优化这段matlab代码：n1 = 1.55; n2 = 1.52; NA = 1.45; lb = 515e-9; d = 10e-6; rp = 1; thetap = pi/4; zp = 0:20e-6:100e-6; a = (n1/n2)*asin(NA); a1 = 0:0.1:pi/2; % a1 = asin(NA/n1); a2 = asin(n1*sin(a1)/n2); k0 = 2*pi/lb; p = k0*rp*n1.*sin(a1); J0 = besselj(0,p); % zp = rp.*cos(thetap); f = -k0*d*(n1*cos(a1)-n2*cos(a2)); as = 2*sin(a2).*cos(a1)./sin(a1+a2); ap = 2*sin(a2).*cos(a1)./(sin(a1+a2).*cos(a1-a2)); I1 = (sqrt(cos(a1)).*sin(a1).*exp(1i*k0*f)).*(as+ap.*cos(a2).*J0.*exp(1i*k0.*zp.*n2.*cos(a2)));

可以考虑使用向量化的方式来优化代码，以减少循环的数量，提高计算速度。具体做法如下：

1. 将 a1 和 a2 的计算向量化，用 linspace 函数代替循环。

2. 将 p 和 J0 的计算向量化，用 meshgrid 函数代替循环。

3. 将 I1 的计算向量化，用点乘代替循环。

改写后的代码如下：

n1 = 1.55;
n2 = 1.52;
NA = 1.45;
lb = 515e-9;
d = 10e-6;
rp = 1;
thetap = pi/4;
zp = 0:20e-6:100e-6;
a = (n1/n2)*asin(NA);
a1 = linspace(0,pi/2,100);
a2 = asin(n1*sin(a1)/n2);
k0 = 2*pi/lb;
[p,J0] = meshgrid(k0*rp*n1.*sin(a1),besselj(0,k0*rp*n1.*sin(a1)));
f = -k0*d*(n1*cos(a1)-n2*cos(a2));
as = 2*sin(a2).*cos(a1)./sin(a1+a2);
ap = 2*sin(a2).*cos(a1)./(sin(a1+a2).*cos(a1-a2));
I1 = (sqrt(cos(a1)).*sin(a1).*exp(1i*k0*f)).*(as+ap.*cos(a2).*J0.*exp(1i*k0.*zp.*n2.*cos(a2)));

这样就可以减少循环的数量，提高计算速度。