Electronics Guide

Silicon Bandgap Characteristics

Silicon has a bandgap of approximately 1.12 eV at room temperature. This value decreases slightly as temperature increases, which contributes to the sharp rise in thermally generated carriers, and therefore in junction leakage, that every silicon device shows when it runs hot. The temperature dependence follows the relationship:

E_g(T) = E_g(0) - (alpha * T²) / (T + beta)

Where E_g(0) is the bandgap at absolute zero, and alpha and beta are the Varshni parameters, which are material-specific. For silicon (alpha approximately 4.73 x 10^-4 eV/K, beta approximately 636 K), the bandgap falls from about 1.17 eV at absolute zero to 1.12 eV at 300 K, shrinking by roughly 0.25 meV per kelvin near room temperature. The physical causes are lattice thermal expansion and, more importantly, the increasing interaction between electrons and lattice vibrations as temperature rises.

Heavy doping narrows the bandgap further. Above roughly 10¹⁸ dopant atoms per cubic centimeter, the impurity levels broaden into a band that merges with the nearest band edge, and carrier-carrier interactions shift the edges toward each other. This bandgap narrowing raises the effective intrinsic carrier concentration in heavily doped regions and is a first-order effect in bipolar transistor emitters, where it limits achievable current gain.

Direct and Indirect Bandgaps

Semiconductors are classified as having either direct or indirect bandgaps based on the relationship between electron energy and momentum in the crystal. In a direct bandgap material like gallium arsenide, the minimum energy state in the conduction band aligns with the maximum energy state in the valence band in momentum space. Electrons can transition between bands by absorbing or emitting photons without changing momentum.

Silicon has an indirect bandgap, meaning the conduction band minimum and valence band maximum occur at different momentum values. Transitions require both a photon and a phonon (lattice vibration) to conserve momentum. Because a three-body process is far less probable than a two-body one, silicon is inefficient at light emission, which is why LEDs and laser diodes typically use direct bandgap materials like gallium arsenide, indium gallium arsenide phosphide, or gallium nitride.

The same distinction governs absorption. A material absorbs photons whose energy exceeds its bandgap, so the long-wavelength cutoff follows from the bandgap directly: wavelength in nanometers is approximately 1240 divided by the bandgap in electron volts. Silicon therefore stops absorbing near 1100 nm, which is why silicon image sensors and solar cells are blind to the 1310 nm and 1550 nm wavelengths used in fiber optics, and why those links rely on indium gallium arsenide detectors instead. Indirect absorption is also weak near the band edge, so a silicon solar cell needs on the order of 100 micrometers of material to absorb red and near-infrared light, whereas a direct-gap gallium arsenide cell absorbs the same light within a few micrometers.

Effective Mass

Electrons and holes in a semiconductor crystal do not behave exactly like free particles. The periodic potential of the crystal lattice affects their motion, which is accounted for by assigning them an effective mass different from the free electron mass. This effective mass depends on the curvature of the energy bands and differs for electrons and holes, and even varies with direction in the crystal.

Silicon illustrates how much structure this single word conceals. Its conduction band has six equivalent minima along the crystal axes, and within each the electron mass is highly anisotropic: about 0.98 times the free electron mass along the valley axis and about 0.19 times the free electron mass transverse to it. Averaging these appropriately yields a conductivity effective mass near 0.26 free electron masses, the figure used in mobility calculations, and a somewhat larger density-of-states effective mass, close to the free electron mass, used to evaluate N_c. The valence band is more complicated still, consisting of overlapping heavy-hole and light-hole bands plus a third band split off by spin-orbit coupling; the resulting conductivity effective mass for holes is about 0.39 free electron masses.

The practical consequence is that holes in silicon are slower than electrons, so n-channel MOSFETs deliver roughly two to three times the current of p-channel devices of the same geometry. CMOS layouts compensate by making p-channel transistors wider. Gallium arsenide, by contrast, has a light electron effective mass near 0.067 free electron masses, which gives it electron mobility several times that of silicon and underpins its use in high-frequency circuits.

N-Type Semiconductors

When silicon is doped with atoms having five valence electrons, such as phosphorus, arsenic, or antimony, the extra electron from each dopant atom is loosely bound and easily ionized into the conduction band. These dopants are called donors because they donate electrons. The resulting material is n-type, with electrons as the majority carriers and holes as minority carriers.

Donor atoms introduce energy levels just below the conduction band edge. In silicon the common donors lie roughly 40 to 55 meV below the edge, with antimony the shallowest, phosphorus in the middle, and arsenic the deepest of the three. Because thermal energy at room temperature is about 26 meV, these levels are shallow enough that virtually all donors are ionized, so the electron concentration approximately equals the donor concentration for typical doping levels. At cryogenic temperatures, ionization becomes incomplete and carriers "freeze out" onto their parent atoms, which is why ordinary silicon devices stop working far below room temperature.

Practical doping concentrations span an enormous range. Lightly doped drift regions in power devices may hold 10¹⁴ atoms per cubic centimeter, general-purpose logic wells fall between 10¹⁶ and 10¹⁸, and contact and source/drain regions are doped above 10¹⁹ to reduce resistance and enable ohmic metal contacts. Resistivity varies by about four orders of magnitude across that doping range, and undoped silicon is far more resistive still, on the order of 10⁵ ohm-centimeters.

P-Type Semiconductors

Doping with atoms having three valence electrons, such as boron, aluminum, or gallium, creates acceptor sites that can capture electrons from the valence band, leaving behind holes. These materials are p-type, with holes as majority carriers. Acceptor levels lie just above the valence band edge; boron, the standard p-type dopant for silicon, sits about 45 meV above it, and like the shallow donors it is essentially fully ionized at room temperature. Boron is preferred in silicon technology because its small atomic size allows high solid solubility with little lattice strain, although that same small size makes it diffuse rapidly and therefore hard to confine in very shallow junctions.

The ability to create both n-type and p-type regions in the same crystal is fundamental to semiconductor device operation. Transistors, diodes, and integrated circuits all depend on controlled junctions between differently doped regions. In modern fabrication, dopants are introduced by ion implantation, which offers precise control of dose and depth, followed by a thermal anneal that repairs implantation damage and moves the dopant atoms onto lattice sites where they become electrically active.

Compensation and Carrier Concentration

When both donors and acceptors are present, they partially compensate each other. The net carrier type depends on which dopant has higher concentration. The carrier concentration is reduced by compensation, and the mobility may also be affected due to increased scattering from ionized impurities.

The relationship between carrier concentrations in thermal equilibrium is governed by the mass action law:

n * p = n_i²

This relationship holds in thermal equilibrium regardless of doping, though it breaks down under illumination or current injection, where excess carriers push the product above n_i². If electrons are the majority carrier with concentration n = N_D (the donor concentration), then the hole concentration is p = n_i² / N_D, which can be many orders of magnitude smaller. Silicon doped with 10¹⁶ donors per cubic centimeter holds only about 10⁴ holes per cubic centimeter, a ratio of a trillion to one. Minority carriers are scarce, yet they control the behavior of bipolar transistors, solar cells, and diode reverse leakage, because their relative changes are enormous even when their absolute numbers are tiny.

Compensation is exploited deliberately as well as tolerated. Every CMOS process builds n-type wells inside p-type substrates and p-type source/drain regions inside n-type wells by over-doping the existing background, so most working regions of a chip are compensated to some degree. Deliberate counter-doping also produces the high-resistivity, nearly intrinsic material used for radiation detectors and some power-device drift layers.

Drift Current

When an electric field is applied to a semiconductor, charge carriers accelerate in response but are frequently scattered by lattice vibrations (phonons) and ionized impurities. This results in a constant average velocity called the drift velocity, proportional to the electric field:

v_d = mu * E

Where mu is the carrier mobility and E is the electric field. Mobility has units of cm²/(V*s) and depends on temperature, doping concentration, and the semiconductor material. For lightly doped silicon at room temperature, electron mobility is roughly 1350 to 1450 cm²/(V*s) and hole mobility is roughly 450 to 500 cm²/(V*s).

Two scattering mechanisms dominate. Lattice or phonon scattering grows stronger as temperature rises and more vibrational modes are excited, so the mobility it permits falls with increasing temperature. Ionized impurity scattering, by contrast, weakens as temperature rises, because faster carriers spend less time deflected by each charged dopant ion, so the mobility it permits increases with temperature. The two combine approximately by Matthiessen's rule, in which the reciprocals of the individual mobilities add. In lightly doped silicon near room temperature, phonon scattering wins and mobility falls with heating; in heavily doped material at low temperature, impurity scattering dominates.

Doping therefore carries a penalty. Raising the donor concentration from 10¹⁵ to 10¹⁹ per cubic centimeter multiplies the carrier count ten-thousandfold but cuts electron mobility by roughly an order of magnitude, so conductivity rises by less than the doping ratio alone would suggest. A further reduction occurs at surfaces: carriers in a MOSFET inversion layer are pressed against the oxide interface by the gate field and scatter from interface roughness and charge, giving effective mobilities well below bulk values.

The drift current density is:

J_drift = q * (n * mu_n + p * mu_p) * E = sigma * E

Where sigma is the conductivity. This is the semiconductor form of Ohm's law.

Diffusion Current

Carriers also move from regions of high concentration to low concentration through diffusion, even without an applied electric field. The diffusion current density for electrons is:

J_diff,n = q * D_n * (dn/dx)

Where D_n is the electron diffusion coefficient. The corresponding hole term carries the opposite sign, J_diff,p = -q * D_p * (dp/dx), because holes carry positive charge while both species flow down their own concentration gradient. The diffusion coefficient is related to mobility through the Einstein relation:

D = (kT/q) * mu

At 300 K, kT/q is about 25.9 mV, commonly rounded to 26 mV and called the thermal voltage. This relation reflects the fundamental connection between random thermal motion (diffusion) and directed motion under a field (drift): both arise from the same scattering process, so a material that impedes one impedes the other in fixed proportion. Applying it to silicon gives an electron diffusion coefficient near 36 cm²/s and a hole diffusion coefficient near 12 cm²/s at room temperature.

Drift and diffusion together, combined with Poisson's equation relating charge density to electric field and with continuity equations that account for generation and recombination, constitute the drift-diffusion model. That system of coupled equations is what technology computer-aided design tools solve numerically, and it remains the workhorse of device simulation despite its classical approximations. Only at the smallest dimensions, where carriers cross the active region in fewer collisions than a diffusive description assumes, do engineers turn to Monte Carlo or quantum transport methods instead.

High-Field Effects

At high electric fields, carrier velocity no longer increases linearly with field. In silicon at room temperature the drift velocity saturates near 1 x 10⁷ cm/s for electrons and around 8 x 10⁶ cm/s for holes, because carriers gain enough energy between collisions to emit optical phonons, and additional field energy then goes into lattice heating rather than into faster motion. Saturation sets in above roughly 10⁴ V/cm, a field reached across a 1-micrometer gap at only 1 volt, so modern nanoscale transistors operate routinely in the velocity saturation regime. This is a central reason why shrinking a transistor no longer delivers proportional gains in drive current.

At even higher fields, carriers may gain enough energy to ionize lattice atoms on impact, creating additional electron-hole pairs that are themselves accelerated, in a multiplying process called impact ionization or avalanche multiplication. Silicon enters this regime near 3 x 10⁵ V/cm. The effect is exploited deliberately in avalanche photodiodes and IMPATT oscillators and provides the controlled clamping action of avalanche-rated power transistors, but in most devices it appears as a destructive breakdown mechanism to be designed around.

Carriers that gain high energy without ionizing anything create a subtler reliability problem. So-called hot carriers can be injected into the gate oxide of a MOSFET, where they become trapped and shift the threshold voltage over time. Limiting hot-carrier damage is one of the reasons supply voltages fell steadily as transistor dimensions shrank.

Junction Formation and the Depletion Region

When p-type and n-type materials are brought together, electrons diffuse from the n-side (where they are plentiful) to the p-side, while holes diffuse in the opposite direction. This diffusion leaves behind ionized dopant atoms: positive donor ions on the n-side and negative acceptor ions on the p-side. These fixed charges create an electric field that opposes further diffusion.

The region depleted of mobile carriers is called the depletion region or space charge region. In equilibrium, the drift current due to the built-in field exactly balances the diffusion current, resulting in zero net current. The built-in potential follows from the doping on each side:

V_bi = (kT/q) * ln(N_A * N_D / n_i²)

For a silicon junction with typical doping this yields roughly 0.6 to 0.8 V at room temperature. The logarithm makes the result insensitive to doping: raising either concentration by a factor of ten adds only about 60 mV. Because n_i appears in the denominator, wider-bandgap materials produce larger built-in potentials, which is why a gallium arsenide diode turns on near 1.2 V and a gallium nitride LED near 3 V while a silicon diode turns on near 0.7 V.

Charge neutrality requires that the depletion region hold equal amounts of positive and negative ionized charge, so it extends farther into the more lightly doped side. A one-sided junction, in which one region is far more heavily doped than the other, therefore places almost the entire depletion region and nearly all of the voltage drop in the lightly doped side. Power devices exploit this deliberately: a thick, lightly doped drift region supports the blocking voltage, and its required thickness sets the device's on-resistance. The width of the depletion region varies as the square root of the total junction voltage, so reverse bias widens it and forward bias narrows it.

Forward and Reverse Bias

Applying a voltage across the junction changes the balance between drift and diffusion. Forward bias (positive voltage on p-side) reduces the barrier, allowing more carriers to diffuse across the junction and producing significant current. The current increases exponentially with voltage according to the ideal diode equation:

I = I_s * (exp(qV/nkT) - 1)

Where I_s is the saturation current (typically 10^-12 to 10^-15 A for small silicon diodes) and n is the ideality factor. An ideality factor near 1 indicates that diffusion of minority carriers across the junction dominates, while a value approaching 2 indicates that recombination within the depletion region dominates, as it commonly does at low current. The exponential is steep: at room temperature, each 60 mV of additional forward bias multiplies an ideal diode's current by ten, which is the basis of the logarithmic converters and bandgap voltage references built from matched junctions.

Reverse bias (negative voltage on p-side) increases the barrier, reducing current to the small saturation current caused by thermally generated minority carriers. Real silicon diodes leak considerably more than the ideal equation predicts, because carriers are also generated at defect states inside the widened depletion region. The depletion region widens under reverse bias, a property exploited in varactor diodes for voltage-controlled capacitance.

Junction Capacitance

The depletion region acts as a parallel-plate capacitor, with the ionized dopants forming the "plates" and the depleted semiconductor as the dielectric. The depletion capacitance depends on voltage:

C_j = C_j0 / (1 - V/V_bi)^m

Where C_j0 is the zero-bias capacitance, V_bi is the built-in potential, and m is the grading coefficient: approximately 0.5 for an abrupt junction and about 0.33 for a linearly graded one. Hyperabrupt profiles, engineered by grading the doping, give exponents above 0.5 and a correspondingly wider tuning range, which is why they are used in voltage-controlled oscillators. Measuring capacitance against reverse bias is also a standard laboratory method for extracting a doping profile, since the slope of 1/C² versus voltage reveals the net doping at the moving depletion edge.

Under forward bias a second and usually larger capacitance appears. Injected minority carriers stored in the neutral regions must be removed before the junction can block again, an effect modeled as diffusion capacitance and observed as the reverse recovery time of a rectifier. Fast-switching diodes shorten it by deliberately reducing carrier lifetime, and Schottky diodes avoid it almost entirely by conducting with majority carriers only.

Junction Breakdown

Under sufficient reverse bias, a junction conducts heavily. Two distinct mechanisms produce this breakdown. Zener breakdown occurs in heavily doped junctions, where the depletion region is thin enough that valence-band electrons tunnel directly into conduction-band states across the gap. Avalanche breakdown occurs in more lightly doped junctions, where the depletion region is wide enough for carriers to accelerate to ionizing energies and multiply.

In silicon, tunneling dominates below roughly 5 V and avalanche above roughly 7 V, with both contributing in between. The two mechanisms have opposite temperature coefficients: the Zener mechanism grows easier with heating, so its breakdown voltage falls, while avalanche multiplication grows harder because hotter lattices scatter carriers before they reach ionizing energy, so its breakdown voltage rises. Near 5 to 6 V the two effects cancel, which is why breakdown diodes in that range are the most temperature-stable. Precision references exploit the same cancellation deliberately, placing a breakdown junction with a positive temperature coefficient in series with forward-biased junctions whose negative coefficient offsets it, and burying the breakdown region beneath the surface where its noise and long-term drift are lower.

Avalanche breakdown is not inherently destructive; devices fail from the heat that the resulting current produces, not from the multiplication itself. Power MOSFETs are therefore rated for a specific avalanche energy, and diodes intended for clamping duty are designed to spread the breakdown current uniformly across the die rather than allow it to concentrate at an edge or corner, where the field is highest.

Wave-Particle Duality

Electrons exhibit both wave and particle properties. In a semiconductor crystal, electron waves interact with the periodic potential of the lattice, and the wavelengths that scatter coherently from the lattice planes are forbidden, leaving gaps in the allowed energy spectrum. Bands and bandgaps are thus a direct consequence of wave interference in a periodic medium, closely analogous to the stopbands of an optical or microwave filter.

The de Broglie wavelength of a thermal electron in silicon is on the order of ten nanometers, which is comparable to the dimensions of modern transistor channels and larger than the thickness of many engineered layers. When a structure becomes smaller than this scale in one or more dimensions, carriers are confined and their energy levels become discrete. Quantum wells, wires, and dots exploit that confinement deliberately, and confinement in the inversion layer of a MOSFET shifts threshold voltage measurably in advanced devices.

Quantum Tunneling

Quantum mechanics allows electrons to penetrate through potential barriers that would be insurmountable according to classical physics. This tunneling effect is important in thin oxide layers, heavily doped p-n junctions (tunnel and Zener diodes), and flash memory, where charge is deliberately driven onto and off an isolated floating gate through an oxide barrier.

Tunneling current rises exponentially as a barrier thins, and that exponential is what ended the long practice of simply making gate oxides thinner with each process generation. As silicon dioxide approached one to two nanometers, gate leakage grew to consume an unacceptable share of chip power. The industry's answer was to replace silicon dioxide with hafnium-based high-permittivity dielectrics, which provide the same electrical coupling with a physically thicker and therefore far less transparent barrier. Intel introduced such a stack at its 45-nanometer node in 2007, and high-permittivity gate dielectrics have been standard in leading-edge logic ever since.

Fermi-Dirac Statistics

Electrons are fermions and obey Fermi-Dirac statistics, which give the probability that a state at energy E is occupied as f(E) = 1 / (1 + exp((E - E_F) / kT)). The Fermi level E_F is the energy at which that probability is exactly one half. When the Fermi level lies more than a few kT below the conduction band edge, the exponential dominates and the expression reduces to the simpler Boltzmann form used in most device analysis. Carrier concentrations then follow directly: n = n_i * exp((E_F - E_i) / kT) and p = n_i * exp((E_i - E_F) / kT), where E_i is the intrinsic Fermi level near midgap.

Doping moves the Fermi level: toward the conduction band in n-type material and toward the valence band in p-type. When doping is heavy enough to push it into a band, the semiconductor becomes degenerate, the Boltzmann approximation fails, and the material behaves increasingly like a metal. This condition is what makes heavily doped contact regions and polysilicon gates conduct well.

The most useful property of the Fermi level is that in thermal equilibrium it is constant everywhere in a connected system. Drawing a single flat Fermi level across a junction between two differently doped regions immediately forces the bands to bend, and the amount of that bending is the built-in potential. The same construction explains metal-semiconductor contacts, predicting whether a given combination yields a rectifying Schottky barrier or an ohmic contact. Under bias or illumination the system leaves equilibrium, and electrons and holes are described by separate quasi-Fermi levels whose separation measures the departure from equilibrium; in a solar cell, that separation is what appears at the terminals as photovoltage.

Recombination Mechanisms

Three main recombination mechanisms occur in semiconductors:

• Radiative recombination: The electron releases energy as a photon. This is the dominant mechanism in direct bandgap materials and is exploited in LEDs and laser diodes.
• Auger recombination: The energy is transferred to another carrier rather than emitted as light. This becomes significant at high carrier concentrations.
• Shockley-Read-Hall (SRH) recombination: Recombination occurs through defect states in the bandgap. This is often the dominant mechanism in indirect bandgap materials like silicon and is why crystal quality affects device performance. Deep states near midgap are the most effective, because they capture both carrier types readily; metallic contaminants such as iron, copper, and gold create exactly such states, which is why fabrication lines guard against parts-per-billion metal contamination.

These mechanisms act in parallel, and the fastest one sets the observed lifetime. Their differing dependence on carrier density has practical consequences. Auger recombination scales with the cube of carrier concentration, so it limits the efficiency of high-brightness LEDs, is a principal suspect in the efficiency droop those devices show at high drive current, and sets the ultimate efficiency ceiling of a silicon solar cell once all defects have been eliminated. Radiative and SRH processes scale less steeply, so SRH dominates in ordinary silicon at low injection.

Deliberate lifetime control is a standard engineering tool. Introducing gold or platinum, or bombarding a wafer with electrons or protons, adds recombination centers that shorten lifetime, speeding the turn-off of rectifiers and thyristors at the cost of higher forward voltage drop. Surfaces are themselves powerful recombination sites, since the abrupt end of the crystal leaves unsatisfied bonds; passivating them with a well-formed oxide or nitride layer is essential to solar cell and image sensor performance.

Optical Generation

A photon with energy exceeding the bandgap can promote an electron across it, creating an electron-hole pair. This is the mechanism behind every photodiode, image sensor, and solar cell. Excess energy above the bandgap is lost as heat within picoseconds as the carrier relaxes to the band edge, which is the fundamental reason a single-junction solar cell cannot exceed roughly a third of the incident sunlight in efficiency, and why multi-junction cells stack materials of differing bandgaps to capture each portion of the spectrum with a better-matched absorber.

Absorption depth varies enormously with wavelength. Blue light is absorbed within a fraction of a micrometer of a silicon surface, while near-infrared light penetrates deep into the wafer. Detectors and cells are designed around this: shallow junctions and good surface passivation for short wavelengths, thick and well-passivated bulk material with light-trapping surface texture for long ones.

Carrier Lifetime and Diffusion Length

The average time a carrier survives before recombining is called the carrier lifetime, typically ranging from nanoseconds to milliseconds depending on material quality and type. The diffusion length is the average distance a carrier diffuses before recombining:

L = sqrt(D * tau)

Where D is the diffusion coefficient and tau is the lifetime. The square root makes diffusion length a forgiving measure: a hundredfold improvement in lifetime buys only a tenfold increase in the distance a carrier travels.

The consequences are direct. A solar cell only collects a carrier if the carrier reaches the junction before recombining, so a wafer must be thinner than the diffusion length or the material must be good enough that the diffusion length exceeds the wafer thickness. High-quality silicon with lifetimes in the hundreds of microseconds gives diffusion lengths from several hundred micrometers to more than a millimeter, comfortably larger than a modern cell wafer of roughly 150 micrometers, which is precisely why cell manufacturers invest so heavily in contamination control and surface passivation. Bipolar transistors depend on the same quantity from the opposite direction: their base must be much thinner than the minority-carrier diffusion length so that injected carriers cross it and reach the collector rather than recombining on the way.