X-Git-Url: https://hackdaworld.org/gitweb/?a=blobdiff_plain;f=potentials%2Ftersoff.c;fp=potentials%2Ftersoff.c;h=87bb1874077c39ca7992030626de689ab1b0e57d;hb=caa3bc828974c35df2462fde737c31c0a618ee4e;hp=0000000000000000000000000000000000000000;hpb=cb177e7c208a85b45d77b09fcada23b62d0248b5;p=physik%2Fposic.git

diff --git a/potentials/tersoff.c b/potentials/tersoff.c
new file mode 100644
index 0000000..87bb187
--- /dev/null
+++ b/potentials/tersoff.c
@@ -0,0 +1,707 @@
+/*
+ * tersoff.c - tersoff potential
+ *
+ * author: Frank Zirkelbach <frank.zirkelbach@physik.uni-augsburg.de>
+ *
+ */
+
+#define _GNU_SOURCE
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <fcntl.h>
+#include <unistd.h>
+#include <math.h>
+
+#include "../moldyn.h"
+#include "../math/math.h"
+//#include "tersoff.h"
+
+/* create mixed terms from parameters and set them */
+int tersoff_mult_complete_params(t_tersoff_mult_params *p) {
+
+	printf("[moldyn] tersoff parameter completion\n");
+	p->S2[0]=p->S[0]*p->S[0];
+	p->S2[1]=p->S[1]*p->S[1];
+	p->Smixed=sqrt(p->S[0]*p->S[1]);
+	p->S2mixed=p->Smixed*p->Smixed;
+	p->Rmixed=sqrt(p->R[0]*p->R[1]);
+	p->Amixed=sqrt(p->A[0]*p->A[1]);
+	p->Bmixed=sqrt(p->B[0]*p->B[1]);
+	p->lambda_m=0.5*(p->lambda[0]+p->lambda[1]);
+	p->mu_m=0.5*(p->mu[0]+p->mu[1]);
+
+	printf("[moldyn] tersoff mult parameter info:\n");
+	printf("  S (A)  | %f | %f | %f\n",p->S[0],p->S[1],p->Smixed);
+	printf("  R (A)  | %f | %f | %f\n",p->R[0],p->R[1],p->Rmixed);
+	printf("  A (eV) | %f | %f | %f\n",p->A[0]/EV,p->A[1]/EV,p->Amixed/EV);
+	printf("  B (eV) | %f | %f | %f\n",p->B[0]/EV,p->B[1]/EV,p->Bmixed/EV);
+	printf("  lambda | %f | %f | %f\n",p->lambda[0],p->lambda[1],
+	                                  p->lambda_m);
+	printf("  mu     | %f | %f | %f\n",p->mu[0],p->mu[1],p->mu_m);
+	printf("  beta   | %.10f | %.10f\n",p->beta[0],p->beta[1]);
+	printf("  n      | %f | %f\n",p->n[0],p->n[1]);
+	printf("  c      | %f | %f\n",p->c[0],p->c[1]);
+	printf("  d      | %f | %f\n",p->d[0],p->d[1]);
+	printf("  h      | %f | %f\n",p->h[0],p->h[1]);
+	printf("  chi    | %f \n",p->chi);
+
+	return 0;
+}
+
+/* tersoff 1 body part */
+int tersoff_mult_1bp(t_moldyn *moldyn,t_atom *ai) {
+
+	int brand;
+	t_tersoff_mult_params *params;
+	t_tersoff_exchange *exchange;
+	
+	brand=ai->brand;
+	params=moldyn->pot1b_params;
+	exchange=&(params->exchange);
+
+	/*
+	 * simple: point constant parameters only depending on atom i to
+	 *         their right values
+	 */
+
+	exchange->beta_i=&(params->beta[brand]);
+	exchange->n_i=&(params->n[brand]);
+	exchange->c_i=&(params->c[brand]);
+	exchange->d_i=&(params->d[brand]);
+	exchange->h_i=&(params->h[brand]);
+
+	exchange->betaini=pow(*(exchange->beta_i),*(exchange->n_i));
+	exchange->ci2=params->c[brand]*params->c[brand];
+	exchange->di2=params->d[brand]*params->d[brand];
+	exchange->ci2di2=exchange->ci2/exchange->di2;
+
+	return 0;
+}
+	
+/* tersoff 2 body part */
+int tersoff_mult_2bp(t_moldyn *moldyn,t_atom *ai,t_atom *aj,u8 bc) {
+
+	t_tersoff_mult_params *params;
+	t_tersoff_exchange *exchange;
+	t_3dvec dist_ij,force;
+	double d_ij,d_ij2;
+	double A,B,R,S,S2,lambda,mu;
+	double f_r,df_r;
+	double f_c,df_c;
+	int brand;
+	double s_r;
+	double arg;
+
+	params=moldyn->pot2b_params;
+	brand=aj->brand;
+	exchange=&(params->exchange);
+
+	/* clear 3bp and 2bp post run */
+	exchange->run3bp=0;
+	exchange->run2bp_post=0;
+
+	/* reset S > r > R mark */
+	exchange->d_ij_between_rs=0;
+	
+	/*
+	 * calc of 2bp contribution of V_ij and dV_ij/ji
+	 *
+	 * for Vij and dV_ij we need:
+	 * - f_c_ij, df_c_ij
+	 * - f_r_ij, df_r_ij
+	 *
+	 * for dV_ji we need:
+	 * - f_c_ji = f_c_ij, df_c_ji = df_c_ij
+	 * - f_r_ji = f_r_ij; df_r_ji = df_r_ij
+	 *
+	 */
+
+	/* constants */
+	if(brand==ai->brand) {
+		S=params->S[brand];
+		S2=params->S2[brand];
+		R=params->R[brand];
+		A=params->A[brand];
+		B=params->B[brand];
+		lambda=params->lambda[brand];
+		mu=params->mu[brand];
+		exchange->chi=1.0;
+	}
+	else {
+		S=params->Smixed;
+		S2=params->S2mixed;
+		R=params->Rmixed;
+		A=params->Amixed;
+		B=params->Bmixed;
+		lambda=params->lambda_m;
+		mu=params->mu_m;
+		params->exchange.chi=params->chi;
+	}
+
+	/* dist_ij, d_ij */
+	v3_sub(&dist_ij,&(aj->r),&(ai->r));
+	if(bc) check_per_bound(moldyn,&dist_ij);
+	d_ij2=v3_absolute_square(&dist_ij);
+
+	/* if d_ij2 > S2 => no force & potential energy contribution */
+	if(d_ij2>S2)
+		return 0;
+
+	/* now we will need the distance */
+	//d_ij=v3_norm(&dist_ij);
+	d_ij=sqrt(d_ij2);
+
+	/* save for use in 3bp */
+	exchange->d_ij=d_ij;
+	exchange->d_ij2=d_ij2;
+	exchange->dist_ij=dist_ij;
+
+	/* more constants */
+	exchange->beta_j=&(params->beta[brand]);
+	exchange->n_j=&(params->n[brand]);
+	exchange->c_j=&(params->c[brand]);
+	exchange->d_j=&(params->d[brand]);
+	exchange->h_j=&(params->h[brand]);
+	if(brand==ai->brand) {
+		exchange->betajnj=exchange->betaini;
+		exchange->cj2=exchange->ci2;
+		exchange->dj2=exchange->di2;
+		exchange->cj2dj2=exchange->ci2di2;
+	}
+	else {
+		exchange->betajnj=pow(*(exchange->beta_j),*(exchange->n_j));
+		exchange->cj2=params->c[brand]*params->c[brand];
+		exchange->dj2=params->d[brand]*params->d[brand];
+		exchange->cj2dj2=exchange->cj2/exchange->dj2;
+	}
+
+	/* f_r_ij = f_r_ji, df_r_ij = df_r_ji */
+	f_r=A*exp(-lambda*d_ij);
+	df_r=lambda*f_r/d_ij;
+
+	/* f_a, df_a calc (again, same for ij and ji) | save for later use! */
+	exchange->f_a=-B*exp(-mu*d_ij);
+	exchange->df_a=mu*exchange->f_a/d_ij;
+
+	/* f_c, df_c calc (again, same for ij and ji) */
+	if(d_ij<R) {
+		/* f_c = 1, df_c = 0 */
+		f_c=1.0;
+		df_c=0.0;
+		/* two body contribution (ij, ji) */
+		v3_scale(&force,&dist_ij,-df_r);
+	}
+	else {
+		s_r=S-R;
+		arg=M_PI*(d_ij-R)/s_r;
+		f_c=0.5+0.5*cos(arg);
+		df_c=0.5*sin(arg)*(M_PI/(s_r*d_ij));
+		/* two body contribution (ij, ji) */
+		v3_scale(&force,&dist_ij,-df_c*f_r-df_r*f_c);
+		/* tell 3bp that S > r > R */
+		exchange->d_ij_between_rs=1;
+	}
+
+	/* add forces of 2bp (ij, ji) contribution
+	 * dVij = dVji and we sum up both: no 1/2) */
+	v3_add(&(ai->f),&(ai->f),&force);
+
+	/* virial */
+	ai->virial.xx-=force.x*dist_ij.x;
+	ai->virial.yy-=force.y*dist_ij.y;
+	ai->virial.zz-=force.z*dist_ij.z;
+	ai->virial.xy-=force.x*dist_ij.y;
+	ai->virial.xz-=force.x*dist_ij.z;
+	ai->virial.yz-=force.y*dist_ij.z;
+
+#ifdef DEBUG
+if(ai==&(moldyn->atom[0])) {
+	printf("dVij, dVji (2bp) contrib:\n");
+	printf("%f | %f\n",force.x,ai->f.x);
+	printf("%f | %f\n",force.y,ai->f.y);
+	printf("%f | %f\n",force.z,ai->f.z);
+}
+#endif
+#ifdef VDEBUG
+if(ai==&(moldyn->atom[0])) {
+	printf("dVij, dVji (2bp) contrib:\n");
+	printf("%f | %f\n",force.x*dist_ij.x,ai->virial.xx);
+	printf("%f | %f\n",force.y*dist_ij.y,ai->virial.yy);
+	printf("%f | %f\n",force.z*dist_ij.z,ai->virial.zz);
+}
+#endif
+
+	/* energy 2bp contribution (ij, ji) is 0.5 f_r f_c ... */
+	moldyn->energy+=(0.5*f_r*f_c);
+
+	/* save for use in 3bp */
+	exchange->f_c=f_c;
+	exchange->df_c=df_c;
+
+	/* enable the run of 3bp function and 2bp post processing */
+	exchange->run3bp=1;
+	exchange->run2bp_post=1;
+
+	/* reset 3bp sums */
+	exchange->zeta_ij=0.0;
+	exchange->zeta_ji=0.0;
+	v3_zero(&(exchange->dzeta_ij));
+	v3_zero(&(exchange->dzeta_ji));
+
+	return 0;
+}
+
+/* tersoff 2 body post part */
+
+int tersoff_mult_post_2bp(t_moldyn *moldyn,t_atom *ai,t_atom *aj,u8 bc) {
+
+	/*
+	 * here we have to allow for the 3bp sums
+	 *
+	 * that is:
+	 * - zeta_ij, dzeta_ij
+	 * - zeta_ji, dzeta_ji
+	 *
+	 * to compute the 3bp contribution to:
+	 * - Vij, dVij
+	 * - dVji
+	 *
+	 */
+
+	t_tersoff_mult_params *params;
+	t_tersoff_exchange *exchange;
+
+	t_3dvec force,temp;
+	t_3dvec *dist_ij;
+	double b,db,tmp;
+	double f_c,df_c,f_a,df_a;
+	double chi,ni,betaini,nj,betajnj;
+	double zeta;
+
+	params=moldyn->pot2b_params;
+	exchange=&(params->exchange);
+
+	/* we do not run if f_c_ij was detected to be 0! */
+	if(!(exchange->run2bp_post))
+		return 0;
+
+	f_c=exchange->f_c;
+	df_c=exchange->df_c;
+	f_a=exchange->f_a;
+	df_a=exchange->df_a;
+	betaini=exchange->betaini;
+	betajnj=exchange->betajnj;
+	ni=*(exchange->n_i);
+	nj=*(exchange->n_j);
+	chi=exchange->chi;
+	dist_ij=&(exchange->dist_ij);
+	
+	/* Vij and dVij */
+	zeta=exchange->zeta_ij;
+	if(zeta==0.0) {
+		moldyn->debug++;		/* just for debugging ... */
+		b=chi;
+		v3_scale(&force,dist_ij,df_a*b*f_c);
+	}
+	else {
+		tmp=betaini*pow(zeta,ni-1.0);		/* beta^n * zeta^n-1 */
+		b=(1+zeta*tmp);				/* 1 + beta^n zeta^n */
+		db=chi*pow(b,-1.0/(2*ni)-1);		/* x(...)^(-1/2n - 1) */
+		b=db*b;					/* b_ij */
+		db*=-0.5*tmp;				/* db_ij */
+		v3_scale(&force,&(exchange->dzeta_ij),f_a*db);
+		v3_scale(&temp,dist_ij,df_a*b);
+		v3_add(&force,&force,&temp);
+		v3_scale(&force,&force,f_c);
+	}
+	v3_scale(&temp,dist_ij,df_c*b*f_a);
+	v3_add(&force,&force,&temp);
+	v3_scale(&force,&force,-0.5);
+
+	/* add force */
+	v3_add(&(ai->f),&(ai->f),&force);
+
+	/* virial */
+	ai->virial.xx-=force.x*dist_ij->x;
+	ai->virial.yy-=force.y*dist_ij->y;
+	ai->virial.zz-=force.z*dist_ij->z;
+	ai->virial.xy-=force.x*dist_ij->y;
+	ai->virial.xz-=force.x*dist_ij->z;
+	ai->virial.yz-=force.y*dist_ij->z;
+
+#ifdef DEBUG
+if(ai==&(moldyn->atom[0])) {
+	printf("dVij (3bp) contrib:\n");
+	printf("%f | %f\n",force.x,ai->f.x);
+	printf("%f | %f\n",force.y,ai->f.y);
+	printf("%f | %f\n",force.z,ai->f.z);
+}
+#endif
+#ifdef VDEBUG
+if(ai==&(moldyn->atom[0])) {
+	printf("dVij (3bp) contrib:\n");
+	printf("%f | %f\n",force.x*dist_ij->x,ai->virial.xx);
+	printf("%f | %f\n",force.y*dist_ij->y,ai->virial.yy);
+	printf("%f | %f\n",force.z*dist_ij->z,ai->virial.zz);
+}
+#endif
+
+	/* add energy of 3bp sum */
+	moldyn->energy+=(0.5*f_c*b*f_a);
+
+	/* dVji */
+	zeta=exchange->zeta_ji;
+	if(zeta==0.0) {
+		moldyn->debug++;
+		b=chi;
+		v3_scale(&force,dist_ij,df_a*b*f_c);
+	}
+	else {
+		tmp=betajnj*pow(zeta,nj-1.0);		/* beta^n * zeta^n-1 */
+		b=(1+zeta*tmp);				/* 1 + beta^n zeta^n */
+		db=chi*pow(b,-1.0/(2*nj)-1);		/* x(...)^(-1/2n - 1) */
+		b=db*b;					/* b_ij */
+		db*=-0.5*tmp;				/* db_ij */
+		v3_scale(&force,&(exchange->dzeta_ji),f_a*db);
+		v3_scale(&temp,dist_ij,df_a*b);
+		v3_add(&force,&force,&temp);
+		v3_scale(&force,&force,f_c);
+	}
+	v3_scale(&temp,dist_ij,df_c*b*f_a);
+	v3_add(&force,&force,&temp);
+	v3_scale(&force,&force,-0.5);
+
+	/* add force */
+	v3_add(&(ai->f),&(ai->f),&force);
+
+	/* virial - plus sign, as dist_ij = - dist_ji - (really??) */
+// TEST ... with a minus instead
+	ai->virial.xx-=force.x*dist_ij->x;
+	ai->virial.yy-=force.y*dist_ij->y;
+	ai->virial.zz-=force.z*dist_ij->z;
+	ai->virial.xy-=force.x*dist_ij->y;
+	ai->virial.xz-=force.x*dist_ij->z;
+	ai->virial.yz-=force.y*dist_ij->z;
+
+#ifdef DEBUG
+if(ai==&(moldyn->atom[0])) {
+	printf("dVji (3bp) contrib:\n");
+	printf("%f | %f\n",force.x,ai->f.x);
+	printf("%f | %f\n",force.y,ai->f.y);
+	printf("%f | %f\n",force.z,ai->f.z);
+}
+#endif
+#ifdef VDEBUG
+if(ai==&(moldyn->atom[0])) {
+	printf("dVji (3bp) contrib:\n");
+	printf("%f | %f\n",force.x*dist_ij->x,ai->virial.xx);
+	printf("%f | %f\n",force.y*dist_ij->y,ai->virial.yy);
+	printf("%f | %f\n",force.z*dist_ij->z,ai->virial.zz);
+}
+#endif
+
+	return 0;
+}
+
+/* tersoff 3 body part */
+
+int tersoff_mult_3bp(t_moldyn *moldyn,t_atom *ai,t_atom *aj,t_atom *ak,u8 bc) {
+
+	t_tersoff_mult_params *params;
+	t_tersoff_exchange *exchange;
+	t_3dvec dist_ij,dist_ik,dist_jk;
+	t_3dvec temp1,temp2;
+	t_3dvec *dzeta;
+	double R,S,S2,s_r;
+	double B,mu;
+	double d_ij,d_ik,d_jk,d_ij2,d_ik2,d_jk2;
+	double rr,dd;
+	double f_c,df_c;
+	double f_c_ik,df_c_ik,arg;
+	double f_c_jk;
+	double n,c,d,h;
+	double c2,d2,c2d2;
+	double cos_theta,d_costheta1,d_costheta2;
+	double h_cos,d2_h_cos2;
+	double frac,g,zeta,chi;
+	double tmp;
+	int brand;
+
+	params=moldyn->pot3b_params;
+	exchange=&(params->exchange);
+
+	if(!(exchange->run3bp))
+		return 0;
+
+	/*
+	 * calc of 3bp contribution of V_ij and dV_ij/ji/jk &
+	 * 2bp contribution of dV_jk
+	 *
+	 * for Vij and dV_ij we still need:
+	 * - b_ij, db_ij (zeta_ij)
+	 *   - f_c_ik, df_c_ik, constants_i, cos_theta_ijk, d_costheta_ijk
+	 *
+	 * for dV_ji we still need:
+	 * - b_ji, db_ji (zeta_ji)
+	 *   - f_c_jk, d_c_jk, constants_j, cos_theta_jik, d_costheta_jik
+	 *
+	 * for dV_jk we need:
+	 * - f_c_jk
+	 * - f_a_jk
+	 * - db_jk (zeta_jk)
+	 *   - f_c_ji, df_c_ji, constants_j, cos_theta_jki, d_costheta_jki
+	 *
+	 */
+
+	/*
+	 * get exchange data 
+	 */
+
+	/* dist_ij, d_ij - this is < S_ij ! */
+	dist_ij=exchange->dist_ij;
+	d_ij=exchange->d_ij;
+	d_ij2=exchange->d_ij2;
+
+	/* f_c_ij, df_c_ij (same for ji) */
+	f_c=exchange->f_c;
+	df_c=exchange->df_c;
+
+	/*
+	 * calculate unknown values now ...
+	 */
+
+	/* V_ij and dV_ij stuff (in b_ij there is f_c_ik) */
+
+	/* dist_ik, d_ik */
+	v3_sub(&dist_ik,&(ak->r),&(ai->r));
+	if(bc) check_per_bound(moldyn,&dist_ik);
+	d_ik2=v3_absolute_square(&dist_ik);
+
+	/* ik constants */
+	brand=ai->brand;
+	if(brand==ak->brand) {
+		R=params->R[brand];
+		S=params->S[brand];
+		S2=params->S2[brand];
+	}
+	else {
+		R=params->Rmixed;
+		S=params->Smixed;
+		S2=params->S2mixed;
+	}
+
+	/* zeta_ij/dzeta_ij contribution only for d_ik < S */
+	if(d_ik2<S2) {
+
+		/* now we need d_ik */
+		d_ik=sqrt(d_ik2);
+
+		/* get constants_i from exchange data */
+		n=*(exchange->n_i);
+		c=*(exchange->c_i);
+		d=*(exchange->d_i);
+		h=*(exchange->h_i);
+		c2=exchange->ci2;
+		d2=exchange->di2;
+		c2d2=exchange->ci2di2;
+
+		/* cosine of theta_ijk by scalaproduct */
+		rr=v3_scalar_product(&dist_ij,&dist_ik);
+		dd=d_ij*d_ik;
+		cos_theta=rr/dd;
+
+		/* d_costheta */
+		tmp=1.0/dd;
+		d_costheta1=cos_theta/d_ij2-tmp;
+		d_costheta2=cos_theta/d_ik2-tmp;
+
+		/* some usefull values */
+		h_cos=(h-cos_theta);
+		d2_h_cos2=d2+(h_cos*h_cos);
+		frac=c2/(d2_h_cos2);
+
+		/* g(cos_theta) */
+		g=1.0+c2d2-frac;
+
+		/* d_costheta_ij and dg(cos_theta) - needed in any case! */
+		v3_scale(&temp1,&dist_ij,d_costheta1);
+		v3_scale(&temp2,&dist_ik,d_costheta2);
+		v3_add(&temp1,&temp1,&temp2);
+		v3_scale(&temp1,&temp1,-2.0*frac*h_cos/d2_h_cos2); /* dg */
+
+		/* f_c_ik & df_c_ik + {d,}zeta contribution */
+		dzeta=&(exchange->dzeta_ij);
+		if(d_ik<R) {
+			/* {d,}f_c_ik */
+			// => f_c_ik=1.0;
+			// => df_c_ik=0.0; of course we do not set this!
+
+			/* zeta_ij */
+			exchange->zeta_ij+=g;
+
+			/* dzeta_ij */
+			v3_add(dzeta,dzeta,&temp1);
+		}
+		else {
+			/* {d,}f_c_ik */
+			s_r=S-R;
+			arg=M_PI*(d_ik-R)/s_r;
+			f_c_ik=0.5+0.5*cos(arg);
+			df_c_ik=0.5*sin(arg)*(M_PI/(s_r*d_ik));
+
+			/* zeta_ij */
+			exchange->zeta_ij+=f_c_ik*g;
+
+			/* dzeta_ij */
+			v3_scale(&temp1,&temp1,f_c_ik);
+			v3_scale(&temp2,&dist_ik,g*df_c_ik);
+			v3_add(&temp1,&temp1,&temp2);
+			v3_add(dzeta,dzeta,&temp1);
+		}
+	}
+
+	/* dV_ji stuff (in b_ji there is f_c_jk) + dV_jk stuff! */
+
+	/* dist_jk, d_jk */
+	v3_sub(&dist_jk,&(ak->r),&(aj->r));
+	if(bc) check_per_bound(moldyn,&dist_jk);
+	d_jk2=v3_absolute_square(&dist_jk);
+
+	/* jk constants */
+	brand=aj->brand;
+	if(brand==ak->brand) {
+		R=params->R[brand];
+		S=params->S[brand];
+		S2=params->S2[brand];
+		B=params->B[brand];
+		mu=params->mu[brand];
+		chi=1.0;
+	}
+	else {
+		R=params->Rmixed;
+		S=params->Smixed;
+		S2=params->S2mixed;
+		B=params->Bmixed;
+		mu=params->mu_m;
+		chi=params->chi;
+	}
+
+	/* zeta_ji/dzeta_ji contribution only for d_jk < S_jk */
+	if(d_jk2<S2) {
+
+		/* now we need d_ik */
+		d_jk=sqrt(d_jk2);
+
+		/* constants_j from exchange data */
+		n=*(exchange->n_j);
+		c=*(exchange->c_j);
+		d=*(exchange->d_j);
+		h=*(exchange->h_j);
+		c2=exchange->cj2;
+		d2=exchange->dj2;
+		c2d2=exchange->cj2dj2;
+
+		/* cosine of theta_jik by scalaproduct */
+		rr=-v3_scalar_product(&dist_ij,&dist_jk); /* -1, as ij -> ji */
+		dd=d_ij*d_jk;
+		cos_theta=rr/dd;
+
+		/* d_costheta */
+		d_costheta1=1.0/dd;
+		d_costheta2=cos_theta/d_ij2;
+
+		/* some usefull values */
+		h_cos=(h-cos_theta);
+		d2_h_cos2=d2+(h_cos*h_cos);
+		frac=c2/(d2_h_cos2);
+
+		/* g(cos_theta) */
+		g=1.0+c2d2-frac;
+
+		/* d_costheta_jik and dg(cos_theta) - needed in any case! */
+		v3_scale(&temp1,&dist_jk,d_costheta1);
+		v3_scale(&temp2,&dist_ij,-d_costheta2); /* ji -> ij => -1 */
+		//v3_add(&temp1,&temp1,&temp2);
+		v3_sub(&temp1,&temp1,&temp2); /* there is a minus! */
+		v3_scale(&temp1,&temp1,-2.0*frac*h_cos/d2_h_cos2); /* dg */
+
+		/* store dg in temp2 and use it for dVjk later */
+		v3_copy(&temp2,&temp1);
+
+		/* f_c_jk + {d,}zeta contribution (df_c_jk = 0) */
+		dzeta=&(exchange->dzeta_ji);
+		if(d_jk<R) {
+			/* f_c_jk */
+			f_c_jk=1.0;
+
+			/* zeta_ji */
+			exchange->zeta_ji+=g;
+
+			/* dzeta_ji */
+			v3_add(dzeta,dzeta,&temp1);
+		}
+		else {
+			/* f_c_jk */
+			s_r=S-R;
+			arg=M_PI*(d_jk-R)/s_r;
+			f_c_jk=0.5+0.5*cos(arg);
+
+			/* zeta_ji */
+			exchange->zeta_ji+=f_c_jk*g;
+
+			/* dzeta_ji */
+			v3_scale(&temp1,&temp1,f_c_jk);
+			v3_add(dzeta,dzeta,&temp1);
+		}
+
+		/* dV_jk stuff | add force contribution on atom i immediately */
+		if(exchange->d_ij_between_rs) {
+			zeta=f_c*g;
+			v3_scale(&temp1,&temp2,f_c);
+			v3_scale(&temp2,&dist_ij,df_c*g);
+			v3_add(&temp2,&temp2,&temp1); /* -> dzeta_jk in temp2 */
+		}
+		else {
+			zeta=g;
+			// dzeta_jk is simply dg, which is stored in temp2
+		}
+		/* betajnj * zeta_jk ^ nj-1 */
+		tmp=exchange->betajnj*pow(zeta,(n-1.0));
+		tmp=-chi/2.0*pow((1+tmp*zeta),(-1.0/(2.0*n)-1))*tmp;
+		v3_scale(&temp2,&temp2,tmp*B*exp(-mu*d_jk)*f_c_jk*0.5);
+		v3_add(&(ai->f),&(ai->f),&temp2); /* -1 skipped in f_a calc ^ */
+		                                  /* scaled with 0.5 ^ */
+
+		/* virial */
+		ai->virial.xx-=temp2.x*dist_jk.x;
+		ai->virial.yy-=temp2.y*dist_jk.y;
+		ai->virial.zz-=temp2.z*dist_jk.z;
+		ai->virial.xy-=temp2.x*dist_jk.y;
+		ai->virial.xz-=temp2.x*dist_jk.z;
+		ai->virial.yz-=temp2.y*dist_jk.z;
+
+#ifdef DEBUG
+if(ai==&(moldyn->atom[0])) {
+	printf("dVjk (3bp) contrib:\n");
+	printf("%f | %f\n",temp2.x,ai->f.x);
+	printf("%f | %f\n",temp2.y,ai->f.y);
+	printf("%f | %f\n",temp2.z,ai->f.z);
+}
+#endif
+#ifdef VDEBUG
+if(ai==&(moldyn->atom[0])) {
+	printf("dVjk (3bp) contrib:\n");
+	printf("%f | %f\n",temp2.x*dist_jk.x,ai->virial.xx);
+	printf("%f | %f\n",temp2.y*dist_jk.y,ai->virial.yy);
+	printf("%f | %f\n",temp2.z*dist_jk.z,ai->virial.zz);
+}
+#endif
+
+	}
+
+	return 0;
+}
+