You are on page 1of 7

Notes on studying the d6 scaling regime in our 1-D model of the capture velocity

To model the capture velocity, we imagine that an atom with speed v is slowed by a 1-D MOT for a distance d. If the atom is stopped in that distance, we say it is captured by the MOT. The capture velocity is the velocity of an atom that is stopped just at the edge of the interaction region. The 1-D MOT is dened by four parameters: the detuning1 , the saturation parameter s, the magnetic gradient b (or dBd/dz), and the stopping distance d. Our goal is to study the capture velocity as a function of these parameters. One can identify some general trends. When the stopping distance is small, the captured atoms only interact with the linear region of the force prole. As a result, the slowing force can be written F = v and the capture velocity depends linearly on the stopping distance. In particular, vc = d/m where is the friction coecient for optical molasses. From H. Metcalfs book, we have 8 k 2 s h . (, s) = 1 + s + 4 2 (1)

Now, the approximation that F = v only holds if stopping distance is suciently small. As d is increased, the MOT begins to capture atoms that interact with the nonlinear region of the force prole. In this regime, vc is no longer proportional to d, but there are several approximations that can be made. In the regime where vc is near the peak of the force prole, a good approximation is vc d. Physically, this amounts to saying that the slowing force on an atom with speed vc is nearly constant. Intutively, this regime occurs when an atom with speed vc mostly interacts with the peak of the force prole so that the force is large and roughly constant during the slowing. At still larger d, it is no longer reasonable to say that the force is nearly constant during the slowing and so vc grows more slowly than d. In this regime, a reasonable approximation is vc dn where n is a parameter to be t2 .
1

In these notes, all frequencies are measured in units of the natural linewidth. Also, the detuning is always to the red of the atomic resonance. In the past, we wrote that vc = a1 d + a2 with a1 and a2 as parameters to be t. This works reasonably well, but it does not lead to a scaling law N dm so I have stopped using this form. This perspective also allows the d3.6 scaling law cited in the literature to emerge naturally in our model.

2 This is the perspective that I presented in my talk at DAMOP. It is illustrated in Fig. 1

FIG. 1: The plot on the left shows the three regimes for vc vs d discussed above. The parameters for the simulation were s = 25, = 2, and b = 30 G/cm. The inset shows the force prole for these parameters. The black points on the force prole illustrate the position of vc on the force prole for several d. Note that at the edge of the vc d regime, the capture velocity is on the peak of the force prole.

We can ask: is it possible to obtain an estimate of the beam size d where the d6 scaling regime ends? A closely related question raised by John: are there natural combinations of the parameters that dene the dierent scaling regimes? In order to answer those questions, one needs a way to dene the edge of the regime where vc d. Looking at Fig. 1, an obvious choice is to dene the transition size dt as the intersection of the best t curves vc = a1 d and vc = a2 d. Another possibility is to say that dt is the length of the interaction region where vc coincides with the peak of the force prole. Based on Fig. 1, I expected that these two denitions would agree, but it turns out that this is not true in all cases. If we dene dt in terms of the peak of the force prole, it is straight forward to come up with an analytic expression for dt . As long as the detuning is not too small and the saturation parameter is not too large, the force prole peaks when the doppler shift is equal to the detuning. In that case, the velocity corresponding to the peak is vp = /k and therefore dt = m (, s) k (2)

3 In general though, one should solve for vp by maximizing F(v). This leads to vp (, s) = k k (s + 1) 1 4 +2 12 2 1+ 1 (1 + s)2 12 (1 + s) 2 + 4 4
2

(3)

(s + 1) (s + 1) +4 + 16 4 2

and the transition size is dt = m vp (, s) . (, s) (4)

Note that our estimates of dt neglect the magnetic eld entirely. This makes some sense because we expect d to be small in the regime where vc d, and as long as d is small, we expect the Zeeman shift to be small so that it will have a small eect on the slowing. I am not sure if this assumption is justied. The next question to consider is whether either Eq. 2 or 4 is a reasonable estimate for the transition size that we observe in numerical simulations. I came up with two methods to quantify dt from our simulations. One way is to use the intersection of the best t curves vc d and vc d. The second method is to say that dt occurs when the simulation deviates from the vc d trendline. Both of these methods involve some arbitrary judgement, but I have not found a better method yet. Using this approach, I estimated dt vs. with s = 25. My results are shown in Fig. 2. It appears that Eq. 4 provides an excellent estimate of the edge of the d6 scaling regime. It is interesting that Eq. 4 appears to be a better estimate of dt when b = 30 G/cm than it does for b = 10 G/cm. I am not sure what to make of this trend. Another observation is that Eq. 4 provides a better estimate of dt when dt is small (i.e. when is small). It is also interesting that when is larger, the regime where vc d appears to continue after vc crosses the peak of the force prole. This is illustrated in Fig. 3. I am not sure what to make of this trend. I thought that it might be due to the magnetic eld, but the case with b = 10 G/cm looks essentially the same so that may not be plausible. Figures 4-7 illustrate this analysis for the cases s=14 and s=5.

FIG. 2: This gure compares several denitions of dt vs. with s = 25. The points show my estimate of dt from numerical simulations with two dierent values of b. The solid lines show the two estimates of dt from Eq. 2 and 4. Both expressions describe the trend of dt vs. , but Eq. 4 provides a better estimate of dt , especially at low .

FIG. 3: The plots on the left illustrates vc vs d for two cases. The plots on the right show the force prole corresponding to these two cases. The black dot on the force prole indicates vc at the edge of the vc d regime (dened by the intersection of the two curves shown on the left. Note that for = 1.8, vc coincides with the peak of the force prole at dt , but when = 3, vc is well past the peak at the transition size.

FIG. 4: This estimate of dt two estimates but at large

gure compares several denitions of dt vs. with s = 14. The points show my from numerical simulations with two dierent values of b. The solid lines show the of dt from Eq. 2 and 4. For < 3, the trends are similar to those shown in Fig. 2, the denition of dt becomes problematic. This is illustrated in Fig. 5

FIG. 5: The plots on the left illustrates vc vs d for two cases. The plots on the right show the force prole corresponding to these two cases. The black dot on the force prole indicates vc at the edge of the vc d regime (dened by the intersection of the two curves shown on the left. Note that for = 3, the shape of the force prole is starting to get distorted so that the vc vs d curve leaves the linear trend line well before it intersects the vc d t.

FIG. 6: This gure compares several denitions of dt vs. with s = 5. The points show my estimate of dt from numerical simulations with two dierent values of b. The solid lines show the two estimates of dt from Eq. 2 and 4. As in Fig. 6, the denition of dt becomes problematic when becomes large enough. This is illustrated in more detail in Fig. 7

FIG. 7: The plots on the left illustrates vc vs d for two cases. The plots on the right show the force prole corresponding to these two cases. The black dot on the force prole indicates vc at the edge of the vc d regime (dened by the intersection of the two curves shown on the left. Note that for = 2.6, the shape of the force prole is quite distorted so that there are almost two dierent linear regimes.

7 Here are a few observations based on Figs. 4-7. In the regime where is small enough, Eq. 4 provides an excellent estimate of the edge of the vc d regime. By comparing Figs. 4 and 6, we can see that small enough depends on s. I think it should be possible to make that more quantitative by dening an eective saturation that combines s and . Something like this is done in the Gibble paper where they dene p =
s , 1+4(/)2

but their p

also contains the Doppler shift (and it would also have the Zeeman shift in our model) so it seems like it is not the right parameter for our purposes. Figure 6 also illustrates that when the eective saturation is too small (i.e. some combination of low s and large ) the analysis presented in Fig. 1 breaks down. There is still a regime where vc d and, at large d, vc still grows ever more slowly with d, but in between there is a new regime. This can be seen most clearly in the lower half of Fig. 7. It is not clear how to dene dt in these cases. This analysis also shows that the intuitive picture where the edge of the d6 scaling regime corresponds to a capture velocity on the peak of the force prole does not hold in all cases. Surprisingly, Eq. 4 still seem to provides a good estimate of dt as long as the force prole has not become too distorted. This can be seen most clearly in the case s = 14, = 3 in Figs. 4 and 5.

You might also like