1240 nm·eV is the product hc/e expressed in convenient units (h = Planck constant, c = speed of light, e = elementary charge). This relationship is the fundamental design equation for LEDs: selecting a semiconductor alloy with a specific bandgap energy directly determines the emission color [1, 2].

4.1AlGaInP (Red-Yellow)

The aluminum gallium indium phosphide (AlGaInP) material system is the basis for the highest-efficiency red, orange, and yellow LEDs. Lattice-matched to GaAs substrates, AlGaInP alloys cover the wavelength range from approximately 560 nm (yellow-green) to 650 nm (deep red) by adjusting the aluminum-to-gallium ratio in the quaternary alloy (Al_xGa_{1-x})_{0.5}In_{0.5}P. Higher aluminum content increases the bandgap and shifts emission toward shorter (yellow) wavelengths; lower aluminum content decreases the bandgap toward longer (red) wavelengths [1, 3].

A critical limitation of AlGaInP is the direct-to-indirect bandgap crossover. As the aluminum fraction increases above approximately x = 0.53, the alloy transitions from a direct to an indirect bandgap, and the internal quantum efficiency drops precipitously. This crossover occurs near 560 nm, creating the well-known "green gap" — the wavelength region between about 530 and 580 nm where neither InGaN nor AlGaInP produces high-efficiency emission. AlGaInP LEDs also suffer a pronounced temperature sensitivity: the bandgap decreases by approximately 0.3–0.5 nm/°C, and the IQE drops by roughly 1% per °C at junction temperatures above 25 °C, making thermal management critical for high-power red/orange LEDs [1, 3, 6].

4.2InGaN (UV-Blue-Green)

The indium gallium nitride (InGaN) system grown on GaN buffer layers on sapphire or silicon carbide substrates is the foundation of all high-efficiency blue, royal blue, cyan, and green LEDs, as well as the blue pump chips used in phosphor-converted white LEDs. InGaN is a direct-bandgap alloy across its entire composition range, with bandgap energies spanning from 3.4 eV (GaN, 365 nm UV) to 0.7 eV (InN, ~1770 nm near-IR). In practice, high-quality InGaN quantum wells are grown with indium fractions up to approximately 30%, covering wavelengths from 365 nm to about 530 nm [1, 3, 4].

InGaN LEDs are remarkable for their tolerance of extremely high dislocation densities (10⁸–10¹⁰ cm⁻²) compared to other III-V systems where dislocation densities above 10⁴ cm⁻² severely degrade radiative efficiency. This tolerance is attributed to carrier localization in indium-rich clusters within the InGaN active region, which traps carriers and prevents them from diffusing to dislocations where they would recombine non-radiatively. However, increasing the indium content to push emission into the green region (above 500 nm) progressively degrades crystal quality and introduces strong piezoelectric fields across the quantum wells, reducing the electron-hole wavefunction overlap and lowering IQE — the other side of the green gap problem [1, 3, 4].

4.3AlGaAs (Near-IR and Red)

Aluminum gallium arsenide (AlGaAs) was the first III-V alloy system to produce commercially successful high-brightness LEDs, beginning in the 1970s. AlGaAs is lattice-matched to GaAs across its entire composition range, enabling high-quality epitaxial growth. The bandgap ranges from 1.42 eV (GaAs, 870 nm) to the direct-indirect crossover at approximately 1.95 eV (Al_{0.45}Ga_{0.55}As, 635 nm). AlGaAs LEDs emit in the near-infrared (850–870 nm for GaAs-rich compositions) and deep red (650–670 nm for higher aluminum fractions) [1, 2].

Near-infrared AlGaAs LEDs at 850 nm and 870 nm remain widely used in fiber optic data links, optocouplers, and remote sensing. However, for visible red emission, AlGaAs has been largely superseded by AlGaInP, which offers higher efficiency, better color purity, and a wider accessible wavelength range. AlGaAs also suffers from significant sensitivity to oxidation — exposed aluminum-containing surfaces degrade over time, limiting long-term reliability unless properly passivated [1, 2, 3].

4.4Emerging and Specialty Materials

Beyond the three mainstream material systems, several emerging and specialty materials address wavelength ranges that are difficult or impossible to reach with conventional alloys. AlGaN-based deep ultraviolet LEDs (UV-B and UV-C below 315 nm) have advanced rapidly for applications in water purification, surface sterilization, and gas sensing. External quantum efficiencies remain below 10% for wavelengths below 280 nm due to challenges in p-type doping of high-aluminum-content AlGaN and poor light extraction from high-refractive-index UV-transparent substrates [3, 7].

In the mid-infrared, InAsSb and InAs/GaSb type-II superlattice structures produce LED emission at wavelengths from 3 to 12 µm for gas sensing applications (detecting CH₄, CO₂, CO, and other molecules by their absorption fingerprints). These devices operate at low efficiency (typically well below 1% WPE) and often require thermoelectric cooling but fill a critical need for compact, electrically modulated mid-IR sources [3, 7].

Micro-LEDs — individual LED chips with dimensions below 100 µm, sometimes as small as 3–5 µm — represent the cutting edge of display technology. Based on standard InGaN (blue/green) and AlGaInP (red) epitaxial structures but requiring novel mass-transfer assembly techniques, micro-LEDs promise display brightness, contrast, and efficiency far exceeding organic LEDs (OLEDs) and liquid crystal displays (LCDs) [4, 7].

5.1Emission Spectrum Shape

The emission spectrum of an LED is determined by the spontaneous emission process in the semiconductor active region. Unlike laser emission, which is concentrated in one or a few extremely narrow longitudinal modes (linewidth < 0.01 nm), LED emission spans a broad spectral band — typically 15 to 60 nm full-width at half-maximum (FWHM) — shaped by the density of states, the Fermi-Dirac distribution of carriers, and the joint density of states for radiative transitions [1, 2].

For a bulk (non-quantum-well) LED, the spontaneous emission rate per unit photon energy is proportional to the product of the optical density of states, the joint density of states for electron-hole transitions, and the occupation probabilities. The resulting spectrum is approximately Gaussian on the high-energy side and has an exponential tail on the low-energy side, producing a slightly asymmetric peak. The spectral width is set primarily by the thermal energy k_BT of the carrier distribution, with higher junction temperatures producing broader emission [1, 2].

The spontaneous emission rate per unit photon energy for a direct-gap semiconductor near the band edge can be expressed as [1, 2]:

Where: h = Planck constant, ν = photon frequency, E_g = bandgap energy, k_B = Boltzmann constant, T = junction temperature (K). The (hν − E_g)^(1/2) term reflects the square-root density of states near the band edge, and the exponential term reflects the thermal distribution of carriers.

5.2FWHM and Spectral Linewidth

The FWHM spectral linewidth of an LED is determined primarily by the thermal spread of carrier energies in the active region. For a simple direct-bandgap material at temperature T, the approximate spectral width in frequency units is [1, 2]:

Where: Δν = FWHM in frequency (Hz), k_B = 1.381 × 10⁻²³ J/K, T = junction temperature (K), h = 6.626 × 10⁻³⁴ J·s. The factor 1.8 arises from the convolution of the square-root density of states with the Boltzmann carrier distribution [1, 2].

Converting from frequency width to wavelength width using the dispersion relation Δλ = λ²Δν/c [1, 2]:

This expression shows that the wavelength FWHM scales as λ², so longer-wavelength LEDs inherently have broader spectral widths in nanometer units — a red LED at 630 nm will have approximately twice the FWHM of a blue LED at 450 nm at the same junction temperature. In practice, quantum well effects, alloy disorder, strain, and internal electric fields modify the spectral width from this simple thermal estimate, but the λ² scaling and the proportionality to temperature remain valid trends [1, 2].

5.3Temperature Dependence of Spectrum

As junction temperature increases, two spectral changes occur simultaneously. First, the bandgap energy decreases (described by the Varshni equation or Bose-Einstein model), shifting the peak emission wavelength toward longer wavelengths — a red-shift of typically 0.1–0.3 nm/°C for InGaN and 0.15–0.5 nm/°C for AlGaInP. Second, the carrier thermal distribution broadens, increasing the FWHM linewidth by approximately 0.04–0.1 nm/°C. For precision spectroscopy applications, junction temperature control to within ±1 °C is necessary to maintain spectral stability better than ±0.5 nm [1, 3, 6].

The temperature-induced red-shift is particularly important when LEDs are used as fluorescence excitation sources or for wavelength-specific applications. A blue LED intended to excite a fluorophore at 470 nm may shift to 475 nm if the junction temperature rises by 30 °C, potentially reducing excitation efficiency if the fluorophore absorption band is narrow. Thermoelectric cooling or active current derating can maintain spectral stability [3, 6].

5.4Current Dependence of Spectrum

Increasing the drive current affects the LED spectrum through two mechanisms. First, higher current raises the junction temperature (increasing heat dissipation), which red-shifts and broadens the spectrum as described above. Second, higher current increases the carrier density in the active region, which can produce a blue-shift due to band-filling effects — higher-energy states in the density of states become populated as the quasi-Fermi levels move deeper into the bands. In InGaN quantum well LEDs, a third effect — screening of the piezoelectric field by injected carriers — also contributes a blue-shift with increasing current [1, 3].

The net spectral shift with current depends on the balance between thermal red-shift and carrier-induced blue-shift. In AlGaInP LEDs, the thermal effect dominates, and increasing current typically produces a net red-shift. In InGaN LEDs, the piezoelectric screening and band-filling effects are strong, and the spectrum often blue-shifts with increasing current at low-to-moderate drive levels before eventually red-shifting at high currents where thermal effects dominate. Typical total wavelength shifts over the operating current range are 2–10 nm [1, 3].

5.5Comparison with Lasers and Lamps

The spectral characteristics of LEDs place them between lasers and lamps in terms of spectral width and coherence. A laser diode operating above threshold typically has a linewidth of 0.001 to 3 nm, depending on the laser type, compared to 15–60 nm for an LED. A tungsten-halogen lamp emits a continuous blackbody-like spectrum spanning hundreds of nanometers. This intermediate spectral width makes LEDs well-suited for applications requiring moderate wavelength selectivity without the expense and complexity of a laser source [1, 2, 5].

For pumping solid-state gain media, the LED spectral width has important consequences. An LED used to pump Nd:YAG must have its emission spectrum overlap with the absorption bands of the Nd³⁺ ion (centered near 808 nm, with approximately 3 nm absorption bandwidth). An LED with 25–40 nm FWHM centered at 808 nm will have only a fraction of its total optical power falling within the narrow absorption band, reducing pump efficiency compared to a laser diode with 2–3 nm linewidth centered precisely on the absorption peak. This spectral mismatch is the primary reason laser diodes have almost entirely replaced LEDs as pump sources for solid-state lasers, despite the lower cost and longer lifetime of LEDs [2, 5].

6.1Internal Quantum Efficiency (IQE)

The internal quantum efficiency (IQE) is the ratio of the number of photons generated inside the semiconductor active region to the number of electrons injected into the device. It quantifies the fraction of carrier recombinations that are radiative rather than non-radiative [1, 2]:

Where: R_rad = radiative recombination rate, R_non-rad = non-radiative recombination rate (Shockley-Read-Hall + Auger + surface). For the best InGaN blue LEDs, IQE exceeds 90% at moderate current densities. For AlGaInP red LEDs, peak IQE values reach 50–70%. For deep-UV AlGaN LEDs below 280 nm, IQE is typically 5–20% due to high dislocation densities and difficulty achieving good p-type conductivity. IQE is difficult to measure directly and is usually inferred from temperature-dependent photoluminescence measurements or efficiency droop analysis [1, 3, 6].

6.2External Quantum Efficiency (EQE)

The external quantum efficiency (EQE) is the ratio of the number of photons emitted from the LED package into free space to the number of electrons injected. It incorporates both the IQE and the light extraction efficiency (LEE) — the fraction of internally generated photons that escape the semiconductor chip rather than being reabsorbed or totally internally reflected [1, 2]:

Light extraction is a major challenge because the high refractive index of III-V semiconductors (n ≈ 2.5 for GaN, n ≈ 3.5 for GaAs) creates a small critical angle for total internal reflection. For a flat GaN-air interface, the critical angle is sin⁻¹(1/2.5) ≈ 23.6°, meaning only photons within a narrow escape cone are transmitted. Without extraction enhancement, only about 4% of photons escape from a single flat surface. Modern high-power LEDs use surface roughening, patterned sapphire substrates, photonic crystals, shaped chips, and encapsulant domes (n ≈ 1.5 silicone) to increase extraction efficiency to 70–90%. The best commercial blue InGaN LEDs achieve EQE values of 60–80% [1, 3, 6].

6.3Wall-Plug Efficiency (WPE)

The wall-plug efficiency (WPE) — also called the power conversion efficiency or electro-optical efficiency — is the ratio of total optical output power to the total electrical input power. It is the most practically useful efficiency metric because it accounts for all loss mechanisms including resistive (I²R) losses in the semiconductor and contacts [1, 2]:

Where: P_opt = total radiant optical power emitted (W), I_f = forward current (A), V_f = forward voltage (V). The WPE is always less than the EQE because some input electrical energy is lost to resistive heating even before carrier injection. The relationship between WPE and EQE is [1, 2]:

The ratio E_g/(eV_f) is always less than or equal to 1 because the forward voltage must exceed the photon energy (in eV) by at least the amount of resistive and series resistance drops. Typical forward voltages are 1.8–2.2 V for red LEDs (E_g ≈ 1.9 eV) and 2.8–3.5 V for blue LEDs (E_g ≈ 2.75 eV) [1, 2].

6.4Luminous Efficacy

For visible-light applications, the luminous efficacy — measured in lumens per watt (lm/W) — is often more relevant than WPE. Luminous efficacy accounts for the spectral sensitivity of the human eye (the photopic luminosity function V(λ), which peaks at 555 nm). The luminous efficacy of radiation (LER) is the conversion factor from radiant watts to lumens for a given spectrum; for a monochromatic 555 nm source, LER = 683 lm/W (the maximum possible). For a phosphor-converted white LED spectrum, typical LER values are 300–350 lm/W [1, 4].

The overall luminous efficacy of the LED source is the product of WPE and LER [1, 4]:

7.1Heat Generation in LEDs

An LED converts a fraction of its input electrical power into light (quantified by the WPE) and dissipates the remainder as heat. For a device drawing P_elec = I_f × V_f of electrical power and producing P_opt of optical power, the heat dissipation is simply P_heat = P_elec − P_opt = P_elec(1 − WPE). For a 1 W blue LED with 40% WPE, 600 mW is dissipated as heat; for a 3 W high-power device at the same efficiency, 1.8 W of heat must be managed. Unlike incandescent lamps where much of the heat escapes as infrared radiation, LED heat is generated entirely within the semiconductor chip and must be conducted away through the package and heatsink [1, 3, 6].

The heat is generated in three locations within the LED structure: (1) non-radiative recombination in the active region (the dominant source), (2) resistive (Joule) heating in the semiconductor bulk and contact regions due to series resistance, and (3) absorption of internally generated photons that fail to escape the chip. All of this heat is deposited in a volume of a few cubic millimeters, resulting in extremely high power densities — typically 10 to 100 W/cm² for high-power LEDs — that must be managed to prevent excessive junction temperature rise [3, 6].

7.2Thermal Resistance and the R_θ Chain

Heat flows from the LED junction to the ambient environment through a series of thermal resistances, analogous to electrical resistances in a series circuit. The total thermal resistance from junction to ambient is the sum of the individual resistances in the thermal path [3, 6]:

Where: R_θ,j-s = junction-to-solder-point thermal resistance (typically 2–10 °C/W for high-power LEDs), R_θ,s-b = solder-to-board resistance (0.5–3 °C/W), R_θ,b-hs = board-to-heatsink resistance including thermal interface material (0.5–2 °C/W), R_θ,hs-a = heatsink-to-ambient thermal resistance (1–20 °C/W depending on heatsink size and airflow). The junction temperature is then [3, 6]:

Where: T_j = junction temperature (°C), T_a = ambient temperature (°C), P_heat = heat dissipation (W).

7.3Junction Temperature Calculation

Calculating the junction temperature for a specific LED operating condition requires knowledge of the thermal resistance chain and the heat dissipation. The procedure is: (1) calculate the electrical input power P_elec = I_f × V_f, (2) calculate the heat dissipation P_heat = P_elec × (1 − WPE), (3) multiply by the total junction-to-ambient thermal resistance, and (4) add the ambient temperature. If the resulting T_j exceeds the manufacturer's maximum rated junction temperature (typically 125–150 °C for high-power LEDs), the operating conditions must be derated — either by reducing the drive current, improving the thermal path, or lowering the ambient temperature [3, 6].

7.4Thermal Effects on Performance

Elevated junction temperature degrades LED performance through several mechanisms. The internal quantum efficiency decreases because non-radiative recombination rates (especially Auger recombination) increase exponentially with temperature, while the radiative rate is relatively temperature-insensitive. For InGaN blue LEDs, the optical output typically drops by 0.2–0.3%/°C near room temperature. For AlGaInP red/amber LEDs, the sensitivity is much greater — 0.5–1.0%/°C — because the direct-indirect bandgap crossover shifts with temperature, progressively quenching radiative recombination [1, 3, 6].

In addition to efficiency reduction, elevated temperature shifts the peak wavelength (red-shift of 0.1–0.5 nm/°C depending on material system), broadens the spectrum, increases the forward voltage slightly (due to increased series resistance), and accelerates degradation mechanisms that reduce useful lifetime. The rated lifetime of an LED (L70 — the time to 70% of initial lumen output) is strongly temperature-dependent, often halving for every 10–15 °C increase in junction temperature above the reference condition [3, 6].

7.5Heatsink Design Principles

The heatsink is the largest thermal resistance in the chain and is the primary element the system designer controls. Natural convection heatsinks (finned aluminum extrusions) provide R_θ,hs-a values of 3–20 °C/W depending on size and fin geometry. Forced-air cooling (fans) can reduce heatsink thermal resistance by a factor of 3–5. For compact, high-power applications, active thermoelectric coolers (TECs) or liquid cooling may be required [3, 6].

Key design guidelines include: (1) use the largest practical heatsink to maximize convective surface area, (2) orient fins vertically for natural convection to exploit buoyancy-driven airflow, (3) minimize thermal interface resistance by using high-conductivity thermal paste or pads (conductivity 1–10 W/m·K), (4) avoid using electrically insulating thermal interface materials unless electrical isolation is required, as they add thermal resistance, and (5) design for the worst-case ambient temperature the system will encounter, not the typical laboratory temperature [3, 6].

8.1Through-Hole (Radial) Packages

The classic through-hole LED package — the 3 mm or 5 mm radial (T-1 or T-1¾) — encapsulates a small LED die in a transparent or diffused epoxy dome with two wire leads. The dome functions as an integral lens, shaping the emission pattern from the native Lambertian distribution into a narrower beam (typically 15° to 60° half-angle, depending on the dome geometry). Through-hole packages dissipate heat primarily through the lead wires, limiting power handling to approximately 70–100 mW. These packages remain widely used for panel indicators, low-power signaling, optocoupler emitters, and prototype circuits, but they are unsuitable for high-flux applications [1, 3].

8.2Surface-Mount (SMD) Packages

Surface-mount device (SMD) LED packages use a ceramic or molded plastic substrate with metal lead pads for reflow soldering to a printed circuit board. Common form factors include PLCC-2 (3528), PLCC-4, and mid-power 2835 and 3030 packages. SMD packages provide better thermal performance than through-hole types because the thermal pad on the bottom of the package conducts heat directly into the PCB copper. Power handling ranges from 100 mW to about 1 W depending on the package size and PCB thermal design. SMD packages dominate the general illumination market for residential and commercial LED lamps and luminaires [1, 3].

8.3High-Power and COB Packages

High-power LED packages are designed for drive currents from 350 mA to over 3 A, producing 1–10 W of optical output per package. These packages use ceramic (alumina or aluminum nitride) substrates with large exposed thermal pads, providing junction-to-solder-point thermal resistances of 1–5 °C/W. The die is typically mounted in flip-chip configuration for optimum thermal extraction. Primary optics (silicone dome lenses) shape the emission pattern, and secondary optics (reflectors, TIR lenses) are added at the luminaire level [3, 6].

Chip-on-board (COB) arrays integrate multiple LED dies directly onto a metal-core printed circuit board (MCPCB) or ceramic substrate, covered by a single phosphor layer and silicone dome. COB modules produce 10–100+ W of optical output from a single compact emitting area (5–30 mm diameter), mimicking a point source for optical design purposes. The dense packing achieves high luminance suitable for spotlights, downlights, and projection applications. Thermal management is critical: the MCPCB must be mounted directly to a large heatsink with minimal thermal interface resistance [3, 6].

8.4Fiber-Coupled Packages

Fiber-coupled LED packages incorporate an optical fiber pigtail aligned and permanently attached to the LED die during manufacturing. For multimode fiber coupling (core diameter 50–600 µm, NA 0.22–0.50), a ball lens or microlens images the LED emitting area onto the fiber face. Typical fiber-coupled LED output powers range from 0.5 to 50 mW depending on the LED type and fiber core diameter. For single-mode fiber coupling — relevant primarily for edge-emitting LEDs and SLEDs — the package uses a precision-aligned lensed fiber or micro-optic assembly housed in a hermetically sealed butterfly package with a thermoelectric cooler and monitor photodiode [1, 3, 5].

Fiber-coupled LEDs are used as broadband sources for optical coherence tomography (OCT), fiber optic gyroscopes, white-light interferometry, and fiber-coupled spectroscopy systems. The fiber pigtail provides a convenient, alignment-free interface to the downstream optical system. Multimode fiber-coupled LEDs are also widely used for fluorescence excitation in microscopy, replacing direct free-space coupling with simpler integration [3, 5].

8.5UV LED Packages

Ultraviolet LEDs present unique packaging challenges. Standard silicone and epoxy encapsulants absorb strongly at wavelengths below 360 nm, precluding their use for UV-B and UV-C devices. UV LED packages use either a flat quartz or fused silica window over the die (no encapsulant) or specialized UV-transparent silicone formulations. The absence of a dome encapsulant reduces light extraction efficiency compared to visible LEDs, contributing to the lower EQE of UV devices. UV-C LEDs often use ceramic submounts with gold or aluminum reflective surfaces to redirect upward-emitted light and improve extraction. Hermetic sealing with nitrogen backfill prevents oxidation of aluminum-containing bond pads and reflectors under UV exposure [3, 7].

9.1Operating Principle

A superluminescent diode (SLED, also abbreviated SLD) is a semiconductor device that combines features of both LEDs and laser diodes. Like a laser diode, it uses a waveguide structure to amplify spontaneous emission as it propagates along the gain medium (amplified spontaneous emission, or ASE). Unlike a laser diode, it is specifically designed to suppress lasing by eliminating optical feedback — typically through angled or anti-reflection-coated facets, curved waveguides, or absorbing sections near one or both ends of the device. The result is a high-brightness, broadband, spatially coherent but temporally incoherent source [1, 5, 8].

The spectral output of a SLED is broader than a laser (typically 20–100 nm FWHM) but narrower than a simple LED, with much higher spectral power density and spatial coherence (brightness) than an LED. The single-pass amplification process produces output powers of 1–50 mW from a single-mode waveguide — 10 to 100 times more power than a comparable LED could couple into a single-mode fiber. The ASE spectrum is approximately Gaussian, centered at the gain peak of the semiconductor, with the spectral width determined by the gain bandwidth of the active material [5, 8].

9.2Key Characteristics

The key performance parameters of SLEDs include output power, spectral bandwidth (FWHM), center wavelength, spectral ripple (residual Fabry-Perot modulation from incomplete feedback suppression), spatial mode quality, and polarization. High-quality SLEDs exhibit spectral ripple below 0.1 dB (ratio of peak-to-valley modulation in the spectrum), indicating effective suppression of cavity feedback. Center wavelengths span from 680 nm to 1600 nm using InGaAs, InGaAsP, and GaAs material systems, with bandwidths from 20 nm (narrower, higher power) to over 100 nm (broader, lower power per unit bandwidth) [5, 8].

The spatial coherence of a SLED is equivalent to a single-mode laser — the output from a single-mode waveguide has full spatial coherence and can be focused to a diffraction-limited spot or coupled efficiently into single-mode fiber. However, the temporal coherence is low (coherence length 5–30 µm for a 50 nm bandwidth SLED at 1310 nm), which means the output does not produce speckle or interference fringes except at path differences comparable to the coherence length. This combination of high brightness and low coherence is the defining advantage of SLEDs [5, 8].

9.3SLED Applications

The primary application of SLEDs is as the light source in optical coherence tomography (OCT), where the broad bandwidth determines the axial resolution of the imaging system (axial resolution ≈ 0.44 λ²/Δλ). A SLED at 1310 nm with 60 nm FWHM provides approximately 12.5 µm axial resolution; a SLED at 840 nm with 50 nm FWHM provides approximately 6.2 µm resolution. OCT is used extensively in ophthalmology (retinal imaging), dermatology, intravascular imaging, and industrial non-destructive testing [5, 8].

SLEDs are also used in fiber optic gyroscopes (FOGs), where the low temporal coherence suppresses parasitic interference from backscattered light and Rayleigh scattering in the sensing fiber coil, improving rotation measurement accuracy. Additional applications include white-light interferometry for surface profilometry, optical spectrum analyzer testing, fiber Bragg grating interrogation, and as seed sources for fiber amplifiers when broader bandwidth than a laser is required [5, 8].

10.1DC Driving and Current Regulation

An LED is a current-driven device: the optical output is proportional to the forward current, and the forward voltage is a consequence of the semiconductor bandgap and series resistance rather than a parameter the user controls independently. The simplest driving method is a series resistor connected between the supply voltage and the LED anode. The resistor value is chosen to set the desired forward current [1, 3]:

Where: R = series resistance (Ω), V_supply = supply voltage (V), V_f = LED forward voltage at the desired current (V), I_f = desired forward current (A). This approach is adequate for low-power indicator LEDs but has poor current regulation because the forward voltage varies with temperature (approximately −2 mV/°C) and manufacturing tolerance. For high-power LEDs, dedicated constant-current driver ICs (switching regulators or linear regulators) maintain current accuracy to within ±1–5% over the full operating temperature range [1, 3].

10.2PWM Driving and Dimming

Pulse-width modulation (PWM) is the preferred method for dimming LEDs because it maintains a constant forward current (and therefore constant forward voltage, spectrum, and color temperature) during the on-phase while controlling the time-averaged output power by varying the duty cycle. The LED switches between full on-current and zero current at a frequency high enough (typically 200 Hz to 20 kHz) to be imperceptible to the human eye and to most photodetectors used in photonics systems [1, 3].

Analog dimming (reducing the DC forward current) is simpler but shifts the LED spectrum and color coordinates because both the peak wavelength and spectral width change with current, as discussed in Section 5.4. For applications where spectral stability is critical — fluorescence microscopy, colorimetry, machine vision — PWM dimming is mandatory. The PWM frequency should be at least 10 times higher than the measurement bandwidth to avoid introducing modulation artifacts into the detected signal [1, 3].

10.3Modulation Bandwidth

The modulation bandwidth of an LED is fundamentally limited by the carrier recombination lifetime τ_c in the active region. The 3 dB electrical bandwidth is [1, 2]:

Typical carrier lifetimes for LED active regions are 1–10 ns, giving 3 dB bandwidths of approximately 15–150 MHz. This is orders of magnitude slower than directly modulated laser diodes (which achieve 10–30 GHz bandwidth) but adequate for many communications and sensing applications. High-brightness illumination LEDs, designed for maximum efficiency rather than speed, have longer carrier lifetimes and lower bandwidth (typically 1–10 MHz). Micro-LEDs with very small active areas have shorter effective lifetimes and have demonstrated bandwidths exceeding 1 GHz, making them promising for visible-light communication (Li-Fi) systems [1, 2, 3].

11.1Matching LED to Application

Selecting an LED for a photonics application requires matching the source characteristics to the application requirements across several dimensions: wavelength, spectral width, optical power, spatial distribution, modulation speed, lifetime, and cost. The following guidelines address the most common application categories [1, 3, 5]:

Fluorescence excitation: Select an LED whose peak wavelength matches the fluorophore excitation maximum. A FWHM of 20–30 nm is typical and usually acceptable because fluorophore absorption bands are broader. High-power surface-emitting LEDs (1–5 W optical) with integrated collimation optics have largely replaced mercury arc lamps for wide-field fluorescence microscopy, offering longer lifetime, lower cost, instant on/off, and intensity control via PWM without mechanical shutters [3, 5, 9].

Machine vision and illumination: Select based on the spectral response of the camera sensor and the reflectance/transmission properties of the object being inspected. Ring lights, bar lights, and backlight panels using arrays of surface-emitting LEDs provide uniform, controllable illumination. NIR LEDs (850–940 nm) are used with silicon sensors for covert illumination; UV LEDs (365–395 nm) excite fluorescence for defect inspection [3].

Spectroscopy sources: Broadband white LEDs serve as compact, stable alternatives to halogen lamps for UV-Vis absorption measurements over limited spectral ranges. For narrowband excitation (Raman spectroscopy, photoluminescence), lasers remain superior due to their monochromatic output. LED sources are increasingly used in portable and field-deployable spectrometers where lamp replacement and power consumption are concerns [5].

Optical communication: Edge-emitting LEDs and SLEDs are used for multimode fiber links at moderate data rates (up to several hundred Mb/s). For higher rates, laser diodes are required. The choice of wavelength follows the fiber attenuation minima: 850 nm for short-reach multimode, 1310 nm for zero-dispersion single-mode, 1550 nm for minimum-attenuation single-mode [1, 2].

11.2Key Specifications to Compare

When comparing LED products from different manufacturers, the following specifications should be evaluated in the context of the application requirements [1, 3, 6]:

Peak wavelength and dominant wavelength: The peak wavelength is the wavelength of maximum spectral intensity. The dominant wavelength (relevant for visible LEDs) is the wavelength that the human eye perceives as the color of the LED, accounting for the spectral shape — it is determined by the intersection of a line from the white point through the LED chromaticity coordinates with the spectrum locus on the CIE chromaticity diagram. For photonics applications, peak wavelength is the relevant parameter; for illumination and display applications, dominant wavelength and color coordinates (CIE x,y or u',v') are more important [1, 4].

Optical power vs luminous flux: Photonics applications require radiant flux (milliwatts or watts) at the specified wavelength. Illumination applications use luminous flux (lumens). Manufacturers of illumination LEDs often specify only lumens, requiring conversion using the spectral power distribution and the photopic luminosity function if radiometric power is needed. Scientific-grade LED suppliers typically provide radiant flux specifications [1, 3].

Viewing angle: The half-power angle (θ_1/2) defines the angular cone within which the intensity falls to 50% of the on-axis value. Lambertian emitters have θ_1/2 = 60°. Narrow-beam LEDs with integrated lenses may have θ_1/2 of 5–15°. The viewing angle determines how effectively the LED output can be collected by downstream optics [1, 3].

Thermal resistance: The junction-to-solder-point thermal resistance R_θ,j-s determines the heatsink requirement. Lower R_θ,j-s allows higher drive currents and higher ambient temperatures for a given maximum junction temperature. This is the most critical specification for high-power LED system design [3, 6].

Lifetime (L70 or L50): LED lifetime is specified as the time to reach a percentage of initial light output — L70 (70% maintenance) is the most common metric for illumination, while L50 (50%) is used for some indicator and scientific applications. Rated lifetimes of 25,000–100,000 hours are common but are based on accelerated testing and extrapolation; the actual lifetime depends strongly on the operating junction temperature and drive current [3, 6].

11.3Common Pitfalls

Ignoring thermal derating: LED datasheets specify optical power and efficiency at a reference junction temperature (typically 25 °C). In real operating conditions with ambient temperatures of 40–60 °C and imperfect thermal management, the junction temperature may reach 80–120 °C, reducing output by 10–30% for InGaN and 30–60% for AlGaInP. System designs must account for thermal derating at the actual operating temperature, not the 25 °C datasheet condition [3, 6].

Confusing radiant power with luminous flux: A 1 W radiant power UV LED (365 nm) produces essentially zero lumens because the human eye has no sensitivity at that wavelength. Conversely, a white LED rated at 200 lumens may produce only 0.6 W of radiant power. Photonics applications must use radiometric units (watts), not photometric units (lumens), when specifying LED output [1, 3].

Neglecting spectral shift with temperature and current: Applications requiring precise wavelength matching (e.g., pumping a narrow absorption line, exciting a specific fluorophore at its peak) must account for the 5–15 nm wavelength shift that occurs over the typical operating current and temperature range. Thermal stabilization or feedback-controlled current adjustment may be necessary [3, 6].

Assuming LED emission is monochromatic: An LED with a "peak wavelength of 630 nm" emits a band spanning approximately 610–650 nm (FWHM ~20 nm). If the application requires monochromatic illumination — for example, as an excitation source in Raman spectroscopy — the LED spectral width is typically too broad, and a laser or a bandpass-filtered LED must be used instead [1, 2].

Over-driving for more power: Increasing the forward current beyond the rated maximum may increase short-term output, but it accelerates degradation mechanisms (electromigration, thermal runaway, phosphor degradation) that dramatically shorten lifetime. The relationship between current and output is also sublinear at high drive levels due to efficiency droop, so doubling the current does not double the output. For more optical power, use a larger die, a multi-die package, or an array rather than over-driving a single device [3, 6].