逆向基體之高功率金氧半場效應電晶體

Abstract

傳統溝槽式閘極高功率金氧半場效電晶體的通道長度受限於源極與汲極的穿透崩潰(punch through)現象，取決於汲極端承受高壓時的空乏區寬度。此外，元件的臨界電壓由通道中最高濃度決定，依傳統製程其基體是從晶片表面以擴散方式形成，因此靠源極端濃度較高而靠汲極端濃度較低。如果要降低臨界電壓而降低源極端的基體濃度，勢必使得汲極端的基體濃度更低，在固定電壓規格下，通道長度必須加長，除使通道阻抗增加外，也會使得製程困難度增加。 本論文提出逆向基體濃度分佈的構想，利用高能量離子植入方式，使得靠汲極端的基體濃度較高而靠近源極端的基體濃度較低。簡化模型計算結果顯示，由於汲極端的濃度較高，空乏區寬度大幅減小，可縮短通道長度，而閘極溝槽深度減少將有利於元件微小化。另外汲極端的高濃度基體當元件工作在飽和區時幾乎都空乏掉，臨界電壓由飽和點的基體濃度決定，所以可同時達成降低臨界電壓以及縮短通道長度的目標，導通電阻也能隨之降低。當臨界電壓降低以及通道長度縮短後，也預期可以得到較大的驅動電流。 從實際測量發現下列重要現象：(1)臨界電壓隨著溝槽深度往下降其標準誤差值會稍稍變大，但此變異尚在可接受範圍內。(2)為了得到更小之導通電阻除縮減溝槽深度外需搭配漂移區厚度之減薄，但如此一來崩潰電壓會降低，於逆向基體之元件設計中上層n－漂移區厚減去溝槽深度需大於0.8um才可得崩潰電壓在30V以上，因此如何取捨端看我們對元件特性之需求為何。(3) 當溝槽深度變淺連帶通道長度變小時，會產生汲極偏壓導致能障下降(DIBL)的現象，在汲極加大偏壓時逆向基體濃度分佈的元件在閘極0V時的汲極漏電流會稍大，不過依然在高功率元件容許的範圍內。 我們所設計的逆向基體分佈元件在元件的垂直方向微小化、導通電流、導通電阻、開關特性上皆可得到不錯的增進和改善。在臨界電壓約1V、崩潰電壓大於30V的規格下，溝槽深度可縮為0.7um、導通電阻降至0.36mohm-cm2(較傳統製程下降約40％)、開關特性之Rds,on•QGS更是比傳統製程低了約70%。因此對溝槽式閘極高功率金氧半場效電晶體的元件特性改善，我們提供了一創新之設計方式。高能量基體植入技術也可以免除傳統基體所需的高溫長時間基體摻雜驅入步驟，對元件可靠度應有幫助。 本論文對基體濃度分佈提出了全新的最佳化調整方法，能有效的縮短通道長度，降低導通電阻，也可以使通道長度和臨界電壓的設計分離。實際應用如加上水平方向縮減元件寬度、增加元件密度，在降低導通電阻上將可得到加成效果。

The channel length of traditional trench gate power MOSFET is confined to punch-through breakdown of source/drain, and it is decided by the depletion width when large drain voltage is applied. Besides, the peak base concentration in channel determines the threshold voltage of the device. The p-body region of traditional process is formed by a boron diffusion from the wafer surface. Thus, the concentration near source side is high and the concentration near drain side is low. If lowering the p-body concentration near source side to reduce the threshold voltage, it will make the p-body concentration near drain side even lower. At the specification for a fixed voltage, the channel length must be increased. Except for increasing the channel resistance, it would make the process harder. In this thesis we invent the retrograde-body profile. It utilizes the high energy implantation to obtain that the body concentration near drain side is high and is low near source side. The calculation results of simple model exhibit that the depletion width is diminished substantially due to the high body concentration near drain side. Therefore we can shorten the channel length and it would be beneficial for device dimension scaling down. In addition, the high body concentration near drain side has almost been depleted as the device in saturated mode. The threshold voltage is determined by the body concentration at saturated-point, so we can decrease the threshold voltage, shorten the channel length and the on-resistance would be lower at the same time. After that we would obtain bigger driving-current. We find some important phenomena from actual measurement as listed below: (1)the standard deviation of threshold voltage would be bigger with the lower trench depth, but the variation is acceptable.(2)Except for diminishing trench depth, we should decrease the thickness of drift region for lower on-resistance. But the breakdown voltage would be lower. In the design of retrograde body, if we want to obtain the breakdown more than 30V, the difference of up-layer n□ drift region and trench depth should be larger than 0.8□m. Thus, the trade-off of on-resistance and breakdown voltage is decided by the need of device performance.(3)We find the phenomenon of drain induce barrier lowing(DIBL) as the channel length is shorter due to the shallower trench depth. When drain bias a large voltage, the leakage current of the retrograde-body device at gate biasing 0V would be larger, but the value is still in the allowable range of power MOSFET. The device of retrograde body profile can obtain nice improvement in device scaling down of perpendicular direction, driving current, on-resistance, and the switching performance. At the specifications for the threshold voltage is 1V and the breakdown voltage is larger than 30V, the trench depth can be shortened to 0.7□m, the on-resistance would be 0.36m□□cm2 (it is lower about 40% than standard process), and the Rds,on•QGS of switching performance is lower about 70% than standard process. So we provide a new design for the improvement of trench gate power MOSFET. Moreover, the high energy implantation would also avoid the drive-in process of body implantation with high temperature and long time which the standard process has to do, and this would be good for the reliability of device. If combined with the optimum drift region thickness, we can decrease the on-resistance about 40% than traditional process. Furthermore, the switching performance has been improved, too. This thesis develops a new modulation on optimum doping distribution of body of trench gate power MOSFET. It can shorten the channel length, lower the on-resistance effectively, and also decouple the design of channel length and threshold voltage. In practical application, if combining it with another method which decreasing the cell pitch to increasing cell density, we would obtain greater effect on reducing on-resistance.