How an MRI Scanner Works: Quantum Spin, Superconducting Magnets, and the Mathematics of Seeing Inside the Human Body

A deeply technical exploration of MRI from first principles — quantum mechanics of proton spin, the Larmor equation, 1.5T/3T superconducting NbTi magnets at…

·

How an MRI Scanner Works: Quantum Spin, Superconducting Magnets, and the Mathematics of Seeing Inside the Human Body

A magnetic resonance imaging scanner is the most complex diagnostic instrument in modern medicine. It combines quantum mechanics, superconductivity, cryogenics at 4.2 Kelvin, radio-frequency engineering at 64–128 MHz, precision electromagnetics with ppm-level homogeneity, and advanced signal processing — all operating simultaneously in a clinical environment where a single mistake could injure a patient or quench a $200,000 superconducting magnet.

Unlike X-ray or CT, which measure attenuation of radiation passing through tissue, MRI manipulates the quantum mechanical spin states of hydrogen nuclei — the protons in water and fat molecules that constitute roughly 63% of the human body by atom count. By placing the body in an extraordinarily strong and uniform magnetic field, exciting those protons with precisely tuned radio-frequency pulses, and then listening to the faint radio signals they emit as they relax back to equilibrium, an MRI scanner reconstructs a three-dimensional map of proton density and chemical environment with sub-millimetre resolution.

This guide covers the entire engineering stack — from the quantum mechanics of a single proton to the Fourier transform that produces the final image. Every equation is rendered as the physics demands it: E = \hbar \omega, not "energy equals Planck's constant times frequency."

1. Quantum Mechanics of Nuclear Spin: The Proton as a Magnetic Dipole

The Spin-½ System

Every hydrogen nucleus — a single proton — possesses an intrinsic quantum mechanical property called spin angular momentum, quantized with spin quantum number I = 1/2. In a magnetic field \mathbf{B}_0 oriented along the z-axis, this spin-½ system has two possible energy eigenstates:

Where:

The two energy levels correspond to the proton's magnetic moment aligned parallel (m_I = +1/2, lower energy) or anti-parallel (m_I = -1/2, higher energy) to the external field. The energy difference between these states:

The Larmor Equation

A photon with exactly this energy difference can induce a transition between spin states. The frequency of this photon — called the Larmor frequency — is the single most important equation in all of MRI:

Or in linear frequency:

For a 1.5 Tesla scanner (the most common clinical field strength): f_0 = 63.87 \text{ MHz}. For 3.0 Tesla: f_0 = 127.74 \text{ MHz}. These frequencies fall in the radio-frequency (RF) range — between FM radio (~100 MHz) and VHF television — which is why MRI uses RF coils and why the scanner room requires a Faraday cage to block external RF interference.

Boltzmann Statistics and Net Magnetization

In a macroscopic sample at thermal equilibrium at temperature T, the populations of the two energy states follow the Boltzmann distribution:

At body temperature (T = 310 K) and B_0 = 1.5 T:

This means that out of every ~2,000,000 protons, only about 7 more are aligned parallel than anti-parallel. This minuscule excess — roughly 3–7 parts per million — is the entire source of the MRI signal. The net magnetization vector \mathbf{M}_0 that results from this population difference:

This is roughly 10^{-6} of the total nuclear magnetization — an extraordinarily weak signal that explains why MRI requires such powerful magnets, sensitive receive coils, and extensive signal averaging.

2. The Superconducting Magnet: 1.5–7 Tesla at 4.2 Kelvin

Why Superconductivity?

The signal-to-noise ratio (SNR) in MRI scales approximately linearly with B_0:

A permanent magnet cannot exceed ~0.5 T in a practical clinical bore size. A resistive electromagnet at 1.5 T would dissipate megawatts of power as I^2R heating in the copper windings and require impossible cooling. The only practical solution for MRI field strengths above ~0.5 T is a superconducting electromagnet — a coil wound from niobium-titanium (NbTi) alloy wire that carries current with zero electrical resistance when cooled below its critical temperature.

NbTi Superconductor Physics

NbTi is a Type-II superconductor with a critical temperature T_c = 9.2 K. Below this temperature, electrons pair into Cooper pairs via phonon-mediated attraction, condensing into a macroscopic quantum state that conducts current without scattering. The critical current density at 4.2 K and 5 T is approximately:

A typical 1.5 T MRI magnet contains approximately 120–150 km of NbTi wire, wound into 6–8 main coils with additional shim coils for field homogeneity. The wire itself is a multifilamentary composite — thousands of NbTi filaments, each ~5–10 μm in diameter, embedded in a copper matrix. The copper serves as a quench protection pathway: if a local region of the superconductor transitions to the normal (resistive) state, the copper provides an alternative current path, preventing catastrophic localized heating.

Magnetic Field Configuration

The main magnet coil produces an approximately uniform magnetic field along the z-axis (the bore direction) over a spherical imaging volume typically 45–50 cm in diameter. The field strength at the center of a solenoid of radius R and length L with N turns and current I:

For a 1.5 T magnet with R \approx 0.5 m and L \approx 1.5 m: NI \approx 2.0 \times 10^6 ampere-turns. The inductance of such a magnet is enormous — typically 50–100 henries — meaning the stored magnetic energy E = \frac{1}{2}LI^2 is on the order of 5–10 megajoules.

Cryogenics: 4.2 Kelvin Liquid Helium

The NbTi coil is immersed in a bath of liquid helium at atmospheric pressure, maintaining a temperature of 4.2 K. The liquid helium vessel (the "helium can") is surrounded by:

  1. A liquid nitrogen shield at 77 K (in older magnets — modern "zero-boiloff" designs eliminate the LN₂ jacket)
  2. Multiple layers of aluminized mylar superinsulation in a high-vacuum cryostat (< 10^{-6} mbar)
  3. A cryocooler coldhead (typically a Gifford-McMahon or pulse-tube refrigerator) that recondenses boiled-off helium gas, achieving near-zero helium consumption

A typical 1.5 T MRI magnet contains 1,500–2,000 litres of liquid helium. The cost of this helium alone is approximately ₹15–25 lakhs ($18,000–30,000) at current market prices. Helium refill is required every 3–5 years for older magnets; modern zero-boiloff systems can operate indefinitely without refill, recondensing evaporated helium through the cryocooler.

Field Homogeneity: Parts Per Million

For imaging, the main magnetic field must be uniform to within 1–5 parts per million (ppm) over the imaging volume — meaning a field variation of less than 5 μT in 1.5 T. Achieving this requires:

  1. Passive shimming: Steel plates or bags of iron filings placed in the magnet bore, correcting lower-order field inhomogeneities
  2. Active shimming: Room-temperature shim coils carrying adjustable DC currents, correcting higher-order spherical harmonic components of the field inhomogeneity (up to 3rd or 5th order)
  3. Superconducting shims: Additional superconducting coils inside the helium vessel, energized during magnet ramp-up and operating in persistent mode

The field inhomogeneity after shimming can be expressed as a spherical harmonic expansion:

Where P_n^m are associated Legendre polynomials and C_{nm} are the shim coefficients minimized during the shimming procedure.

3. Radio-Frequency Excitation: Flipping Spins

The Rotating Frame and the Bloch Equations

The quantum mechanical description of a single spin is exact but impractical for describing the macroscopic magnetization \mathbf{M}(t) that generates the measurable signal. The classical Bloch equations describe the time evolution of \mathbf{M} in the presence of both the static field \mathbf{B}_0 and an applied RF field \mathbf{B}_1(t):

Where T_1 is the longitudinal (spin-lattice) relaxation time and T_2 is the transverse (spin-spin) relaxation time — both tissue-specific constants that provide the primary contrast mechanisms in MRI.

The B₁ Field and Flip Angle

To generate a signal, the magnetization must be tipped away from the z-axis into the transverse (xy) plane. This is achieved by applying a circularly polarized RF magnetic field \mathbf{B}_1(t) oscillating at the Larmor frequency, oriented perpendicular to \mathbf{B}_0:

In the rotating frame — a coordinate system rotating at \omega_0 about the z-axis — the effective field is simply B_1 \hat{\mathbf{x}}', and the magnetization precesses about this field at frequency \omega_1 = \gamma B_1.

The flip angle \alpha achieved by an RF pulse of duration \tau:

For a rectangular pulse: \alpha = \gamma B_1 \tau. A 90° pulse tips the magnetization entirely into the transverse plane; a 180° pulse inverts it.

For a typical clinical MRI at 1.5 T, B_1 \approx 10-20 μT (microtesla), and a 90° pulse duration is typically 2–5 milliseconds. The RF power deposited in the body is quantified by the Specific Absorption Rate (SAR) — regulated to <2 W/kg (whole body) and <3.2 W/kg (head) by IEC standards.

4. Spatial Encoding: Gradient Coils and the Mathematics of Slice Selection

If all protons in the body experience the same magnetic field B_0, they all resonate at the same Larmor frequency — and the received signal contains no spatial information. To create an image, the magnetic field must be made spatially varying so that position is encoded in frequency and phase.

Gradient Coil Physics

Three orthogonal sets of gradient coils produce linear magnetic field variations in the x, y, and z directions:

Where \mathbf{G}(t) = (G_x(t), G_y(t), G_z(t)) is the time-varying gradient vector. The Larmor frequency becomes position-dependent:

The gradient coils are typically designed as Maxwell pairs (for z-gradients) and Golay coils (for x and y gradients). A Maxwell pair consists of two coaxial circular loops with opposite currents, separated by \sqrt{3}R (where R is the radius), producing a linear field gradient along the z-axis at the centre:

Gradient strengths in clinical MRI range from 20–80 mT/m, with slew rates (rate of change of gradient) of 100–200 T/m/s. These rapidly switched gradients produce loud acoustic noise — up to 110–120 dB — due to Lorentz forces on the gradient coil conductors within the strong static field:

Slice Selection

To image a specific slice perpendicular to the z-axis, a slice-selective RF pulse is applied simultaneously with a z-gradient G_z. The RF pulse is amplitude-modulated with a sinc function profile — the Fourier transform of the desired rectangular slice profile:

The slice thickness \Delta z is determined by the RF bandwidth \Delta f and the gradient strength:

For example, with \Delta f = 2 kHz and G_z = 10 mT/m:

5. K-Space and Image Reconstruction

The Signal Equation

After RF excitation, the precessing transverse magnetization induces a voltage in the receive coil via Faraday's law. The received signal from a volume element dV at position \mathbf{r} with spin density \rho(\mathbf{r}):

Defining the k-space vector:

The signal equation simplifies to:

This is the fundamental equation of MRI: the acquired signal S(\mathbf{k}) is the spatial Fourier transform of the proton density \rho(\mathbf{r}). The image is recovered by the inverse Fourier transform:

K-Space Trajectory

The gradients \mathbf{G}(t) control the trajectory through k-space. A standard 2D spin-echo sequence traverses k-space line by line:

  1. Slice-selective 90° pulse + G_z → excites the slice
  2. Phase-encoding gradient G_y applied for duration \tau_{pe} → positions the trajectory at k_y = (\gamma/2\pi) G_y \tau_{pe}
  3. Frequency-encoding (readout) gradient G_x applied during signal acquisition → sweeps k_x from -k_{x,\max} to +k_{x,\max}
  4. The process repeats for N_y different G_y values, acquiring N_y lines of k-space

The k-space step size determines the field of view (FOV):

The maximum k-space extent determines the spatial resolution:

Where N is the number of acquired k-space points in that dimension. A typical clinical brain MRI with FOV = 24 cm and matrix = 256 × 256 gives in-plane resolution of approximately 0.94 \times 0.94 mm.

The 2D Discrete Fourier Transform

In practice, the image reconstruction is performed via the 2D discrete Fourier transform (DFT), typically implemented as a Fast Fourier Transform (FFT):

The magnitude image (what radiologists view) is:

The phase image \angle I(m,n) contains valuable information about magnetic field inhomogeneity, flow, and susceptibility — and is used in advanced techniques like phase-contrast angiography and susceptibility-weighted imaging.

6. Contrast Mechanisms: T₁, T₂, and Proton Density

Different tissues produce different signal intensities because they have different relaxation times. The signal intensity in a spin-echo sequence as a function of sequence parameters:

Where:

Tissue · T₁ at 1.5T (ms) · T₂ (ms) · Proton Density

White matter · ~780 · ~90 · 0.70

Grey matter · ~920 · ~100 · 0.83

Cerebrospinal fluid (CSF) · ~4,000 · ~2,000 · 1.00

Fat · ~260 · ~80 · 0.90

Muscle · ~870 · ~47 · 0.75

T₁-Weighted Imaging

Short TR (\ll T_1 of most tissues) + short TE (\ll T_2):

Tissues with short T₁ (fat) appear bright. Tissues with long T₁ (CSF) appear dark. Excellent anatomical detail.

T₂-Weighted Imaging

Long TR (\gg T_1) + long TE (comparable to T₂):

Tissues with long T₂ (CSF, edema, inflammation) appear bright. Pathological tissue is often highlighted.

7. Signal, Noise, and the Engineering Limits

Johnson-Nyquist Noise

The fundamental noise source in MRI is thermal (Johnson-Nyquist) noise in the receive coil and the patient. In a coil of resistance R at temperature T, the RMS noise voltage:

Where \Delta f is the receiver bandwidth. At body temperature and typical receiver bandwidth of 16–64 kHz, this translates to noise floors of approximately 10–100 nanovolts.

Signal-to-Noise Ratio

The SNR for a voxel of volume \Delta V in a single acquisition:

Parallel Imaging

Modern MRI uses multiple receive coil elements (8, 16, 32, or even 64 channels) in phased arrays. Each coil element is sensitive to a different spatial region, providing additional spatial encoding that can be exploited to reduce scan time:

The geometry factor (g-factor) quantifies the SNR penalty of parallel imaging:

A well-designed 32-channel head coil at 3 T can achieve g \approx 1.0-1.2 in central brain regions.

8. The Complete Scanner: From Helium Fill to Clinical Image

System Architecture

Subsystem · Function · Key Specifications

Superconducting magnet · Generate uniform B₀ field · 1.5 T / 3.0 T, < 5 ppm homogeneity over 45 cm DSV

Cryogenic system · Maintain 4.2 K · 1,500–2,000 L liquid helium, zero-boiloff cryocooler

Gradient coils · Spatial encoding · 40–80 mT/m, 200 T/m/s slew rate

Gradient amplifiers · Drive gradient coils · 1–2 kA peak current, ±2 kV

RF power amplifier · Generate B₁ excitation pulses · 15–35 kW peak power at 64/128 MHz

RF receive chain · Detect microvolt-level NMR signals · 8–64 channel phased array, 16-bit ADC at 1–5 MS/s

Faraday cage · Shield from external RF interference · > 90 dB attenuation at Larmor frequency

Patient table · Position patient at isocentre · ±1 mm positional accuracy

Reconstruction computer · FFT, filtering, display · GPU-accelerated, < 1 second per slice

A Complete Spin-Echo Pulse Sequence Timing

A single 2D spin-echo acquisition with 256 phase-encoding steps at TR = 500 ms:

With parallel imaging (R = 2): T_{scan} = 64 seconds. A standard clinical brain protocol (T₁, T₂, FLAIR, DWI) takes approximately 15–25 minutes total scan time.

9. The Future: 7T, 10.5T, and Ultra-High Field MRI

Research scanners now operate at 7 T (Siemens Terra, GE SIGNA 7T) and even 10.5 T (University of Minnesota). At these field strengths:

The fundamental trade-off remains — higher field = higher SNR = higher cost. A 7 T scanner costs approximately ₹40–60 crores (5–7 million) versus ₹8–12 crores (1–1.5 million) for a 1.5 T system. The liquid helium volume scales approximately with B_0^2, and the fringe field — the region outside the magnet where the field exceeds 5 gauss (0.5 mT) — extends much farther, demanding larger shielded rooms and stricter access control.

An MRI scanner is not a camera that takes pictures of anatomy. It is a quantum measurement device that manipulates nuclear spins, encodes spatial information in frequency and phase, and reconstructs images through Fourier transformation — all while maintaining a magnetic field a hundred thousand times stronger than Earth's, cooled to within four degrees of absolute zero, inside a hospital where patients with metal implants, pacemakers, and anxiety must be kept safe. The fact that this machine exists, works reliably across tens of thousands of installations worldwide, and produces diagnostic images that save millions of lives every year — that may be the greatest engineering achievement in the history of medicine.

Last updated: July 2026

Sources: Haacke, Brown, Thompson, Venkatesan — "Magnetic Resonance Imaging: Physical Principles and Sequence Design" (Wiley, 2014); Bernstein, King, Zhou — "Handbook of MRI Pulse Sequences" (Elsevier, 2004); Siemens Healthineers and GE Healthcare technical documentation; NEMA MS-5 Standards for MR magnet safety; IEC 60601-2-33 Safety standard for MR equipment.

More FabFlow blog posts