update qeq, matrix invertion method and projected gradient method werâ�e added, units of some paramters were unified

Author: gust-07

docs/user_guide/4.3ADMPQeqForce.md

@@ -184,7 +184,7 @@ To create a ADMP QEQ potential function, one needs to do the following in python
 hamilt = Hamiltonian('forcefield.xml')
 pot = hamilt.createPotential(dmfftop, nonbondedCutoff=rc*unit.nanometer, nonbondedMethod=app.PME,
                             ethresh=1e-3, neutral=True, slab=True, constQ=True,
-                            const_list=const_list, const_vals=const_val,
+                            const_list=const_list, const_vals=const_val, part_const=True
                             has_aux=True
                              )
 ```
@@ -206,6 +206,7 @@ A few keywords in `createPotential` would impact the behavior of the function:
 * `const_vals`: the corresponding constraint valu on each residue.
 
 * `ethresh`: the accuracy of PME, used to setup the size of the meshgrid of PME.
+* `part_const`: bool type, whether there are constraint groups or not.
 * `has_aux`: bool type, whether to intruduce auxilliary parameters, equilibrated atomic charges, etc.
 
 ## 3. ADMPQeqGenerator Doc
@@ -252,6 +253,9 @@ The backend of the ADMP QEQ energy is an `ADMPQeqForce` object. It contains the
 * `e_sr`: short range damping energy function
 * `e_site`: on site energy function
 * `coul_energy`: point charge interaction energy function
+* `icount`: times to run get_energy function
+* `hessinv_stride`: strides to update charges by matrix inversion method
+* `qupdate_stride`: strides to update charges
 
 ***METHOS*** 
 
@@ -278,8 +282,6 @@ The backend of the ADMP QEQ energy is an `ADMPQeqForce` object. It contains the
           Np: interacting pair indices within cutoff distance
       eta:
           N: the atomic length damping parameter
-      ds:
-          Np: topology distance of pairs
       buffer_scales:
           N * 3: buffer scales on pairs
       mscales:
@@ -288,14 +290,14 @@ The backend of the ADMP QEQ energy is an `ADMPQeqForce` object. It contains the
       energy: charge related coulombic energy plus constraint energy      
   ```
 
-* `E_grads`:
-
- Same as `E_full`, only return the concatenation of gradients of funtion `E_ful` on the first two paramters.
-
-* `get_energy`:
+* `E_hession`: 
+  ```
+  Same as `E_full`, only return hession matrix of `E_full`.
+  ```
 
+* `get_int_energy`: 
   ```
-  This contains charge equilbrant process, and retruns total energy after charge equilibration.
+  This contains charge equilbrant process by matrix inversion method, and retruns total energy after charge equilibration.
   
   Input:
       positions:
@@ -318,3 +320,30 @@ The backend of the ADMP QEQ energy is an `ADMPQeqForce` object. It contains the
       energy: total energy after charge equilibration
       aux: auxilliary parameters, including equilibrated atomic charges and lagmt 
   ```
+
+* `get_proj_grad`: 
+  ```
+  Function to get the value and projected gradients.
+
+   Input:
+      func:
+          function: position
+      constriant_matrix:
+          N * Na: N constrained groups
+      has_aux:
+          bool type: whether to return auxiliaries.
+  ```      
+
+* `get_step_energy`: 
+  ```
+  This contains charge equilbrant process by matrix inversion method or projected gradient method, it depends on step count and `hessinv_stride`, and retruns total energy after charge equilibration.
+
+   The inputs are the as `E_full`.
+  ```      
+
+* `get_energy`: 
+  ```
+  This contains charge equilbrant process and total energy calculation, retruns total energy every step and update charges by every strides of `qupdate_stride`.
+
+   The inputs are the as `E_full`.
+  ```