Patent Filed - July 2026

The First Bismuth-Layer Ternary Processor

Revolutionary ternary logic architecture {-1, 0, +1} on a topological Bi2Se3 layer, with a native ternary instruction set (T-ISA) and integrated memristor memory.

Bi2Se3
Core technology

Ternary Architecture & Bismuth Layer

A fundamental breakthrough in digital computing

-0+

Ternary Logic {-1, 0, +1}

Unlike binary limited to 2 states, ternary logic encodes 3 states per trit. A register of n trits stores 3n values versus 2n in binary - that is log2(3)/log2(2) - 1.585x more information per state unit.

-1NEG
0ZERO
+1POS
Bi

Bismuth Layer Bi2Se3

Bismuth selenide (Bi2Se3) is a topological insulator: insulating in bulk but conducting on the surface via protected edge states. Low contact resistance, thermal stability up to 270 C, quantum spin Hall effect.

270 CThermal stability
QSHQuantum Hall effect
Bi2Se3Topological insulator
< 1mOContact resistance

Die Cross-Section

Binary vs Ternary

Parameter Binary Ternary Bismuth
States per digit 2 (0, 1) 3 (-1, 0, +1)
Info per state 1 bit 1.585 bits
Logic gates AND, OR, NOT MIN, MAX, INV, NAND3
Power consumption 1x (reference) 0.67x estimated
Die area 1x (reference) 0.73x estimated
Logic

Ternary Logic Gate Set

A complete set of hardware-wired gates operating on {-1, 0, +1} states

Inputs on columns / rows; color = output state (-1 purple, 0 gray, +1 gold)

On-die storage

Integrated Memristor Memory

Non-volatile ternary on-die RAM

Memristors exploit 3 distinct resistance levels (R_low, R_mid, R_high) to natively encode trits {-1, 0, +1}. Integrated directly on-die next to compute cores, they eliminate the classic memory bottleneck.

Ternary Memristor Cell

TOP ELECTRODE R_lo (-1) R_mid (0) R_hi (+1) BOT ELECTRODE IinGND
Heterogeneous system

Specialized Compute Units

TVX - TPU - GPU - NPU - APU - VPU - ISP - SEC - DSP

TERNIUM T1 DIE - TRIANGULAR THERMAL DESIGN apex = hot point - wide base = optimized thermal dissipation

Triangular Design: Advantages & Trade-offs

Why a triangular die instead of a square one - a full analysis of heat dissipation and its trade-offs

Pros

    Cons

      Ternary APU - Built-in Functions

      The ternary audio/signal processing unit embeds a broad range of DSP functions wired in hardware, leveraging 3-state resolution for extended dynamic range and low latency.

      Cutting-Edge Platform & Connectivity

      Every technology of the latest market processors, reimagined in ternary

      T-ISA

      Ternary Instruction Set

      Native {-1, 0, +1} machine instructions wired on-die to maximize ternary compute strength

      2-tryte (6-trit) encoding

      Compact encoding: each instruction fits in 2 trytes (6 trits) - ~9.5 equivalent binary bits - giving ~40% higher code density vs binary.

      Numerical engine

      Advanced Computing, Mathematics & Physics

      Quadratic - Trigonometry - Complex Numbers - Maths & Physics

      -

      Interactive Ternary Equations

      Select an equation to visualize its dedicated animation

      Complete Mathematical Library

      Hardware acceleration across every domain of numerical computing

      Integrated Physics Engine

      Hardware acceleration across the major fields of physics

      Bio-computing

      Bio-Inspired & Biological Computing

      Biological units that push the processor beyond its limits

      Ternary logic {-1, 0, +1} naturally mirrors biological processes (inhibition / rest / excitation). Ternium embeds bio-inspired co-processors that accelerate genomics, molecular simulation and organic neural networks.

      Ternary DNA Encoding

      Each DNA base maps onto trit states for native genomic processing

      DNA to Ternary Encoder

      Click the bases to build a sequence and watch it encode into trits in real time

      2 trits/base - 33% more compact than 2-bit binary encoding + parity

      Biology is naturally ternary

      Living signals have three states - exactly like {-1, 0, +1} logic

      Ternary Spiking Neuron

      The membrane potential crosses two thresholds: below the inhibitory threshold it encodes -1, at rest 0, above the excitatory threshold +1. A single ternary neuron thus replaces two binary neurons.

      Circuit level

      Transistor Networks

      How the processor's functions are wired at the transistor level

      Ternary logic relies on multi-threshold transistors (MVT). By combining MOSFETs with low and high threshold voltages, each gate produces three output levels: 0 V for -1, V_dd/2 for 0, and V_dd for +1. Select an input level to see which transistors conduct.

      STI

      Standard Ternary Inverter (STI)

      Two PMOS (pull-up) and two NMOS (pull-down) with shifted thresholds invert the three levels: -1 to +1, 0 to 0, +1 to -1.

      Input level

      Output +1 - V_dd
      Execution flow

      Compiler & Pipeline Optimizations

      TERN-GCC and 12-stage pipeline architecture

      12-Stage Ternary Pipeline

      TERN-GCC Compiler

        Memristor Cache Hierarchy

        Cache Size Latency Position
        Process

        Design & Fabrication

        From the silicon-bismuth wafer to the finished ternary processor - step by step

        One line for the whole range

        Every Ternium product follows these 8 steps, tuned to its target

        Benchmarks

        Performance Comparison

        Ternium T1 vs Binary Silicon

        Development

        Roadmap

        From prototype to production

        Evidence

        Evidence Framework and Confidence

        Claims in this document combine established semiconductor engineering with forward-looking architecture projections.

        High confidence

        Core ternary logic mathematics, known fabrication steps, and baseline performance-model methodology.

        Medium confidence

        Projected architecture-level efficiency and density benefits under constrained implementation assumptions.

        Medium-Low confidence

        Industrial-scale adoption timelines and ecosystem migration across full software/hardware stacks.

        References

        References and Source Families

        Reference families include ternary computing literature, ISA/compiler optimization sources, memristor research, and process/fabrication engineering documents.

        ← Back to HumanityA Maxware project