Definition, alternating-current pathway, and category boundaries
A transcranial alternating current stimulator establishes a closed circuit through two or more scalp electrodes and applies low-intensity current that periodically reverses direction to investigate or modulate brain-network oscillations. A common output is approximately sinusoidal with a near-zero mean, but nonsinusoidal, amplitude-modulated, phase-modulated, and multifrequency waveforms are also possible. Scalp, skull, cerebrospinal fluid, and brain tissue jointly determine current distribution, and substantial current may be shunted across the scalp. The presence of a programmed frequency and an intracranial electric field does not mean neurons necessarily synchronize at that frequency.
tACS is one form of transcranial electrical stimulation. It differs from approximately constant-current tDCS, random-spectrum tRNS, slow-oscillatory direct-current stimulation with a DC offset, and TMS, which induces current with pulsed magnetic fields. Amplitude-modulated stimulation using a kilohertz carrier to create a low-frequency envelope also differs in electrodes, field distribution, and evidence requirements from transcranial temporal interference intended to form an interference envelope at depth. Cranial electrotherapy stimulation is a product category in some regulatory systems and may use several waveforms; the CES name alone does not identify a device as tACS.
Alternating stimulation ordinarily does not trigger action potentials directly. A subthreshold periodic field can instead change the probability of neuronal timing and network phase relationships. Noninvasive means that no craniotomy or implant is required, not that the current lacks physiological, visual, or skin effects. Research instruments, prescription medical systems, home devices, and consumer brainwave-synchronization headsets can differ in output accuracy, purpose, supervision, and regulatory status. Similar appearance or a shared frequency label does not establish equivalence.
Waveform, amplitude, frequency, phase, and dose
A complete dose specifies waveform shape, frequency, amplitude convention, DC offset, starting phase, interchannel phase difference, modulation depth, carrier and envelope, total harmonic distortion, electrode material and area, conductive medium, montage, ramps, duration, intersession spacing, and treatment course. Amplitude must be identified as peak, peak-to-peak, or root mean square. The same statement of 2 mA can imply materially different peaks when conventions differ. A label such as alpha, delta, or high-frequency stimulation is not reproducible because individual peak frequency, waveform, and spatial field also matter.
In a two-electrode circuit, currents at the two terminals have opposite direction. A multichannel system should conserve the instantaneous sum of delivered and return currents. In-phase or anti-phase stimulation must identify whether the comparison concerns electrode currents, fields at the targets, or the predicted relationship between two brain regions, because changing channel phase can also change field magnitude and distribution. Every return electrode contributes to the intracranial field and is not an irrelevant ground. A channel-map error, timing delay, or polarity reversal can turn the intended protocol into another stimulus.
Average current density and charge at an electrode describe aspects of the skin interface but do not directly represent the cortical field. Hair, skin hydration, skull thickness, skull defects, atrophy, lesions, and cerebrospinal-fluid spaces alter current pathways. A finite-element model can assist montage design but depends on image segmentation, conductivity assumptions, and actual electrode position; its field map is not an in-vivo measurement. A high-frequency carrier, multiple electrodes, or individualized modeling does not automatically mean deeper, more focal, or more effective stimulation.
Mechanisms, measurement artifacts, and evidence interpretation
Major hypotheses propose that a periodic external field produces phase pulling, resonance, or phase-dependent plasticity in endogenous oscillations. Effects depend on baseline frequency, intensity, brain state, and task and may also be mediated indirectly through scalp nerves, retina, or other peripheral pathways. A post-stimulation EEG-band change does not prove neuronal synchronization during stimulation, and behavioral improvement may reflect sensory cues, expectation, repeated practice, or natural fluctuation. A flickering phosphene can result from current spreading to the eye or optic pathway and is not a dose marker proving that an occipital cortical target was reached.
When EEG is recorded during stimulation, the alternating output is usually far larger than the neural signal and introduces the fundamental, harmonics, amplifier nonlinearities, movement, and electrode-interface artifacts. A residual peak at the stimulation frequency after mathematical cleaning can still be processing residue and cannot be called neural entrainment without further validation. A closed-loop system triggered from real-time EEG phase must characterize filter group delay, phase-estimation error, sampling clocks, stimulator latency, and artifact recovery. An offline simulation or software target phase is not the patient's actual intracranial phase.
Clinical evidence and authorization must be checked separately by condition, jurisdiction, product, waveform, and protocol. Professional guidance and randomized trials can identify research or clinical promise for particular symptoms, but an evidence recommendation does not authorize all equipment and a result at one frequency does not transfer to another. Studies report full parameters, blinding credibility, attrition, concomitant care, and adverse events and distinguish subjective scales, objective physiology, and patient-important outcomes. Marketing must not convert an EEG-band change into a claim of cure or cognitive enhancement.
Patient screening, electrode preparation, and visual risk
Before treatment, a qualified professional confirms diagnosis, objective, applicable labeling, and alternatives and examines electrode sites for open wounds, infection, dermatitis, sunburn, recent surgery, prominent scars, allergy, or impaired sensation. History includes epilepsy or unexplained loss of consciousness, stroke, brain injury, tumor, skull defects or reconstruction, intracranial metal, severe headache, and neurosurgery. Pacemakers, defibrillators, deep-brain or vagus-nerve stimulators, cochlear implants, infusion pumps, and other active implants require review. Suitability depends on information for both devices, location, and clinical judgment.
Assessment also covers migraine, photosensitivity or flash-triggered symptoms, eye disease, pregnancy, psychosis, mania or hypomania, suicide risk, substance use, medication changes, severe sleep deprivation, and the ability to report discomfort. A tACS phosphene is typically a visual sensation without external light, but a first or unusual visual symptom cannot automatically be attributed to the device. Persistent monocular flashes, a visual-field deficit, eye pain, or a neurological symptom requires cessation and appropriate clinical assessment.
Each session verifies identity, prescription and software version, electrodes and medium, montage, channels, amplitude convention, frequency, phase, and health changes. Hair is parted and intact skin cleaned as instructed without sharp tools or aggressive abrasion. Sponges or gel are prepared uniformly: dryness can cause high resistance and hotspots, whereas excess fluid, dripping, or a conductive bridge between adjacent electrodes changes the field. After fixation, full-surface contact and strain-free leads are confirmed and position recorded. A total impedance value cannot demonstrate uniform contact at every point.
Stimulation delivery, adverse effects, and emergency response
Before output, the system completes self-tests, channel and return balance, contact-status, and local emergency-stop checks and ramps smoothly to the prescribed waveform. During treatment, staff or the validated workflow monitor itching, tingling, warmth, burning sensation, headache, dizziness, nausea, fatigue, attention change, facial sensation, and phosphenes and ask about mood or neurological symptoms. Patient movement, electrode lift, heavy sweating, or lead strain requires a pause and reassessment. Electrodes must not be moved or re-wetted during output, and frequency must not be improvised or another stimulator added.
Mild uniform erythema and transient sensation differ from persistent pain, blistering, broken skin, or a burn shaped by an electrode edge. A focal intense hot sensation, persistent or distressing flashes, reduced vision, a new neurological deficit, altered consciousness, suspected seizure, severe headache, marked agitation, signs of mania, or suicide risk requires immediate cessation, electrode removal, and clinical assessment. Stimulation outside the authorized labeling should not occur while driving, operating machinery, or alone without access to assistance.
Adverse-event records include onset, severity, duration, parameters, device and electrode lot, contact readings, intervention, and outcome. Extensive safety experience with conventional low-intensity research sessions does not establish safety for a higher amplitude, a different amplitude convention, a kilohertz carrier, prolonged or densely repeated exposure, a homemade circuit, or an understudied population. Sleep-delivered research also addresses electrode displacement, medium drying, cable entanglement, arousal, and sleep-stage misclassification under a reviewed monitoring and stopping plan.
Closed-loop and home governance, and device quality management
Closed-loop tACS may adapt frequency and phase from an individual peak, real-time EEG, or sleep stage, but the algorithm must define input quality, artifact suppression, confidence, latency, loss-of-lock criteria, and safe fallback. With unreliable signals, it should stop adaptation or stimulation rather than track the artifact. Model or algorithm updates can change actual patient dose and require version control, validation data, and clinical change assessment; artificial-intelligence optimization does not bypass the prescription.
A home system locks settings within a range users cannot freely modify and enforces daily and weekly session limits, minimum spacing, restart rules, and missed-session handling. The workflow provides placement training, skin checks, timely help, and adverse-event reporting. Network loss, phone failure, cloud outage, or portal unavailability leads to a predefined safe state. A synchronized log is not proof of correct placement, and an automated rating scale is not suicide-crisis monitoring. Emergency stop works locally, and unapproved chargers or accessories should not be connected during stimulation.
Procurement and acceptance testing cover amplitude convention and accuracy, frequency, phase and interchannel synchronization, DC offset, waveform distortion, ramps, timing, return balance, compliance voltage, contact detection, open- and short-circuit protection, emergency stop, electrical and electromagnetic compatibility, skin-contact materials, battery, and software security. Products integrating EEG or closed-loop control are tested with known signals for timestamps, filtering delay, artifact saturation, and recovery. Changes to electrodes, medium, headgear, cables, firmware, applications, or algorithms require change control and revalidation.
Sleep evidence, treatment boundaries, and quality conclusion
tACS for chronic insomnia has positive signals and clear negative findings. A small randomized trial of medial-parietal alpha-frequency stimulation reported a higher response rate during follow-up, but the sample was limited, the confidence interval was wide, later response proportions declined, and a corrigendum followed publication. Another 87-person study enrolled people with insomnia symptoms who did not meet insomnia-disorder criteria. Insomnia scores and diaries improved in both active-frequency groups and sham, without a significant between-group difference, and actigraphy did not improve; the authors also noted that the high carrier could obscure the target frequencies.
A recent meta-analysis pooled only four randomized trials and reported advantages for subjective sleep, sleep-onset latency, and total sleep time, but protocols differed and heterogeneity was very high for some outcomes. A broader systematic review of sleep electrophysiology likewise judged tES effects heterogeneous and inconsistent, with few significant findings and substantial risk of bias. Changes in alpha, delta, or slow-wave power therefore do not establish a standard insomnia treatment, and a study's frequency, carrier, or nominal current cannot be copied across devices. Chronic insomnia still requires proper evaluation and access to established care such as cognitive behavioral therapy for insomnia and management of comorbidities.
tACS does not splint the upper airway, provide ventilation, or abolish obstructive respiratory events. It cannot replace testing and disease-specific treatment such as positive airway pressure for obstructive sleep apnea and should not be an unsupervised sleep- or cognitive-optimization tool for healthy users. Authorized clinical use or compliant research prespecifies subjective and objective outcomes, blinding, follow-up, stopping criteria, and attribution of phosphenes and skin events. Conclusions remain specific to product, waveform, montage, and population; sharing a name with an EEG rhythm is not proof of effective neural synchronization or clinical benefit.