Repository navigation
Releases
v5.4.3
v5.4.2
5.4.2 (2026-11-18)
Bug Fixes
- a Grains and phases run writes its counts like the other sources (#486) (e08097b), closes #480
- saved disk-centre labels come back after the dataset's folder moved (#489) (a595cbe)
- take CuboKit 0.12.1 and DiffractoKit 0.18.1 (SiGe Bragg origin, ellipse on crystalline rings) (#492) (d7ccc0f)
- take EMSuiteUI 0.12.4 (run bar follows its count, EDS Auto ID Kα rule, sidebar rows press to VoiceOver) (#491) (0e7d228)
- take FormatoKit 0.18.0, SpectroKit 0.16.0 and EMSuiteUI 0.13.0 (DM4 read memory, Auto ID near misses) (#493) (2c16eb6)
v5.4.1
v5.4.0
5.4.0 (2026-11-16)
Features
- an orientation map per phase for phase mapping from the structures (#393) (ef013da)
- claimed disks for a phase map made from the structures (#419) (39ff0b8)
- closing, opening another file or quitting asks before it stops a running analysis (#401) (93bea93)
- Grains and phases (no matrix), the best phase per position with its reliability (#420) (bfe0f01)
- Keep in memory by default when the cube fits a quarter of the RAM (#407) (f60e6c2)
- Keep in memory holds the cube at its stored bit depth (#392) (23e2822)
- Materials Project lookup through the agent door (#382) (4389fbe)
- the agent door reaches Settings, Clear Calibration, Remove Fine-Tuned Model, Reopen Without This Session, Release cube and the demo (#380) (4ce2014)
- the agent door reaches the EDS room: get_eds, set_eds, run_eds, and run_task quantifies a spectrum image alone (#375) (1374f67)
- the agent door sets the view: pane zoom, display orientation and mirror, fit overlay, Same scale (#384) (7891e2d)
- type the R–Q rotation and transpose in Prepare and set_calibration, as py4DSTEM's QR_rotation and QR_flip (#417) (1267cc5)
Bug Fixes
- a crafted sidecar crop or RGBA map is refused or clipped instead of trapping (#383) (cbfe69d)
- a dataset switch during a residency or open await publishes nothing onto the new dataset (#395) (ea91348)
- a malformed precipitate class map is refused with a sentence instead of trapping (#387) (97c9641)
- a phase map with the matrix mostly not indexed names the origin and the distortion fit (#394) (c52f54b), closes #298
- a release run in a fresh tag checkout creates its dmgbuild venv (70d9208)
- a scattered centre-of-mass origin points to Origin from Bragg Peaks (#436) (f273a35)
- a sidecar write returns its own lineage omission instead of leaving it in a global (#399) (44f8b1a)
- a stale parallax depth index names no depth instead of trapping (#389) (718f5f3)
- a strain map computed without the ellipse says so beside its Quantitative badge (#403) (e9ddb81)
- a typed-zones phase map records the excitation slab of each phase (#433) (6643cf1)
- a virtual-detector run lands at most once (#400) (77f5744)
- Build Kernel keeps an origin set by hand (#432) (4937049)
- closing a dataset window releases its resident cube and the file at once (#429) (a6b9545)
- crafted 4D files are refused instead of trapping (#379) (db18fdd)
- DPC guidance no longer asks for an origin that is already set (#431) (b732011)
- File › Open offers .hspy files, which the HDF5 reader opens (#463) (ee4c87e)
- Grains and phases offers no claimed-disks overlay instead of asking for a new run (#456) (b2cd3a1)
- Grains and phases offers no claimed-disks overlay instead of asking for a new run (#462) (b2cd3a1)
- hold back Grains and phases for 5.4.0 (#475) (081021b)
- NaN or empty data in DPC, parallax, fit overlays and Quick Look no longer traps (#385) (70c0234)
- NaN, infinite or empty data in DPC, parallax, fit overlays and Quick Look is refused or drawn black instead of trapping (70c0234)
- one open at a time, claimed before the open's first await (#405) (65b3772)
- phase mapping progress moves through both passes (#381) (288ac3b)
- preparing the learned detector loads a package once however many callers wait (#398) (30a2df6)
- real-data acceptance pins follow the py4DSTEM disk-detection defaults (#444) (35ed281)
- Settings > Agents > Folder > Choose… opens a folder panel (#457) (3321d67)
- Settings > Agents > Folder > Choose… opens a folder panel (#465) (3321d67)
- the ACOM full-scan estimate counts a monoclinic phase's 4x library (#418) (d2c0eea)
- the agent door refuses an open during a run, naming the run (#409) (def0b0a)
- the agent door's busy, failure and open answers agree with the window (#378) (10e4642)
- the agent door's tool layer reads bounded files, creates files exclusively, finds a moved Agent folder, quotes agent text (#376) ([f527167](https://github.com/macEMSuite/mac4DSTE...
v5.3.1
mac4DSTEM 5.3.1 fixes wrong or misleading results.
- Placing a virtual detector no longer moves the origin. You now set the origin on purpose in Prepare.
- Phase mapping refuses an unmeasured origin, a typed orientation relationship covers all its variants, and Fit Distortion fits the right matrix.
- Calibrate Q reads diamond-type rings correctly.
- Opening a dataset writes nothing to its disk, and a new file no longer shows the last file's EDS maps.
- A tidier status bar, keyboard and VoiceOver fixes, and a new update signing key.
5.3.1 (2026-11-14)
Bug Fixes
- a cube from another file no longer shows the previous file's EDS maps (#327) (345776b), closes #322
- a file the app may not read is refused as unreadable; StaleTaskJourney opens its fixture in place (#359) (bf36437)
- a new dataset starts with an empty phase list (#317) (20b4414)
- a typed orientation relationship stands for all its variants in the matrix (#358) (8c81635)
- an empty virtual detector and an origin with no measured position are refused (#343) (37e71e7)
- Calibrate Q from crystal reads diamond [110] rings from {111} (double-diffraction shells, DiffractoKit 0.16.0) (#352) (28daf6b)
- calibrated Bragg vectors keep the count of positions where detection failed (#328) (8042c20), closes #302
- fine-tuning scales a windowed pattern's dose as inference does (#335) (fc654dd)
- Fit Distortion from Spots fits Phase mapping's matrix, not the last added phase (#316) (e174530)
- opening a dataset writes nothing on its volume (read-only Recents bookmark) (#329) (4d23857)
- phase mapping refuses an origin that is not a measured beam centre (#346) (7e14151)
- placing a virtual detector no longer moves the origin; Prepare sets a centre by hand (#360) (899a12a)
- release: new Sparkle update key after the keychain loss (#287) (9252fb1), closes #285
- sidebar task rows and Workspace › Next/Previous Task choose the task through accessibility and the keyboard (#319) (5db2ba1)
- the task bar shows one ellipsis, no rate, and a spaced Cancel (#362) (d9ac2cf)
- the UI test runner keeps one signature in gate and drive runs (#331) (0b35b25)
- Thronsen A parity pin moves to the 98.59 % main reads (accepted drift) (#354) (aae9cc8)
- Thronsen A parity pin moves to the 98.59 % main reads (accepted, #345) (aae9cc8)
v5.3.0
5.3.0 (2026-11-12)
Features
- agent door pattern views: a cut of the pattern, the pattern without disk circles, raw counts around a pixel (#234) (3a4e3d5)
- agent door pattern views: a cut of the pattern, the pattern without disk circles, raw counts around a pixel (#248) (3a4e3d5)
- agent door phase list: add a phase from a CIF, mark the matrix, set a phase's excitation slab, read the last run (#251) (f32331b)
- agent door sets the remaining disk detection fields: subpixel, upsampling, pattern smoothing, absolute threshold, reference peak (#235) (9fa0f91)
- agent door sets the remaining disk detection fields: subpixel, upsampling, pattern smoothing, absolute threshold, reference peak (#241) (9fa0f91)
- agent door setup notes, and a scan map frame an agent can find (#237) (3077449)
- agent door setup notes, and a scan map frame an agent can find (#249) (3077449)
- agent door workflow guide: how to work through a scan and judge each step (#236) (0c985da)
- agent door workflow guide: how to work through a scan and judge each step (#247) (0c985da)
- agent door: an Agent folder for batches: an agent lists, opens and saves results in one folder the person chooses (#205) (324bf74)
- agent door: Claude and other agents on this Mac look at mac4DSTEM's windows over MCP (#157) (1b325f7)
- agent door: get_scan_map, the scan map in scan coordinates, so an agent picks a precipitate's position (#213) (52520e7)
- agent door: probe kernel tools: an agent reads and sets the probe radius, builds the kernel and picks the detector (#212) (4b470b9)
- agent door: run and tune: an agent moves to a scan position, tunes disk detection, and presses Run and Stop (#202) (e85eb30)
- agent door: select_task, and Phase mapping's settings for an agent to read and set (#215) (5743f93)
- agent door: set_calibration and run_calibration, Prepare's calibration for an agent (#216) (662d153)
- Check for Updates… in the website download (Sparkle) (#182) (a5cd2fb)
- Count variants on a phase map made from the structures (#253) (1b7addc)
- Fit Distortion from Spots in Prepare, for a scan of one crystal (#220) (b08c0ba)
- Map Phases finds the orientations from the structures; typed zone axes become an Advanced switch (#243) (0270842)
- open Merlin's multi-file save and EMPAD pairs of other names, asking once for the folder under the sandbox (#183) (61f5a4d)
- Settings shows how much the learned detector's model caches hold (#181) (1c7921f)
- the agent door reaches every analysis (#255) (011762c)
- the Al–Mg–Si preset leaves the Add Phase menu (#250) (6cb9b24)
Bug Fixes
- a Cancel at the companion folder prompt is a status line, not an error, and a granted folder is remembered across launches (#187) (70816d0)
- a cleared or typed Q scale stays cleared on a rewind, and ptychography reads stale after a new origin or Q (#154) (39064d6)
- a file opened from Finder fills the empty window instead of opening a second one (#166) (45ad430)
- a HyperSpy file's Q and R pixel scale are read from its axes (#214) (a9585d2)
- a reopened session has its imported phases and phase-mapping settings back (#152) (be6c77c)
- a reopened session's phase-mapping run is named in the Phases section (#149) (8e9f7aa)
- agent door answers agree: a point in numbers, one subpixel name, can_run, unreadable sidecar named (#233) (18a1168)
- agent door answers agree: a point in numbers, one subpixel name, can_run, unreadable sidecar named (#240) (18a1168)
- agent door says when a kernel or probe radius moves a disk setting, and logs every agent action (#232) (929b655)
- agent door says when a kernel or probe radius moves a disk setting, and logs every agent action (#239) (929b655)
- agent door: the pattern picture shows the detected disks as a Retina pane does (#188) (76cf9c6)
- Calibrate Q from crystal asks for an approximate pixel size when the rings fit two shells (#177) (52b32ef)
- Calibrate Q from crystal lets the file's pixel size decide between two ring assignments (#171) (0261651)
- Calibrate Q from crystal uses DiffractoKit 0.12.1 and asks which ring when two assignments fit equally (#170) (89c5729)
- composition per object is fitted with the method the room's fit ran with (#172) (a129335)
- composition per object is one pooled fit of the object's summed spectrum (#158) ([d72...
v5.2.0
5.2.0 (2026-11-05)
Features
- Origin from Bragg Peaks, an opt-in origin for crystals whose disks pull the centre of mass (#139) (8266df0)
Bug Fixes
- Known variants shows the detector-scale warning and Scale to This Detector (#101) (4b78487), closes #100
- read a .mib's .hdr and an EMPAD pair under the App Sandbox (#126) (6347b09)
- remember files the sandbox grants read-only (recent-file bookmark 256) (#132) (18eb06a)
- the object table and Precipitates section say which elements the composition covers (#112) (ab34b5b)
- the object table and Precipitates section say which elements the per-object composition covers (ab34b5b)
- the object table says "no scan scale" only when the scan has no scale (#105) (3cbcdb9), closes #104
- the object table says to show wt% or at% when the spectrum image is quantified without composition maps (#102) (354d585)
- the object table shows the composition and strain columns beside the phase (#109) (05448ae)
- the object table shows the composition and strain columns beside the phase, wide enough for their text (05448ae)
- the synthetic probe kernel's radius can be typed in Bragg Disks (#116) (de6e5f7)
- the toolbar is busy while the EDS room fits and checks (#135) (67bda36)
v5.1.1
New
- Parallax and ptychography results are saved with your session and come back when you reopen it.
- The EDS room is now shared with macTEM, so it looks and works the same in both apps.
Improved
- Results follow py4DSTEM more closely: circular scan regions, ACOM orientations and phase mapping.
- Beam-centre fitting now refuses unreliable fits, for example on very noisy data or under a beam stop, instead of using them.
- EDX quantification is more accurate in some cases, so results can differ slightly from 5.0.
- Numbers use your system's decimal mark.
Fixed
- Sessions reopen more reliably, keeping Bragg disks, origin and calibration.
- A room no longer shows an outdated result after you switch rooms.
- The download now works: 5.1.0 shipped without an installer.
Requires macOS 26 or later on Apple silicon.
v5.1.0
Download 5.1.1: this version has no installer.
New
- Parallax and ptychography results are saved with your session and come back when you reopen it.
- The EDS room is now shared with macTEM, so it looks and works the same in both apps.
Improved
- Results follow py4DSTEM more closely: circular scan regions, ACOM orientations and phase mapping.
- Beam-centre fitting now refuses unreliable fits, for example on very noisy data or under a beam stop, instead of using them.
- EDX quantification is more accurate in some cases, so results can differ slightly from 5.0.
- Numbers use your system's decimal mark.
Fixed
- Sessions reopen more reliably, keeping Bragg disks, origin and calibration.
- A room no longer shows an outdated result after you switch rooms.
Requires macOS 26 or later on Apple silicon.
mac4DSTEM 5.0.1
v5.0.0 as described below, notarized (v5.0.0 was tagged but never released: its build was refused because the Quick Look extensions lacked the hardened runtime).
Requires macOS 27 on Apple Silicon, as v4.5.0. A major release: mac4DSTEM now runs on the six macEM kits at exact released tags
(CuboKit 0.9.0, DiffractoKit 0.10.0, SpectroKit 0.12.0, FormatoKit 0.11.0, ImagoKit 0.8.0, MacEMUI 0.3.0) and gains an EDS room,
the same room macTEM 1.0 ships, joined to the 4D scan.
- The EDS room (Spectroscopy › EDS) replaces the Spectroscopy room built during v5 development. It is macTEM 1.0's room: maps over
the spectrum, Auto ID with its reasons in the results footer, the results table with net counts and at% and wt% each ± σ, pins
(kept regions) with their own CSV, line profiles, composition maps (Cliff-Lorimer, zeta, cross section), and Export (Results CSV,
Method JSON, Spectrum CSV and MSA, Regions CSV, Maps). mac4DSTEM adds what the 4D scan brings: with a cube and a stated
registration (a GMS joint 4D + EDS file) the scan's phases, diffraction groups and not-indexed pixels are regions in a Source row;
Composition by Class fits every phase pool; the precipitate objects take their composition from the room's maps; each Quantify is
a step in the session's lineage and a row in Results. Auto ID refines the energy axis before it proposes, as macTEM does. - What changes for you. The room's controls sit where macTEM's do (the control placement standard). The at%/wt% unit menu is
gone (both columns show); the Profile and Table tabs are now the profile inset and the pins table; pins are not limited to three.
Steps recorded by earlier v5 development builds still read in Results and restore their method.
The full list of changes is in CHANGELOG.md. Download mac4DSTEM.dmg (or the versioned copy), open it, and drag mac4DSTEM to Applications.