This paper describes the direct measurement of the relation between the worked layers in quartz-crystal plates and their effect on the frequency stability. The rectangular fundamental 5-MHz A T-cut specimen plates lapped and polished with emery powder and cerium oxide, respectively, were placed in gap-type holders and con nected with oscillator circuits, and their frequency stabilities were measured. The frequency stabilities are evaluated by the fractional frequency deviation to the nominal frequency ratio, power spectral density, and the Allan variance. The main results are: 1) The frequency drift due to the worked layer is dealt with as a relaxation process. It can be approximated by a sum of one to three terms of an exponential function.2) The depth of the worked layer has a large effect not only on the long-term frequency stability but on the medium term in the Fourier frequency range of 10~3 to 10~" 7 Hz. 3) It does not have a dominant effect on the short-term stability with an averaging time of less than 10 s. I. INTRODUCTION I T IS CONSIDERED that the worked layer in the quartzcrystal blank caused by the lapping and polishing process has a large influence on frequency stability, when the crystal units are used in oscillators. With respect to this layer, many studies have been made from the viewpoint of the precision polishing of the thin quartz plate [1]. Vig and others show in detail the effect of lapping and polishing on the quartz plates and their cleaning methods by examining with the electron scanning microscope and surface analyzing instruments, etc. [2], [3]. The authors have also reported on the polished layer [4] and the refractive index distribution normal to the polished surface of the fused quartz observed by ellipsometry [5]. However, most of the studies have placed their emphasis on the obser vation of the surface roughness and the measurement of the depth of the layer. Measurement has not been made on the correlation between the frequency stability and the processing condition. In this paper, the authors try to clearly show such a correlation. Crystal units were made by using the same gap-type holders and crystal blanks having the same geometry and angle of cut, finished with various grain-size abrasives. Crystal oscillators were made by using the same type oscillator circuit and ovens for each specimen. The long-, medium-, and short-term fre quency stabilities were measured with the measure oidf/f, the power spectral density, and the Allan variance, respectively, Manuscript received July 7, 1980. The authors are with the Research f a 5 MHz x.r U.9mm ye= 0.334 mm Y z. = 16.0 mm β= 35 β 15'