Tissues and phantoms

The materials SAR is computed in — frequency-dependent body tissues and tissue-simulating liquids, each with its source — and the parametric phantoms a 3D view draws them as.

SAR is computed in tissue: a lossy dielectric whose material states a density. circuitRF ships a library of body tissues and tissue-simulating liquids, each record citing its source, and six phantoms — parametric bodies of nested layers — to put them in. Your own segmented anatomy comes in through Voxel and mesh import.

Research and engineering use

circuitRF's SAR is for research and engineering. It is not a compliance tool, and the shipped values are reference data for research use, not a claim that any phantom or liquid meets a standard.

The tissue library

Add Tissue Materials copies tissue-materials.cmat beside the technology and names it as one of its libraries (see Materials). It is on the technology editor's Materials tab, and at the top of the 3D view's Materials dialog (a phantom layer's New Material…, or Assign Material…), where it lists the tissues at once to assign and names the library when you press the dialog's OK. Either way the technology is left unsaved until you save it. It holds:

Records	Values	Density
16 body tissues: skin (dry, wet), fat, muscle, cortical and cancellous bone, brain grey and white matter, cerebrospinal fluid, blood, vitreous humour, lung (inflated, deflated), heart, liver, kidney	the 1996 parametric (Cole-Cole) model of Gabriel, Lau and Gabriel, Phys. Med. Biol. 41, 2271–2293, evaluated by circuitRF at any frequency	ICRU Report 44 for fat, muscle, cortical bone, blood, brain and deflated lung; the other eight state none
Head and body tissue-simulating liquids	target values from 150 MHz to 5.8 GHz, FCC OET Bulletin 65 Supplement C (Edition 01-01), interpolated linearly between rows, as the bulletin prescribes, and refused outside them	1000 kg/m³, as the bulletin states
Phantom shell (circuitRF's own)	εr 3.7, lossless: circuitRF's own value, inside the bounds the same bulletin gives for a shell (εr and tanδ below 5.0 and 0.05)	none: a shell is never tissue
Saline gel (implant heating)	σ 0.47 S/m, εr 80, from the published implant-heating literature its record cites	water's
openEMS's example tissues	the head phantom at 298 MHz, the body cylinder at 128 MHz and the dipole example's head at 1 GHz, exactly as openEMS's scripts state them	the same scripts

A tissue with no density cannot carry SAR: a SAR setup whose region holds one is refused, naming it. No free source gave a density that could be checked for both skins, cancellous bone, cerebrospinal fluid, vitreous humour, inflated lung, heart, liver and kidney, so those records state none rather than an unchecked number. Add one, with its source, to your own copy. Thermal properties (conductivity, specific heat, perfusion, metabolic heat) are empty for every tissue, for the same reason.

openEMS's example tissues are not circuitRF's. openEMS's scripts label their phantoms' values "at 298 MHz" and "at 128 MHz", but they are not the Gabriel model's values there — the 298 MHz skull is εr 11.4, σ 0.29 S/m, where cortical bone is εr 13.4, σ 0.082 S/m. They are kept as the scripts state them so a replication of those examples compares like with like; for your own work, use the Gabriel records.

One frequency per run

A tissue's εr and σ move strongly with frequency, and an FDTD run sees each material as one constant pair. So a 3D run evaluates every frequency-dependent material once: at the setup's first SAR frequency when it computes SAR, otherwise at the sweep's centre. The run's notes give the frequency and each value — "Frequency-dependent materials are evaluated once, at 1.75 GHz (the SAR frequency) … Head tissue-simulating liquid εr 40.08, σ 1.371 S/m" — and circuitrf explain view.c3d --materials lists every material as the solver will receive it.

Selecting a tissue in the Materials editor shows its Frequency dependence card: εr and σ at a frequency you type (298M, 1.9 GHz), and both plotted against frequency. The model or the table is in the record, and is edited in the file.

Phantoms

3D ▸ Draw ▸ Phantom adds one:

Phantom	Layers, outermost first	Starts as
Head (3 ellipsoids)	skin, skull, brain	openEMS's head phantom: semi-axes 100, 80, 100 mm, shells 9 mm
Body (3 cylinders)	skin, bone, tissue	openEMS's body phantom: radius 100 mm, 260 mm long, shells 8 and 10 mm
Flat Phantom	shell, liquid: the liquid in an open-topped box, the shell on its floor and four walls	a 2.0 mm shell and 150 mm of body liquid over 296 × 196 mm
Generic Head (circuitRF's own)	ear-left, ear-right, shell, liquid	an ellipsoid shell of head liquid, semi-axes 100, 78, 115 mm, a 2.0 mm shell, and a flat ear spacer each side, 4.0 mm thick and 25 mm in radius
Sphere, Box	one homogeneous body	radius 50 mm; a 100 mm cube

One object. A phantom moves, turns, hides and is left out of the solve as one; the tree lists its layers under it, and Properties shows the dimensions its shape uses and one material per layer. In the solve each layer is a solid named <phantom>/<layer>, so a SAR region can name a layer (head/brain) or the whole phantom (head).
The innermost wins where layers overlap, so the brain is brain, not skin. Every layer is solved as a lossy dielectric.
The flat phantom follows the bulletin where it gives a value: a shell of 2.0 ± 0.2 mm and a liquid 15.0 ± 0.5 cm deep. The bulletin gives no floor size, so the 300 × 200 mm outside is circuitRF's own, chosen to be at least ¾ by ½ of a wavelength from 900 MHz up. Its named faces are floor (the shell's outside) and liquid (the liquid's surface through the shell, the plane a system check's spacing is measured from).
The generic head is circuitRF's own shape and is named so, so that nobody takes it for a standard's mannequin. x is to the front, y ear to ear (left is +y), z up. The spacers' 4.0 mm comes from the same bulletin; their 25 mm radius is circuitRF's own. Its named faces are ear-left and ear-right, each spacer's flat face: the face a handset is placed against (Phone SAR).
Voxelized. A head or body phantom can be given to openEMS as voxels of a stated pitch (Voxel size in Properties; 0 is the smooth layers): each voxel takes the innermost layer holding its centre. openEMS's own example phantoms are 2.5 mm voxels of exactly these ellipsoids, so a voxelized phantom reproduces them cell for cell. The 3D view, Palace and the thermal solver still draw and mesh the smooth layers, and the run says so.
Flatten Phantom (its menu) replaces it with one ordinary solid per layer, <phantom>_<layer>, in a group named after it; the solve is the same. It is refused on a voxelized phantom, whose layers would become smooth.

What each object is in the file, field by field: circuitrf reference 3d.