Mathematical Models of Sleep — Process S, Circadian Rhythm and the Two-Process Model

Why do you feel sleepy after 16 hours awake regardless of the time of day, yet feel alert at 11 PM but desperately tired at 3 AM? Because two independent biological processes — a homeostatic sleep drive and a circadian alertness signal — interact to create the human sleep-wake cycle. Alexander Borbély's two-process model, proposed in 1982, captures this interaction with remarkable predictive power using nothing more than exponential functions and a sinusoid.

1. The Two-Process Framework

The Borbély model postulates that sleep-wake behaviour is controlled by two separable biological processes:

Process S (Sleep homeostasis): a sleep-promoting drive that builds exponentially during wakefulness and dissipates exponentially during sleep. Neurobiological correlate: adenosine accumulation in the basal forebrain during wakefulness (caffeine works by blocking adenosine A1 and A2A receptors).
Process C (Circadian clock): an approximately 24-hour oscillation in alerting/arousal driven by the suprachiasmatic nucleus (SCN) in the hypothalamus, entrained by light via retinal photoreceptors. Acts as an internal clock independent of sleep history.

Sleep occurs when Process S (pressure to sleep) exceeds an upper threshold modulated by Process C; wake occurs when Process S drops below a lower threshold. The two thresholds themselves oscillate with the circadian rhythm, creating a "window of opportunity" for sleep that aligns with night.

2. Process S — Homeostatic Sleep Pressure

Process S is modeled as an exponential with different time constants for waking and sleeping:

During wakefulness: dS/dt = (S_max − S) / τ_w (exponential rise toward S_max) S(t) = S_max − (S_max − S₀) · e^{−t/τ_w} During sleep: dS/dt = (S_min − S) / τ_s (exponential decay toward S_min) S(t) = S_min + (S₀ − S_min) · e^{−t/τ_s} Typical parameters: τ_w ≈ 18.2 h (waking time constant) τ_s ≈ 4.2 h (sleep time constant — faster recovery than buildup) S_max = 1.0, S_min = 0.0 (normalized) After a normal 16h wake / 8h sleep: S rises from ~0.2 to ~0.68 during the day S falls from ~0.68 back to ~0.18 during the night

The neurophysiological markers of Process S are slow-wave activity (SWA, 0.5–4.5 Hz delta power) in the EEG during NREM sleep: SWA is highest at the start of sleep when S is highest, and decreases through the night as S dissipates — exactly as predicted by the model.

3. Process C — The Circadian Oscillator

Process C is modeled as a damped sinusoidal oscillation with a near-24-hour period:

C(t) = C_m + A \cdot cos(2π(t - φ)/T) T \approx 24.2 h (human free-running period in constant darkness) φ = phase alignment (peak alertness midday; nadir ~4-6 AM) A = amplitude of oscillation C_m = midpoint The Kronauer-Jewett mathematical model is more detailed: dx/dt = π/9 \cdot (x_c - (4/3)x³ - (128/105)(π/24 \cdot ω - 0.465)) \cdot (x + x_c) / B dx_c/dt = π/9 \cdot B \cdot (μ(x_c + 0.465 - 304/105\cdotx_c³) - x) where B includes light input weighting and ω = intrinsic frequency

The SCN generates the circadian signal even in constant darkness (it acts as a self-sustaining oscillator). Light resets the phase — morning light advances the clock (earlier timing); evening light delays it. The timing of the circadian nadir around 4 AM explains why shift workers driving home at that hour have peak drowsiness crash risk.

Chronotypes: individual variation in circadian phase determines "morning larks" (phase-advanced, peak alertness earlier) vs "night owls" (phase-delayed). The distribution is normally shaped but the late chronotype is more common in teens (a genuine biological shift in adolescent SCN phase) — contributing to why early school start times impair learning.

4. Sleep-Wake Transitions

In the Borbély model, the thresholds for sleep onset (H_u, upper) and wake onset (H_l, lower) are modulated by the circadian process:

H_u(t) = h_u + A_u \cdot cos(2π(t - φ_u)/T) (upper threshold) H_l(t) = h_l + A_l \cdot cos(2π(t - φ_l)/T) (lower threshold) Sleep onset: when S(t) reaches H_u(t) Wake onset: when S(t) falls to H_l(t) The sinusoidal thresholds create: - A "forbidden zone for sleep" in the late afternoon (circadian alertness peak) - A "sleep maintenance zone" from midnight to 6 AM - The "wake maintenance zone" resisting sleep in the evening despite rising S

The circadian system actively promotes wakefulness in the evening — counteracting rising Process S to keep humans alert until after dark. Edgar et al. (1993) confirmed this with lesioning studies: SCN destruction eliminated the daily sleep-wake cycling, leaving rats with constant short fragmented sleep bouts determined only by Process S.

5. Ultradian REM Cycles

Within a night's sleep, the brain cycles through NREM and REM stages with an ultradian period of approximately 90 minutes. This cycling is not modeled by the two-process model — it requires separate mechanisms:

NREM stages (N1→N2→N3): N3 (slow-wave sleep, SWS) dominates early in the night when Process S is highest. SWS is associated with memory consolidation (declarative memory) and growth hormone secretion.
REM sleep: increases in later cycles as the night progresses and Process S dissipates. Associated with emotional memory processing, narrative dreaming, and motor memory. Suppressed by adenosine (which explains why caffeine late in the day reduces REM duration).

Typical sleep architecture (8h night): Cycle 1 (hr 0-1.5): N2 30min, N3 45min, REM 15min Cycle 2 (hr 1.5-3): N2 35min, N3 25min, REM 30min Cycle 3 (hr 3-4.5): N2 40min, N3 5min, REM 45min Cycle 4 (hr 4.5-6): N2 45min, N3 0min, REM 60min Cycle 5 (hr 6-8): N2 45min, REM 75min

6. Sleep Deprivation and Recovery

The two-process model accurately predicts cognitive performance during partial and total sleep deprivation. Sustained wakefulness elevates Process S monotonically; performance on psychomotor vigilance tests (PVT) decreases in proportion.

Alertness/performance proxy: W(t) = S(t) - C(t) (simplified "cognitive wake drive") When W is low (high S, low alertness signal from C): sleepy, impaired. When W is high: alert. After 24h without sleep: S \approx 0.85 (near maximum) Cognitive impairment equivalent to 0.1% blood alcohol concentration Recovery sleep: S initially decaysquickly (τ_s \approx 4h) with elevated SWA — the brain "repaying sleep debt" by intensifying slow-wave sleep.

Chronic partial sleep restriction: sleeping 6h/night for two weeks produces the same total impairment as 24–48h total deprivation, but subjects mostly stop noticing their own impairment — subjective sleepiness plateaus while objective performance continues to decline. A key prediction of the two-process model validated in controlled studies (Van Dongen et al., 2003).

7. Jet Lag and Shift Work

Jet lag occurs when the internal circadian phase (Process C) is misaligned with the external environmental cues (light-dark cycle at new location). The SCN re-entrains at a rate of ~1 hour per day eastward and ~1.5 hours per day westward — meaning a 9-hour eastward flight (e.g., New York to London) takes roughly 9 days to fully re-entrain.

The asymmetric recovery (westward easier than eastward) occurs because:

The free-running period is ~24.2h (slightly longer than 24h) — naturally drifts later.
Eastward travel requires Phase Advance (earlier timing) — fighting the natural drift.
Westward travel requires Phase Delay (later timing) — aligning with the natural drift.

Shift workers who rotate between day and night shifts never fully re-entrain. Their Process C remains synchronized to daylight while they must sleep during the day — explaining elevated rates of metabolic syndrome, cardiovascular disease, and cancer attributable to chronic circadian disruption.

Mathematical "circadian misalignment" can be quantified as the phase difference |φ_clock − φ_environment| integrated over time, predicting cumulative health consequences — an active area of chronobiology research.

🧠 Open Disease Spread Simulation →