展示HN：构建一款由核衰变驱动的智能手机：电力工程
在过去二十年中制造的每一款智能手机都假设电池每天充电一次。在我开始讲解计算之前，请先思考一下：如果手机从来不需要插电，实际会发生什么变化？
这就是我正在构建的产品。其电源采用贝塔伏特技术——通过放射性同位素的衰变持续产生电力，且没有运动部件、没有燃烧、也不需要充电基础设施。
我所描述的设备并不是一款采用不同电源的现有智能手机，而是从零开始围绕当前贝塔伏特技术所能支持的功率预算设计的设备。第一天的能力与2026年的旗舰手机相比，确实有限。但随着电源架构的进步，这一差距将逐渐缩小。
源物理学：
氚（H-3）通过贝塔衰变转变为氦-3。
释放的电子最大能量为18.6 keV，平均能量约为5.7 keV。中微子携带其余能量，无法回收。贝塔谱是连续的——电子能量的统计分布直至终点——这就是为什么在功率计算中，平均能量而非最大能量是相关数据的原因。
用于功率计算的氚活度：1居里（Curie）的氚每秒有3.7 × 10¹⁰次衰变。在每次衰变5.7 keV的平均能量下，1居里的氚释放的能量约为1.6 × 10⁻¹⁹ J/eV × 5,700 eV × 3.7 × 10¹⁰ dis/s = 约33.8 mW的总贝塔能量。这是可回收功率的上限——转换效率决定了多少能转化为电能。
氚的比活度为9,650 Ci/g。因此，一克氚释放的总贝塔能量约为33.8 mW/Ci × 9,650 Ci/g = 326 W/g。商业贝塔伏特电池的转换效率为1-4%，电输出约为3.3-13 W/g，但氚气体分散，实际电池包括基材，因此每单位体积的功率密度远低于每克氚的功率密度。目前商业产品中，实际组装电池的功率密度在1-10 mW/cm²之间。
位于迈阿密的City Labs在NRC许可下商业生产氚硅贝塔伏特电池。他们当前的电池在典型负载下的有效接触面积范围内产生50-300 μW/cm²的功率。以150 μW/cm²作为规划目的的中间值是合理的。
一个多层堆叠电池模块，每层有35 cm²的有效接触面积，共四层——总堆叠体积大致相当于传统电池组——可产生约35 cm² × 4层 × 150 μW/cm² = 21 mW的总输出。考虑到互连损耗和电池间的差异，模块端口的净输出约为15-18 mW的连续功率。
保守估计——较少的层数、较低产量的电池——大约在5-8 mW之间。我正在围绕5 mW的连续功率作为底线设计第一天的架构，目标是随着电池技术的成熟达到15 mW。
衰变曲线由半衰期决定：
P(t) = P₀ × (1/2)^(t/12.32)
在t = 5年时：P = P₀ × 0.755。在t = 8年时：P = P₀ × 0.629。在t = 12年时：P = P₀ × 0.499。设备在第12年时的输出为其启动输出的一半。这个输出在多年之前就可以高精度预测。没有电池的衰老曲线如此干净。
衰变曲线与升级周期：
一款以5 mW启动的设备在第12年达到第一天的能力底线。历史上没有任何消费电子设备是围绕十年的可预测、透明的衰减设计的，而不是由软件过时驱动的两年更换周期。这款设备就是。
我在寻找的：
我是创始人。我正在进行一轮种子前融资，以签署两位联合创始人：一位具有贝塔伏特或核电池硬件经验的人，能够领导原型项目。如果你是那个人，或者你认识这样的人，我希望能与你交谈。
欲了解完整的文章和技术白皮书，或直接与我联系，请访问shelvin.com。
查看原文
Every smartphone built in the last twenty years assumes the same thing: the battery will be charged once a day. Before I get to the math, sit with this for a moment: what would actually change if a phone never needed to be plugged in?<p>That is what I am building. The power source is betavoltaic — electricity generated continuously from the decay of a radioactive isotope, with no moving parts, no combustion, no charging infrastructure.<p>The device I am describing is not a current smartphone with a different power source. It is a device designed from the ground up around a power budget that current betavoltaic technology can sustain. Day One capability is meaningfully more limited than a 2026 flagship phone. That gap closes as the power architecture advances.<p>The source physics:<p>Tritium (H-3) decays by beta emission to helium-3.<p>The emitted electron carries a maximum energy of 18.6 keV, with a mean energy of approximately 5.7 keV. The neutrino carries the rest and is unrecoverable. The beta spectrum is continuous — a statistical distribution of electron energies up to the endpoint — which is why mean energy, not maximum, is the relevant figure for power calculations.<p>Tritium activity for power calculation: 1 Curie of tritium is 3.7 × 10¹⁰ disintegrations per second. At 5.7 keV mean energy per disintegration, 1 Ci of tritium releases approximately 1.6 × 10⁻¹⁹ J/eV × 5,700 eV × 3.7 × 10¹⁰ dis/s = approximately 33.8 mW of total beta energy. This is the upper bound on recoverable power — conversion efficiency determines how much becomes electricity.<p>Tritium's specific activity is 9,650 Ci/g. One gram of tritium therefore releases approximately 33.8 mW/Ci × 9,650 Ci/g = 326 W/g of total beta energy. Commercial betavoltaic cells at 1–4% conversion efficiency yield approximately 3.3–13 W/g electrical output — but tritium gas is diffuse and the actual cell includes a substrate, so power density per unit volume of assembled cell is far lower than per gram of tritium. Practical assembled cell power densities in current commercial products run 1–10 mW/cm².<p>City Labs in Miami produces tritium-on-silicon betavoltaic cells commercially under NRC license. Their current cells produce in the range of 50–300 μW/cm² of active junction area under typical loading. A mid-range figure of 150 μW/cm² is defensible for planning purposes with current commercial cells.<p>A multi-layer stacked cell module with 35 cm² of active junction area per layer and four layers — a total stack volume roughly comparable to a conventional battery pack — yields approximately 35 cm² × 4 layers × 150 μW/cm² = 21 mW total output. Accounting for interconnect losses and cell-to-cell variation, net output at the module terminals: approximately 15–18 mW continuous.<p>The conservative number — fewer layers, lower-yield cells — lands around 5–8 mW. I am designing the Day One architecture around 5 mW continuous as the floor, with 15 mW as the target as cell technology matures.<p>The decay curve is governed by the half-life
P(t) = P₀ × (1/2)^(t/12.32)<p>At t = 5 years: P = P₀ × 0.755. At t = 8 years: P = P₀ × 0.629. At t = 12 years: P = P₀ × 0.499. The device at year 12 produces half its launch output. That output is predictable to high precision years in advance. No battery ages on a curve this clean.<p>The decay curve and upgrade cycle:<p>A device launched at 5 mW reaches the Day One capability floor at year 12. No consumer electronics device in history has been designed around a decade of predictable, transparent degradation rather than a two-year replacement cycle driven by software obsolescence. This one is.<p>What I'm looking for:<p>I'm the founder. I'm raising a pre-seed round to sign two co-founders: one with betavoltaic or nuclear battery hardware experience who can lead the prototype program. If you're that person, or you know them, I'd like to talk.<p>For the full essay and technical whitepaper, or to contact me directly, visit shelvin.com.