I tried all the ways to calculate the particle alpha, and set shaderResource to the draw process, in the renderdoc, the screenDepthTexture is always no Resource.
You’re probably trying to use the same depth buffer texture in two stages of the pipeline at the same time: read in pixel shader to compute softness, and use the same texture as a depth render target in output merger stage.
This is not going to work. When you call OMSetRenderTargets, the pixel shader binding is unset for the resource view of that texture.
An easy workaround is making a copy of your depth texture with CopyResource, and bind the copy to the input of the pixel shader. This way your pixel shader can read from there, while output merger stage uses another copy as depth/stencil target.
In the future, to troubleshoot such issues use D3D11_CREATE_DEVICE_DEBUG flag when creating the device, and read debug output in visual studio. RenderDoc is awesome for higher level bugs, when you’re rendering something but it doesn’t look right. Your mistake is incorrect use of D3D API. That debug layer in Windows SDK is more useful than RenderDoc for such things.