@RebelYoung 出现负值不代表计算结果是负的。数值溢出也会成负值。就是大到极限了。 温度的 min max 最后一步最大值不但跳增,而且离上一个坐标点很远,这很明显有问题啊。一个时间步至少增长了4000K,你的时间步还那么的小。这种增长率很可能是数值溢出了。 如果问题描述搭的附件太多,会看起来很难成一体,即使尽力地暗逻辑描述。一般这种状况的信息是像引用文献那样表述的。下面是个例子,自由软件基金会的公开邮件。不过因为是纯文本格式的,所以网址是直接贴的,你可以用超链接。 这样主体信息和逻辑就不会被附件打断了。现在看你的帖子有种听微信语音的感觉 This spring, as the time for planning our [biannual appeal][1] came around, we discussed the difficult time all of us are experiencing: charities like us, the free software community, and every individual. And it led us to consider why people from all walks of life cherish user freedom. [1]: https://www.fsf.org/appeal?pk_campaign=frspring2020&pk_source=collab1 The socially distant, digital way in which we are carrying on our work and private lives is affecting our software freedom. Globally, decisions to transition to an online and remote life were made with less consideration than we normally put into them, giving proprietary corporations access to parts of our lives we normally protect. Lately, we have been pointing to grim examples of [bulk surveillance][2] and [privacy violations][3] in the realms of education and [communication][4] to help everyone understand why this fight is so important. [2]: https://www.fsf.org/blogs/community/remote-education-does-not-require-giving-up-rights-to-freedom-and-privacy [3]: https://www.fsf.org/bulletin/2019/spring/its-not-just-about-privacy [4]: https://www.fsf.org/blogs/community/better-than-zoom-try-these-free-software-tools-for-staying-in-touch