An observatory does not have to take a picture of the sky
Thousands of light-sensitive spheres hang beneath Antarctic ice. At two other observatories, a laser travels between mirrors inside long vacuum tubes. Both instruments use light. Does that make them ordinary optical telescopes with unusual shapes?
No. An electromagnetic telescope receives photons from a celestial object; a neutrino detector reconstructs a particle’s arrival from the consequences of its interaction; a gravitational-wave detector measures changes in spatial distances caused by a passing wave. Light inside an instrument can be a secondary signal or a measuring ruler, rather than a message from a star. Neutrinos and gravitational waves are not further bands of the electromagnetic spectrum beyond radio or gamma rays.
The word “telescope” sometimes has a broad meaning: an instrument for astronomical observation. Here we will identify the physical messenger each time. We will examine IceCube and LIGO: what arrives, how a record forms, and how researchers check the inference. Useful prerequisites are the independent axes of telescope classification and different bands and detection methods. You can skip the mathematical branches. IceCube: the detection principle, LIGO: what the observatory measures.
A different messenger reveals a different part of an event
Light interacts with matter: it is emitted, absorbed and scattered. This lets us investigate composition and temperature, but it can hide an interior region. Neutrinos interact far less often and can carry information from conditions opaque to photons. That penetrating ability has a price: most neutrinos cross a detector without leaving a record. We need a large volume of matter and a way to notice a rare interaction.
Gravitational waves provide another kind of information: the dynamics of masses, such as the approach and merger of compact objects. They are neither a stream of neutrinos nor “light with an extremely long wavelength.” The detector looks for a patterned change in relative distances, rather than a source’s color. Combining these channels is called multi-messenger astronomy: different physical carriers test different parts of one explanation.
Historical milestones show that this is already observational practice. The last string of the baseline IceCube array was lowered on December 18, 2010, and its first fully instrumented IC86 run began on May 13, 2011; data had also been collected during construction. On September 14, 2015, the two LIGO detectors recorded GW150914: the first direct detection of gravitational waves, from merging black holes. The paper appeared in February 2016: the announcement and the event have different dates. IceCube timeline, GW150914 primary publication.
First identify what actually arrives from space
Similar photosensors do not imply identical astronomy. In an ordinary optical telescope, a celestial photon travels through the optics to the detector. In IceCube, the recorded photon originated in terrestrial ice. In LIGO, a local laser emitted it. This difference determines what the data can tell us.
| Messenger | Interaction | Record | Inference |
|---|---|---|---|
| Photon · EM | Optics → absorption in detector | Position, time, charge or energy | Brightness, spectrum, image |
| Neutrino | Rare interaction → charged particles → light in ice | Pulse times and sizes across many DOMs | Direction and energy with uncertainty |
| Gravitational wave | Distance changes → phase of local laser | Photocurrent and control signals | Calibrated strain and source parameters |
A detector record is not yet an image of the source. A physical model connects signals with direction, energy or merger parameters; calibration establishes the scale; estimating the background tells us how convincing an event is. Here, “seeing” means obtaining checked evidence, rather than necessarily forming a photograph.
IceCube sees light from secondary charged particles
A neutrino is electrically neutral. It does not leave a continuous glowing line in IceCube simply by flying past the spheres. A rare interaction produces secondary particles. A charged particle moving faster than the phase velocity of light in ice emits Cherenkov light. It does not exceed the speed of light in vacuum: light propagates more slowly in matter.
In one class of events, a muon travels a long distance and leaves an elongated light pattern: a track. Other interactions produce a more compact cascade pattern. Sensors record the timing and size of light pulses; their distribution allows direction and energy to be estimated. A secondary particle’s direction does not always coincide exactly with the original neutrino’s direction, and light can scatter or be absorbed on its way. IceCube: neutrinos, secondary light and reconstruction.
The diagram uses a dashed line for the neutrino before its interaction, a solid line for an illustrative charged track, and blue rays for direct secondary photons. The small spheres represent six hypothetical selected sensors. They are not photographing a star through the ice.
Dashed: neutrino; rust track: charged particle; blue segments: direct secondary photons; circles: hypothetical DOMs.
Branch · With formulasOptional detail; the main explanation continues below.
Why arrival times form a consistent pattern
Take our simplified example: homogeneous nondispersive ice with index n = 1.32, and a charged particle moving along the x axis at a speed approximating c. Time starts at the vertex x_(v) = − 300 m. A sensor has coordinate x and perpendicular distance r from the track. A direct photon must leave at a position x_(e) that satisfies the Cherenkov-angle condition:
cosθ = (1)/(n),; x_(e) = x − (r)/(√((n)^(2) − 1)).
The particle first reaches x_(e), then the photon travels to the sensor at speed c/n. Adding those two times gives:
t = (x − x_(v) + r√((n)^(2) − 1))/(c).
In our geometry, the sensor at x = 0, situated 100 m above the track, receives a direct photon about 1288 ns after the time origin. Moving such a sensor 125 m along the track adds about 417 ns. One event produces mutually related times, rather than just a count of illuminated spheres.
This is an explanatory model, not IceCube’s reconstruction algorithm. Setting n_(phase) = n_(group) deliberately removes dispersion: in real ice, the phase index determines the angle, whereas the group index determines pulse timing. Ice layers, scattering, sensor acceptance and the randomness of emission are omitted. These calculated numbers are not a measured event. Primary discussion of phase and group velocity, eq.2 and §2.
The sphere is a sealed measurement module
The baseline Digital Optical Module, DOM, contains a photomultiplier, high-voltage supply, digitizing electronics, its own clock and calibration LEDs. A photon releases an electron at the photocathode; a multiplication cascade inside the photomultiplier turns a weak optical signal into an electrical pulse. Electronics record its waveform and time and send the data to the surface. A cable carries power and communications: there is no separate GPS receiver inside every sphere.
The glass housing must transmit the relevant light, withstand pressure and avoid excess background from radioactive impurities. The baseline design used borosilicate glass with low potassium and other radioactive trace-element content. Clear silicone gel between the lower housing and photomultiplier provides optical contact and support; it does not “amplify neutrinos.” A mechanical harness carries the string load around the glass sphere. Primary DOM description: §2.1–2.2, Fig.3–4.
- Photon in ice
- Transparent glass
- Optical gel
- Photocathode → electron
- PMT → amplified pulse
- Electronics → digital waveform
- Clock → local time
- Cable → surface data
RAPCal ↔ surface time base ↔ UTC
LED flashes → check timing and medium
Assembly begins long before drilling
The baseline DOM production description explains how the photomultiplier was accurately positioned in the lower glass hemisphere. Gel was mixed and degassed under partial vacuum to remove bubbles; after the photomultiplier was inserted, the gel was allowed to cure. Electronic boards were assembled separately, the sealed cable penetrator was installed, and the hemispheres were joined and sealed. Production required control of static electricity, temperature and humidity. Otherwise, a tiny contact or optical-layer defect would remain deep in the ice for years.
Before shipment, modules underwent temperature cycling and cold soaking in dark freezer laboratories. Tests covered booting, electronics, dark noise, gain, timing response and linearity; external light sources supplied known stimuli. Modules were tested again at the South Pole to identify shipping damage. These are published tests by the manufacturers and collaboration; this article includes no physical test of our own. Assembly and Final Acceptance Testing, §2.3.
In the ice, a hot-water drill creates a deep hole, reaming it to the required profile while the drill is raised. Before the water refreezes, the team successively attaches modules to a string, connects the cable and lowers the system. Surveying, pressure readings and distance checks establish initial coordinates. Once frozen in, the modules cannot simply be pulled up for repair. Production control, drilling pace and the short Antarctic summer season therefore form one engineering problem. Drilling and deployment, §5.1–5.3.
Both the clocks and the ice need calibration
The South Pole supplies a large volume of deep, dark ice and a research station’s infrastructure. Compression of deep layers removes the main effect of air bubbles, while overlying ice shields part of the surface background. The site was chosen for its measurement medium and the ability to instrument it, rather than a special neutrino “window” found only over Antarctica. IceCube: why the South Pole.
A kilometer of ice is not a perfectly uniform, transparent glass block. Dust layers and other optical variations change absorption and scattering. Mistaking a late photon for a direct one changes the reconstructed direction. LED flashes from known modules and ice studies help separate the medium’s properties from those of an event. Flashes also refine the modules’ relative positions.
DOM clocks drift. RAPCal regularly sends a pulse down the cable and a reply back up, recording timestamps at both ends. Reciprocity of the transmission path and matching pulse shapes connect local time to the common surface time base, and then to UTC. Calibration LEDs check the entire path through to the recorded photon time. Merely switching all modules on together would not suffice. Time base and verification, §3.3; geometry, §5.3.
There is also a physical background: cosmic rays produce atmospheric muons and neutrinos. A track in the ice does not by itself establish a cosmic neutrino source. Selection uses direction, energy, event-start geometry and outer-module information; the appropriate combination depends on the analysis. This is a statistical problem, rather than “every sphere means a discovery.”
The familiar 5160 in-ice DOMs on 86 strings, at depths of 1450–2450 m, describe baseline IceCube. On February 12, 2026, the operator reported the completed deployment of the Upgrade: five new dense strings with over 600 new sensors and calibration instruments. New module types should not be reduced to the old “one sphere, one photomultiplier” picture. Our diagram explains a principle, rather than the complete current inventory. Baseline detector, actual 2026 expansion.
LIGO compares two arms using an internal laser
Imagine two perpendicular distances between nearly free test masses. A gravitational wave can change those lengths differently. In a simple favorable orientation, one arm lengthens slightly as the other shortens, before the changes reverse. The actual response depends on arrival direction, polarization and wave frequency. This is a change in spacetime geometry, rather than sound pushing the tubes through air.
In a basic Michelson interferometer, a partly transmitting beam splitter divides a laser beam in two. Each part travels along its arm, reflects from the end mirror and returns. When recombined, the waves reinforce or weaken one another depending on their phase difference: the relative positions of their crests. A photodetector measures the resulting power. Changing the arm-length difference changes phase, making the optical record sensitive to test-mass motion. LIGO: the basic interferometer.
| ΔL / λ | Dark output p_d | Other output p_b |
|---|---|---|
| 0 | 0.00 | 1.00 |
| 0.125 | 0.50 | 0.50 |
| 0.25 | 1.00 | 0.00 |
| 0.5 | 0.00 | 1.00 |
The laser is a local measuring ruler. A celestial object did not send a light beam down the tubes. Unlike an ordinary telescope, this instrument is not turned toward a star to collect its photons.
Branch · With formulasOptional detail; the main explanation continues below.
The round trip, phase and the dark port
Let ΔL = L_(x) − L_(y) be the one-way arm-length difference and λ the laser wavelength. After traveling out and back, the optical-path difference is 2ΔL. For an ideal simple instrument:
ϕ = (4πΔL)/(λ),; p_(d) = (sin)^(2) ((ϕ)/(2)),; p_(b) = (cos)^(2) ((ϕ)/(2)).
p_(d) and p_(b) are the fractions of input power in the two output ports of an ideal instrument with 50/50 splitting, no loss and properly matched fields. They add to one: light that disappears from one output through interference does not disappear from the energy balance. At ΔL = 0, one port is dark; at ΔL = λ/4, it reaches maximum brightness. One full cycle corresponds to changing ΔL by λ/2.
Near a perfectly dark port, power depends quadratically on a small displacement and does not distinguish its sign. This formula therefore cannot stand for LIGO’s entire readout. In Advanced LIGO, a small operating-point offset supplies a reference optical field for DC readout; additional cavities and controls change the full response. The formula explains interference, rather than calibrating real data. Advanced LIGO: §2.3 and §4.8.
The real instrument adds cavities and control
LIGO has two main sites: Hanford, Washington, and Livingston, Louisiana, roughly 3000 km apart. Each has two arms about 4 km long. Site selection sought a large, sufficiently flat area, quiet seismic conditions and accessible infrastructure; a large separation helps distinguish local disturbances and compare arrival times. LIGO: choosing the sites. Advanced LIGO uses a main laser wavelength of 1064 nm. These are parameters of a particular design, rather than universal requirements for every gravitational-wave detector. Operator’s site description, primary specification, Table1.
In each arm, an additional input mirror and the end mirror form a Fabry–Perot cavity: light accumulates length information through repeated circulation. Power-recycling mirrors return useful optical power, while signal recycling shapes the response to signal changes. This is a frequency-dependent wave-resonance system, rather than a photon always making the same fixed number of bounces.
Photodiodes and feedback keep cavities and mirrors at their operating point. An output mode cleaner removes unwanted spatial and frequency components before the main readout. Our simple diagram therefore shows only the arm-comparison mechanism: it omits much of Advanced LIGO’s optics. Actual optical configuration, Fig.1, §2 and §4.8.
Branch · With formulasOptional detail; the main explanation continues below.
What a relative change around 10⁻²¹ means
Take the long-wavelength limit: the wave arrives perpendicular to the arm plane, with its plus polarization aligned with the arms. For a small dimensionless strain, h, one arm changes by about + hL/2, the other by − hL/2. The differential change is then:
ΔL = hL, h = (ΔL)/(L).
For our chosen L = 4000 m and h = (10)^( − 21), this gives ΔL = 4 × (10)^( − 18) m. A single round trip of light at λ = 1064 nm would acquire a phase shift of about 4.72 × (10)^( − 11) rad. Its travel-time difference would be about 2.67 × (10)^( − 26) s. Nobody starts a stopwatch with that time step: interference is measured and the signal reconstructed through a calibrated response.
This is an assigned scale, rather than an estimate of a particular event or the full cavity response. A different direction or polarization, or a wave period no longer long compared with the light-travel time, requires the appropriate model. Nor should we imagine everything, including the light “ruler,” stretching identically and automatically cancelling the measurement: the instrument compares the phase of light propagating between test masses. Advanced LIGO: differential readout, LIGO: light as a ruler and arm changes.
Mirrors, suspensions and vacuum work together
Advanced LIGO’s test masses are large fused-silica mirrors; the 2015 specification gives a mass of 40 kg. The blank is shaped and polished, large-scale deviations are corrected with an ion beam, and a multilayer dielectric coating is deposited. Alternating silica and tantalum-oxide layers, with titanium oxide added to the test masses’ tantalum-oxide layers, provide the required reflection. This is unlike the ordinary metal film of an amateur mirror: absorption, scattering and mechanical coating losses all affect noise. Surfaces and coatings undergo optical metrology during production. Optical materials, fabrication and metrology, §4.3.
Suspensions isolate mirrors from ground motion. Test masses use four-stage pendulums; an adjacent reaction chain allows controlled forces to be applied. The bottom silica assembly uses separate silica ears bonded to the masses and drawn fibers; the fibers are annealed and laser-welded. Geometry, strength and internal friction matter together. Active seismic isolation and suspensions complement one another, without making a mirror motionless at every frequency. Suspensions and isolation, §4.4–4.5.
Vacuum removes most gas from the optical path, reducing its effect on phase and protecting clean surfaces from contamination. Steel tubes are welded, cleaned, baked and pumped down; pumps and monitoring maintain the operating condition. Vacuum does not eliminate thermal motion of materials, quantum fluctuations of light or vibrations transmitted through solid structures. The longest tube alone is therefore not a sensitive detector. LIGO: making and maintaining the vacuum system.
A photocurrent change needs calibration and background checks
A photodiode supplies an electrical record, rather than immediately producing a graph of h(t). Researchers must know the frequency response of the optics, electronics and controls. Since feedback itself moves the mirrors, reconstructing an external differential change uses both error and control signals. Reading only the residual at one sensor would not give the entire wave.
The primary Advanced LIGO description presents several cross-checking calibration methods: interference fringes referenced to the laser wavelength, known frequency modulation, and a photon calibrator. The latter uses a separate modulated laser to exert known radiation pressure on a mirror. Two symmetrically offset beams reduce unwanted torque. This checks the relation between an applied influence and recorded response; it does not create a genuine astrophysical gravitational wave. Calibration, §4.9.
Seismometers, magnetometers, microphones and other auxiliary sensors help identify terrestrial disturbances. Agreement between distant detectors and the expected signal shape strengthens the evidence, but noise, calibration reliability and the statistical rate of accidental coincidences still matter. Successful measurement combines optics, mechanics and analysis, rather than merely a more powerful laser. Environmental monitoring, §4.10, GW150914 significance estimate.
A network narrows the search area rather than drawing a point
A wave reaches distant stations at different times. The difference depends on the projection of their separation onto the propagation direction. The same delay between two stations is compatible with many sky directions. Additional stations, relative amplitudes and polarization response narrow the area, but its size also depends on data quality.
This matters for ordinary telescopes: a gravitational-event alert supplies an area to search for a light counterpart. Field of view and response speed can matter more than the highest magnification. Neutrino alerts also provide a direction with uncertainty; one particle does not always uniquely identify its galaxy.
Baseline A → B: 3000 km
| Propagation vector n | Delay B−A · ms |
|---|---|
| (0.500, 0.866, 0.000) | 5.003 |
| (0.500, 0.000, 0.866) | 5.003 |
| (0.500, -0.866, 0.000) | 5.003 |
Maximum |τ|: 10.007ms. Toward the source: −n.
Branch · With formulasOptional detail; the main explanation continues below.
One delay leaves a whole family of directions
Let \mathbf b be the vector from station A to B, and \mathbf n the unit propagation vector of a plane wave. Their arrival-time difference is:
\begin{aligned} \tau&=t_B-t_A=\frac{\mathbf b\cdot\mathbf n}{c},\\ |\tau|&\leq\frac{|\mathbf b|}{c}. \end{aligned}
The direction toward the source is opposite: -\mathbf n. In our example, \mathbf b=(3000,0,0)\,\mathrm{km} and \mathbf n=(1/2,\sqrt3/2,0). The result is 5.003 ms, with a maximum absolute delay of 10.007 ms for this baseline. Rotating the transverse component of \mathbf n around the baseline while keeping its first coordinate 1/2 does not change the delay. On the celestial sphere this defines a circle of possible directions; timing uncertainty broadens it into a band.
This is our Euclidean geometry with ideal synchronized stations, rather than exact Hanford/Livingston coordinates or a probability map for a real event. It explains why one time difference is insufficient. Actual analysis also uses a signal model, network characteristics and measurement uncertainties.
GW170817: different channels tested one event
On August 17, 2017, LIGO and Virgo observed the gravitational signal GW170817, consistent with two neutron stars merging. About 1.7 s after the merger time, Fermi GBM independently detected gamma-ray burst GRB170817A. Less than 11 hours later, the 1M2H team used the Swope telescope to discover an optical counterpart in NGC4993; other teams independently found it too. Follow-up observations traced its changing color and fading, and later its X-ray and radio emission. Joint primary publication, abstract and chronology.
The gravitational signal constrained the system’s dynamics and masses; light identified the galaxy and helped investigate ejected matter and subsequent evolution. These are complementary measurements, rather than several identical photographs. The 1.7 s difference alone does not mean one messenger simply traveled more slowly: the emissions could have originated at different times in the source.
The reported searches found no neutrino candidates consistent with this source. This constrains models and search sensitivity, rather than proving that the source emitted no neutrinos. IceCube’s participation in a joint paper does not turn a non-detection into a detection.
Not every astronomical “eye,” then, collects light from an object. Follow messenger → interaction → record → inference. An optical telescope, a cubic kilometer of ice and a laser interferometer can study the same sky while answering different questions. For further reading, see gamma-ray telescope construction and converting a high-energy photon into a record: there the original messenger remains electromagnetic, even when there is no ordinary mirror-formed image.